/ EN
2026-07-24 00:00 United Kingdom Papers Foundations & Methods Translated from EN

Deep Learning Localizes Epileptogenic Zones

Summary The researchers developed a deep learning architecture that analyzes multichannel interictal intracranial EEG to localize epileptogenic zones without requiring seizure-period recordings or manual annotation. The model combines a Morlet-wavelet temporal Transformer with a spatial attention encoder. It was evaluated with 50.5 hours of recordings from 161 patients and 17,012 channels at 7 independent centers. Leave-one-center-out validation produced a pooled AUROC of 0.778 with a 95% confidence interval of 0.748 to 0.808, and discrimination remained above chance at every held-out center. The results indicate performance comparable to established electrophysiological baselines across centers and implantation modalities, but prospective clinical validation is still needed. This medRxiv preprint has not been peer reviewed.
Why it matters The model tests cross-center generalization on a large multichannel intracranial EEG dataset and uses interictal activity for localization, providing a substantial basis for AI-assisted presurgical assessment.

BCIwiki (bciwiki.com) — Developmental Neuroscience, Great Ormond Street Institute of Child Health, University College London, London, UK. posted the study as a preprint on medRxiv on July 24, 2026. The team developed a deep learning architecture for multichannel interictal intracranial EEG to generate hypotheses about epileptogenic-zone location. The model combines a Morlet-wavelet temporal Transformer with a spatial attention encoder.

The model was evaluated with 50.5 hours of intracranial EEG from 161 patients across 17,012 channels at 7 independent centers. Leave-one-center-out validation yielded a pooled AUROC of 0.778 with a 95% confidence interval of 0.748 to 0.808, with above-chance discrimination at every held-out center. The researchers report performance comparable to established electrophysiological baselines across centers and implantation modalities. However, the retrospective design and absence of structural imaging or effective-connectivity priors limit the conclusion, and prospective validation is still required. This study is a preprint and has not been peer reviewed.

Compiled by BCIwiki from public sources

Sources · 1
medrxiv.org 2026-07-24

University College London timeline

2026-08 Stanford's Palanker Wins Defense Health Agency Award for PRIMA Retinal Implant 2026-08 P300 BCI Reads Silently Chosen Digits in a Granada Classroom 2026-07 UCL Workshop Syncs EEG, Eye Tracking, ECG and GSR in a Flight Simulator All entries →
Read original ↗
Suggest a correction Revisions · none / EN
© 2026 BCIwiki.com Digest Topics Tips Subscribe Revisions About