The current pain estimation methodologies in the medical industry have some shortcomings due to the subjective nature of pain measurement and the labor involved in current measurement techniques. This Major Qualifying Project (MQP) addresses these problems by using Machine Learning and the UNBC McMaster Shoulder Pain dataset to develop an Automatic Pain Estimation system from patients’ facial expressions. The UNBC McMaster dataset contains images of individuals with shoulder pain and their pain scores at both the frame and sequence level. This project currently focuses on extracting spatial features from the images and applying neural networks of various designs to estimate Prkachin and Solomon Pain Intensity (PSPI) scores. Our results show that an ensemble model consisting of Vision Transformers (ViT) combined with Linear Regression shows the most promising results with a Pearson Correlation Coefficient (PCC) of 0.6456 and Root Mean Squared Error (RMSE) of 1.1286 compared to ground-truth pain scores.

• 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

• Goyal, Suryansh

Publisher

• Worcester Polytechnic Institute

Identifier

• 49826
• E-project-030622-123141

Advisor

• Whitehill, Jacob Richard

Year

• 2022

Date created

• 2022-03-06

Resource type

• Major Qualifying Project

Major

• Computer Science

Rights statement

• In Copyright