Neuronal network architecture and temporal lobe epilepsy: A connectome-based and machine learning study

B. C. Munsell, Guorong Wu, S. Keller, J. Fridriksson, B. Weber, M. Styner, Dinggang Shen, L. Bonilha

Research output: Chapter in Book/Report/Conference proceedingChapter


This chapter provides an overview of two different clinical studies that apply machine learning techniques that create computational models capable of identifying abnormal network connections in the structural brain connectomes of patients with temporal lobe epilepsy (TLE). In particular, using only the structural connectome we introduce two new computational approaches aimed at predicting: (1) the surgical treatment outcome of patients with TLE, and (2) the naming impairment performance of patients with TLE. In both studies, prediction frameworks are trained to identify abnormal network connection patterns (ie, biomarkers) by applying supervised learning techniques to brain network features based on edge or node graph measures derived exclusively from the structural connectome. Furthermore, the performance of the proposed prediction frameworks is able to predict treatment outcomes in epilepsy with similar accuracy as compared with "expert-based" clinical decision, or is able to predict naming impairment outcomes that are very similar to real outcomes as observed on standard language tests.

Original languageEnglish
Title of host publicationMachine Learning and Medical Imaging
PublisherElsevier Inc.
Number of pages22
ISBN (Electronic)9780128041147
ISBN (Print)9780128040768
Publication statusPublished - 2016 Aug 9
Externally publishedYes


  • Computational modeling
  • Diffusion tensor imaging (DTI)
  • Machine learning
  • Naming impairment prediction
  • Structural brain connectome
  • Temporal lobe epilepsy (TLE)
  • Treatment outcome prediction

ASJC Scopus subject areas

  • Engineering(all)


Dive into the research topics of 'Neuronal network architecture and temporal lobe epilepsy: A connectome-based and machine learning study'. Together they form a unique fingerprint.

Cite this