Development of a Seizure Matching System for Clinical Decision Making in Epilepsy Surgery

Abstract

AbstractBackground and ObjectivesThe proportion of patients becoming seizure-free after epilepsy surgery has stagnated. Large multi-center stereo-electroencephalography datasets can potentially allow comparing a new patient to past similar cases and make clinical decisions with the knowledge of how similar cases were treated in the past. However, the complexity of these evaluations makes the manual search for similar patients in a large database impractical. We aim to develop an automated system that electrographically and anatomically matches seizures from a patient to those in a database. In addition, since we do not know what features define seizure similarity, particularly considering the various stereo-electroencephalography implantation schemes, we evaluate the agreement and features among experts in classifying seizure similarity.MethodsWe utilized SEEG seizures from consecutive patients who underwent stereo-electroencephalography for epilepsy surgery. Eight international experts evaluated seizure-pair similarity using a four-level similarity score through a graphical user interface. As our primary outcome, we developed and validated an automated seizure matching system by employing a leave-one-expert-out approach. Secondary outcomes included the inter-rater agreement and features for classifying seizure similarity.Results320 SEEG seizures from 95 patients were utilized. The seizure matching system achieved an area-under-the-curve of 0.82 (95% CI, 0.819-0.822), indicating its feasibility. Six distinct seizure similarity features were identified and proved effective: onset region, onset pattern, propagation region, duration, extent of spread, and propagation speed. Among these features, the onset region showed the strongest correlation with expert scores (Spearman’s rho=0.75,p<0.001). Additionally, the moderate inter-rater agreement confirmed the practicality of our approach: for the four-level classification, median agreement was 73.9% (interquartile range, 7%), and beyond-chance Gwet’s kappa was 0.45 (0.16); for the binary classification of similar vs. not related, agreement stood at 71.9% (4.7%) with a kappa of 0.46 (0.13).DiscussionWe demonstrate the feasibility and validity of a stereo-electroencephalography seizure matching system across patients, effectively mirroring the expertise of epileptologists. This novel system can identify patients with seizures similar to that of a patient being evaluated, thus optimizing the treatment plan by considering the treatment and the results of treating similar patients in the past, potentially resulting in an improved surgery outcome.