Emotional Brain State Classification on fMRI data using 3D Residual Neural Networks
Video
Team Information
Team Members
Xiaofu He, Assistant Professor, Department of Psychiatry, Vagelos College of Physicians and Surgeons
Maxime Tchibozo, Data Science Institute, Columbia University (Alumni)
Zijing Wang, Department of Statistics, Columbia University
Donggeun Kim, New York State Psychiatric Institute
Abstract
Emotional brain states classification of fMRI data aims to classify brain activity patterns in regions of interest (ROIs) - such as amygdala and prefrontal cortex - which are activated during an emotion task. Brain state classification is also referred to as cognitive state classification, or brain decoding, and has been used in real-time fMRI neuro-feedback research and Brain-Computer Interfaces (BCI) to successfully assist individuals with motor disabilities in interacting with their environment. However, the task of selecting informative features from the whole 3D brain fMRI image for classifiers is still challenging. In this project, we propose a brain state classification model based on a 3D adaptation of the Residual Neural Network (ResNet), which can classify each repetition time (TR) fMRI data with promising accuracy, while avoiding the selection of ROIs.
Contact this Team
Contact: Xiaofu He (use form to send email)
