diff --git a/README.md b/README.md index bfb18e3..531fd6a 100644 --- a/README.md +++ b/README.md @@ -6,11 +6,24 @@ Thesis defended on December 1st, 2022. This thesis investigates how to robustly estimate time-varying functional connectivity (TVFC), a construct in neuroimaging research that looks at changes in functional coupling (correlation between time series) between brain regions during a functional magnetic resonance imaging (fMRI) scan, and how it can be used as a lens through which to study depression as a functional disorder. -Unfortunately, the field of TVFC is still riddled with uncertainty, especially regarding its estimation. This is mainly due to the absence of a ground truth. Without resolving this first, the value of any study, including this depression study, is significantly undermined and conclusions made therein less trustworthy. Therefore, I propose a novel and principled method for estimating TVFC, based on the Wishart process (WP), a covariance matrix stochastic process that has recently become computationally tractable, and introduce a comprehensive benchmarking framework based on machine learning principles to make sure it performs better than existing methods in the field. These benchmarks include simulations, subject phenotype prediction, test-retest studies, brain state analyses, external task prediction, and a range of qualitative method comparisons. Furthermore, I introduce a benchmark based on cross-validation, that can be run on any data set. The WP model is found to outperform other common estimation methods, such as sliding-windows (SW) approaches and dynamic conditional correlation (DCC). - -Returning to the depression study, several differences are found between depressed and healthy control cohorts. The study is run on thousands of participants from the UK Biobank, yielding unprecedented statistical power and robustness. I investigate connectivity between individual brain regions as well as functional networks (FNs). Depressed participants show decreased global connectivity, and increased connectivity instability (as measured by the temporal characteristics of estimated TVFC). By defining multiple depression phenotypes, I find that brain dynamics are affected especially when patients have been professionally diagnosed or indicated to be depressed during their fMRI scan, but were less or not at all affected based on self-reported past instances and genetic predisposition. I show that choosing a different TVFC estimation method would have changed our scientific conclusions. This sensitivity to seemingly arbitrary researcher choices highlights the need for robust method development and the importance of community-approved benchmarking. +Unfortunately, the field of TVFC is still riddled with uncertainty, especially regarding its estimation. +This is mainly due to the absence of a ground truth. +Without resolving this first, the value of any study, including this depression study, is significantly undermined and conclusions made therein less trustworthy. +Therefore, I propose a novel and principled method for estimating TVFC, based on the Wishart process (WP), a covariance matrix stochastic process that has recently become computationally tractable, and introduce a comprehensive benchmarking framework based on machine learning principles to make sure it performs better than existing methods in the field. +These benchmarks include simulations, subject phenotype prediction, test-retest studies, brain state analyses, external task prediction, and a range of qualitative method comparisons. +Furthermore, I introduce a benchmark based on cross-validation, that can be run on any data set. +The WP model is found to outperform other common estimation methods, such as sliding-windows (SW) approaches and dynamic conditional correlation (DCC). + +Returning to the depression study, several differences are found between depressed and healthy control cohorts. +The study is run on thousands of participants from the UK Biobank, yielding unprecedented statistical power and robustness. +I investigate connectivity between individual brain regions as well as functional networks (FNs). +Depressed participants show decreased global connectivity, and increased connectivity instability (as measured by the temporal characteristics of estimated TVFC). +By defining multiple depression phenotypes, I find that brain dynamics are affected especially when patients have been professionally diagnosed or indicated to be depressed during their fMRI scan, but were less or not at all affected based on self-reported past instances and genetic predisposition. +I show that choosing a different TVFC estimation method would have changed our scientific conclusions. +This sensitivity to seemingly arbitrary researcher choices highlights the need for robust method development and the importance of community-approved benchmarking. I wrap up this thesis with a discussion of results and how this style of work fits into the bigger picture of neuroscientific research, reflect on what has been learned about depression, and posit promising directions for future work. + ## Source code for experiments https://github.com/OnnoKampman/neuro-dynamic-covariance @@ -25,6 +38,8 @@ This document is generated using `Latexmk` version 4.77 and Biber version 2.17. You may need to clear the cache by running `rm -rf $(biber --cache)` if you encounter issues with compiling. +I use Skim to view the generated PDF. + ## Inspiration [1] https://github.com/kks32/phd-thesis-template diff --git a/appendix/03_extra_benchmarking_results.tex b/appendix/03_extra_benchmarking_results.tex index 7d3dc61..534c5c1 100644 --- a/appendix/03_extra_benchmarking_results.tex +++ b/appendix/03_extra_benchmarking_results.tex @@ -331,3 +331,66 @@ \subsection{Trivariate} Means and standard deviations are shown across $T = 200$ trials. }\label{fig:results-sim-d3s-200-no-noise-all-correlation-matrix-RMSE} \end{figure} + + +\clearpage +\section{HCP: Extra subject measure predictions}\label{appendix:hcp-more-results} +%% + + +\begin{figure}[h] + \centering + \includegraphics[width=\textwidth]{fig/hcp/d15/subject_measure_prediction/social-emotional/morphometricity_all_TVFC_summary_measures} + \caption{ + HCP benchmark subject social-emotional measures prediction morphometricity scores (with standard error). + Run on TVFC summary measures of mean (top), variance (middle), and rate-of-change (bottom row). + sFC is added for reference to the TVFC mean plot. + }\label{fig:results-subject-measures-prediction-social-emotional} +\end{figure} + + +\begin{figure}[h] + \centering + \includegraphics[width=\textwidth]{fig/hcp/d15/subject_measure_prediction/other/morphometricity_all_TVFC_summary_measures} + \caption{ + HCP benchmark subject other measures prediction morphometricity scores (with standard error). + Run on TVFC summary measures of mean (top), variance (middle), and rate-of-change (bottom row). + sFC is added for reference to the TVFC mean plot. + }\label{fig:results-subject-measures-prediction-other} +\end{figure} + + +%% +\clearpage +\section{HCP: Brain states}\label{appendix:hcp-brain-states} +%% + + +\begin{figure}[ht] + \centering + \includegraphics[width=\textwidth, trim={0.0cm 0cm 0.0cm 0cm}, clip]{fig/hcp/d15/brain_states/k03/brain_states_SVWP_joint} + \caption{ + Brain states extracted from SVWP TVFC estimates on HCP data ($D = 15$ time series). + Brain states across the four scans look very similar. + }\label{fig:hcp-results-brain-states-svwp} +\end{figure} + + +\begin{figure}[ht] + \centering + \includegraphics[width=\textwidth, trim={0.0cm 0cm 0.0cm 0cm}, clip]{fig/hcp/d15/brain_states/k03/brain_states_DCC_joint} + \caption{ + Brain states extracted from DCC TVFC estimates on HCP data ($D = 15$ time series). + Brain states are low contrast in general, and the four scans look very similar. + This aligns with our intuition when looking at the DCC model predictions. + }\label{fig:hcp-results-brain-states-dcc} +\end{figure} + + +\begin{figure}[ht] + \centering + \includegraphics[width=\textwidth, trim={0.0cm 0cm 0.0cm 0cm}, clip]{fig/hcp/d15/brain_states/k03/brain_states_SW_cross_validated} + \caption{ + Brain states extracted from SW-CV TVFC estimates on HCP data ($D = 15$ time series). + }\label{fig:hcp-results-brain-states-sw-cv} +\end{figure} diff --git a/ch/1_Introduction/0_Introduction.tex b/ch/1_Introduction/0_Introduction.tex index fb59b6e..f7b771f 100644 --- a/ch/1_Introduction/0_Introduction.tex +++ b/ch/1_Introduction/0_Introduction.tex @@ -3,7 +3,7 @@ \chapter{Introduction}\label{ch:introduction} \info[inline]{Paragraph: Direct summary of what this thesis is about.} This thesis introduces novel approaches for robust estimation of \gls{tvfc} from \gls{fmri} neuroimaging data. -\Gls{tvfc} is a construct that studies the time-varying nature of interaction between brain regions. +\Gls{tvfc} is a construct that studies the time-varying nature of the interaction between brain regions. Estimation approaches are compared and evaluated to existing ones through a proposed benchmarking framework. % Afterwards, we use the best performing method to investigate how brain dynamics differ between depressed and healthy (control) subjects. @@ -29,7 +29,7 @@ \chapter{Introduction}\label{ch:introduction} Some look at small circuits of neurons, some look at whole-brain structures, some investigate emergent brain waves, and some try to model human brain function with animal or computational models and speculate how these relate to human brain function (sometimes referred to as comparative cognition). Neuroscientists have long considered that the brain is modular in its organization~\parencite{Prinz2006}. Many would therefore specialize in a particular brain \gls{roi} to study its function and relationship to other regions. -Even though the view of the brain as a collection of entirely distinct modules has been abandoned, this tradition persists to some degree, and is often still (approximately) valid~\parencite{Genon2018}. +Even though the view of the brain as a collection of entirely distinct modules has been abandoned, this tradition persists to some degree and is often still (approximately) valid~\parencite{Genon2018}. Moreover, contemporary neuroscientists have argued for a more sophisticated view of \emph{hierarchy} in such modularity. % The division between psychology and neuroscience has blurred and merged in recent years. @@ -60,20 +60,20 @@ \chapter{Introduction}\label{ch:introduction} \info[inline]{Paragraph: Explain our point-of-view of the human brain.} The particular point-of-view of the human brain in this work is the following. -We abstract away implementational level details, and consider the human brain as a complex, dynamic system organ that is divided into distinct yet interacting regions.\footnote{Brain regions and components (or `nodes', borrowing from graph theoretic jargon) can be defined in several ways, which will be addressed later.} +We abstract away implementational level details and consider the human brain as a complex, dynamic system organ that is divided into distinct yet interacting regions.\footnote{Brain regions and components (or `nodes', borrowing from graph theoretic jargon) can be defined in several ways, which will be addressed later.} We will not go into depth on any neurobiological signatures of behavior and disorders. % This characterization is based on a historical trend. Early modern neuroscientists discovered that the brain can be \emph{segregated} into distinct cortical (and subcortical) regions with distinct functions. Naturally, it followed that neuroscientists became interested in how these regions connect and communicate with one another. -This is often referred to as the `functional' architecture of the brain, in contrast with the better understood \emph{structural} or \emph{anatomical} brain architecture. -Higher-order cognition and complex behavior is made possible by the spatiotemporal integration, re-organization, and segregation of brain regions~\parencite{Deco2011}. +This is often referred to as the `functional' architecture of the brain, in contrast with the better-understood \emph{structural} or \emph{anatomical} brain architecture. +Higher-order cognition and complex behavior are made possible by the spatiotemporal integration, re-organization, and segregation of brain regions~\parencite{Deco2011}. Over the years a plethora of studies has painted a picture of the brain as a complex, distributed, and adaptive network of anatomical and functional segregation and integration that re-organizes itself at different time scales to process information and address a given task. Brain constituent parts also exhibit complex system structures, such as modular and hierarchical topology~\parencite{Meunier2009, Deco2015}. Human brains have been found to rely on higher degrees of synergistic interactions compared to nonhuman primates, highlighting their importance to complex cognition~\parencite{Luppi2022}. % -Anatomical organization and connectivity in the brain has been studied by a range of imaging methods, including \gls{mri}, which uses powerful magnetic fields and radio waves to produce high resolution images of the inside of the body, and \gls{dti}, which images axons in white matter tracts. -Functional interactions, often referred to as `connectivity' too, can be characterized using neuroimaging methods such as \gls{fmri}~\parencite{Soares2016}, \gls{pet}, which requires an injection of positron emitting isotopes, \gls{nirs}, \gls{eeg}, and \gls{meg}~\parencite{Rossini2019}. +Anatomical organization and connectivity in the brain has been studied by a range of imaging methods, including \gls{mri}, which uses powerful magnetic fields and radio waves to produce high-resolution images of the inside of the body, and \gls{dti}, which images axons in white matter tracts. +Functional interactions, often referred to as `connectivity' too, can be characterized using neuroimaging methods such as \gls{fmri}~\parencite{Soares2016}, \gls{pet}, which requires an injection of positron-emitting isotopes, \gls{nirs}, \gls{eeg}, and \gls{meg}~\parencite{Rossini2019}. Crucially, such connectivity is usually not directly observed, and needs to be estimated. \info[inline]{Paragraph: Discuss confusing terminology of `connectivity'.} diff --git a/ch/1_Introduction/1_Functional_connectivity.tex b/ch/1_Introduction/1_Functional_connectivity.tex index 9f5c2d8..87215f8 100644 --- a/ch/1_Introduction/1_Functional_connectivity.tex +++ b/ch/1_Introduction/1_Functional_connectivity.tex @@ -8,7 +8,7 @@ \section{Functional connectivity} Such connectivity depends on statistical dependencies between activity or \emph{activation} in brain regions, which can be measured by various neuroimaging modalities, most commonly \gls{fmri}, \gls{eeg}~\parencite[e.g.][]{Tagliazucchi2012, Chang2013}, and \gls{meg}~\parencite[e.g.][]{Baker2014, Vidaurre2018}.\footnote{As discussed later, modeling dependencies between random variables is also an important problem in machine learning and statistics.} Sometimes multiple concurrent modalities are used, such as \gls{fmri} and \gls{eeg} (but it is not possible to combine \gls{fmri} with \gls{meg}). % -Neuroimaging methods can be divided into those that capture \emph{structural} and those capture \emph{functional} signals. +Neuroimaging methods can be divided into those that capture \emph{structural} and those that capture \emph{functional} signals. The former aims to map out the location and molecular properties of brain tissue. The latter aims to capture more dynamic signals, related to the activity of brain tissue. Structural and functional (including \gls{fc}) analyses are complementary in building a holistic understanding of the brain, as they capture complementary (and disparate) information~\parencite{Lang2012}. @@ -59,7 +59,7 @@ \section{Functional connectivity} Preprocessing pipelines, which take raw \gls{fmri} data and output node time series of interest, impact study results and conclusions~\parencite{Caballero-Gaudes2017}. As such, preprocessing steps also heavily influence subsequently extracted \gls{fc}~\parencite{Aquino2022}. -An example of an impactful preprocessing step is \gls{gsr}, which can improve spatial localization of networks, but can introduce artificial anticorrelations into the data~\parencite{Murphy2009}. +An example of an impactful preprocessing step is \gls{gsr}, which can improve the spatial localization of networks, but can introduce artificial anticorrelations into the data~\parencite{Murphy2009}. \info[inline]{Paragraph: Describe fMRI functional connectivity.