Extracting interpretable features from large scale clinical EEGs using unsupervised learning

Gupta, Teja Borra

Permalink

https://hdl.handle.net/2142/120453 Copy

Description

Title

Extracting interpretable features from large scale clinical EEGs using unsupervised learning

Author(s)

Gupta, Teja Borra

Issue Date

2023-05-03

Director of Research (if dissertation) or Advisor (if thesis)

Varatharajah, Yogatheesan

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• EEG
• neurological disorders, tensor decomposition
• unsupervised learning
• interpretability
• variational autoencoder

Abstract

Analyzing clinical electroencephalograms (EEG) is crucial for diagnosing and monitoring neurological disorders. However, manual expert review is not scalable and is prone to errors. Thus, more efficient and reliable methods are needed. Current methods rely on two-dimensional decompositions such as principal component analysis (PCA) or indepedent component anal ysis (ICA) and deep learning methods such as autoencoders (AE) and self-supervised learning (SSL). However, these methods do not retain the naturalstructure of the data and not easily interpretable. To overcome these limitations, we propose using tensor decomposition (TD) to extract interpretable and clinically useful features from EEGs. Tensor decomposition retains the natural structure of the data and provides a more efficient and reliable alternative to traditional approaches. Additionally, to address the lack of expressivity with tensor decomposition, we explore ways to incorporate tensor decomposition with the variational autoencoder framework.

Graduation Semester

2023-05

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/120453

Copyright and License Information

Copyright 2023 Teja Gupta

Owning Collections