Seminar: Reinforcement learning to predict sensor signals during walking in animal models and humans

Title: Towards Human–AI Safety
Speaker: Ashley Dalrymple, Assistant Professor, Dept of Biomedical Engineering at the U
2/27/24 – 2pm in WEB L102

Abstract: Walking is how we navigate our environments, explore, and get from place-to-place. After injury to the nervous system, such as with paralysis caused by spinal cord injury or stroke, walking is often impaired. Regaining the ability to walk is of high importance to people with paralysis. Currently, clinical rehabilitation methods can only take a patient so far in terms of the recovery of function. We aim to augment the rehabilitation using electrical stimulation and/or powered orthoses. However, the extent of paralysis and remaining muscle function varies with different types of injuries and across individuals, because most often the paralysis is incomplete. Therefore, any intervention aiming to improve mobility should use a personalized approach to accommodate the differences in residual function. In this talk, I will discuss how reinforcement learning can be used to predict signals recorded from various body-worn sensors during walking. I will first show how this was initially done in an animal model of hemiplegia (one-sided paralysis) and used to control stimulation of the spinal cord. I will then discuss how we plan to use a similar reinforcement learning approach in people to predict changes in walking terrain, so that we can provide adaptive and personalized control of a powered orthosis.

Bio: Dr. Ashley Dalrymple is currently an Assistant Professor in the Departments of Biomedical Engineering and Physical Medicine and Rehabilitation at the University of Utah. She completed her BSc in Electrical and Biomedical Engineering and her PhD in Neuroscience at the University of Alberta. She then went on to complete a postdoc at the Bionics Institute in Melbourne, Australia. Dr. Dalrymple moved to Pittsburgh to complete additional postdoc training, first at the University of Pittsburgh and then at Carnegie Mellon University. During that time, she obtained a pilot grant from the National Center of Neuromodulation for Rehabilitation and was awarded a prestigious fellowship from the Canadian Institutes of Health Research. The focus of Dr. Dalrymple’s work involves the use of implanted and non-invasive neural interfaces to restore function, particularly the rehabilitation of walking, restoring sensation, and reducing pain. Her research spans basic science, proof-of-concept, pre-clinical, and clinical studies.