Individual-Specific Connectome Fingerprint Based Classification of Temporal Lobe Epilepsy.

Non-lesional temporal lobe epilepsy (TLE) is a syndrome of epilepsies that have no clear morphological change and cannot be diagnosed by structural imaging. TLE has been found extensive disruption in cortical functional connectome based on inaccurate traditional group-level atlas. In the present study, we utilized a well-recognized individualized functional parcellation method which achieve more accurate definition of functional areas compare to traditional group-level atlas to identify functional regions of interest (ROIs). Based on the individualized ROIs, we constructed individual-specific connectome fingerprint. By the aid of machine learning algorithm, we extract core features of individualized functional connectome and fed it into support vector machine to classify TLE patients. We parcellated the cerebral cortex to individual-specific functional networks and further identified 73 homologous functional ROIs utilized in the subsequent group-level analysis. The individualized connectome achieved better performance in the classification model in terms of multiple evaluating metrics compare to connectome fingerprint based on group-level atlas. This study verified the feasibility of individual-level parcellation method in the application of TLE classification and may provide potential imaging biomarkers for TLE.