} In the context of \gls{fmri}, `connectivity' is typically (i.e.~traditionally) characterized as the Pearson correlation coefficient between brain region \gls{bold} measurement component time series~\parencite{Zalesky2012}. diff --git a/ch/1_Introduction/2_Functional_networks.tex b/ch/1_Introduction/2_Functional_networks.tex index f6509d1..340207a 100644 --- a/ch/1_Introduction/2_Functional_networks.tex +++ b/ch/1_Introduction/2_Functional_networks.tex @@ -7,7 +7,7 @@ \section{Functional networks}\label{sec:functional-brain-networks} Cognitive tasks are not just performed by isolated brain regions, but rather by such networks, i.e.~linked collections of brain regions~\parencite{Bressler2010}. % These networks have been identified from \gls{fmri} \gls{bold} signals from scanning subjects presented with external tasks. -Moreover, it has been shown that such large-scale networks exist at rest, and that these strongly resemble those found in task paradigms~\parencite{Smith2009}.\footnote{In neuroimaging, `rest' usually refers to the absence of an external stimulus or task, leaving mental activity relatively unconstrained. Signals measured at rest are sometimes referred to as `intrinsic' or `spontaneous', but rather confusingly in neuroimaging this sometimes refers to an actual signal of interest and sometimes to just noise.} +Moreover, it has been shown that such large-scale networks exist at rest and that these strongly resemble those found in task paradigms~\parencite{Smith2009}.\footnote{In neuroimaging, `rest' usually refers to the absence of an external stimulus or task, leaving mental activity relatively unconstrained. Signals measured at rest are sometimes referred to as `intrinsic' or `spontaneous', but rather confusingly in neuroimaging this sometimes refers to an actual signal of interest and sometimes to just noise.} In such cases, these networks are referred to as \glspl{rsn} or \glspl{icn}. However, this name can cause confusion, because networks with similar extents can be found during task executions. Therefore, we opt to simply use the term `functional network'~\parencite[FN; see also][]{Finn2021}. diff --git a/ch/1_Introduction/5_Functional_connectivity_and_depression.tex b/ch/1_Introduction/5_Functional_connectivity_and_depression.tex index ea52f14..982df64 100644 --- a/ch/1_Introduction/5_Functional_connectivity_and_depression.tex +++ b/ch/1_Introduction/5_Functional_connectivity_and_depression.tex @@ -7,7 +7,7 @@ \section{Functional connectivity and depression}\label{sec:fc-depression} But what do we mean by depression? How does depression affect the brain? And more specifically, how does depression affect \gls{fc} in the brain? -Can \gls{fc} be used to assign credit or discredit to varies theories of depression? +Can \gls{fc} be used to assign credit or discredit to various theories of depression? In this thesis we argue that depression is a particularly good disease to study through the lens of \gls{tvfc}. \gls{fc} has the potential of offering new diagnostic value in neuropsychiatric disorders, where typical \gls{fmri} activations are often small~\parencite{Fornito2012}. The rest of this section reviews the current understanding of what depression is, why it is important to study it, what subtypes exist, what symptoms typically occur, how it affects the brain, and how it affects \gls{fc} in the brain. @@ -33,7 +33,7 @@ \subsection{What is depression?}\label{subsec:depression} Two common ways of diagnosing (i.e.~categorizing or classifying) depression are based on standard diagnostic (category-based) frameworks: the \gls{dsm} and the \gls{icd}.\footnote{The \gls{icd} defines mental disorders as ``clinically recognizable set of symptoms or behaviors associated in most cases with distress and with interference with personal functions''~\parencite{WHO1992}.} The \gls{dsm} criteria for major depression are shown in Box~\ref{box:depression}. We shall return to these criteria in \cref{subsec:cohort-stratification} for defining participant cohorts. -The \gls{icd} criteria are similar, requiring three criteria to be met: persistent low mood or loss of interests; happening for most of the time on most days for at least two weeks; and experiencing four of more other symptoms like disturbed sleep, concentration, appetite, guilt, self-blame, low self-esteem, agitation, and/or suicidal thoughts. +The \gls{icd} criteria are similar, requiring three criteria to be met: persistent low mood or loss of interests; happening for most of the time on most days for at least two weeks; and experiencing four or more other symptoms like disturbed sleep, concentration, appetite, guilt, self-blame, low self-esteem, agitation, and/or suicidal thoughts. More broadly, depression endophenotypes\footnote{Endophenotypes, or `intermediate phenotypes', refer to heritable traits used to more robustly define behavioral symptoms into phenotypes. Similar terms are `biological marker' or \emph{biomarker} and `subclinical trait', although these are typically not used to refer to genetic components.} and cardinal symptoms include anhedonia, anergia, anxiety, rumination, changes in appetite and sleep patterns, strong and persistent feelings of guilt and grief, and, most tragically, self-injury~\parencite{Goldstein2014, Pizzagalli2014}. Although core symptoms are typically present, depression is not a consistent syndrome with a fixed set of symptoms. In fact, \textcite{Fried2015} found over 1,000 unique symptoms in a cohort of about 3,700 patients~\parencite[see also][]{Fried2015b}. @@ -78,14 +78,14 @@ \subsection{What is depression?}\label{subsec:depression} Genetic risk for \gls{mdd} is polygenic, meaning a variety of genes are involved, and the exact mechanisms are yet to be uncovered~\parencite{Hyman2014}. This is likely due to the heterogeneity of depressive symptoms as well. Moreover, much of depression risk may be due to other genetic factors. -Higher overall cognitive function, for example, could lead to higher socioeconomic status, which in turn could lead to healthier diet and increased sense of safety and control in the world (which in turn have been linked to lower depression risk). +Higher overall cognitive function, for example, could lead to higher socioeconomic status, which in turn could lead to a healthier diet and an increased sense of safety and control in the world (which in turn have been linked to lower depression risk). Genetic risk has often been described as influencing cognitive biases and thus \emph{resilience} to stressors. The most important take-away from genetic studies is that genes are about vulnerability and resilience to depression and not about inevitability. % In a large cohort study, \textcite{Garcia-Gonzalez2017} failed to find a robust genetic contribution to treatment response. \info[inline]{Paragraph: Describe cognitive effects of MDD.} -Before looking at the brain and impacts of \gls{mdd} on \gls{fc}, we give an overview of changes in cognition and behavior. +Before looking at the brain and the impacts of \gls{mdd} on \gls{fc}, we give an overview of changes in cognition and behavior. These will be referred to in \cref{sec:ukb-discussion}. Deficits in memory systems, attention, learning, processing speed, and decision-making are common among \gls{mdd} patients. \textcite{Rock2014} found especially executive function\footnote{In neuroscience, \textbf{executive function} generally refers to functions related to planning, focus, sticking with instructions, and multi-tasking~\parencite{Banich2009}.}, memory, and attention affected by \gls{mdd}. @@ -113,11 +113,11 @@ \subsection{What is depression?}\label{subsec:depression} Importantly, knowing what works to treat depression and what does not can also shed light on what the condition entails. Treatment options for \gls{mdd} generally are pharmacological intervention and/or one of the many types of (psycho)therapy~\parencite{Otte2016}. Antidepressant medication is usually meant to increase the concentration of a certain neurotransmitter in the brain, most commonly serotonin. -In the case of serotonin these antidepressants are called \gls{ssri}. +In the case of serotonin, these antidepressants are called \gls{ssri}. Interest in therapies using psilocybin~\parencite{Carhart-Harris2016, Luppi2021, Daws2022, Singleton2022} and ketamine~\parencite{Krystal2019, Kotoula2021} has spiked in recent years as well, but the jury is still out on the efficacy of such treatments. Depression is seen as decreasing brain state entropy, which psychedelics can alleviate, as it lies on the opposite side of a spatiotemporal dynamics spectrum~\parencite{Vohryzek2022}. % -There is an intense debate in society at large how best to treat depression, some pointing predominantly to medication, and others to social approaches. +There is an intense debate in society at large on how best to treat depression, some pointing predominantly to medication, and others to social approaches. Part of this controversy stems from the fact that depression is still poorly understood, and it is likely grouping together many types of depression. For example, medication seems to work well for some but has no effect on others. The latter are sometimes called `treatment-resistant', but they may well suffer from a different subtype of depression, where neurobiologically distinct domains are collapsed into a simple diagnostic index. @@ -131,9 +131,9 @@ \subsection{Depression and neuroimaging}\label{subsec:fc-neuroimaging} %% Neuroimaging has the potential to offer unique insight into the mechanisms of depression. -A lot of neuroimaging can be considered as a mapping exercise. +A lot of neuroimaging can be considered a mapping exercise. Starting from how the brain is anatomically characterized, to mapping literal wiring diagrams such as white matter tracts, but also including the \gls{fc} mapping exercise as we study here. -Once we have a good map, it is natural to ask whether we can find individual landmarks that are unique to disease. +Once we have a good map, it is natural to ask whether we can find individual landmarks that are unique to a disease. Structural scans have shown that the neuroanatomy in depressed patients is affected~\parencite{Drevets2000}. Multiple brain region volumes are either increased or decreased~\parencite{Sacher2012, Schmaal2020}. @@ -145,7 +145,7 @@ \subsection{Depression and neuroimaging}\label{subsec:fc-neuroimaging} A recent large (1809 participants) study found very modest predictive power of (univariate) neuroimaging modalities (\gls{mri}, \gls{dti}, \gls{rs-fmri}, and \gls{tb-fmri}) of \gls{mdd}~\parencite{Winter2022}. They found environmental factors such as social support and childhood maltreatment to have much more predictive power. Similar sentiments were echoed in \textcite{Nour2022}. -At present, neuroimaging plays little to no role in clinical decision making~\parencite{Kapur2012}. +At present, neuroimaging plays little to no role in clinical decision-making~\parencite{Kapur2012}. %% \subsection{Functional connectivity in psychiatric disorders}\label{subsec:fc-depression} @@ -165,10 +165,10 @@ \subsection{Functional connectivity in psychiatric disorders}\label{subsec:fc-de Another data point that highlights the promise of studying \gls{mdd} through the perspective of \gls{fc} is that such connectivity and networks have been found to re-normalize after anti-depressants treatment. \info[inline]{Paragraph: Discuss FNs in neurological and psychiatric disorders.} -\Glspl{fn} are often affected with neuropsychiatric disorders, even if their individual brain region constituents appear normal. +\Glspl{fn} are often affected by neuropsychiatric disorders, even if their individual brain region constituents appear normal. Such disorders are therefore increasingly studied as \emph{network} disorders~\parencite[see][for a review on depression]{Mulders2015}. Instead of a single brain region not functioning properly, there is an aberration in the integration and segregation of brain regions. -These changes in \gls{fn} are believed to contribute or be caused by cognitive changes from mental illness. +These changes in \gls{fn} are believed to contribute to or be caused by cognitive changes from mental illness. Many mental illness conditions have been postulated to occur with large-scale disruptions, driven by neurotransmitter dysfunction, of whole-brain systems. Even though whole-brain \glspl{fn} are found to be highly similar across groups with or without a range of mental illnesses, the subtle differences that \emph{do} occur are meaningful in the sense that they are predictive of diagnosis~\parencite{Spronk2020}. This makes intuitive sense as well: a piano does not need major disruption to ruin a classical piece, one key being out of tune is sufficient~\parencite[see also][for a discussion of small effect sizes]{Paulus2019}. @@ -180,24 +180,21 @@ \subsection{Functional connectivity in psychiatric disorders}\label{subsec:fc-de The exact makeup of these networks varies across studies. A rough overview of each network is provided here (the more precise implementational details will be provided in \cref{subsec:ukb-fn-analysis}). -The \gls{dmn} primarily consists of \gls{mpfc} and \gls{pcc}, as well as the (para)hippocampal areas, precuneus (cortex), and angular gyrus~\parencite{Andrews-Hanna2010}. -It is often described as the neurological basis for `the self', and is attributed functions like self-referential thinking~\parencite{Sheline2009}, cognitive flexibility~\parencite{Vatansever2016}, mind-wandering, memory processing and rumination, theory of mind, emotion regulation, and as storage of autobiographical information. -It it connected to the \gls{amg} and \gls{hpc}~\parencite{Andrews-Hanna2014}. -\unsure{How should we define our DMN?} +The \gls{dmn} primarily consists of the \gls{mpfc} and \gls{pcc}, as well as the (para)hippocampal areas, precuneus (cortex), and angular gyrus~\parencite{Andrews-Hanna2010}. +It is often described as the neurological basis for `the self' and is attributed to functions like self-referential thinking~\parencite{Sheline2009}, cognitive flexibility~\parencite{Vatansever2016}, mind-wandering, memory processing and rumination, theory of mind, emotion regulation, and as storage of autobiographical information. +It is connected to the \gls{amg} and \gls{hpc}~\parencite{Andrews-Hanna2014}. The \gls{cen} primarily consists of the lateral \gls{pfc}, posterior parietal cortex (PPC), \gls{dlpfc} (especially middle frontal gyrus), \gls{dmpfc}, and posterior parietal regions~\parencite{Rogers2004}. It is associated with cognitive processes and functions, like working memory and attention. -\unsure{How should we define our CEN?} The \gls{sn} primarily consists of the \gls{ai} and (dorsal) \gls{acc}, with some adding the \gls{amg}, frontoinsular cortex, temporal poles, and striatum~\parencite{Seeley2007, Menon2010, Beck2016}. -\unsure{How should we define our SN?} The \gls{sn} is a key network in cognitive flexibility~\parencite{Dajani2015}. \info[inline]{Paragraph: Discuss TVFC in neurological and psychiatric disorders.} What about the dynamics of \gls{fc}? The promise of \gls{tvfc} has been highlighted more recently as well in neurodegenerative conditions~\parencite{Filippi2019}. More relevant information is contained in \gls{tvfc} compared to \gls{sfc}. -\gls{tvfc} may be especially relevant for dynamical brain disorders like schizophrenia~\parencite{Jin2017}. +\gls{tvfc} may be especially relevant for dynamic brain disorders like schizophrenia~\parencite{Jin2017}. \info[inline]{Paragraph: Discuss graph topology in neurological and psychiatric disorders.} Graph topology and network neuroscience have also been suggested to shed more light on neurological and psychiatric conditions~\parencite{Fornito2013}. diff --git a/ch/2_Robust_estimation_of_TVFC/1_Established_methods_and_baselines.tex b/ch/2_Robust_estimation_of_TVFC/1_Established_methods_and_baselines.tex index cf3e276..8919953 100644 --- a/ch/2_Robust_estimation_of_TVFC/1_Established_methods_and_baselines.tex +++ b/ch/2_Robust_estimation_of_TVFC/1_Established_methods_and_baselines.tex @@ -58,14 +58,14 @@ \subsection{Sliding-windows functional connectivity}\label{subsec:sliding-window The benefits of using \gls{sw} include that it is well-established, computationally cheap, and simple to implement. \info[inline]{Paragraph: Discuss variations on sliding-windows functional connectivity.