Enhancing AI-CDSS with U-AnoGAN: Tackling Data Imbalance

Clinical Decision Support Systems (CDSS) have evolved significantly over the years to support healthcare professionals in making informed decisions about patient care. With the advent of AI, including machine learning and natural language processing, has enhanced the predictive power and precision of CDSS. For instance, it enables accurate disease prediction, diagnosis, treatment recommendations, and adverse event detection. However, one significant hurdle in applying AI algorithms to CDSS (AI-CDSS) is the issue of data imbalance and black box, especially for rare diseases or underrepresented demographic groups. AI models trained on imbalanced data may favor the majority class and poorly predict the minority class, impacting diagnosis and treatment decisions. Nevertheless, the proposed model, U-AnoGAN was trained using masks derived from normal data and calculated anomaly scores for the Covid-19 and pneumonia datasets. The U-AnoGAN surpasses the existing AnoGAN-related algorithms while enhancing interpretability through visualization of abnormal regions, and it augments diagnostic accuracy. Furthermore, it successfully addresses the daunting challenge of the data imbalance problem since it required only normal data. Since the healthcare sector wrestles with the escalating intricacy and volume of data, the role of AI-integrated CDSS becomes increasingly vital. The U-AnoGAN significantly bolsters the predictive prowess of AI-CDSS, paving the way for more precise, timely diagnoses with better visualization which could overcome the black box problem. In essence, the proposed model holds tremendous potential for diagnostic capabilities, elevates patient care with cutting-edge AI tools, and fosters more accurate and effective decision-making in healthcare environments.