Authors :
Presenting Author: Carlota Pagès Portabella, PhD – Universitat Politècnica de Catalunya
Manel Vila-Vidal, PhD – Universitat Politècnica de Catalunya; Mariam AlKhawaja, MD – Hospital Clínic; Estefanía Conde-Blanco, MD – Hospital Clínic; Mar Carreño, MD – Hospital Clínic; Pedro Roldán, MD – Hospital Clínic; Jordi Rumià, MD – Hospital Clínic; Antonio Donaire, MD – Hospital Clínic; Adrià Tauste Campo, PhD – Universitat Politècnica de Catalunya
Rationale:
The standard pre-surgical diagnostic procedure in drug-resistant epilepsy typically involves visual inspection of iEEG recordings to identify epileptogenic regions to resect. This procedure is time-consuming and still provides incomplete diagnosis, leaving those patients not eligible for surgery with no treatment option. In this context, we developed a web-based computational tool to help clinicians identify SOZs and plan resective surgery with more confidence.Methods:
We designed BrainFocus iEEG (BF), a cloud-based software based on an own-developed automatic SOZ detection algorithm (Vila-Vidal et al., 2017, Clin Neurophysiol, 128 (6), 977-985, doi.org/10.1016/j.clinph.2017.03.040; Vila-Vidal et al., 2020, NeuroImage, 208, 116410, doi.org/10.1016/j.neuroimage.2019.116410) that reveals the spectral signatures of emerging seizure-specific onset patterns. For each seizure, our algorithm calculates mean power activations and, based on these, identifies time-frequency windows where locally enhanced oscillations are maximized. Regions with enhanced Global Activation (GA) (the magnitude of the highest activations) and Spatial Spread (SS) (the spread of spectral activations) across recording sites are labeled as part of the SOZ. We retrospectively validated the core algorithm with 30 intracranially-monitored patients against the standard diagnostic information and reached an average sensitivity and specificity of ~.9. The cloud version of the algorithm has been improved according to the needs of clinical users and its performance is currently under further validation.Results:
Overall, clinical users have valued BF informative and intuitive to use. First, users can easily upload seizure files (in .edf format) and configure the parameters for analysis. Then, the software allows exploration of the original iEEG, removal of artifacted channels and features a seizure activation map and activation plot to inspect the relevant patterns more exhaustively. For a specific pair of GA and SS thresholds (that the user can configure), BF automatically highlights those contacts that most likely lie or are connected to the SOZ. BF also features a summary table of the relevant time-frequency onset patterns for each detected contact.Conclusions:
The use of BrainFocus can assist clinicians in the diagnosis of epileptic patients with iEEG, allowing SOZ validation in simple cases and revealing relevant epileptogenic patterns that might be invisible to the human eye in challenging cases.
Funding:
CPP, MVV and ATC were supported by the Spanish Ministry of Science and Innovation (PID2020-119072RA-I00, MCIN/AEI/) and by AGAUR (Catalan Agency for Management of University and Research Grants, grant no. 2021PROD00090 and grant no. 2021INNOV00027).