This Proof of Concept (PoC) translates key results from the ERC‑COG SPEEDY project into a bedside workflow for presurgical language mapping in drug‑resistant epilepsy. Electrical Stimulation Mapping (ESM) is the clinical gold standard to avoid resecting eloquent cortex, but it is time‑consuming, fatiguing, and often yields noninformative negative stimulations when language is not engaged. SPEEDY-Clinic combines a patient‑friendly, self‑paced bedside acquisition system for naturalistic listening (podcasts/audiobooks) during intracranial EEG (iEEG) monitoring, together with an automated iEEG analysis pipeline. The pipeline extracts acoustic and hierarchical linguistic predictors (e.g., phoneme/syllable/word; entropy/surprise from large language models), dissociates acoustic from linguistic processing, and outputs individualized contact-level maps with confidence indices plus a small set of targeted ESM confirmation probes. A beta prototype has already been implemented and tested in >30 patients in research conditions, providing strong feasibility evidence. Over 18 months we will:
(O1) consolidate the prototype into a clinically deployable system (tablet app, automated iEEG analysis pipeline, and clinician report) and deliver standard operating procedures (SOPs) covering use, safety, data governance, and analysis;
(O2) prospectively validate clinical utility at AP-HM by comparison with routine ESM (contact-level concordance; mapping efficiency; patient-reported burden);
(O3) demonstrate portability via site-readiness testing and first replication with our Québec partner (CHU de Québec, Université Laval).
The expected outcome is a realistic, standardized and transferable route towards evidence-based language mapping that reduces patient burden and supports safer surgical decisions.
