Automated Pain Assessment from Facial Expressions of Elderly People using Machine Learning

This thesis presents the development of an advanced machine learning model designed to accurately assess pain levels in dementia patients residing in elderly care homes. The project, conducted in collaboration with Sentigrate, a start-up focused on data science company, aims to create a predictive model that assigns pain scores ranging from 0 (no pain) to 6 (maximum pain) based on facial expressions. The research employs computer vision techniques, primarily convolutional neural networks, to extract meaningful features from facial images. A comparative study of various predictive techniques is conducted to determine the most effective approach. This project addresses the critical issue of inadequate pain management in dementia patients due to communication challenges. The objective is to provide an objective pain assessment tool that will significantly improve pain management strategies and enhance the quality of life for dementia patients in elderly care settings. The findings of this research have the potential to transform elderly care practices, offering valuable insights into pain management and contributing to the broader field of healthcare technology.