Applications of machine learning for peripheral artery disease diagnosis and management: A systematic review

Peripheral artery disease (PAD) is a chronic condition caused by atherosclerosis, leading to arterial narrowing and obstruction, primarily in the lower extremities. This results in reduced blood flow and increases the risk of loss of limbs and mortality. Early diagnosis is essential for preventing complications and improving patient outcomes. Machine learning (ML), a subset of artificial intelligence, provides a non-invasive and efficient approach not only for diagnosing PAD but also for guiding management strategies by analyzing large datasets and identifying complex patterns. This systematic review explores the application of ML algorithms in PAD diagnosis and management, focusing on data types, whether numerical or non-numerical, features, performance metrics, software tools, and predicted outcomes. A comprehensive literature search in PubMed, Scopus, and Web of Science identified 30 relevant studies published between 2014 and 2024. The reviewed studies span various machine learning domains such as regression, classification, and clustering. These studies have utilized different techniques, including neural networks, both fully connected and convolutional, ensemble learning, and deep learning. A risk of bias assessment was performed across five domains to evaluate study reliability. Findings indicate that clinical records were the primary data source in approximately 50 % of studies. Random forest was the most frequently used algorithm for PAD analysis. ML models were applied to both diagnostic and risk assessment datasets, demonstrating their versatility. The overall risk of bias assessment revealed that 50 % of studies exhibited low risk across all domains. These findings highlight the potential of ML in enhancing PAD diagnosis and management.