} -Many variations on this method exist and how to best implement \gls{sw} methods has been studied extensively~\parencite[see e.g.][]{Mokhtari2019, Vergara2019}. +Many variations of this method exist and how to best implement \gls{sw} methods has been studied extensively~\parencite[see e.g.][]{Mokhtari2019, Vergara2019}. One popular way to reduce sensitivity to outliers is \emph{tapered} \gls{sw}. This approach assigns less weight to observations further away from the center of the window~\parencite{Allen2014, Lindquist2014}. Effectively this just changes the shape of the window from rectangular/square into e.g.~a Gaussian curve. % Rules of thumb have been established regarding window length~$w$ and necessary data preprocessing steps. % -The lack of consistency across studies in implementation of \gls{sw} is sub-optimal as it makes comparison across studies harder. +The lack of consistency across studies in the implementation of \gls{sw} is sub-optimal as it makes comparison across studies harder. Furthermore, stacking heuristics does not scale. Typically, this situation is when we should start applying machine learning techniques~\parencite[][built this case beautifully]{Zinkevich2015}. @@ -78,11 +78,11 @@ \subsection{Sliding-windows functional connectivity}\label{subsec:sliding-window \info[inline]{Paragraph: Discuss our particular implementation.} In all experiments and benchmarks that follow in this thesis we implement a standard \gls{sw} approach to mimic a typical (often non-technical) investigator interested in using the construct of \gls{tvfc} to study the brain. -Researchers are recommended to use a window length between 30 and 60 seconds~\parencite{Shirer2012}. +Researchers are recommended to use a window length of between 30 and 60 seconds~\parencite{Shirer2012}. Both of these window lengths are implemented to test the limit cases. We implement the rectangular (non-tapered) window, with a step size of a single volume. We follow the rule of thumb proposed by \textcite{Leonardi2015} and high-pass filter the data to remove frequency components below $\frac{1}{w}$ before running the \gls{sw} algorithm.\footnote{\textcite{Smith2012} and \textcite{Hutchison2013} made similar suggestions.} -Zeros are padded to the start and end of node time series to allow for computing the correlation coefficients around those locations. +Zeros are padded to the start and end of the node time series to allow for computing the correlation coefficients around those locations. %% \subsection{Multivariate GARCH} @@ -113,7 +113,7 @@ \subsection{Multivariate GARCH} \mathbf{y}_n = \mathbf{\Sigma}_n^{\frac12} \mathbf{\eta}_n, \end{equation} where $\mathbf{\eta}_n$ is an i.i.d. white noise vector process. -As this is an autoregressive process, $\mathbf{\Sigma}_n$ is only conditioned on all observation up until time $n - 1$. +As this is an autoregressive process, $\mathbf{\Sigma}_n$ is only conditioned on all observations up until time $n - 1$. It assumes that the variances of the individual node time series follow a vector autoregression (VAR) process. % Our job as the modeler is then to model $\mathbf{\Sigma}_n$, for which several options exist. diff --git a/ch/2_Robust_estimation_of_TVFC/3_The_Wishart_process.tex b/ch/2_Robust_estimation_of_TVFC/3_The_Wishart_process.tex index 7d1181d..3d81f16 100644 --- a/ch/2_Robust_estimation_of_TVFC/3_The_Wishart_process.tex +++ b/ch/2_Robust_estimation_of_TVFC/3_The_Wishart_process.tex @@ -3,7 +3,7 @@ \section{The Wishart process}\label{sec:wishart-process} %%%%% \info[inline]{Paragraph: Introduce Wishart process and its history and application.} -In this section we introduce the \gls{wp} to the task of estimating \gls{tvfc}. +In this section, we introduce the \gls{wp} to the task of estimating \gls{tvfc}. Described by \textcite{Bru1991} as matrix generalizations of square Bessel processes, the \gls{wp} is a stochastic matrix process consisting of (for our intents and purposes) covariance matrices. \textcite{Wilson2010} described a more modern and generalized version of this stochastic process. % @@ -47,7 +47,7 @@ \subsection{Wishart process model definition} %% \info[inline]{Paragraph: Define Wishart process model construction.} -For the \gls{wp} definition, we follow notations from~\textcite{Heaukulani2019}. +For the \gls{wp} definition, we follow the notations from~\textcite{Heaukulani2019}. Let $Y$ := ($\mathbf{y}_n$, $1 \leq n \leq N$) denote a sequence of measurements in $\mathbb{R}^D$. That is, $\mathbf{y}_n = [y_{n,1}, \dots, y_{n,D}]$. In \gls{fmri} analyses, $N$ refers to the number of time steps or scan \emph{volumes}, and $D$ refers to the number of node time series (e.g.~number of brain regions for which a characteristic \gls{bold} signal time series is determined). @@ -58,7 +58,7 @@ \subsection{Wishart process model definition} Even though we expect \gls{fmri} data to be organized in a grid-like fashion of regular time intervals of a single \gls{tr}, this model flexibility could still be useful. For example, it allows for naturally leaving out a measurement due to an artifact. From the Bayesian perspective, we just so happen to make \emph{observations} at fixed intervals (giving the impression of a state-space structure), but this may not reflect the underlying process. -For computational ease-of-use we consider $X$ (the scan frame times) in the fixed interval of [0, 1] during training and prediction. +For computational ease-of-use, we consider $X$ (the scan frame times) in the fixed interval of [0, 1] during training and prediction. Estimates are then scaled back to the scan time frame. We let the conditional likelihood of observations be multivariate Gaussian: @@ -173,7 +173,7 @@ \subsection{Variational Wishart processes} \operatorname{KL}(q(F)~\|~p(F | Y)) = \mathbb{E}_{q(F)}[\log q(F)] - \mathbb{E}_{q(F)}[\log p(F, Y)] + \log p(Y) \end{equation} In \gls{vi} we optimize the \gls{elbo}, which is equivalent to this \gls{kl} term up to an added constant: -\begin{equation} \label{eq1} +\begin{equation} \begin{split} ELBO & = \mathbb{E}_{q(F)}[\log p(F, Y)] - \mathbb{E}_{q(F)}[\log q(F)] \\ & = \mathbb{E}_{q(F)}[\log p(Y|F)] - \operatorname{KL}(q(F)~\|~p(F)). @@ -272,7 +272,7 @@ \subsection{Implementation details} Or, as with the Gibbs kernel, kernel parameters could themselves be a function of input features $\mathbf{x}$ (e.g.~time). In fact, this is one of the exciting aspects of the \gls{wp}. We can characterize time series through these kernels. -Smoothness of correlation structure can be expressed by such kernel functions for example~\parencite{Fyshe2012, Fox2015, Foti2019}. +The smoothness of correlation structures can be expressed by such kernel functions for example~\parencite{Fyshe2012, Fox2015, Foti2019}. Kernels can be combined too; sums and products of kernels are also valid kernels. As such more expressive kernels can be designed~\parencite{Gonen2011} and domain knowledge incorporated. Kernel choice requires trial-and-error, although some efforts have been made to automate this process~\parencite[see e.g.][]{Steinruecken2019}. @@ -300,7 +300,7 @@ \subsection{Implementation details} This means that the amount of code to write is minimal and consists mainly of implementing a (customized) likelihood function. For these reasons, this black-box implementation~\parencite{Ranganath2014} is simple and fast compared to proposed inference routines based on \gls{mcmc}. % -In fact, this is crucial insight. +In fact, this is a crucial insight. While we acknowledge the \gls{wp} as a complex model, and thus incur its accompanying cost, in return we get a favorable optimization routine that is robust and relatively straightforward. Contrast this with the \gls{sw} approach, which is trivial in its description, but not straightforward in its implementation, with researchers facing many (arbitrary) implementational decisions to make. Parameters are updated through gradient descent with Adam~\parencite{Kingma2015} with an initial learning rate of $\alpha = 0.001$. diff --git a/ch/2_Robust_estimation_of_TVFC/4_Extracting_TVFC-based_features_and_biomarkers.tex b/ch/2_Robust_estimation_of_TVFC/4_Extracting_TVFC-based_features_and_biomarkers.tex index 7f6eb15..4a767d5 100644 --- a/ch/2_Robust_estimation_of_TVFC/4_Extracting_TVFC-based_features_and_biomarkers.tex +++ b/ch/2_Robust_estimation_of_TVFC/4_Extracting_TVFC-based_features_and_biomarkers.tex @@ -49,7 +49,7 @@ \subsection{TVFC summary measures}\label{subsec:tvfc-summary-measures} This summary measure captures how smooth a time series is over time (i.e.~the smoothness of the estimated \gls{fc} time series in our case). It is more informative of \gls{fc} frequency amplitudes and is akin to \gls{fc} `variability' as described in \textcite{Allen2014}. To illustrate its relevance, this summary measure can distinguish two sine waves with identical mean and variance, yet oscillating at different frequencies (see e.g. \cref{fig:synthetic-covariance-structures}). -As such it is complementary to the other two summary measures. +As such, it is complementary to the other two summary measures. %% \subsection{Brain states and related metrics}\label{subsec:brain-states} @@ -64,10 +64,10 @@ \subsection{Brain states and related metrics}\label{subsec:brain-states} These states can be insightful on their own or can be used as extracted features. For example, \textcite{Rashid2016} showed that schizophrenia patients spend more time in low-contrast states. One exciting aspect of the construct of brain states is that they can be synthesized across species and imaging modalities (e.g.~with microstates in \gls{eeg}, see \textcite{Allen2014} for further discussion). -As such they can serve as a common language to bridge multiple levels of neuroscientific research. +As such, they can serve as a common language to bridge multiple levels of neuroscientific research. Brain states are typically extracted either by using $k$-means clustering on estimated \gls{tvfc}~\parencite[see e.g.][]{Allen2014, Abrol2016, Zhi2018, Hakimdavoodi2020} or using models (usually \glspl{hmm}, see \cref{subsec:state-based-models}) that have this states assumption baked into their definition~\parencite{Lurie2020}. -In such latter cases we do not estimate the full \gls{tvfc} tensor. +In such latter cases, we do not estimate the full \gls{tvfc} tensor. Thus, certain information is lost in the process. Here we extract brain states from estimated \gls{tvfc} using the original $k$-means algorithm described in \textcite{Lloyd1982}. @@ -79,7 +79,7 @@ \subsection{Brain states and related metrics}\label{subsec:brain-states} % Perhaps surprisingly, most studies using brain states to summarize the brain's activity find a relatively small number of distinct states, for example three in \textcite{Choe2017, Dini2021} and 12 in \textcite{Vidaurre2017}. -Even though the \gls{wp} is not constrained in estimating \gls{tvfc} in grid-like fashion, we do this for extracting brain states to allow for better comparison to other methods. +Even though the \gls{wp} is not constrained in estimating \gls{tvfc} in a grid-like fashion, we do this for extracting brain states to allow for better comparison to other methods. However, we could estimate as many covariance matrices at as high of a temporal resolution as we like for this task. % Furthermore, some studies have relaxed the assumption that participants are assigned to a single state and learn a weight vector over all basis states instead for each time step~\parencite{Leonardi2014}. diff --git a/ch/2_Robust_estimation_of_TVFC/5_The_benchmarking_framework.tex b/ch/2_Robust_estimation_of_TVFC/5_The_benchmarking_framework.tex index 9556ff3..b679bfe 100644 --- a/ch/2_Robust_estimation_of_TVFC/5_The_benchmarking_framework.tex +++ b/ch/2_Robust_estimation_of_TVFC/5_The_benchmarking_framework.tex @@ -9,7 +9,7 @@ \section{The benchmarking framework}\label{sec:benchmark-framework} When a single optimization target or `learning task' is missing, it is common practice to define a suite of \emph{benchmarks}. Each benchmark frames method selection as a prediction task and competition~\parencite{Breiman2001, Shmueli2010, Bzdok2018, Khosla2019, Poldrack2020, Tejavibulya2022}. Such benchmarks need to be uniquely domain specific. -A collection of benchmarks then paints a rich picture about which methods are more sensitive or specific, what the failure modes of each method are, and ultimately lead to practical guidelines on which method should be used in each real-life situation.\footnote{In many machine learning sub-fields benchmarks are framed as clearly defined targets, such as as image classification accuracy. A more apt comparison may be \gls{nlp}, where optimization targets are not straightforward, and a \emph{range} of desired targets are evaluated per model~\parencite[see e.g.][]{Bommasani2021}.} +A collection of benchmarks then paints a rich picture of which methods are more sensitive or specific, what the failure modes of each method are, and ultimately leads to practical guidelines on which method should be used in each real-life situation.\footnote{In many machine learning sub-fields benchmarks are framed as clearly defined targets, such as as image classification accuracy. A more apt comparison may be \gls{nlp}, where optimization targets are not straightforward, and a \emph{range} of desired targets are evaluated per model~\parencite[see e.g.][]{Bommasani2021}.} This is a live process, and insights and approaches are updated as time goes by. % In fact, this shift in focus toward \emph{predictive} methods is increasingly argued for in neuroscience, especially for translational work to clinic practice~\parencite{Yarkoni2017, Leenings2022, Voytek2022}. @@ -34,8 +34,8 @@ \subsection{Simulations benchmarks}\label{subsec:simulation-benchmarks} %% \info[inline]{Paragraph: Describe general idea behind simulation benchmarks.} -In \textbf{simulation benchmarks}, time series are generated by a process with known, pre-specified underlying covariance structure~\parencite[see e.g.][]{Sakoglu2010, Lindquist2014, Hindriks2016, Shakil2016, Lan2017, Monti2017, Taghia2017, Thompson2018, Warnick2018, Li2019b, Ebrahimi2020}. -Such specified covariance structure can be deterministic or random. +In \textbf{simulation benchmarks}, time series are generated by a process with a known, pre-specified underlying covariance structure~\parencite[see e.g.][]{Sakoglu2010, Lindquist2014, Hindriks2016, Shakil2016, Lan2017, Monti2017, Taghia2017, Thompson2018, Warnick2018, Li2019b, Ebrahimi2020}. +Such a specified covariance structure can be deterministic or random. % The prediction task is then to `reconstruct' this ground truth covariance structure from the simulated observations. Performance is measured as the difference between estimated \gls{tvfc} and ground truth values. @@ -80,7 +80,7 @@ \subsection{Resting-state fMRI benchmarks} \info[inline]{Paragraph: Describe general idea behind rs-fMRI benchmarks.} In benchmarks based on \gls{rs-fmri} data, it is common to look at the predictive power of estimated \gls{tvfc}. -This could be behavioral performance on some task, subject measures and phenotypes, or cognitive states. +This could be the behavioral performance on some task, subject measures and phenotypes, or cognitive states. In the absence of an external stimulus, it is thought that \gls{rs-fmri} can illuminate the brain's functional architecture. It has been demonstrated that individual differences are embedded in such covariance structures. In fact, in a so-called `fingerprint' analysis, \textcite{Finn2015} showed that an individual's covariance structure or `connectome' is unique. @@ -88,9 +88,9 @@ \subsection{Resting-state fMRI benchmarks} \info[inline]{Paragraph: Describe idea behind subject measure prediction benchmarks.