Quantification of Behavioral Parkinsonian Symptoms in Rodents via Motion Data and Machine Learning
Author(s)
Gadekar, Amogh
Advisor(s)
Holder, Mary
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Parkinson’s Disease (PD) is a progressive neurodegenerative disease in which motor skills, control, and coordination are lost. While researchers are exploring PD and potential novel treatment options by studying animal models of PD in rodents, a strong evaluation system is needed to determine the success of the treatments and/or if the rodents were lesioned successfully. Current research studies using animal models have used qualitative tests such as the Limb-Use Asymmetry Test; however, as technology evolves, modern tools like machine learning could be used to produce a quantitative representation of behaviors which allows for further statistical analysis and comparisons between different stages of the rat’s condition. With this goal, this project endeavors to use machine learning and the animal pose estimation software DeepLabCut (DLC) to quantify behavioral analysis of rodents under the is 6-hydroxydopamine (6-OHDA) model using video data collected from rats in their pre-operated, post-operated, and treatment. Currently, we have already observed how using a DLC-trained model of 30 videos can identify quantitative trends corresponding to PD-like symptoms when 6-OHDA lesioning is successful (corroborated by histology of staining for tyrosine hydroxylase positive cells in the substantia nigra and striatum). Moving forward, we are further analyzing animal pose estimation data with Python libraries such as VAME and PyRat to classify and identify behavior clusters and metrics. Exploration with this technique in observing other movement disorders is also of interest, such as studying the epileptic mouse model with kainic acid (KA). Future implications of the project involve contribution to the development of reliable and convenient observations that concur with clinical professionals and only require easily accessible equipment and training. Additionally, the use of artificial intelligence and machine learning could detect abnormal involuntary movements which are not easily observed qualitatively.
Sponsor
Date
Extent
Resource Type
Text
Resource Subtype
Undergraduate Research Option Thesis