Localization of epileptic seizure focus by computerized analysis of fMRI recordings

Abstract

By computerized analysis of cortical activity recorded via fMRI for pediatric epilepsy patients, we implement algorithmic localization of epileptic seizure focus within one of eight cortical lobes. Our innovative machine learning techniques involve intensive analysis of large matrices of mutual information coefficients between pairs of anatomically identified cortical regions. Drastic selection of pairs of regions with biologically significant inter-connectivity provides efficient inputs for our multi-layer perceptron (MLP) classifier. By imposing rigorous parameter parsimony to avoid overfitting, we construct a small-size MLP with very good percentages of successful classification.

Deep learning in neuroimaging of epilepsy

“…DL models for diagnosis and lateralization are also based on functional neuroimaging (Table 2). Three studies used resting-state fMRI (rs-fMRI) [45][46][47] and two [48,49] used interictal [ 18 F]FDG-PET imaging.…”

“…Two other DL models based on fMRI have been proposed for EZ lateralization with high accuracy values. In one [46], an FFNN was trained with 18 functional connectivity coefficients reaching an accuracy of 89%, higher than that obtained by SVM algorithm (71%) with the same dataset. The other [47] used rs-fMRI time series to train a CNN achieving 90% accuracy.…”

“…Briefly, Tables 1 and 2 overview 10 articles that developed DL models for diagnosis and/or lateralization in different types of epilepsy [40][41][42][43][44][45][46][47][48][49]. These models were based on functional or structural imaging and in most images were preprocessed to obtain metrics used as input.…”

“…Whether the multimodal approach could improve the performance of DL models is unknown. Some investigations based on connectivity, both structural [41,43] and functional [45][46][47], identified the relevant connections in the classification of patients. However, the role of these potential connectivity biomarkers remains to be investigated and validated in future studies.…”