} Typically, the estimated \gls{tvfc} (or features derived from it) are viewed as extracted \emph{biomarkers}. These are fed to a regressor or classifier to predict either non-clinical~\parencite[see e.g.][]{Taghia2017, Li2019a} or clinical~\parencite[see e.g.][]{Filippi2019, Du2021} subject measures and phenotypes. -Methods which do better at these prediction tasks are then said to have preserved more useful information. +Methods that do better at these prediction tasks are then said to have preserved more useful information. % -In a similar data-driven spirit, \textcite{Li2019a} argued that the controversial data preprocessing step of \gls{gsr} \emph{should} be included, since it increases the predictive power of subsequently extracted networks. +In a similar data-driven spirit, \textcite{Li2019a} argued that the controversial data preprocessing step of \gls{gsr} \emph{should} be included since it increases the predictive power of subsequently extracted networks. \info[inline]{Paragraph: Introduce test-retest robustness studies and frame as benchmark.} Test-retest robustness studies~\parencite{Noble2019}, although usually not explicitly described as predictive tasks, can be viewed as looking at the predictive power of a first \gls{rs-fmri} scan to predict which subsequent scan belongs to the same subject~\parencite{Fiecas2013, Choe2017, Abrol2017, Zhang2018, Elliott2020}. @@ -144,9 +144,9 @@ \subsection{The imputation benchmark}\label{subsec:imputation-benchmark} % Train and test splits can be done in many ways. Each of these tests for something slightly different. -We could leave out a single data point as test set, a collection of data points, or data points at the end of the time series (which would test forecasting performance). +We could leave out a single data point as a test set, a collection of data points, or data points at the end of the time series (which would test forecasting performance). % -In our experiments we split up data under a \gls{leoo} scheme (i.e.~an equally sized train and test set split). +In our experiments, we split up data under a \gls{leoo} scheme (i.e.~an equally sized train and test set split). % Determining test location estimates is non-trivial due to the difference in nature of the \gls{tvfc} estimation methods considered. Since the \gls{wp} is not tied to a certain lattice as its training input or test output, predicting at unobserved data points follows naturally. diff --git a/ch/3_Benchmarking_TVFC_estimation/1_Material_and_methods.tex b/ch/3_Benchmarking_TVFC_estimation/1_Material_and_methods.tex index c4d14b8..81b0e10 100644 --- a/ch/3_Benchmarking_TVFC_estimation/1_Material_and_methods.tex +++ b/ch/3_Benchmarking_TVFC_estimation/1_Material_and_methods.tex @@ -20,7 +20,7 @@ \subsubsection{Synthetic covariance structures}\label{subsec:synthetic-covarianc The coupling of \gls{bold} time series is, in fact, still a black box to a large degree. In lieu of known covariance structure we test models against a battery of possible and reasonably exhaustive (synthetic) structures that may be encountered in an \gls{fmri} scan. The covariance structures studied are shown in \cref{fig:synthetic-covariance-structures}; -null covariance (node time series are uncorrelated during the entire scan)\improvement{Discuss importance of adding null model}, +null covariance (node time series are uncorrelated during the entire scan), a constant (i.e.~static) covariance of $\sigma_{ij} = 0.8$, periodic covariance structures (a slowly oscillating sine wave defined by one period, and a fast one with three periods) that model transient changes in coupling, a stepwise covariance that models two sudden (large) change points in covariance, @@ -55,9 +55,9 @@ \subsubsection{Data generation} All time series are subsequently individually normalized to have mean zero and unit standard deviation. The number of time series ($D$) we expect to see in practice depends on the experimental design and research question at hand. -In most applications more than two nodes or components are studied. +In most applications, more than two nodes or components are studied. Some studies even consider voxel time series directly, in which case $D$ can be in the hundreds of thousands. -Here, we test on pairwise (i.e.~\emph{bivariate}; $D = 2$) data, as well as on a trivariate ($D = 3$) data sets. +Here, we test on pairwise (i.e.~\emph{bivariate}; $D = 2$) data, as well as on trivariate ($D = 3$) data sets. Ideally, we want to study \gls{tvfc} estimation performance per method \emph{as a function of} dimensionality. The trivariate case serves as an intermediate step toward scaling up to higher dimensions. All simulation experiments are repeated $T = 200$ times to ensure robustness while balancing computational cost. @@ -108,7 +108,7 @@ \subsubsection{Data generation} Therefore, we study the cases of $N \in \{120, 200, 400, 1200\}$ data points per time series. We report results for $N = 400$, because this is the closest to the values of $N$ in \cref{ch:ukb} and thus most representative. Results for other values of $N$ are retired to \cref{appendix:more-benchmarking-results}. -It is important to be aware of the typical trade-off where scans with higher \glspl{tr} consist of fewer data points, although the increase in number of slices per scanning volume can result in higher spatial resolution and reduce autocorrelation effects and other sources of noise~\parencite{Amaro2006, Iranpour2015, Yoo2018, McDowell2019}. +It is important to be aware of the typical trade-off where scans with higher \glspl{tr} consist of fewer data points, although the increase in the number of slices per scanning volume can result in higher spatial resolution and reduce autocorrelation effects and other sources of noise~\parencite{Amaro2006, Iranpour2015, Yoo2018, McDowell2019}. We also note that Bayesian methods typically perform better on smaller data sets. Even though we strive to find a robust method that can be used in any setting, we consider the option that some methods may work better with smaller but less noisy data sets and other methods may perform better with larger, noisy data sets. This style of analysis is reminiscent of the multiverse analysis~\parencite{Steegen2016}, as discussed in more detail in \cref{subsec:robustness}. @@ -120,7 +120,7 @@ \subsubsection{Noise addition and hybrid simulations} The human nervous system is subject to various sources of noise, such as Brownian motion in synapses and ion channels~\parencite{Faisal2008}. \gls{fmri} scanners are imperfect machines and introduce further sources of noise as well. These account for system noise and observational noise. -Taken together this makes \gls{fmri} data (in)famously noisy~\parencite[see also][for analysis and biophysical simulations of impact of noise and delay]{Deco2009}. +Taken together this makes \gls{fmri} data (in)famously noisy~\parencite[see also][for analysis and biophysical simulations of the impact of noise and delay]{Deco2009}. The noise is both spatially and temporally correlated. To make our benchmark more robust, all experiments are repeated on the data sets described above, but with added noise. @@ -134,7 +134,6 @@ \subsubsection{Noise addition and hybrid simulations} Two types of added noise are studied. The simplest case (white noise) can be considered like thermal noise in \gls{fmri} scanners. -\unsure{Do we still want to report the white noise results?} Secondly, in the \emph{hybrid} simulations, we use an \gls{rs-fmri} data set to add noise to the synthetically generated activation data. We use data from the \gls{hcp} data set as described in more detail in \cref{subsec:data-hcp}. If we take time series from brain regions (or \gls{ica} components) from different subjects and add them to the synthetic signal, we can assume that no additional covariance structure is added. @@ -274,13 +273,13 @@ \subsubsection{Data: Human Connectome Project}\label{subsec:data-hcp} Each subject undergoes four scans in total, divided into two consecutive scans on two separate days. We will refer to these scans as 1A, 1B, 2A, and 2B. We only include subjects for which all of these four scans are available, resulting in 812 subjects in total. -Scans were acquired with a \gls{tr} of 0.72 seconds and voxel size of 2~mm isotropic. +Scans were acquired with a \gls{tr} of 0.72 seconds and a voxel size of 2~mm isotropic. Each scan contains $N = 1200$ images and is 15 minutes long. % Data was preprocessed by the \gls{hcp} team according to \textcite{Smith2013a} with a (minimal) preprocessing pipeline using \gls{fsl}~\parencite{Jenkinson2012} and FreeSurfer~\parencite{Fischl2012}. The pipeline is described in more detail in \textcite{Jenkinson2002, Glasser2013, Smith2013b}. Importantly, noise components have been filtered out during preprocessing using \gls{ica}+FIX~\parencite{Salimi2014, Griffanti2014}. -This ensured that none of the components we use can be considered as a noise component. +This ensured that none of the components we use can be considered a noise component. All time series were also demeaned, with variance normalized~\parencite{Beckmann2004}. All \gls{ica} components for each subject and for each scan were individually standardized to have zero mean and unit variance. For more data preprocessing details, we refer the reader to the \gls{hcp} release manual. @@ -307,10 +306,10 @@ \subsubsection{Subject measure prediction benchmark} \textcite{Sabuncu2016} defined this global metric of morphometricity as the proportion of phenotypic variation that can be explained by macroscopic brain morphology (i.e.~inter-subject anatomical variation). The score has a value between 0 and 1, with 1 explaining all variance. Their original study looked at anatomical variation between subjects. -Therefore, in their context this score is a measure of the anatomical signature of a certain trait. +Therefore, in their context, this score is a measure of the anatomical signature of a certain trait. Scores are computed as follows. -We posit for following model: +We posit the following model: \begin{equation} \mathbf{y} = \mathbf{X} \mathbf{\beta} + \mathbf{a} + \mathbf{\epsilon}, \label{eq:lme} @@ -330,12 +329,12 @@ \subsubsection{Subject measure prediction benchmark} \info[inline]{Paragraph: Define subject-by-subject similarity matrix.} How should we define the similarity $\mathbf{K}_a$ between subjects? -In our analysis we re-purpose the morphometricity analysis and consider subject `similarity' as indicated by whole-brain estimated \gls{tvfc}. +In our analysis, we re-purpose the morphometricity analysis and consider subject `similarity' as indicated by whole-brain estimated \gls{tvfc}. Instead of using the full $N \times D \times D$ estimated covariance structure, we use the $D \times D$ dimensional summary measures of it. The summary measures are the mean, variance, and rate-of-change of \gls{tvfc} estimates, as discussed in \cref{subsec:tvfc-summary-measures}. This can be viewed as a feature or biomarker extraction step. We take the lower triangular values as a vector of size $\frac{D(D-1)}{2}$, excluding the diagonal (self-connections) and duplicate values (since the matrices are symmetric). -Following \textcite{Sabuncu2016}, distance between subjects is defined by a Gaussian kernel, that is $k(\textbf{x}, \textbf{x}') = \exp(-(x-x')^2)$. +Following \textcite{Sabuncu2016}, the distance between subjects is defined by a Gaussian kernel, that is $k(\textbf{x}, \textbf{x}') = \exp(-(x-x')^2)$. \info[inline]{Paragraph: Describe our subject measures.} The variance component model relates \gls{tvfc} estimates to just a single phenotype or subject measure. @@ -349,23 +348,23 @@ \subsubsection{Subject measure prediction benchmark} % Age and gender are included as nuisance variables in our model, that is $\mathbf{X}$ in \cref{eq:lme}. Some subject measures were missing from the raw data. -We filled missing values with mean imputation. +We filled in missing values with mean imputation. % -Furthermore, the same analysis is run on social-emotional and a range of other subject measures (including personality). -However, these are considered as less relevant, and their results have been moved to \cref{appendix:hcp-more-results}. +The same analysis is run on social-emotional and a range of other subject measures (including personality). +However, these are considered less relevant, and their results have been moved to \cref{appendix:hcp-more-results}. %% \subsubsection{Test-retest robustness benchmark} %% -Apart from a performance metric that captures how well methods can estimate covariance structures, we can also compare method \emph{robustness} across scans from the same subject. +Apart from a performance metric that captures how well methods can estimate covariance structures, we can compare method \emph{robustness} across scans from the same subject. Such robustness is another desired property of any \gls{tvfc} estimation method. -The idea here is that a method that predicts consistent subject attributes across scans should be more robust and would have captured some aspect of individual differences. +The idea here is that a method that predicts consistent subject attributes across scans is more robust and has captured some aspect of individual differences. Here we broadly follow \textcite{Choe2017} and their test-retest study for \gls{tvfc} estimates in \gls{rs-fmri} data. They studied the same data set, albeit an older and smaller version thereof. % -Their idea was to test reliability of \gls{tvfc} summary measures derived from \gls{tvfc} methods. +Their idea was to test the reliability of \gls{tvfc} summary measures derived from \gls{tvfc} methods. In their study they looked at \gls{sw}, tapered \gls{sw}, and \gls{dcc} (i.e.~\gls{mgarch}) methods. They evaluated these methods on two publicly available \gls{rs-fmri} data sets suitable for test-retest studies: the Multimodal MRI Reproducibility Resource (Kirby Data) and the \gls{hcp} Data. The results for both data sets were found to be very consistent. @@ -457,7 +456,7 @@ \subsubsection{Data: Rockland visual task}\label{subsec:rockland-data} The Rockland data set consists of 286 subjects alternatingly either fixating in the dark or seeing a checkerboard visual pattern consecutively (20 seconds in duration of both stimuli and rest periods). We only consider subjects between the ages of 18 and 35. The checkerboard (moving) pattern is designed to continuously stimulate the visual cortex, while avoiding any adaptation effects (i.e.~reduced response after prolonged stimulation). -As further motivation for using this design, \textcite{Di2015} has shown that \gls{fc} is not static across such visual stimuli, and that \gls{fc} between higher and lower visual regions is affected in a consistent manner with stimulus on- and offset. +As further motivation for using this design, \textcite{Di2015} have shown that \gls{fc} is not static across such visual stimuli and that \gls{fc} between higher and lower visual regions is affected in a consistent manner with stimulus on- and offset. % Two data sets are available, one collected with a \gls{tr} of 1.4~seconds ($N = 98$) and one with a \gls{tr} of 0.645~seconds ($N = 240$). We only use the latter in these experiments. diff --git a/ch/4_TVFC_and_depression/1_Data_cohorts_and_parcellations.tex b/ch/4_TVFC_and_depression/1_Data_cohorts_and_parcellations.tex index f31a7b2..d487ef7 100644 --- a/ch/4_TVFC_and_depression/1_Data_cohorts_and_parcellations.tex +++ b/ch/4_TVFC_and_depression/1_Data_cohorts_and_parcellations.tex @@ -22,7 +22,7 @@ \subsection{Data overview} Questions regarding depressive symptoms (administered on touchscreens) were only added to the initial assessment protocol for the final two recruitment years (for 172,751 participants in total). More generally, given the sheer size and long collection duration, not all data fields are available for all participants. % -In an early descriptive epidemiological study, \textcite{Smith2013c} found \emph{probable} prevalence rates of 6.4\% for single lifetime episode of major depression, 12.2\% for recurrent major depression (moderate), and 7.2\% for recurrent major depression (severe). +In an early descriptive epidemiological study, \textcite{Smith2013c} found \emph{probable} prevalence rates of 6.4\% for a single lifetime episode of major depression, 12.2\% for recurrent major depression (moderate), and 7.2\% for recurrent major depression (severe). They noted that this is in line with other large population studies, thus underscoring the validity and representativeness of this data set (for depression studies at least). The richness in data included in this biobank presents an unprecedented opportunity to understand the interaction of mood disorders such as \gls{mdd} with genetic, lifestyle, and environmental risk factors and influences. All data used in this work has been fetched on the 1st of March 2021. @@ -33,7 +33,7 @@ \subsection{Data overview} Data collection was standardized across scanning facilities. All source images were acquired with a voxel resolution of $2.4 \times 2.4 \times 2.4$ mm and a \gls{te} of 39 ms, for a duration of 6 minutes and a \gls{tr} of 0.735 seconds, resulting in $N = 490$ volumes per scan (for the majority of participants). Participants with fewer volumes than this were discarded. -For those with more volumes than this the time series were truncated to this length. +For those with more volumes than this, the time series were truncated to this length. Data preprocessing was done by Richard Bethlehem and team at the Department of Psychiatry. \info[inline]{Paragraph: Describe data collection timeline.