Deep Learning for Rodent Behavioral Analysis – UROP Spring Symposium 2022

Deep Learning for Rodent Behavioral Analysis

photo of presenter

James Kessler

Pronouns: He/Him

Research Mentor(s): Daniel Leventhal
Co-Presenter:
Research Mentor School/College/Department: Neurology and Biomedical Engineering / Medicine
Presentation Date: April 20
Presentation Type: Oral5
Session: Session 4 – 2:40pm – 3:30 pm
Room: Breakout room 4
Authors: Daniel Leventhal, James Kessler
Presenter: 4

Abstract

Ever since Parkinson’s disease was discovered in 1817, it has been an active field of research in the medical community. As a neurological disease that inhibits motor skills due to a loss of cells that produce dopamine, doctors, neuroscientists, and psychologists alike have all tried to tackle the issue. In our study of Parkinson’s, we taught rats to reach for food pellets on command, and then recorded the results from when their dopamine receptors were restricted. In order to accelerate the process of data analysis, we utilized unsupervised learning algorithms called DeepLabCut and B-SOiD in order to track the motion of the rats and categorize their movements into different categories in an automated fashion rather than manually. This allowed us to process more data at a much quicker pace. We found that as dopamine was restricted, the rats’ depth perception and motor coordination harshly declined, but as soon as they regained normal levels of dopamine, they were able to successfully complete the task once again. From our results, we gathered that humans with Parkinson’s can be similarly treated so that if we can find a dopamine “sweet spot,” their disease will be much less of a hindrance on their daily lives.

Presentation link

Biomedical Sciences, Interdisciplinary

lsa logoum logo