Add figure

Author: tsbinns

paper/Overview.png

@@ -49,18 +49,16 @@ Various forms of information can be extracted from neural timeseries data. Of th
 
 Analysis of phase-amplitude coupling, time delays, and non-sinusoidal waveshape characteristics provide important mechanistic insights into interneuronal communication [@Canolty2010;@Silchenko2010;@Sherman2016]. Studies of these features in neural timeseries data have been used to investigate core nervous system functions such as movement and memory, including their perturbation in disease [@deHemptinne2013;@Cole2017;@Bazzigaluppi2018;@Binns2024]. However, traditional methods for analysing this information have critical limitations that hinder their utility. In contrast, the bispectrum - the Fourier transform of the third order moment [@Nikias1987] - can be used for phase-amplitude coupling [@Zandvoort2021], non-sinusoidal waveshape [@Bartz2019], and time delay analyses [@Nikias1988], overcoming many of the limitations associated with traditional methods.
 
-Despite these benefits, the bispectrum has seen relatively little use in the field of neuroscience, in part due to the lack of an accessible, easy-to-use toolbox tailored to the analysis of electrophysiology data. Code written in MATLAB exists for electrophysiological analyses (see e.g., [github.com/sccn/roiconnect](https://github.com/sccn/roiconnect), [github.com/ZuseDre1/AnalyzingWaveshapeWithBicoherence](https://github.com/ZuseDre1/AnalyzingWaveshapeWithBicoherence)), however it is spread across multiple repositories, and often not in the form of a toolbox. Furthermore, use of this code requires a paid MATLAB license, limiting its accessibility. Code for computing the bispectrum can also be found written in the free-to-use Python language (e.g., @Stingray), however these implementations are not tailored to the analysis of electrophysiology data, limiting their use for neuroscience research. The `PyBispectra` package aims to address these limitations by providing a single, comprehensive toolbox for analysing phase-amplitude coupling, time delays, and non-sinusoidal waveshape characteristics in electrophysiology data with the bispectrum, including tutorials to facilitate an understanding of these analyses in the context of neuroscience research \autoref{fig:overview}. Data formats follow conventions from popular electrophysiological signal processing packages like `MNE-Python` [@Gramfort2013], and helper functions are provided as wrappers around `MNE-Python` and `SciPy` [@Virtanen2020] tools to facilitate data processing prior to bispectral analyses. Additional plotting tools are provided to visualise and aid the interpretation of results.
+Despite these benefits, the bispectrum has seen relatively little use in the field of neuroscience, in part due to the lack of an accessible, easy-to-use toolbox tailored to the analysis of electrophysiology data. Code written in MATLAB exists for electrophysiological analyses (see e.g., [github.com/sccn/roiconnect](https://github.com/sccn/roiconnect), [github.com/ZuseDre1/AnalyzingWaveshapeWithBicoherence](https://github.com/ZuseDre1/AnalyzingWaveshapeWithBicoherence)), however it is spread across multiple repositories, and often not in the form of a toolbox. Furthermore, use of this code requires a paid MATLAB license, limiting its accessibility. Code for computing the bispectrum can also be found written in the free-to-use Python language (e.g., @Stingray), however these implementations are not tailored to the analysis of electrophysiology data, limiting their use for neuroscience research. The `PyBispectra` package aims to address these limitations by providing a single, comprehensive toolbox for analysing phase-amplitude coupling, time delays, and non-sinusoidal waveshape characteristics in electrophysiology data with the bispectrum \autoref{fig:overview}, including tutorials to facilitate an understanding of these analyses in the context of neuroscience research. Data formats follow conventions from popular electrophysiological signal processing packages like `MNE-Python` [@Gramfort2013], and helper functions are provided as wrappers around `MNE-Python` and `SciPy` [@Virtanen2020] tools to facilitate data processing prior to bispectral analyses. Additional plotting tools are provided to visualise and aid the interpretation of results.
 
-![Overview of the `PyBispectra` toolbox.\label{fig:overview}](Overview.pdf)
+![Overview of the `PyBispectra` toolbox.\label{fig:overview}](Overview.png)
 
 # Features
 
 ## Phase-amplitude coupling
 
 Phase-amplitude coupling refers to the interaction between the phase of a lower frequency oscillation and the amplitude of a higher frequency oscillation, either within or across signals. It has been posited as a mechanism for the integration of neural information across spatiotemporal scales [@Canolty2010], with pathological alterations in coupling seen in neurological disorders [@deHemptinne2013;@Bazzigaluppi2018]. Common methods for quantifying phase-amplitude coupling include the Canolty approach [@Canolty2006] and modulation index [@Tort2010], which involve bandpass filtering signals in the frequency bands of interest and using the Hilbert transform to extract the phase and amplitude information. This approach has several limitations. First, the bandpass filters require precise properties that are not readily apparent, with incorrect filter properties smearing information across a broad frequency range, hindering the analysis of frequency-specific interactions [@Zandvoort2021]. Second, the Hilbert transform is a relatively demanding procedure, contributing to a high computational cost of the analysis. Finally, when analysing interactions between different signals, spurious coupling estimates can arise due to interactions within each signal which cannot be easily corrected for [@PellegriniPreprint]. In contrast, the bispectrum captures phase-amplitude coupling whilst overcoming these limitations. Specifically, coupling information is captured without requiring bandpass filtering, preserving the spectral resolution and reducing the risk of misinterpreting results [@Zandvoort2021]. Furthermore, bispectral analysis is computationally efficient, relying on the computationally cheap Fourier transform. Finally, spurious across-signal coupling estimates can be corrected for using bispectral antisymmetrisation [@Chella2014;@PellegriniPreprint]. `PyBispectra` provides the tools for performing bispectral phase-amplitude coupling analysis, with options for antisymmetrisation as well as normalisation using the threenorm [@Shahbazi2014], bounding coupling scores between 0 and 1 to enhance interpretability. Altogether, the bispectrum is a robust and computationally efficient approach for phase-amplitude coupling analysis, overcoming key limitations of traditional methods.
 
-\autoref{fig:overview}
-
 ## Time delay analysis
 
 Time delay analysis provides insight into the physical connections between brain regions by identifying latencies of information transfer between signals, complementing structural analyses using electrophysiology data [@Silchenko2010;@Binns2024]. A traditional approach for time delay analysis is cross-correlation, quantifying the similarity of signals at a set of time lags. However, there a two key limitations to this approach, including its limited robustness to noise in the data [@Nikias1988], as well as a vulnerability to spurious zero time lag interactions arising due to volume conduction and source mixing in the sensor space [@Chella2014]. On the other hand, the bispectrum is resilient to Gaussian noise sources in the data [@Nikias1988], and the process of antisymmetrisation can also be used to correct for spurious zero time lag interactions [@Chella2014]. `PyBispectra` provides tools for time delay analysis using the bispectrum, with options for antisymmetrisation, offering a robust method for the analysis of time delays in electrophysiology data.