Reconstruction of Resting State FMRI Using LSTM Variational Auto-Encoder on Subcortical Surface to Detect Epilepsy

Yunan Wu, Pierre Besson, Emanuel A. Azcona, S. Kathleen Bandt, Todd B. Parrish, Aggelos K. Katsaggelos

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Functional MRI offers unique insights for the characterization and presurgical evaluation of people with epilepsy (PWE). In this paper, we develop a graph-based variational auto-encoder (gVAEs) to 1) learn the patterns of resting state functional MRI (rsfMRI) within the brain's subcortical structures in healthy subjects and 2) reconstruct it in PWE to identify findings unique to patients with epilepsy. The gVAE was enriched with Sequential Long Short Term Memory (LSTM) and perceptual loss to learn temporal rsfMRI features and smooth the reconstructed signals. Using a cross-validation approach on healthy controls, our best model yielded an average spatial correlation of 0.791 and an average temporal correlation of 0.793. When applied to PWE, the average and spatial correlation decreased to 0.752 and 0.750 respectively. Our findings pave the path to the development of a whole brain data-driven tool that may be valuable for the characterization of abnormalities within the epileptic brain. This may advance our understanding as to how these abnormalities are related to the location of seizure onset and can inform the care of patients with epilepsy.

Original languageEnglish (US)
Title of host publicationISBI 2022 - Proceedings
Subtitle of host publication2022 IEEE International Symposium on Biomedical Imaging
PublisherIEEE Computer Society
ISBN (Electronic)9781665429238
DOIs
StatePublished - 2022
Event19th IEEE International Symposium on Biomedical Imaging, ISBI 2022 - Kolkata, India
Duration: Mar 28 2022Mar 31 2022

Publication series

NameProceedings - International Symposium on Biomedical Imaging
Volume2022-March
ISSN (Print)1945-7928
ISSN (Electronic)1945-8452

Conference

Conference19th IEEE International Symposium on Biomedical Imaging, ISBI 2022
Country/TerritoryIndia
CityKolkata
Period3/28/223/31/22

Keywords

  • LSTM
  • epilepsy
  • graph-based Variational auto-encoder
  • rsfMRI reconstruction

ASJC Scopus subject areas

  • Biomedical Engineering
  • Radiology Nuclear Medicine and imaging

Fingerprint

Dive into the research topics of 'Reconstruction of Resting State FMRI Using LSTM Variational Auto-Encoder on Subcortical Surface to Detect Epilepsy'. Together they form a unique fingerprint.

Cite this