MASA

Project Description

MASA (Machine Learning Assisted Swallowing Assessment) is an innovative project designed to leverage the power of machine learning (ML) to transform the way swallowing assessments are performed, with an aim to improve quality of care. The primary goal of this project is to harness ML capabilities in spectral analysis of human voice during swallowing assessments to classify the assessment state as per the TOR-BSST standards, a recognized tool for assessing swallowing disorders particularly after acute stroke.

This project involves the application of advanced techniques such as Convolutional Neural Networks (CNNs) initially, with plans to expand to Visual Transformers (ViT) in the future. The use of these cutting-edge technologies is aimed at capturing and understanding the intricate nuances of human voice during swallowing assessments that are potentially missed in traditional assessments.

The algorithms developed under MASA aim to bring efficiency, accuracy and scalability in the assessment process, potentially enabling clinicians to make more informed decisions regarding patient treatment and management.

Project Objectives

Develop a robust machine learning model capable of performing spectral analysis on human voice during swallowing assessments. Validate the model’s performance against TOR-BSST labelling. Improve the quality and efficiency of swallowing assessments, particularly in the context of post-acute stroke care. Explore the incorporation of advanced technologies such as Visual Transformers to enhance the model’s capabilities. Getting Started Refer to the Getting Started Guide for instructions on how to install, run, and use this project.