Toward Predicting Peripheral Artery Disease Treatment Outcomes Using Non-Clinical Data

Peripheral Artery Disease (PAD) significantly impairs quality of life and presents varying degrees of severity that correctly identifying would help choose the proper treatment approach and enable personalized treatment approaches. However, the challenge is that there is no single agreed-on measure to quantify the severity of a patient with PAD. This led to a trial-and-error approach to deciding the course of treatment for a given patient with PAD. This study uses non-clinical data, such as biomechanical data and advanced machine-learning techniques, to detect PAD severity levels and enhance treatment selection to overcome this challenge. After analyzing biomechanical data from 42 healthy controls and 65 patients with PAD before and after treatment and correlating it with other measures such as quality of life questionnaires, our findings reveal that ground reaction forces (GRF) features emerged as robust indicators of PAD severity. The GRF Propulsive Peak, in particular, demonstrated high accuracy (0.909) in quantifying PAD severity and is used to develop a straightforward metric for assessing PAD severity. This severity metric is used to gauge the outcome of a specific PAD treatment by comparing the severity before and after the treatment. Machine-learning models were then developed to predict such post-treatment outcomes effectively from the patient non-clinical data before treatment. This approach showed promise in predicting the effectiveness of a treatment for a patient with PAD before performing it, highlighting the potential of machine learning models in revolutionizing PAD treatment strategies. Our findings lay the groundwork for a more data-driven, patient-centric approach to PAD management, optimizing treatment strategies for better patient outcomes.