*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title: Knowledge-guided Multi-modal Data Integration for Interpretable EEG Seizure Detection
Committee:
Dr. D. Wang, Advisor
Dr. Fekri, Chair
Dr. Anderson
Abstract: The objective of this proposed research is to develop a novel pipeline for knowledge-guided multi-modal data integration for interpretable EEG seizure detection. Electroencephalogram (EEG) are widely used to detect seizures in clinical practices. As clinical diagnosis is time-consuming, expensive, and having low inter-rater agreement, automatic seizure detection is critical in computer-aided diagnosis. Despite recent progress using shallow features or deep learning models for seizure detection, the classification performance is far from satisfying. This proposed research consists of three phases. In Phase 1, I will learn word embeddings on clinical notes and pretrain the deep neural networks on EEGs from an external dataset to learn medical domain knowledge. In Phase 2, I will work with TUH EEG Seizure Corpus (TUSZ) dataset to do multi-task learning with the deep neural network model. The learned EEG features and multi-module word embedding representation will be integrated for seizure detection. In Phase 3, I will improve deep learning outcome interpretability by developing novel time-series EEG saliency map and scalp attention map. I will work on the proposed work between February 2021 and February 2022.
