Shallow Learning for Predictive Blood Test Anomaly Detection: Case Study for Rheumatic Diseases

The emergence of Artificial Intelligence (AI) over the past three decades has given rise to numerous innovative and practical solutions, leading to a transformative revolution in clinical diagnostics and healthcare processes. While blood tests serve as vital diagnostic tools in modern medicine, there has been a surge in leveraging AI-driven solutions to enhance the accuracy and efficiency of blood test analysis and streamline the monitoring process. This study proposes and applies a shallow learning method to create a predictive anomaly detection in blood tests, focusing on rheumatic disease patients undergoing Disease-Modifying Antirheumatic Drugs (DMARDs) treatment. Employing a stacked ensemble Machine Learning (ML) model including gradient boosting, support vector machine, and shallow neural networks, the developed methods identify irregularities in blood test data and provide forecasts for individual patients. With significant performance measures, this study highlights shallow learning potential in improving clinical decision support for patient-level blood test management and monitoring.