} @@ -67,12 +67,12 @@ \subsection{Cohort stratification}\label{subsec:cohort-stratification} \info[inline]{Paragraph: Describe general participant filters.} Before going into specific depression phenotype definitions, several general filters were run across all participants. % -Firstly, we only select participants between~40 and~64 years old (at the time when the scan was taken), to avoid including co-morbidities and changes in brain structure and function to do with old age.\footnote{There is a trade-off between sample size and sample homogeneity in this case.} +Firstly, we only select participants between~40 and~64 years old (when the scan was taken), to avoid including co-morbidities and changes in brain structure and function to do with old age.\footnote{There is a trade-off between sample size and sample homogeneity in this case.} This reduced the number of (broadly) eligible participants to include from~44,083 to~21,877. Secondly, following \textcite{Howard2020}, any participant that had been diagnosed with schizophrenia, a personality disorder, and/or bipolar disorder was filtered out. These diagnoses were taken from \gls{icd} data fields as well as Data-Field~20544. This further reduced the number of eligible participants to~21,675. -Another factor to consider are cardiovascular disorders~\parencite{Whooley2013}, such as hypertension. +Another factor to consider is cardiovascular disorders~\parencite{Whooley2013}, such as hypertension. \Gls{bold} signals are based on blood flow, so such conditions may bias our findings. However, this information is not used in our cohort stratification. General clinical and demographic characteristics of all such broadly eligible participants are shown in \cref{tab:ukbiobank-cohorts}. @@ -101,7 +101,7 @@ \subsubsection{Diagnosed lifetime occurrence (depressive trait analysis)} %% \info[inline]{Paragraph: Describe diagnosed lifetime occurrence phenotype.} -For this first lifetime occurrence (a.k.a.~\emph{history} or \emph{instance}) phenotype two cohorts are selected from all eligible participants: an \gls{mdd} cohort and a \gls{hc} cohort. +For this first lifetime occurrence (a.k.a.~\emph{history} or \emph{instance}) phenotype two cohorts are selected from all eligible participants: an \gls{mdd} cohort and an \gls{hc} cohort. This cohort is based on medical diagnoses of \gls{mdd}, which can be found in Data-Field 41270. We select participants that have at any point in their lives been diagnosed with \gls{icd} codes F320--F323, F328--F329 (single depressive episodes), F330--F334, F338, and/or F339 (recurrent depressive episodes). As such we do not distinguish between single or recurrent episodes (and thus depression severity). @@ -109,14 +109,14 @@ \subsubsection{Diagnosed lifetime occurrence (depressive trait analysis)} Moreover, participants that were taking antidepressants were filtered out from the \gls{hc} cohort. % The male/female ratios of these cohorts show a large discrepancy.\footnote{Higher reported depression incidence for women was to be expected~\parencite{Albert2015, Bogren2018}. The prevalence of depression decreases after the age of 65, however, and becomes similar across sex~\parencite{Bebbington2003}. This is likely influenced by female prevalence of depression peaking around hormonal changes (puberty, prior to menstruation, following pregnancy, and perimenopause).} -Therefore, the control cohorts are subsampled to match the depressed cohort not only in size, but also in sex ratio. +Therefore, the control cohorts are subsampled to match the depressed cohort not only in size but also in sex ratio. %% \subsubsection{Self-reported lifetime occurrence (depressive trait analysis)} %% \info[inline]{Paragraph: Describe self-reported lifetime occurrence phenotype.} -For this second lifetime occurrence phenotype we again select two cohorts from all eligible participants: a depressed cohort (we avoid the term \gls{mdd} here due to a lack of professional diagnosis) and a \gls{hc} cohort. +For this second lifetime occurrence phenotype we again select two cohorts from all eligible participants: a depressed cohort (we avoid the term \gls{mdd} here due to a lack of professional diagnosis) and an \gls{hc} cohort. We broadly follow \textcite{Howard2020} for this analysis and use the self-reported lifetime instance depression phenotype definition based on the \gls{cidi-sf} \parencite{Kessler1998} as described and defined by \textcite{Davis2020}.\footnote{The scoring criteria from \textcite{Davis2020} are equivalent to the \gls{dsm} criteria for \gls{mdd}. See \cref{sec:fc-depression} for more details.} This inventory was part of the follow-up \gls{mhq} sent out to participants. We again note that this phenotype indicates a \emph{lifetime} instance measure of depression. @@ -125,7 +125,7 @@ \subsubsection{Self-reported lifetime occurrence (depressive trait analysis)} \info[inline]{Paragraph: Discuss inherent data limitations.} The relevant online follow-up questionnaires were sent out to participants with valid email addresses and completed in 2016, whereas the scans were taken any time between 2014 and 2018. -This means that some participants filled it out before the scan, whereas others did so afterwards. +This means that some participants filled it out before the scan, whereas others did so afterward. Consequently, some individuals may have gotten depressed for the first time after filling out the questionnaire, but before or during their scan. Moreover, for those who reported ever having been depressed, some may have been so while in the scanner, whereas for others it was a long time ago. Unfortunately, we have no surefire way of separating out these groups. @@ -192,7 +192,7 @@ \subsubsection{Polygenic risk score (depression risk analysis)} Subjects with high \glspl{prs} may well have never experienced any depressive episode, and vice versa. As we reviewed in \cref{subsec:depression}, the genetics of depression is still an active field with many open questions~\parencite{Ormel2019}. We know that depression risk is heritable to some degree, but depressive episodes typically still need a trigger, such as an adverse life event. -As before, changing our depression phenotype inherently changes our study focus and scope of subsequent conclusions. +As before, changing our depression phenotype inherently changes our study focus and the scope of subsequent conclusions. \info[inline]{Paragraph: Discuss overlap between cohorts.} As such, how much correspondence is there between this genetics-based phenotype and the three aforementioned (diagnosed and self-reported, based on actual life experiences and symptoms) depression phenotype definitions? @@ -230,7 +230,7 @@ \subsection{Brain regions of interest} An influential concept in neuroanatomy is that the human brain is in fact made up of three brains. The concept of this \emph{triune} brain was introduced by \textcite{Maclean1985}. It posits from an evolutionary perspective that the human brain consists of a primal (`reptilian'; including basal ganglia and brain stem structures that help with the `plumbing' and regulatory side of bodily homeostasis), limbic (`mammalian' or `emotional'; involved with critical emotional skills required for social animals), and neomammalian (`rational'; responsible for higher and more complex cognitive function and regulation of emotions) brain.\footnote{The terms `reptilian' and `mammalian' should, of course, not be taken literally. Reptiles were never ancestors to mammals; our evolutionary lines diverged over 300 million years ago~\parencite{Striedter2019}.} -The primal and limbic systems are more ancient than the (evolutionarily speaking) newer neocortex, and their physiology and anatomy is therefore categorically different as well. +The primal and limbic systems are more ancient than the (evolutionarily speaking) newer neocortex, and their physiology and anatomy are therefore categorically different. For example, limbic areas such as the \gls{hpc} are typically comprised of three layers, whereas cortical areas typically have five or six layers. Theories of depression often pertain to such functional descriptions, often suggesting aberrant and dysregulated emotional, limbic, and reward processing~\parencite{Akiskal1973}. It makes intuitive sense that depression would affect the regions and circuits involved with these functions, as opposed to the visual cortex, for example (although even such regions may very well be affected). @@ -255,7 +255,7 @@ \subsection{Brain regions of interest} It was first described by Karl Friedrich Burdach in 1822~\parencite{Burdach1826}. It lies in the midbrain, next to the \gls{hpc}, and is connected to many other brain regions. As a brain structure it is made up of about 13 nuclei.\footnote{In neuroanatomy, a \textbf{nucleus} refers to any cluster of neurons, where such neurons have similar functions and connections to other nuclei.} -The \gls{amg} is especially involved in processing of emotions and memories related to fear, threats, aggression, and pain~\parencite{Thompson2017b}. +The \gls{amg} is especially involved in the processing of emotions and memories related to fear, threats, aggression, and pain~\parencite{Thompson2017b}. It also assigns value and emotional meaning to memories and decisions. This makes it a prime contestant for relevant brain regions. It is theorized that the \gls{amg} is dysregulated and hyperactive in patients with \gls{mdd}. @@ -263,7 +263,7 @@ \subsection{Brain regions of interest} It has been found that the \gls{amg} has decreased connectivity with a range of other brain regions with \gls{mdd}~\parencite{Tang2013, Ramasubbu2014}. Even though the \gls{amg} is known to have three functionally distinct subdivisions, we consider it as one whole brain region here. The \gls{amg} is a relatively small brain region. -\textcite{Brabec2010} found the average \gls{amg} in their sample to be 1,240--1,630~$mm^3$ in size per hemisphere (depending on measurement method, with no significant interhemisphere or intersex differences). +\textcite{Brabec2010} found the average \gls{amg} in their sample to be 1,240--1,630~$mm^3$ in size per hemisphere (depending on the measurement method, with no significant interhemisphere or intersex differences). The volume of the \gls{amg} has been shown to shrink with recurrent major depression~\parencite{Sheline1998}. @@ -293,9 +293,9 @@ \subsection{Brain regions of interest} The \gls{hpc} is another subcortical region, located deep in the temporal lobe, and is also part of the limbic system. -It has long been known to be involved with learning, memory, and replay of memories (consolidation). +It has long been known to be involved with learning, memory, and the replay of memories (consolidation). More recently it has also been implied with emotional behavior and spatial navigation. -The \gls{hpc} is a vulnerable brain region, and is one of the earlier and most severely affected brain areas with neurodegenerative disorders such as \gls{ad}. +The \gls{hpc} is a vulnerable brain region and is one of the earlier and most severely affected brain areas with neurodegenerative disorders such as \gls{ad}. Hippocampal volumes differ across hemispheres. \textcite{McHugh2007} found human hippocampal volumes of 3,480~$\pm$~430~$mm^3$ and 3,680~$\pm$~420~$mm^3$ for left and right \gls{hpc}, respectively. Macaque primate as well as human studies have found decreased hippocampal volume with depression~\parencite{Campbell2004, Malykhin2010, Brown2014, Schmaal2016}. @@ -311,15 +311,15 @@ \subsection{Brain regions of interest} More recent work showed that it also processes \emph{social} context. The insula, (also known as insular cortex) is part of the \gls{sn} and the limbic system~\parencite{Uddin2017}. -It is known to be involved with modulation of emotional processing. +It is known to be involved with the modulation of emotional processing. It has been linked to salience detection, self-awareness, consciousness, interoception, pain processing, and addiction as well~\parencite{Menon2010}. -This is highly related to its function of controlling regulation of the sympathetic and parasympathetic systems. +This is highly related to its function of controlling the regulation of the sympathetic and parasympathetic systems. It also controls awareness of hunger, pain, and fatigue, emotions related to homeostasis that are dysregulated in \gls{mdd}. Especially the \gls{ai} half is relevant~\parencite{Pasquini2020}, and in this work we only consider this as an \gls{roi}. -The \gls{ofc} is part of the \gls{pfc} and believed to be a nexus for sensory integration. +The \gls{ofc} is part of the \gls{pfc} and is believed to be a nexus for sensory integration. It is involved with emotional and reward-related behavior~\parencite{Kringelbach2005}. -In a review work, \textcite{Stalnaker2015} finds most evidence for its function in credit and value assignment. +In a review work, \textcite{Stalnaker2015} found most evidence for its function in credit and value assignment. This relates to decision-making as well. The \gls{pcc}, as part of the cingulate cortex, is located centrally in the brain. @@ -334,10 +334,10 @@ \subsection{Brain regions of interest} This region is mostly linked to executive function (e.g.~planning, reasoning, cognitive flexibility~\parencite{Dajani2015}, and working memory). It also plays a role in mood regulation. -The \gls{acc} is involved with salience and attention, as well as management of pain and emotions. +The \gls{acc} is involved with salience and attention, as well as the management of pain and emotions. It has connections to both the limbic system and prefrontal areas. Due to its importance in mood regulation, it is one of the most common sites of \gls{dbs} for affective disorders~\parencite{Drevets2008}. -The \gls{acc} is often sub-divided into a rostral or ventral part and a caudal or dorsal part~\parencite[see][for more clarification on subdivisions]{Stevens2011}. +The \gls{acc} is often subdivided into a rostral or ventral part and a caudal or dorsal part~\parencite[see][for more clarification on subdivisions]{Stevens2011}. The \gls{mpfc} is another \gls{pfc} subdivision, a part of the ventromedial \gls{pfc}. It is believed to be involved in introspection (the \gls{mpfc} has one of the highest baseline metabolic rates at rest). @@ -352,7 +352,7 @@ \subsection{Brain regions of interest} This atlas contains 180 regions per hemisphere. We merge all brain regions across both hemispheres, assuming lateralization does not play a significant role. Each of the brain regions of interest consists of a collection of subregions from this atlas. -We take a weighted (by number of voxels per subregion) average to obtain our final time series. +We take a weighted (by the number of voxels per subregion) average to obtain our final time series. Note that parcellation choices are often controversial, see also \textcite{Arslan2018, Bryce2021} for a comparison of parcellation methods. Subregion specifics (including full names and volumes) are shown in \cref{tab:ukbiobank-brain-regions}. @@ -385,7 +385,7 @@ \subsection{Brain region edges of interest} Connectivity measures or interaction indicators vary as well. Depression phenotypes are typically different across studies. For example, most studies include participants that were depressed \emph{during} the scan. -Moreover, samples sizes for these studies were often small (see \cref{ch:discussion} for further discussion). +Moreover, sample sizes for these studies were often small (see \cref{ch:discussion} for further discussion). %% \subsubsection{Static functional connectivity edges affected in depression} @@ -427,10 +427,10 @@ \subsubsection{Time-varying functional connectivity edges affected in depression \textcite{Demirtas2016} studied `global synchronization' and temporal stability of \gls{fc} in 27 patients and 27 \glspl{hc}. They find that \gls{sfc} is \emph{increased} within the \gls{dmn} with \gls{mdd}, and \gls{fc} variability \emph{decreased} between \gls{dmn} and \gls{cen}. -\textcite{Wise2017} used a \gls{sw} (40 seconds in length, created using a Gaussian kernel with a standard deviation of 8 seconds) analysis and found increased variance between \gls{pcc} and \gls{mpfc}, the core nodes of the \gls{dmn}. +\textcite{Wise2017} used an \gls{sw} (40 seconds in length, created using a Gaussian kernel with a standard deviation of 8 seconds) analysis and found increased variance between \gls{pcc} and \gls{mpfc}, the core nodes of the \gls{dmn}. They linked this finding to correlation with rumination, a core symptom of depression. -\textcite{AlonsoMartinez2020} characterized \gls{fc} dynamics through brain states, and found depressed subjects have overall lower connectivity between \gls{dmn} regions and regions outside of it. +\textcite{AlonsoMartinez2020} characterized \gls{fc} dynamics through brain states and found depressed subjects have overall lower connectivity between \gls{dmn} regions and regions outside of it. Especially the precuneus was found to be implied. They also demonstrated how important it is to take the temporal dynamics of \gls{fc} into account. @@ -465,7 +465,7 @@ \subsection{Functional network analysis}\label{subsec:ukb-fn-analysis} \info[inline]{Paragraph: Introduce functional networks analysis.