Date of Award

2021

Degree Type

Dissertation

Degree Name

Doctor of Philosophy in Electrical Engineering

Specialization

Biomedical Engineering

Department

Electrical, Computer, and Biomedical Engineering

First Advisor

Yalda Shahriari

Abstract

Human brain studies that quantify neural functions using neuroimaging techniques have many applications related to neurological disorders, including characterizing symptoms, identifying biomarkers, and enhancing existing brain computer interface (BCI) systems. The first major goal of this dissertation is to quantify the neural functions associated with neurological impairments, specifically in amyotrophic lateral sclerosis (ALS), using two neuroimaging modalities, electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS), that respectively characterize electrical and hemodynamic neural functions. The next major goal is to integrate these modalities using state-of-the-art techniques including time-frequency based decompositions and functional and directional connectivity methods, and to use the quantified neural functions to classify different brain states through leading edge techniques, including information theory based fused feature optimization and deep learning based automatic feature extraction. In this dissertation, we explored the non-motor neural alterations in ALS patients reflected by simultaneously recorded EEG-fNIRS data both during task performance and in the resting state. Our results revealed significant neural alterations in ALS patients compared to healthy controls. Moreover, these neural signatures were used to classify data as coming from ALS patients versus healthy controls. For this purpose, we used mutual information-based fused feature optimization for EEG-fNIRS to select the best features from all the extracted neural markers, which considerably improved classification performance in classifying data as from people with ALS vs. healthy controls based on mental workload. These results support the idea of using complementary features from fused EEG-fNIRS in neuro-clinical studies for the optimized decoding of neural information, and thus, improving the performance of relevant applications, including BCIs and neuro-pathological diagnosis. In addition, we examined our findings in motor imagery classification, another fundamental processing step in applying BCIs for people with neurological disorders, including ALS patients. To do this, we proposed a convolutional neural network-based classification architecture for automatic feature extraction from EEG-fNIRS data, which outperformed conventional classification methods using manually extracted features. These outcomes suggest promising improvements in BCI performance using multimodal EEG-fNIRS and deep learning classifiers with automatic feature extraction, which can be utilized in clinical applications for people with neurological disorders including ALS patients. These findings can be further developed to automate the optimal quantification of neural functions in neurological disorders, with less dependence on prior knowledge, and thereby facilitate BCIs and other clinical applications for patients with neurological disorders.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.