Georgia Institute of Technology
Bridging the Gap from Research to Real-World Implementation: Deep Learning-Based Dynamical Systems Modeling Approaches to Overcome Practical Barriers for Intracortical Brain Computer Interfaces
Abstract
dc:description.abstractFor individuals with neuromotor impairments, intracortical brain-computer interfaces (iBCIs) can restore movement and communication capabilities by recording neural activity and decoding it to a control signal for an intended output. Despite high interest amongst potential users, very few individuals have had access to use an iBCI in the last 25 years. The work in this dissertation aims to address practical considerations of iBCI use in the hopes of improving future device availability and adoption. To do so, we leveraged powerful deep-learning models of neural population dynamics to develop approaches that may solve three primary obstacles to iBCI translation. First, we used neural dynamics models to improve the stability of iBCI decoder performance over long timescales, reducing the need for lengthy daily recalibration procedures and shortening device setup time. In addition, we developed a set of standardized datasets and metrics to benchmark decoder stabilization approaches and released a platform for evaluating future modeling innovations. Next, we proposed a modeling approach to reduce the power requirements of neural signal acquisition and transmission while maintaining high decoding accuracy. This effort may be useful for wireless iBCIs, which are largely preferred by potential users but have design constraints regarding battery life. Lastly, we applied neural dynamics models to investigate the neural basis of corrective movements and used our findings to develop a decoding strategy that may improve iBCI accuracy during precise movements when mistakes and subsequent corrections are common. Our work proposes solutions to practical barriers in iBCI translation from research settings to everyday use and establishes a framework for further improvements using neural dynamics models.
Degree
thesis:*- Level thesis:degree_level
- Doctoral
- Department dc:contributor.department
- Biomedical Engineering (Joint GT/Emory Department)
- Grantor dc:publisher
- Georgia Institute of Technology
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Karpowicz, Brianna M.
- Advisor dc:contributor.advisor
-
- Pandarinath, Chethan
- Committee members dc:contributor.committeemember
-
- Ting, Lena H
- Wu, Anqi
- Choi, Hannah
- Miller, Lee E
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/1853/80097