} In our second analysis type, we study brain \emph{networks} instead of individual brain \glspl{roi}. -As reviewed in \cref{subsec:fc-depression}, depression has also been considered as a network disorder. +As reviewed in \cref{subsec:fc-depression}, depression has also been considered a network disorder. % The three networks considered (and described in more detail in \cref{subsec:fc-depression}) in this study are visualized in \cref{fig:ukb-brain-functional-networks}. The regions and respective sub-regions of these networks are shown in \cref{tab:ukbiobank-functional-networks}. diff --git a/ch/4_TVFC_and_depression/2_Material_and_methods.tex b/ch/4_TVFC_and_depression/2_Material_and_methods.tex index 2014744..4cb6f08 100644 --- a/ch/4_TVFC_and_depression/2_Material_and_methods.tex +++ b/ch/4_TVFC_and_depression/2_Material_and_methods.tex @@ -11,7 +11,7 @@ \subsection{TVFC estimation method} Based on our benchmarking work in the previous chapter, how should we now estimate \gls{tvfc}? The goal of the benchmarking framework was to pick the best method available. -Based on \cref{ch:benchmarking}, at the time of writing the evidence points toward the \gls{wp} approach for being the best overall \gls{tvfc} estimation method. +Based on \cref{ch:benchmarking}, at the time of writing the evidence points toward the \gls{wp} approach as being the best overall \gls{tvfc} estimation method. However, if new methods are developed that outperform the \gls{wp} model on these benchmarks, or if additional benchmarks paint a different picture, the experiments in this current chapter should be re-run and conclusions updated accordingly. More specifically, we pick \gls{svwp} for computational efficiency, as $N > 400$. @@ -52,12 +52,12 @@ \subsection{Cohort comparison} To assess the significance of contrasts between cohorts, we apply methods from the Neyman-Pearson (i.e.~standard, orthodox) school of statistical hypothesis testing.\footnote{See \textcite{Dienes2008} for an excellent and complete overview of the various schools of thought.} % We run two-sample $t$-tests on the \gls{tvfc} summary measures across all subjects for the edges of interest. -To correct for multiple comparisons, we run both the Bonferroni (which is often considered too conservative) as well as the Benjamini-Hochberg \gls{fdr} method~\parencite{Benjamini1995}, using $\alpha = 0.05$ as significance threshold (as is traditional). +To correct for multiple comparisons, we run both the Bonferroni (which is often considered too conservative) as well as the Benjamini-Hochberg \gls{fdr} method~\parencite{Benjamini1995}, using $\alpha = 0.05$ as the significance threshold (as is traditional). All figures show significance indications after the latter correction. These \gls{mht} corrections are considered separately for each \gls{tvfc} summary measure, as they are considered independent from one another. % As we anticipate small effect sizes (see \cref{subsec:fc-depression}), we report Cohen's $d$ where appropriate~\parencite{Cohen1988, Lakens2013}. -To get a ball-park interpretation of effect size, \textcite{Cohen1988} suggested `small' ($d = 0.2$), `medium' ($d = 0.5$), and `large' ($d = 0.8$) as effect size labels. +To get a ballpark interpretation of effect size, \textcite{Cohen1988} suggested `small' ($d = 0.2$), `medium' ($d = 0.5$), and `large' ($d = 0.8$) as effect size labels. However, he already noted that these are rather arbitrary and should only be used if findings are so novel that they cannot be compared to prior findings. Relating effect sizes to previous findings is indeed its most useful application. Unfortunately, not everyone reports these. diff --git a/ch/4_TVFC_and_depression/3_Results.tex b/ch/4_TVFC_and_depression/3_Results.tex index 348b79b..8ed6c78 100644 --- a/ch/4_TVFC_and_depression/3_Results.tex +++ b/ch/4_TVFC_and_depression/3_Results.tex @@ -94,7 +94,7 @@ \subsection{Diagnosed lifetime occurrence} \end{figure} -\Cref{fig:ukb-results-dlo-roi-cohort-comparison-edges-of-interest-wp} shows \gls{svwp} \gls{tvfc} estimates summary measures for the edges of interest between our brain region of interest for both cohorts. +\Cref{fig:ukb-results-dlo-roi-cohort-comparison-edges-of-interest-wp} shows \gls{svwp} \gls{tvfc} estimates summary measures for the edges of interest between our brain regions of interest for both cohorts. % We find the same global decrease with \gls{mdd} in connectivity strength for all edges. As expected, the mean \gls{tvfc} estimates are almost identical to the \gls{sfc} estimates. diff --git a/ch/4_TVFC_and_depression/4_Discussion.tex b/ch/4_TVFC_and_depression/4_Discussion.tex index 0e2d9e5..2e9a888 100644 --- a/ch/4_TVFC_and_depression/4_Discussion.tex +++ b/ch/4_TVFC_and_depression/4_Discussion.tex @@ -56,7 +56,7 @@ \subsection{Varying results across depression phenotypes}\label{subsec:variation Moreover, the connectivity rate-of-change edges are not affected for the depressed state cohorts. These differences are considered minor. However, the brain state analysis finds a more interesting contrast between these two paradigms. -We found that hippocampal-prefrontal connections are more involved with lifetime history of depression, and anterior insular-prefrontal connections more involved with current depressive episodes. +We found that hippocampal-prefrontal connections are more involved with a lifetime history of depression, and anterior insular-prefrontal connections are more involved with current depressive episodes. %% \subsection{Comparison to prior studies} @@ -65,7 +65,7 @@ \subsection{Comparison to prior studies} While we compare directly to prior work, not all studies are created equally. Moreover, many previous findings may not replicate. In fact, there are many reasons why many neuroimaging depression study results may be false or inflated~\parencite{Flint2021}. -The considerable number of studies looking to use neuroimaging to separate \gls{mdd} patients from \glspl{hc} still lack cohesion. +The considerable number of studies looking to use neuroimaging to separate \gls{mdd} patients from \glspl{hc} still lacks cohesion. Sample sizes are often considered a key issue~\parencite{Varoquaux2018, Szucs2020, Libedinsky2022, Marek2022}. For \gls{mdd} estimation from structural \gls{mri} for example, \textcite{Flint2021} showed that small sample sizes can dramatically inflate the predictive power of models. We consider the large sample sizes in all benchmarks and experiments a major strength of the work in this thesis. @@ -99,7 +99,7 @@ \subsection{Choice of TVFC estimation method} Therefore, we carefully conclude that \gls{sw}-based methods would return more false positives rather than false negatives. Based on our simulations benchmarking in \cref{ch:benchmarking}, this may come as no surprise. % -However, all estimation methods find an increased rate-of-change for the \gls{dlpfc}--\gls{acc} edge, indicating that this finding is robust to \gls{tvfc} estimation method choice. +However, all estimation methods find an increased rate-of-change for the \gls{dlpfc}--\gls{acc} edge, indicating that this finding is robust to the \gls{tvfc} estimation method choice. % Furthermore, all other methods do indeed find significantly increased \gls{tvfc} variance between all \glspl{fn}, in contrast to the \gls{svwp} estimates. As we have seen before, \gls{svwp} estimates are smooth compared to the other methods' estimates. @@ -107,7 +107,7 @@ \subsection{Choice of TVFC estimation method} For the self-reported lifetime occurrence analysis, the \gls{dcc} methods do not find any significant alterations with depression across all edges and \gls{tvfc} summary measures. % -All \gls{sw} methods find exactly the same single affected edge (\gls{hpc}--\gls{ai}), except for the 30-seconds \gls{sw} estimate that finds the rate-of-change to also be significantly affected for this edge. +All \gls{sw} methods find exactly the same single affected edge (\gls{hpc}--\gls{ai}), except for the 30-second \gls{sw} estimate that finds the rate-of-change to also be significantly affected for this edge. % Overall, this shows that the reduced \gls{sfc} with depression for this edge is reasonably robust across \gls{tvfc} estimation method and should be studied further as an affected brain region connection. % @@ -116,7 +116,7 @@ \subsection{Choice of TVFC estimation method} The \gls{sw-cv} approach only returns the two estimate means contrasts for the same two between-network connectivities as the \gls{svwp}. For the self-reported depressed state analysis, the \gls{dcc} methods predict a distinct set of edges whose mean \gls{tvfc} is affected in depression compared to both the \gls{sfc} and \gls{svwp} estimates. -Furthermore, the joint approach returns many significantly different edges for the rate-of-change summary measure, where the pairwise implementation returns none (like the \gls{svwp} estimates). +Furthermore, the joint approach returns many significantly different edges for the rate-of-change summary measure, whereas the pairwise implementation returns none (like the \gls{svwp} estimates). The \gls{sw-cv} estimates are broadly like the \gls{svwp} estimates, except for a curious rate-of-change \gls{pha}--\gls{mpfc} edge. % For the \gls{fn} analysis, all other \gls{tvfc} estimation methods find the same cohort contrasts. @@ -125,5 +125,5 @@ \subsection{Choice of TVFC estimation method} For the \gls{prs} analysis, none of the other methods' estimates are significantly different for any of the edges, just like the \gls{sfc} and \gls{svwp} estimates. The null finding is robust across \gls{tvfc} estimation methods. -Lastly, we note that any estimation method may return spurious structure or fail to detect certain structures. +Lastly, we note that any estimation method may return spurious structures or fail to detect certain structures. However, since we average across participants in cohorts, some of these failure modes may be hidden from our results. diff --git a/ch/5_Discussion/2_Future_work.tex b/ch/5_Discussion/2_Future_work.tex index 6340844..80a75b9 100644 --- a/ch/5_Discussion/2_Future_work.tex +++ b/ch/5_Discussion/2_Future_work.tex @@ -258,7 +258,7 @@ \subsubsection{Improved labels} One commonly reported challenge in the study of depression is its heterogeneity of symptoms. In fact, many have suggested that depression is more of an umbrella term for various diseases. We consider this one of the major shortcomings of this work. -Using imperfect diagnostic labels limits any analysis (including simple comparisons of populations in terms of neural and cognitive and processes) and practical utility. +Using imperfect diagnostic labels limits any analysis (including simple comparisons of populations in terms of neural and cognitive processes) and practical utility. The \gls{dsm} diagnostic labels were meant as a rough clinical indication, not as a carefully designed label to find biomarkers with~\parencite{Fried2022b}. Future work would build on top of depression subtype definitions~\parencite{Tokuda2018}. @@ -276,7 +276,7 @@ \subsubsection{Specificity and applications} As we have discussed briefly already, this study, like many others, relies on broad depression phenotypes and thus contains large disorder heterogeneity. Moreover, broad brain networks are typically studied. -These study characteristics may lead to a lack in specificity. +These study characteristics may lead to a lack of specificity. Such specificity may be important in certain circumstances. For example, a clinician would rather know a precise treatment target, instead of broadly knowing that the \gls{dmn} is affected. One big drawback of more fine-grained studies~\parencite[e.g.][]{Klein-Flamp2022} is that \gls{fmri} signals become noisier when we look at smaller brain region parcellations. diff --git a/main.tex b/main.tex index f04a756..fd5622d 100644 --- a/main.tex +++ b/main.tex @@ -1,56 +1,53 @@ \documentclass[ - % draft, % comment for final version: uncomment to skip loading figures - % abstract, % generate only title page and abstract page for student registry submission + % draft, % comment for final version (both ELECTRONIC and HARDBOUND): uncomment to skip loading figures a4paper, - 12pt, % should be 11pt or 12pt - dvipsnames, - twoside, % uncomment for final version: different left and right side pages - openright % uncomment for final version: start new chapters on right hand side page + 12pt, % should be 11pt or 12pt + dvipsnames, % for extra color options + twoside, % uncomment for final version (both ELECTRONIC and HARDBOUND): different left and right side pages + openright % uncomment for final version (both ELECTRONIC and HARDBOUND): start new chapters on right hand side page ]{report} %%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%% -\input{misc/headings} % uncomment for fancy layout +\input{misc/headings} % for fancy layout \input{misc/preamble} \input{misc/glossaries} %%%%%%%%%%%%%%%%%%%%%%%%%%% -\newcommand\optionalindent{} % uncomment for fancy layout +\newcommand\optionalindent{} % for fancy layout %% \begin{document} \pagenumbering{roman} %%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%% +% \includepdf{misc/deposit_and_copying_declaration.pdf} % uncomment only for HARDBOUND version +% \clearpage \input{misc/title_page} \thispagestyle{empty} \input{misc/declaration} \clearpage \thispagestyle{empty} -\input{misc/acknowledgements} \input{misc/abstract} +\input{misc/acknowledgements} %%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%% \tableofcontents %% \clearpage -% \addtocontents{lof}{\protect\addcontentsline{toc}{chapter}{List of Figures}} -\addcontentsline{toc}{chapter}{List of Figures} \listoffigures %% \clearpage \thispagestyle{empty} -\listoftables\addcontentsline{toc}{chapter}{List of Tables} +\listoftables %% \clearpage \thispagestyle{empty} -\addcontentsline{toc}{chapter}{Acronyms} % \printglossary[type=\acronymtype,title=List of Abbreviations] \printnoidxglossaries %%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%% -\renewcommand\optionalindent{\tabto{1.2cm}} % uncomment for fancy layout +\renewcommand\optionalindent{\tabto{1.2cm}} % for fancy layout \newpage \pagenumbering{arabic} -\pagestyle{fancy} % uncomment for final version: add fancy headers and footers -% \linenumbers +\pagestyle{fancy} % add fancy headers and footers %% \input{ch/1_Introduction/0_Introduction} \input{ch/1_Introduction/1_Functional_connectivity} @@ -94,14 +91,12 @@ \appendix %% \input{appendix/03_extra_benchmarking_results} -\input{appendix/hcp_extra_results} \input{appendix/04_ukb_with_other_methods} %% %%%%%%%%%%%%%%%%%%%%%%%%%%% \clearpage -% \nolinenumbers -\renewcommand\optionalindent{} % uncomment for fancy layout? -\renewcommand{\bibname}{References} % changes the default name `Bibliography` -> `References' -\addcontentsline{toc}{chapter}{References}\printbibliography +\renewcommand\optionalindent{} % for fancy layout +\renewcommand{\bibname}{References} % changes the default name `Bibliography` -> `References' +\printbibliography[heading=bibintoc] %%%%%%%%%%%%%%%%%%%%%%%%%%% \end{document} diff --git a/misc/abstract.tex b/misc/abstract.tex index f70103b..4bba48f 100644 --- a/misc/abstract.tex +++ b/misc/abstract.tex @@ -1,4 +1,14 @@ -\chapter*{Abstract} +% uncomment for ELECTRONIC simple version +% \chapter*{Abstract} + +% uncomment for both ELECTRONIC and HARDBOUND final version +\cleardoublepage +\begin{center} + \vspace*{2em} + { \Large {\bfseries Robust time-varying functional connectivity estimation and its relevance for depression} \par} + {{\large \vspace*{1em} Onno Pepijn Kampman} \par} + \vspace{2em} +\end{center} This thesis investigates how to robustly estimate \gls{tvfc}, a construct in neuroimaging research that looks at changes in functional coupling (correlation between time series) between brain regions during a \gls{fmri} scan, and how it can be used as a lens through which to study depression as a functional disorder. @@ -13,7 +23,7 @@ \chapter*{Abstract} Returning to the depression study, several differences are found between depressed and healthy control cohorts. The study is run on thousands of participants from the UK Biobank, yielding unprecedented statistical power and robustness. I investigate connectivity between individual brain regions as well as \glspl{fn}. -Generally, depressed participants show decreased global connectivity, and increased connectivity instability (as measured by the temporal characteristics of estimated \gls{tvfc}). +Depressed participants show decreased global connectivity, and increased connectivity instability (as measured by the temporal characteristics of estimated \gls{tvfc}). By defining multiple depression phenotypes, I find that brain dynamics are affected especially when patients have been professionally diagnosed or indicated to be depressed during their \gls{fmri} scan, but were less or not at all affected based on self-reported past instances and genetic predisposition. I show that choosing a different \gls{tvfc} estimation method would have changed our scientific conclusions. This sensitivity to seemingly arbitrary researcher choices highlights the need for robust method development and the importance of community-approved benchmarking. diff --git a/misc/glossaries.tex b/misc/glossaries.tex index 9b55572..86708b5 100644 --- a/misc/glossaries.tex +++ b/misc/glossaries.tex @@ -1,4 +1,5 @@ -\usepackage[acronym]{glossaries} +% The glossaries package should be loaded after the hyperref package. +\usepackage[acronym,toc]{glossaries} % \makeglossaries \makenoidxglossaries @@ -7,12 +8,11 @@ \newacronym{ad}{AD}{Alzheimer's disease} \newacronym{ai}{AI}{anterior insula} \newacronym{amg}{AMG}{amygdala} -\newacronym{ann}{ANN}{Artificial Neural Network} \newacronym{bmi}{BMI}{body mass index} \newacronym{bold}{BOLD}{blood oxygenation level dependent} -\newacronym{cap}{CAP}{coactivation pattern} +% \newacronym{cap}{CAP}{coactivation pattern} \newacronym{cbm}{CBM}{cerebellum} \newacronym{cen}{CEN}{central executive network} \newacronym{cidi-sf}{CIDI-SF}{Composite International Diagnostic Interview Short Form} @@ -23,9 +23,8 @@ \newacronym{dcc}{DCC}{dynamic conditional correlation} \newacronym{dfc}{dFC}{dynamic functional connectivity} \newacronym{dlpfc}{dlPFC}{dorsolateral prefrontal cortex} -\newacronym{dmpfc}{dmPFC}{dorsomedial prefrontal cortex} \newacronym{dmn}{DMN}{default mode network} -\newacronym{dnn}{DNN}{deep neural network} +\newacronym{dmpfc}{dmPFC}{dorsomedial prefrontal cortex} \newacronym{dsm}{DSM}{Diagnostic and Statistical Manual of Mental Disorders} \newacronym{dti}{DTI}{diffusion tensor imaging} @@ -41,7 +40,7 @@ \newacronym{fdr}{FDR}{false discovery rate} \newacronym{fmri}{fMRI}{functional magnetic resonance imaging} \newacronym{fn}{FN}{functional network} -\newacronym{fov}{FOV}{field of view} +% \newacronym{fov}{FOV}{field of view} \newacronym{fsl}{FSL}{FMRIB Software Library} \newacronym{fwhm}{FWHM}{full-width at half-maximum} @@ -54,8 +53,8 @@ \newacronym{hc}{HC}{healthy controls} \newacronym{hcp}{HCP}{Human Connectome Project} -\newacronym{hpc}{HPC}{hippocampus} \newacronym{hmm}{HMM}{hidden Markov model} +\newacronym{hpc}{HPC}{hippocampus} \newacronym{hrf}{HRF}{hemodynamic response function} \newacronym{ica}{ICA}{independent component analysis} @@ -70,8 +69,8 @@ \newacronym{lme}{LME}{linear mixed effects} \newacronym{m1}{M1}{primary motor cortex} -\newacronym{mae}{MAE}{mean absolute error} -\newacronym{maoi}{MAOI}{monoamine oxidase inhibitor} +% \newacronym{mae}{MAE}{mean absolute error} +% \newacronym{maoi}{MAOI}{monoamine oxidase inhibitor} \newacronym{mcmc}{MCMC}{Markov Chain Monte Carlo} \newacronym{mdd}{MDD}{major depressive disorder} \newacronym{meg}{MEG}{magnetoencephalography} @@ -79,14 +78,14 @@ \newacronym{mhq}{MHQ}{mental health questionnaire} \newacronym{mht}{MHT}{multiple hypothesis testing} \newacronym{mle}{MLE}{maximum likelihood estimate} -\newacronym{mni}{MNI}{Montreal Neurological Institute} +% \newacronym{mni}{MNI}{Montreal Neurological Institute} \newacronym{mpfc}{mPFC}{medial prefrontal cortex} \newacronym{mri}{MRI}{magnetic resonance imaging} -\newacronym{mse}{MSE}{mean squared error} +% \newacronym{mse}{MSE}{mean squared error} -\newacronym{na}{NA}{noradrenaline} -\newacronym{nac}{NAc}{nucleus accumbens} -\newacronym{nirs}{NIRS}{near infrared spectroscopy} +% \newacronym{na}{NA}{noradrenaline} +% \newacronym{nac}{NAc}{nucleus accumbens} +\newacronym{nirs}{NIRS}{near-infrared spectroscopy} \newacronym{nlp}{NLP}{natural language processing} \newacronym{ofc}{OFC}{orbitofrontal cortex} @@ -98,14 +97,14 @@ \newacronym{pet}{PET}{positron emission tomography} \newacronym{pfc}{PFC}{prefrontal cortex} \newacronym{pha}{PHA}{parahippocampal area} -\newacronym{phc}{PHC}{parahippocampus} +% \newacronym{phc}{PHC}{parahippocampus} \newacronym{prs}{PRS}{polygenic risk scores} \newacronym{ptsd}{PTSD}{post-traumatic stress disorder} -\newacronym{rct}{RCT}{randomized controlled trial} +% \newacronym{rct}{RCT}{randomized controlled trial} \newacronym{reml}{ReML}{restricted maximum likelihood} \newacronym{rl}{RL}{reinforcement learning} -\newacronym{rms}{RMS}{Root Mean Square} +% \newacronym{rms}{RMS}{Root Mean Square} \newacronym{rmse}{RMSE}{root mean square error} \newacronym{rnn}{RNN}{recurrent neural network} \newacronym{roi}{ROI}{region of interest} @@ -113,7 +112,7 @@ \newacronym{rsn}{RSN}{resting-state network} \newacronym{sfc}{sFC}{static functional connectivity} -\newacronym{sm}{SM}{subject measure} +% \newacronym{sm}{SM}{subject measure} \newacronym{sn}{SN}{salience network} \newacronym{snr}{SNR}{signal-to-noise ratio} \newacronym{ssri}{SSRI}{selective serotonin reuptake inhibitor} @@ -135,4 +134,4 @@ \newacronym{who}{WHO}{World Health Organization} \newacronym{wp}{WP}{Wishart process} -\newacronym{serotonin}{5-HT}{serotonin} +% \newacronym{serotonin}{5-HT}{serotonin} diff --git a/misc/headings.tex b/misc/headings.tex index 5cb01fb..279cac2 100644 --- a/misc/headings.tex +++ b/misc/headings.tex @@ -1,13 +1,8 @@ %!TEX root=main.tex -% http://mirror.ox.ac.uk/sites/ctan.org/macros/latex/contrib/titlesec/titlesec.pdf -% https://tex.stackexchange.com/questions/338049/how-do-fonts-work-in-latex -% https://tex.stackexchange.com/questions/68745/possible-values-for-fontseries-and-fontshape - +% The tabto package is used to align titles horizontally regardless of the width of the number of the item which precedes the title. \usepackage{titlesec} \usepackage{tabto} -% The tabto package is used to align titles horizontally regardless of the width -% of the number of the item which precedes the title. % CHAPTERS \titleformat{\chapter}[display] % shape @@ -37,3 +32,8 @@ % {\thesubsubsection} % label text % {.5em} % label-to-title distance % {\optionalindent} % pre-title + +% REFERENCES +% http://mirror.ox.ac.uk/sites/ctan.org/macros/latex/contrib/titlesec/titlesec.pdf +% https://tex.stackexchange.com/questions/338049/how-do-fonts-work-in-latex +% https://tex.stackexchange.com/questions/68745/possible-values-for-fontseries-and-fontshape diff --git a/misc/preamble.tex b/misc/preamble.tex index 9308a83..0cd9ceb 100644 --- a/misc/preamble.tex +++ b/misc/preamble.tex @@ -1,7 +1,7 @@ % Returns the width of the current document in pts. % This can be useful when generating figures, so aspect ratios are preserved. % \showthe\textwidth -% Currently 360.0pt, independent from margin, but this could change still! +% Currently 360.0pt, independent from margin. % The inputenc package is ignored with utf8 based engines. % \usepackage[utf8]{inputenc} @@ -12,7 +12,6 @@ % BOXES % %%%%%%%%% -% For adding e.g. Box 1. \usepackage{tcolorbox} \newtcolorbox[auto counter]{mybox}[2][]{ float, @@ -22,28 +21,35 @@ #1 } -% https://en.wikibooks.org/wiki/LaTeX/Colors -% e.g. define a color, then add colback=my-blue -% \definecolor{my-blue}{cmyk}{0.80, 0.13, 0.14, 0.04, 1.00} +% REFERENCES +% https://en.wikibooks.org/wiki/LaTeX/Colors +% e.g. define a color, then add colback=my-blue +% \definecolor{my-blue}{cmyk}{0.80, 0.13, 0.14, 0.04, 1.00} -%%%%%%%%%%% -% GENERAL % -%%%%%%%%%%% +%%%%%%%% +% MATH % +%%%%%%%% \usepackage{amsfonts} % blackboard math symbols \usepackage{amsmath} \usepackage{amsthm} -%% + +%%%%%%%%%%% +% GENERAL % +%%%%%%%%%%% + \usepackage[toc,page]{appendix} +\usepackage[nottoc]{tocbibind} % the nottoc option removes Contents from Contents %% \usepackage{booktabs} % professional-quality tables \usepackage[font=small,labelfont=it]{caption} % small is 11pt when normalsize is 12pt %% -\usepackage{cleveref} % must be loaded after amsmath +\usepackage{cleveref} % CAUTION: must be loaded after amsmath %% -% \usepackage{mathptmx} % Times New Roman +% \usepackage{mathptmx} % use Times New Roman \usepackage[T1]{fontenc} % use 8-bit T1 fonts %% +% The printing company recommended equal margins on all pages; so no binding offset. \usepackage[ left=30mm, right=30mm, @@ -51,16 +57,16 @@ bottom=30mm ]{geometry} \usepackage{graphicx} -\usepackage{lettrine} % for dropped capital letters (2 lines high) +% \usepackage{lettrine} % for dropped capital letters (2 lines high) \usepackage{mathtools} %% \usepackage{microtype} % microtypography (improves visual appearance) \usepackage{nicefrac} % compact symbols for 1/2, etc. +% \usepackage{pdfpages} % for including PDF pages \usepackage{setspace} % define line spacing in paragraph %% \usepackage[labelfont=bf,textfont=normalfont]{subcaption} % allows for subplots %% -\usepackage{lipsum} % Dummytext \usepackage{xargs} % Use more than one optional parameter in a new commands \usepackage{url} % simple URL typesetting @@ -73,19 +79,23 @@ \fancyhead{} \fancyfoot{} -% \fancyhead[RE]{Chapter \thechapter} -% \fancyhead[RO]{\rightmark} - -% \fancyhead[LE]{\thepage} % uncomment for two-sided version -% \fancyhead[RO]{\thepage} % uncomment for two-sided version +% \fancyhead[LE]{\thepage} % uncomment for final (HARDBOUND) two-sided version +% \fancyhead[RO]{\thepage} % uncomment for final (HARDBOUND) two-sided version +\fancyhead[R]{\thepage} % uncomment for one-sided draft and final (ELECTRONIC) two-sided version -\fancyhead[R]{\thepage} % uncomment for one-sided version (draft) +% REFERENCES +% https://www.overleaf.com/learn/latex/How_to_Write_a_Thesis_in_LaTeX_(Part_2)%3A_Page_Layout -% \fancyfoot[LE,RO]{\thepage} +%%%%%%%%% +% OTHER % +%%%%%%%%% \input{misc/todo} \usepackage{lineno} -% \doublespacing \onehalfspacing % \linespread{1.25} % this is equal to 1.5 linespacing in Microsoft Word +% \doublespacing + +\author{Onno Pepijn Kampman} +\date{September 2022} diff --git a/misc/references.tex b/misc/references.tex index 9405131..767734c 100644 --- a/misc/references.tex +++ b/misc/references.tex @@ -3,39 +3,41 @@ backend=biber, % biber is the default, can use bibtex or bibtex8 as well backref=true, % whether to add citation page(s) to References bibstyle=authoryear, % can set to 'draft' when writing + bibwarn=true, block=space, % none, space, par, nbpar (does not allow page breaks), or ragged citestyle=authoryear, % citation style, e.g. Kampman (2022) - doi=true, % whether to print DOIs in References + doi=true, % whether to print DOIs in References, true for ELECTRONIC version, should be false for HARDBOUND version isbn=false, % whether to print International Standard Serial Number (ISSN) in References mincitenames=1, maxcitenames=3, minbibnames=2, maxbibnames=10, sorting=nyt, % 'count' for sorting in order of number of times cited - url=false, % whether to print URLs in References - bibwarn=true + url=false % whether to print URLs in References ]{biblatex} \addbibresource{references.bib} +% TODO: how can we print a newline before the DOI? +% \DeclareFieldFormat*{doi}{\printunit{\newline}#1} % this works and prints the correct DOI, but without the proper formatting +% \DeclareFieldFormat*{doi}{\printunit{\newline}} % uncomment for HARDBOUND version, this works by replacing the DOI with a newline + \renewcommand*{\nameyeardelim}{\addcomma\space} % add comma between author and year \renewcommand*{\bibfont}{\small} % smaller font size for References % The hyperref package transforms citations into hyperlinks. \usepackage[ - citecolor=blue, + citecolor=Gray, % `Gray' for ELECTRONIC version, `Black' for HARDBOUND version colorlinks, - linkcolor=MidnightBlue, + linkcolor=MidnightBlue, % `MidnightBlue' for ELECTRONIC version, `Black' for HARDBOUND version pdfborder={0 0 0}, pdfauthor={Onno Pepijn Kampman}, - pdfkeywords={functional connectivity, time-varying functional connectivity, Wishart process, depression}, - pdfsubject={This thesis studies robust estimation of time-varying functional connectivity using Wishart processes, and its utility for understanding depression.}, + pdfkeywords={functional connectivity, time-varying functional connectivity, machine learning, Wishart process, depression, PhD thesis, University of Cambridge}, + pdfsubject={This thesis studies robust estimation of time-varying functional connectivity using Wishart processes and its utility for understanding depression.}, pdftitle={Robust time-varying functional connectivity estimation and its relevance for depression}, - urlcolor=blue + urlcolor=RoyalBlue % `RoyalBlue' for ELECTRONIC version, `Black' for HARDBOUND version ]{hyperref} -% \usepackage[printonlyused,withpage]{acronym} - % REFERENCES % https://mirrors.concertpass.com/tex-archive/macros/latex/contrib/biblatex/doc/biblatex.pdf diff --git a/misc/title_page.tex b/misc/title_page.tex index ef37c5b..f21a6e8 100644 --- a/misc/title_page.tex +++ b/misc/title_page.tex @@ -1,14 +1,14 @@ \begin{titlepage} \begin{center} - \vspace*{6mm} + \vspace*{4mm} \Huge \textbf{Robust time-varying functional connectivity estimation and its relevance for depression} \vspace{19mm} - \includegraphics[width=0.28\textwidth]{misc/queens_college_arms_without_crest} + \includegraphics[width=0.25\textwidth]{misc/queens_college_arms_without_crest} \vspace{12mm} @@ -25,7 +25,7 @@ This thesis is submitted for the degree of\\ \textit{Doctor of Philosophy} - \vspace{19mm} + \vspace{32mm} Queens' College \hspace*{\fill} September 2022 diff --git a/misc/todo.tex b/misc/todo.tex index 73e6b89..79e7dbe 100644 --- a/misc/todo.tex +++ b/misc/todo.tex @@ -1,5 +1,5 @@ \usepackage[ - obeyDraft, % if enabled, the todo notes will only show when running a draft + obeyDraft, % uncomment for final version (both ELECTRONIC and HARDBOUND); if enabled, the todo notes will only show when running a draft colorinlistoftodos, prependcaption, textsize=tiny % recommended: footnotesize, scriptsize, or tiny (the smallest) diff --git a/references.bib b/references.bib index f344fd0..ff2fd83 100644 --- a/references.bib +++ b/references.bib @@ -304,10 +304,8 @@ @article{Bassett2011 year = {2011} } @article{Bassett2013, - archivePrefix = {arXiv}, author = {Bassett, Danielle S. and Porter, Mason A. and Wymbs, Nicholas F. and Grafton, Scott T. and Carlson, Jean M. and Mucha, Peter J.