Neuroscience and computational modeling have a symbiotic relationship, with discoveries in each field inspiring the other. This paper explores the relationship between visual stimuli from the Natural Scenes Dataset and functional magnetic resonance imaging (fMRI) activity in distinct brain regions of interest (ROI). Valuable information can be extracted from images using components of various pretrained vision models known as feature extractors. This extracted information can be used by neural networks to predict how each ROI will respond to stimuli. Observing the patterns and behaviors, we identified specific regions of interest corresponding to categories identified by a classification model. This study found that YOLOv8n for classification and ResNet50 for feature extraction work best alongside linear regression. We then analyzed the patterns among the categories to identify which classes have similar activations.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Mani, Shivali
• Morales, Jose
• McNicholas, Alison

Subject

• Project-Based Learning
• Innovation
• Computing
• Health

Publisher

• Worcester Polytechnic Institute

Identifier

• E-project-042624-090213
• 121867

Stichwort

• Regression
• Grad-CAM
• fMRI
• Neuroscience
• Computer Vision

Advisor

• Zhang, Ziming

Year

• 2024

UN Sustainable Development Goals

• Not Specified

Date created

• 2024-04-26

Resource type

• Major Qualifying Project

Major

• Data Science
• Computer Science

Source

• E-project-042624-090213

Rights statement

• In Copyright