Localization of Epileptic Seizure Focus by Computerized Analysis of fMRI Recordings

doi:10.21203/rs.3.rs-32204/v2

Download PDF

Research

Localization of Epileptic Seizure Focus by Computerized Analysis of fMRI Recordings

https://doi.org/10.21203/rs.3.rs-32204/v2

This work is licensed under a CC BY 4.0 License

Journal Publication

published 31 Oct, 2020

Read the published version in Brain Informatics →

You are reading this latest preprint version

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.

Medical Informatics

Time Series

Deep Learning

Mutual Information

fMRI

Epilepsy

Seizure Focus

Table 1: Lobe diagnosis of the cohort. The 5 classes are not perfectly disjoint because 4 patients were ambiguously diagnosed by clinicians and thus belong to 2 classes simultaneously (RT/C5, RF/RT, LF/LT, LT/RF).

Lobe	Frequency
Group 1 - Right Frontal (RF)	10
Group 2 - Left Frontal (LF)	8
Group 3 - Right Temporal (RT)	7
Group 4 - Left Temporal (LT)	11
Group 5 - Others: Right & Left Parietal + Right & Left Occipital (C5)	6

Table 2: Automatic classification accuracy in a leave-one-out design based on 18 strongly discriminating inter-connectivity coefficients at cortex regions scale

Patient Class	RF	LT	RT	LF	Other
Accuracy	82%	90.5%	100%	87%	90%

Table 3: Classification accuracy in leave-one-out design based on 18 highly discriminating Intra-connectivity Coefficients

Patient Class	RF	LT	RT	LF	Other
Accuracy	70%	80%	100%	87%	100%

Table 4: Classification accuracy for SVM in leave-one-out design based on 18 highly discriminating connectivity Coefficients

Patient Class	RF	LT	RT	LF	Other
Accuracy (Inter-connectivity)	%59	%64	%83	%78	%69
Accuracy (Intra-connectivity)	%57	%62	%73	%71	%64

Download PDF

Journal Publication

published 31 Oct, 2020

Read the published version in Brain Informatics →

Editorial decision: Accept
19 Oct, 2020
Editor assigned by journal
16 Oct, 2020
Submission checks completed at journal
15 Oct, 2020
Editor invited by journal
15 Oct, 2020

You are reading this latest preprint version

Localization of Epileptic Seizure Focus by Computerized Analysis of fMRI Recordings

Status:

Journal Publication

Version 2

Abstract

Figures

Full Text

Tables

Status:

Journal Publication

Version 2