}, doi = {10.1063/1.4790830}, - eprint = {1206.4358}, journal = {Chaos}, number = {1}, title = {{Robust detection of dynamic community structure in networks}}, @@ -390,11 +388,9 @@ @inproceedings{Bell2021 year = {2021} } @inproceedings{Bell2022, - archivePrefix = {arXiv}, author = {Bell, Samuel J. and Kampman, Onno P. and Dodge, Jesse and Lawrence, Neil D.}, booktitle = neurips, doi = {10.48550/arXiv.2206.05985}, - eprint = {2206.05985}, title = {{Modeling the machine learning multiverse}}, year = {2022} } @@ -440,10 +436,8 @@ @article{Berman2011 year = {2011} } @article{Betzel2016, - archivePrefix = {arXiv}, author = {Betzel, Richard F. and Fukushima, Makoto and He, Ye and Zuo, Xi Nian and Sporns, Olaf}, doi = {10.1016/j.neuroimage.2015.12.001}, - eprint = {1511.06352}, journal = ni, pages = {287--297}, title = {{Dynamic fluctuations coincide with periods of high and low modularity in resting-state functional brain networks}}, @@ -451,12 +445,10 @@ @article{Betzel2016 year = {2016} } @incollection{Betzel2022, - archivePrefix = {arXiv}, author = {Betzel, Richard F.}, booktitle = {Connectomic Deep Brain Stimulation}, doi = {10.1016/b978-0-12-821861-7.00002-6}, editor = {Horn, Andreas}, - eprint = {2010.01591}, pages = {25--58}, title = {{Network neuroscience and the connectomics revolution}}, year = {2022} @@ -501,9 +493,8 @@ @article{Bollerslev1986 year = {1986} } @article{Bommasani2021, - archivePrefix = {arXiv}, author = {Bommasani, Rishi and Hudson, Drew A. and Adeli, Ehsan and Altman, Russ and Arora, Simran and von Arx, Sydney and Bernstein, Michael S. and Bohg, Jeannette and Bosselut, Antoine and Brunskill, Emma and Brynjolfsson, Erik and Buch, Shyamal and Card, Dallas and Castellon, Rodrigo and Chatterji, Niladri and Chen, Annie and Creel, Kathleen and Davis, Jared Quincy and Demszky, Dora and Donahue, Chris and Doumbouya, Moussa and Durmus, Esin and Ermon, Stefano and Etchemendy, John and Ethayarajh, Kawin and Fei-Fei, Li and Finn, Chelsea and Gale, Trevor and Gillespie, Lauren and Goel, Karan and Goodman, Noah and Grossman, Shelby and Guha, Neel and Hashimoto, Tatsunori and Henderson, Peter and Hewitt, John and Ho, Daniel E. and Hong, Jenny and Hsu, Kyle and Huang, Jing and Icard, Thomas and Jain, Saahil and Jurafsky, Dan and Kalluri, Pratyusha and Karamcheti, Siddharth and Keeling, Geoff and Khani, Fereshte and Khattab, Omar and Koh, Pang Wei and Krass, Mark and Krishna, Ranjay and Kuditipudi, Rohith and Kumar, Ananya and Ladhak, Faisal and Lee, Mina and Lee, Tony and Leskovec, Jure and Levent, Isabelle and Li, Xiang Lisa and Li, Xuechen and Ma, Tengyu and Malik, Ali and Manning, Christopher D. and Mirchandani, Suvir and Mitchell, Eric and Munyikwa, Zanele and Nair, Suraj and Narayan, Avanika and Narayanan, Deepak and Newman, Ben and Nie, Allen and Niebles, Juan Carlos and Nilforoshan, Hamed and Nyarko, Julian and Ogut, Giray and Orr, Laurel and Papadimitriou, Isabel and Park, Joon Sung and Piech, Chris and Portelance, Eva and Potts, Christopher and Raghunathan, Aditi and Reich, Rob and Ren, Hongyu and Rong, Frieda and Roohani, Yusuf and Ruiz, Camilo and Ryan, Jack and R{\'{e}}, Christopher and Sadigh, Dorsa and Sagawa, Shiori and Santhanam, Keshav and Shih, Andy and Srinivasan, Krishnan and Tamkin, Alex and Taori, Rohan and Thomas, Armin W. and Tram{\`{e}}r, Florian and Wang, Rose E. and Wang, William and Wu, Bohan and Wu, Jiajun and Wu, Yuhuai and Xie, Sang Michael and Yasunaga, Michihiro and You, Jiaxuan and Zaharia, Matei and Zhang, Michael and Zhang, Tianyi and Zhang, Xikun and Zhang, Yuhui and Zheng, Lucia and Zhou, Kaitlyn and Liang, Percy}, - eprint = {2108.07258}, + doi = {10.48550/arXiv.2108.07258}, journal = {arXiv}, pages = {1--214}, title = {{On the opportunities and risks of foundation models}}, @@ -570,10 +561,8 @@ @article{Brown2014b year = {2014} } @article{Bru1991, - archivePrefix = {arXiv}, author = {Bru, Marie-France}, doi = {10.1007/BF01259552}, - eprint = {1201.3256}, journal = {Journal of Theoretical Probability}, number = {4}, pages = {725--751}, @@ -898,11 +887,9 @@ @article{Cronbach1955 %%%%%%%%%%%%%%%%%%%%%%% @inproceedings{Dadashkarimi2021, - archivePrefix = {arXiv}, author = {Dadashkarimi, Javid and Karbasi, Amin and Scheinost, Dustin}, booktitle = {International Conference on Medical Image Computing and Computer-Assisted Intervention}, doi = {10.1007/978-3-030-87199-4_28}, - eprint = {2107.01303v1}, organization = {Springer}, pages = {293--302}, title = {Data-driven mapping between functional connectomes using optimal transport}, @@ -1145,9 +1132,8 @@ @thesis{Duvenaud2014 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% @article{Ebrahimi2020, - archivePrefix = {arXiv}, author = {Ebrahimi, Mohammadreza and Calarco, Navona and Campbell, Kieran and Hawco, Colin and Voineskos, Aristotle and Khisti, Ashish}, - eprint = {2006.05572v2}, + doi = {10.48550/arXiv.2006.05572}, journal = {arXiv}, title = {{Time-resolved fMRI shared response model using Gaussian process factor analysis}}, year = {2020} @@ -1348,10 +1334,8 @@ @article{Fischl2012 year = {2012} } @article{Flint2021, - archivePrefix = {arXiv}, author = {Flint, Claas and Cearns, Micah and Opel, Nils and Redlich, Ronny and Mehler, David M. A. and Emden, Daniel and Winter, Nils R. and Leenings, Ramona and Eickhoff, Simon B. and Kircher, Tilo and Krug, Axel and Nenadic, Igor and Arolt, Volker and Clark, Scott and Baune, Bernhard T. and Jiang, Xiaoyi and Dannlowski, Udo and Hahn, Tim}, doi = {10.1038/s41386-021-01020-7}, - eprint = {1912.06686}, journal = npp, number = {8}, pages = {1510--1517}, @@ -1396,19 +1380,16 @@ @article{Fox2007 year = {2007} } @article{Fox2011, - archivePrefix = {arXiv}, author = {Fox, Emily B. and West, Mike}, - eprint = {1107.5239}, + doi = {10.48550/arXiv.1107.5239}, journal = {arXiv}, title = {{Autoregressive models for variance matrices: Stationary inverse Wishart processes}}, year = {2011} } @inproceedings{Fox2015, - archivePrefix = {arXiv}, author = {Fox, Emily B. and Dunson, David B.}, booktitle = jmlr, doi = {10.5555/2789272.2912079}, - eprint = {1101.2017}, number = {1}, pages = {2501--2542}, title = {{Bayesian nonparametric covariance regression}}, @@ -1554,9 +1535,8 @@ @article{Gershman2010 year = {2010} } @article{Gershman2019, - archivePrefix = {arXiv}, author = {Gershman, Samuel J.}, - eprint = {1901.07945}, + doi = {10.48550/arXiv.1901.07945}, journal = {arXiv}, title = {{What does the free energy principle tell us about the brain?}}, year = {2019} @@ -1827,10 +1807,9 @@ @book{Hastie2009 year = {2009} } @inproceedings{Heaukulani2019, - archivePrefix = {arXiv}, author = {Heaukulani, Creighton and {van der Wilk}, Mark}, booktitle = neurips, - eprint = {1906.09360}, + doi = {10.48550/arXiv.1906.09360}, pages = {4584--4594}, title = {{Scalable Bayesian dynamic covariance modeling with variational Wishart and inverse Wishart processes}}, volume = {32}, @@ -1846,10 +1825,9 @@ @article{Heine2012 year = {2012} } @inproceedings{Hensman2013, - archivePrefix = {arXiv}, author = {Hensman, James and Fusi, Nicol{\`{o}} and Lawrence, Neil D.}, booktitle = {Uncertainty in Artificial Intelligence - Proceedings of the 29th Conference, UAI 2013}, - eprint = {1309.6835}, + doi = {10.48550/arXiv.1309.6835}, pages = {282--290}, title = {{Gaussian processes for big data}}, year = {2013} @@ -1873,20 +1851,17 @@ @article{Ho2021 year = {2021} } @inproceedings{Hoffman2015, - archivePrefix = {arXiv}, author = {Hoffman, Matthew D. and Blei, David M.}, booktitle = jmlr, - eprint = {1404.4114}, + doi = {10.48550/arXiv.1404.4114}, pages = {361--369}, title = {{Structured stochastic variational inference}}, volume = {38}, year = {2015} } @article{Holme2012, - archivePrefix = {arXiv}, author = {Holme, Petter and Saram{\"{a}}ki, Jari}, doi = {10.1016/j.physrep.2012.03.001}, - eprint = {1108.1780}, journal = {Physics Reports}, number = {3}, pages = {97--125}, @@ -2218,10 +2193,8 @@ @article{Kessler1998 year = {1998} } @article{Khosla2019, - archivePrefix = {arXiv}, author = {Khosla, Meenakshi and Jamison, Keith and Ngo, Gia H. and Kuceyeski, Amy and Sabuncu, Mert R.}, doi = {10.1016/j.mri.2019.05.031}, - eprint = {1812.11477}, journal = {Magnetic Resonance Imaging}, pages = {101--121}, title = {{Machine learning in resting-state fMRI analysis}}, @@ -2238,18 +2211,16 @@ @article{Kim2021 year = {2021} } @inproceedings{Kingma2014, - archivePrefix = {arXiv}, author = {Kingma, Diederik P. and Welling, Max}, booktitle = iclr2, - eprint = {1312.6114}, + doi = {10.48550/arXiv.1312.6114}, title = {{Auto-encoding variational Bayes}}, year = {2014} } @inproceedings{Kingma2015, - archivePrefix = {arXiv}, author = {Kingma, Diederik P. and Ba, Jimmy}, booktitle = iclr, - eprint = {1412.6980}, + doi = {10.48550/arXiv.1412.6980}, title = {Adam: A method for stochastic optimization}, year = {2015} } @@ -2438,10 +2409,8 @@ @article{Lakens2013 year = {2013} } @article{Lan2017, - archivePrefix = {arXiv}, author = {Lan, Shiwei and Holbrook, Andrew and Elias, Gabriel A. and Fortin, Norbert J. and Ombao, Hernando and Shahbaba, Babak}, doi = {10.1214/19-BA1173}, - eprint = {1711.02869}, journal = {Bayesian Analysis}, number = {4}, pages = {1199--1228}, @@ -2811,11 +2780,9 @@ @article{Matsui2019 year = {2019} } @inproceedings{Matthews2016, - archivePrefix = {arXiv}, author = {Matthews, Alexander G. de G. and Hensman, James and Turner, Richard E. and Ghahramani, Zoubin}, booktitle = {Artificial Intelligence and Statistics}, - doi = {10.48550/arxiv.1504.07027}, - eprint = {1504.07027}, + doi = {10.48550/arXiv.1504.07027}, pages = {231--239}, title = {On sparse variational methods and the Kullback-Leibler divergence between stochastic processes}, organization = {PMLR}, @@ -2864,9 +2831,8 @@ @article{Megevand2014 year = {2014} } @article{Mehler2018, - archivePrefix = {arXiv}, author = {Mehler, David Marc Anton and Kording, Konrad Paul}, - eprint = {1812.03363}, + doi = {10.48550/arXiv.1812.03363}, journal = {arXiv}, title = {{The lure of misleading causal statements in functional connectivity research}}, year = {2018} @@ -2882,10 +2848,8 @@ @article{Menon2010 year = {2010} } @article{Meunier2009, - archivePrefix = {arXiv}, author = {Meunier, David and Lambiotte, Renaud and Fornito, Alex and Ersche, Karen D. and Bullmore, Edward T.}, doi = {10.3389/neuro.11.037.2009}, - eprint = {1004.3153}, journal = fninf, number = {OCT}, title = {{Hierarchical modularity in human brain functional networks}}, @@ -2986,10 +2950,9 @@ @book{Murphy2023 year = {2023} } @inproceedings{Murray2010, - archivePrefix = {arXiv}, author = {Murray, Iain and Adams, Ryan Prescott and MacKay, David J. C.}, booktitle = jmlr, - eprint = {1001.0175}, + doi = {10.48550/arXiv.1001.0175}, pages = {541--548}, title = {{Elliptical slice sampling}}, volume = {9}, @@ -2999,10 +2962,8 @@ @inproceedings{Murray2010 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% @article{Nakajima2017, - archivePrefix = {arXiv}, author = {Nakajima, Jouchi and West, Mike and others}, doi = {10.1214/17-BJPS364}, - eprint = {1606.08292}, journal = {Brazilian Journal of Probability and Statistics}, volume = {31}, number = {4}, @@ -3048,9 +3009,8 @@ @article{Nichols2017 year = {2017} } @article{Nielsen2016, - archivePrefix = {arXiv}, author = {Nielsen, S{\o}ren F. V. and Madsen, Kristoffer H. and R{\o}ge, Rasmus and Schmidt, Mikkel N. and M{\o}rup, Morten}, - eprint = {1601.00496v2}, + doi = {10.48550/arXiv.1601.00496}, title = {{Nonparametric modeling of dynamic functional connectivity in fMRI data}}, url = {http://mloss.org/software/view/205/}, year = {2016} @@ -3197,11 +3157,10 @@ @article{Park2018 year = {2018} } @article{Parkes2020, - archivePrefix = {arXiv}, author = {Parkes, Linden and Satterthwaite, Theodore D. and Bassett, Danielle S.}, - eprint = {2006.04728}, + doi = {10.48550/arXiv.2006.04728}, title = {{Towards precise resting-state fMRI biomarkers in psychiatry: Synthesizing developments in transdiagnostic research, dimensional models of psychopathology, and normative neurodevelopment}}, - journal = {arXiv preprint arXiv:2006.04728}, + journal = {arXiv}, year = {2020} } @article{Parrish2000, @@ -3472,10 +3431,9 @@ @article{Ramasubbu2014 year = {2014} } @inproceedings{Ranganath2014, - archivePrefix = {arXiv}, author = {Ranganath, Rajesh and Gerrish, Sean and Blei, David M.}, booktitle = jmlr, - eprint = {1401.0118}, + doi = {10.48550/arXiv.1401.0118}, pages = {814--822}, title = {{Black box variational inference}}, volume = {33}, @@ -3640,10 +3598,8 @@ @article{Sakoglu2010 year = {2010} } @article{Salimans2013, - archivePrefix = {arXiv}, author = {Salimans, Tim and Knowles, David A.}, doi = {10.1214/13-BA858}, - eprint = {1206.6679}, journal = {Bayesian Analysis}, number = {4}, pages = {837--882}, @@ -3825,10 +3781,8 @@ @article{Shirer2012 year = {2012} } @article{Shmueli2010, - archivePrefix = {arXiv}, author = {Shmueli, Galit}, doi = {10.1214/10-STS330}, - eprint = {1101.0891}, journal = {Statistical Science}, number = {3}, pages = {289--310}, @@ -4420,10 +4374,9 @@ @article{Voytek2022 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% @inproceedings{Wagstaff2012, - archivePrefix = {arXiv}, author = {Wagstaff, Kiri L.}, booktitle = icml29, - eprint = {1206.4656}, + doi = {10.48550/arXiv.1206.4656}, pages = {529--534}, title = {{Machine learning that matters}}, volume = {1}, @@ -4492,14 +4445,14 @@ @phdthesis{Wilk2019 } @article{Wilk2020, author = {{van der Wilk}, Mark and Dutordoir, Vincent and John, ST and Artemev, Artem and Adam, Vincent and Hensman, James}, - journal = {arXiv preprint arXiv:2003.01115}, + doi = {10.48550/arXiv.2003.01115}, + journal = {arXiv}, title = {A framework for interdomain and multioutput Gaussian processes}, year = {2020} } @article{Wilkinson2021, - archivePrefix = {arXiv}, author = {Wilkinson, William J. and Solin, Arno and Adam, Vincent}, - eprint = {2103.10710}, + doi = {10.48550/arXiv.2103.10710}, journal = {arXiv}, title = {{Sparse algorithms for Markovian Gaussian processes}}, year = {2021} @@ -4515,19 +4468,16 @@ @article{Willinger2022 year = {2022} } @inproceedings{Wilson2010, - archivePrefix = {arXiv}, author = {Wilson, Andrew Gordon and Ghahramani, Zoubin}, booktitle = {Proceedings of the 27th Conference on Uncertainty in Artificial Intelligence, UAI 2011}, doi = {10.48550/arXiv.1101.0240}, - eprint = {1101.0240}, title = {Generalised Wishart processes}, year = {2010} } @inproceedings{Wilson2012, - archivePrefix = {arXiv}, author = {Wilson, Andrew Gordon and Knowles, David A. and Ghahramani, Zoubin}, booktitle = icml29, - eprint = {1110.4411}, + doi = {10.48550/arXiv.1110.4411}, pages = {599--606}, title = {{Gaussian process regression networks}}, volume = {1}, @@ -4548,10 +4498,8 @@ @thesis{Wilson2014 year = {2014} } @article{Winter2022, - archivePrefix = {arXiv}, author = {Winter, Nils R. and Leenings, Ramona and Ernsting, Jan and Sarink, Kelvin and Fisch, Lukas and Emden, Daniel and Blanke, Julian and Goltermann, Janik and Opel, Nils and Barkhau, Carlotta and Meinert, Susanne and Dohm, Katharina and Repple, Jonathan and Mauritz, Marco and Gruber, Marius and Leehr, Elisabeth J. and Grotegerd, Dominik and Redlich, Ronny and Jansen, Andreas and Nenadic, Igor and N{\"{o}}then, Markus M. and Forstner, Andreas and Rietschel, Marcella and Gro{\ss}, Joachim and Bauer, Jochen and Heindel, Walter and Andlauer, Till and Eickhoff, Simon B. and Kircher, Tilo and Dannlowski, Udo and Hahn, Tim}, doi = {10.1001/jamapsychiatry.2022.1780}, - eprint = {2112.10730}, journal = {JAMA Psychiatry}, title = {{Quantifying deviations of brain structure and function in major depressive disorder across neuroimaging modalities}}, year = {2022} @@ -4730,10 +4678,8 @@ @article{Zarghami2020 year = {2020} } @article{Zeidman2019, - archivePrefix = {arXiv}, author = {Zeidman, Peter and Jafarian, Amirhossein and Corbin, Nad{\`{e}}ge and Seghier, Mohamed L. and Razi, Adeel and Price, Cathy J. and Friston, Karl J.}, doi = {10.1016/j.neuroimage.2019.06.031}, - eprint = {1902.10597}, journal = ni, pages = {174--190}, title = {{A guide to group effective connectivity analysis, part 1: First level analysis with DCM for fMRI}}, @@ -4751,10 +4697,8 @@ @article{Zhang2011 year = {2011} } @article{Zhang2016, - archivePrefix = {arXiv}, author = {Zhang, Zitong and Telesford, Qawi K. and Giusti, Chad and Lim, Kelvin O. and Bassett, Danielle S.}, doi = {10.1371/journal.pone.0157243}, - eprint = {1512.07596}, journal = pone, number = {6}, title = {{Choosing wavelet methods, filters, and lengths for functional brain network construction}},