Each neuron in the primary motor cortex (M1) is like a musician in an orchestra, contributing to a larger harmony under the constraint of a “neural manifold”—a geometric score describing the correlated signals produced by the neural musicians that drive movement. Despite the widespread recognition of the importance of M1 manifolds, the connection between their geometry and motor behavior remains unclear. Is the simple geometry a fundamental characteristic of M1, or is it an artifact resulting from the constrained laboratory tasks typically used in these investigations? Answering this question may enhance our understanding of how M1 orchestrates various movements. It also has practical implications for intracortical brain-computer interfaces (iBCIs), tools that translate neural activity into control of external devices for individuals with paralysis. My first study focused on understanding the geometric properties of neural manifolds using simulated neural signals. In my second study, I used this methodology to compare the geometry of M1 manifolds of monkeys during naturalistic, unconstrained behaviors to that during laboratory-based behaviors. Surprisingly, the geometry of M1 manifolds was only slightly more complex during naturalistic behaviors. Finally, I used M1 manifolds to improve the “decoders” of iBCIs, the brain-to-behavior maps that typically allow the control of external devices, to instead enable the control of the user’s own muscles. I used two strategies based on the neural manifold concept: first, a decoder that mapped neural signals in the M1 manifold of an individual with quadriplegia directly to the muscle activity of a monkey, and second, adapting a decoder built solely from monkey data for human use. My findings 1) show that the simple geometry of M1 manifolds is not just an artifact of laboratory constraints but rather a computational strategy that provides a harmonized tune for reliable movement control across tasks and 2) illuminate the potential of iBCIs to empower individuals with paralysis to regain control of their own muscles, even on the real-world stage outside the laboratory.

Creator

• Altan, Ege

DOI

• https://doi.org/10.21985/n2-kdyy-nj03

Subject

• Biomedical engineering
• Neurosciences

Language

• en

Alternate Identifier

• etdadmin_upload_1001502
• http://dissertations.umi.com/northwestern:16658

Keyword

• neural networks
• neural manifolds
• motor control
• machine learning
• brain-computer interface

Date created

• 2023-01-01

Resource type

• Dissertation

Rights statement

• In Copyright