Home
Welcome to the Semiology-Visualisation-Tool wiki!
- read the README
- Don't forget to pull before making changes.
This software preprocesses the Semio2Brain Database spreadsheet which has extracted thousands of patient-level seizure semiology localisation and lateralistion information from published literature, and converts them to a visualisation of the brain regions involved in surgery (epileptogenic zone), sEEG (seizure onset zone) or multimodal concordance (irritative, functional deficit and seizure onset zones).
In brief, it cleans the DataFrame, pivots occurrences of semiology, allowing for semiology_dictionary taxonomy replacement regex searches (SemioDict YAML), maps the documented localisations to gif parcellations and can scale these mappings using different scalers/transformers. Data can be filtered based on the ground-truth used to label the localisation and lateralisation, Concordance, stereo-EEG/electrical stimulation and Epilepsy-Topology.
The slicer module then takes the output of this module (in the format of gif parcellations and patient numbers/intensities) to allow a 3D visualisation of semiology to complete the Semiology Visualisation Tool.
The spreadsheet of semiology descriptions, shown below, can be downloaded from here
- Master branch - for download and use with 3D-Slicer for 3D-visualisations
- Profiling branch - #208 - see the GitHub branch original pull request for details on how to profile
- MegaAnalysis-March2020 branch - original verbose module for development and progress stats prior to integration with 3D Slicer with GUI ** (thenineteen jupyter notebook backwards compatibility)
- Mappings-calibration-starting-at-v-1.1.0 - for manual visual calibration of mapping descriptions of brain regions to GIF parcellations
- Dec 2018 project conception (JD, AM, GR)
- Jan 2019 - Aug 2020 Data collection (GR, AM)
- Aug 2019 first working module using jupyter notebook (AM)
- March/April 2020 first integrated SVT software with 3D slicer (AM, FPG)
- Aug 2020 complete database integrated
- 2021 Bayesian inference (AM)
We integrated our Semio2Brain database, taxonomy, brain region mappings to gif MNI space parcellations, ground truths and Bayesian filters (for exclusion of patients from studies that preselected patients based on prior knowledge of the region of seizure onset, to mitigate the publication bias that favours temporal lobe epilepsy) in a Python module for the 3D Slicer program, to create a novel user-friendly and open-source Semiology Visualisation Tool (SVT) in the form of a GUI. This allows 3D-brain visualisations of semiologies and their simultaneous combinations. Instead of idenitifying symptomatogenic zones, we therefore visualise the most likely original sources of the seizures from initial or most prominent semiology.
GUI options include filtering paediatric cases and determining the laterality of your patient's dominant hemisphere, ticking checkboxes for your patient's semiology and determining the laterality of the semiology.
Further details on the above can be found in the description of the (backend) mega_analysis module below.
v1.2.3 3rd March 2021: remove trailing sums from database (19th Aug 2020, 310 studies) v 1.0.0 (early Aug 2020) 100% of the Data: 4,649 patients' data from 312 included original journal papers
Until the 5th Aug 2020 the Beta version was used, named syst_review_single_table - 47% (146/312 included studies integrated)
We curated the largest patient-level database of 4649 unique patients' semiologies from 312 studies, yielding over ten thousand lateralising and localising datapoints for initial or most prominently reported seizure semiologies. These patient-level seizure-semiology data-points (where 1 point corresponds to 1 patient, presenting with a particular seizure semiology) were extracted from selected peer-reviewed journal publications, if at least one of the following ground-truth criteria regarding the certainty of lateralisation and/or localisation was satisfied:
- Post-operative seizure freedom (ILAE 1,2 = Engel Ia,Ib; but also Engel I if not otherwise specified), confirmed at a minimum follow-up of 12 months;
- Invasive EEG recording and/ or electrical stimulation, mapping seizure semiology;
- Multi-modal concordance between brain imaging & neurophysiology (e.g. PET, SPECT, MEG, EEG, fMRI) in pointing towards a highly probable epileptogenic zone.
The data are also tagged to allow Bayesian filtering of patient semiologies from studies which preselected patients with known brain seziure-region zones (whether this be epileptogenic zones or seizure onset zones) in order to allow mitigation of publications bias which favours temporal lobe epilepsy (e.g. of invasive EEG electrode targets) to allow filtering out specific paper/patient-level priors:
- Epilepsy Topology (ET): when the paper selects sample of patients based on their established epileptogenic zone (site of surgical resection) or seizure onset zone (neurophysiological/anatomical), and describes the related seizure semiology - e.g. papers looking at TLE, FLE, OLE;
- Spontaneous Semiology (SS): when the paper pre-selects a sample of patients based on their seizure semiology (e.g. nose-wiping, gelastic, ictal kissing), or reports on a cohort of unselected patients with epilepsy, or pre-selects based on other non-topological factors (specific techniques or conditions e.g. FCD) and provides details of epileptogenic zone localisation/lateralisation
- Cortical Electrical Stimulation (CES/ES): when the paper describes the semiology elicited by electrical brain stimulation, in the context of pre-/ intra-surgical functional mapping.
- If new data is to be added, must include paediatric under 7 label, and if postictal, must be typed "postictal" and not in any other format.
See our Wiki page for more details on the semiology categories and their semantic synonyms.
- 3rd March 2021: merged ictal speech and dysphasia
- 21st Jan 2021 v1.1.1 --> v1.2.1 --> v1.7.1: 36 semiologies, 7 postictals, 1 no semiology on stimulation to mirror Sankey diagram and SVT version
- Aug 2020:v 1.0.1 56+1 semiology categories
- v 1.0.0, 55+1 semiologies
- Up to May 2020: v 0.9.1 (47 semiologies)
3. Resources: lateralisation and localisation mappings - v 1.1.9 (Aug 2020, 103 localisation labels, Calibrations for minor versions)
until 5th Aug - v 1.0.8 (55 localisation labels) To query this database, we developed a taxonomy of 55 semiological terms (+ 1 "no semiology" for cortical stimulation studies), mapped reported categorical brain regions to 103 localising cerebral atlas labels and 5 lateralising possibilities (contra- or ipsi- lateral, dominant or non-dominant hemisphere and bilateral).
The mega_analysis module cleans the DataFrame of data, pivots occurences of semiology, allowing for semiology_dictionary taxonomy replacement regex searches, maps the documented localisations to gif parcellations and can scale these mappings using different scalers/transformers.
Data can be filtered based on the above ground truths, Bayesian priors and other exclusions.