Date of Award

2026

Degree Type

Thesis

Degree Name

Master of Science in Electrical Engineering (MSEE)

Specialization

Biomedical Engineering

Department

Electrical, Computer, and Biomedical Engineering

First Advisor

Reza Abiri

Abstract

Severe upper-limb paralysis following spinal cord injury or stroke disrupts the pathways that translate motor intention into movement while leaving cortical planning networks intact. This thesis investigates whether macroscopic, noninvasive EEG can decode upper-limb and dexterous hand movements to support intention-driven brain-machine interface (BMI) control, through two complementary strategies: grasp planning and execution decoding and dexterous individual finger movement decoding.

Using a novel vision-based grasping platform that temporally isolates grasp planning from execution, low-frequency theta oscillations (4-7 Hz) were found to carry grasp-specific information during the planning phase, achieving 75.3% precision-versus-power classification accuracy before movement onset, compared to 61.1% for the traditional MRCP approach, while higher-frequency activity emerged only during execution and reflected sensorimotor integration rather than grasp discrimination. Expanding on this first approach, the second strategy focused on restoring hand dexterity by decoding individual finger movements from EEG, moving beyond fixed grasp categories toward a broader space of hand configurations. Using a multimodal platform that pairs EEG with simultaneously recorded hand kinematics, functional connectivity between electrodes was modeled with graph-based deep learning to learn the finger-specific network patterns underlying individual finger movements.

Overall, these findings demonstrate that grasp planning and dexterous individual finger movements can be decoded from noninvasive EEG, advancing toward scalable, movement-free neuroprosthetic control for individuals with severe motor impairment.

Download

Available for download on Sunday, September 10, 2028

Share

COinS