Highlights from our recent work, presented by Dr. Vera Gramigna at the Explainable AI for Biomedical Images and Signals Special Session of the 32nd Italian Workshop on Neural Networks (WIRN) 2024!

📄 Title of the Paper: Bringing Explainability to the Prediction of Gait Disorders from Ground Reaction Force (GRF): A Machine Learning Study on the GaitRec Dataset

Our research focuses on improving the prediction of gait disorders by analyzing ground reaction force (GRF) patterns using advanced machine learning techniques. We leveraged the GaitRec dataset, which includes GRF measurements from individuals with various musculoskeletal conditions, to develop a model that can distinguish between healthy controls and those with gait disorders.

What makes our work unique is the use of Explainable Boosting Machines (EBMs). Unlike traditional “black-box” models, EBMs provide transparency by showing which specific features of the gait data contribute to the predictions. This not only enhances the model’s accuracy but also allows clinicians to understand the reasoning behind each prediction, making AI tools more trustworthy and easier to integrate into clinical practice.

Key results:

• Our model achieved an accuracy of 88.2% in predicting gait disorders.
• We identified that specific frames of the right vertical GRF were crucial in distinguishing between healthy individuals and those with gait disorders.
• The model’s explainability also revealed potential areas of improvement, such as the challenge in accurately classifying healthy controls, likely due to the diversity within the gait disorder category.

This work is a step forward in combining AI with clinical expertise, paving the way for more precise and understandable diagnostic tools in healthcare.

#AI #MachineLearning #GaitAnalysis #ExplainableAI #BiomedicalEngineering #WIRN2024