Renee Mapa '21
2020 SOAR Profile
Computer Vision Tracking Framework for Parkinson’s Disease Model Rats
Major and Minor: Cognitive Neuroscience
Hometown: Saylorsburg, PA
Project Advisor(s): Dr. Jeffrey Bush
Briefly describe your project.
Using a combination of computer vision and machine learning techniques, Dr. Jeffrey Bush and I created a computer software that uses computer vision to identify a rat and track its location and movements without any physical markers or harnesses for identification. As opposed to other computer vision rat-tracking projects, we applied this software to track the actions and locations of Parkinson’s Disease (PD) rat models and control rats. Through a series of filters and algorithms, we attempted to collect quantitative data to identify differences between the actions of the control rats and PD rat models. By tracking the actions and movements of the rats while performing foot-fault tests, we obtained insight into movement patterns that may be missed by the human eye. We used the filtered images as training input for the machine learning model which would identify and count when the rats would make a foot fault. Through the quantitative analysis of the acquired data between the observable differences in the rat groups, we were able to supplement the qualitative data and explore specific aspects of PD.
Describe the origin of your project. (E.g., did you pitch the idea and choose a faculty member, or did they come to you with an idea?)
Dr. Bush and I had been wanting to work together for a while. His research primarily focused on image processing on different types of neuronal dendrites. Since I was one of the only students at the time interested in both neuroscience and computer science, we decided to collaborate on a project that would explore how the two can be integrated to further the research being done in both fields.
What’s the best part about working with your faculty mentor? What valuable insights have they brought to your project?
Due to Dr. Bush’s extensive experience and knowledge on the subject, he was able to introduce me to a whole new field of neuroscience and computer science research that prepared me for future courses and potential career opportunities. Before SOAR, I had never considered a future in computational neuroscience, but through this experience, I discovered how interesting of a field it is.
What has been your biggest obstacle so far?
COVID-19 was our biggest obstacle, as we were working remotely and were unable to get into the lab to get our own image and video footage of the rat groups. However, we were able to overcome this by utilizing Dr. Cecilia Fox’s footage from previous years.
What has been your biggest takeaway from this experience?
The fields of neuroscience and computer science are young and growing quickly. Taking an interdisciplinary approach can be beneficial as each field can significantly contribute to the other. The possibilities for the future are endless, and it is very exciting to be a part of it.
What was the result of your project?
After separating each of the trial videos and cropping out the unnecessary, distracting information, we separated each video into individual frames to obtain our input data to start training the machine learning model. Through trial and error, we identified the best combination of smoothing and edge detection filters to isolate the crucial patterns in the rat’s body and feet for accurate and consistent identification. To prepare the data for the algorithm, we wrote a function that would layer all these filters and stack the arrays in sequence along a third axis. With this function, we were able to achieve more reproducibility on any dataset to try various transformations to see which combination yields the best results. With this filtered input, we began the training process by feeding the model the input along with an identifying label. The model can then pick up on these patterns and classify them into one of the two labeled groups, which would be either the presence or absence of a foot-fault. For the future, we would play around with different footage as the input to determine what the optimal angle or background would be to yield the most accurate machine learning algorithm.
In your own words, how do you feel about being awarded this opportunity? Why should other students take advantage of the SOAR program at É«ÖÐÉ«?
The privilege of working one-on-one with a professor as an undergrad is an experience that students at É«ÖÐÉ« should not overlook. You will gain invaluable knowledge, research skills, and professional experience in a way that cannot be attained from the classroom alone.
Now that SOAR is over, do you plan to expand upon your research? If so, how?
Dr. Bush and I continued our collaboration into the fall semester as an independent study course on computational neuroscience. Although we did not expand on our computer vision and machine learning research, we continued to explore how the two fields of computer science and neuroscience can be integrated through basic neuron models and artificial neural networks.
Have you, or do you plan to present this research outside the SOAR presentations? If so, where? Be specific, if possible.
I have not presented my research outside of the SOAR presentations.