DiaXplain: A transparent and interpretable artificial intelligence approach for Type-2 diabetes diagnosis through deep learning

Artificial intelligence has become a pivotal element in the healthcare industry, and it is used in healthcare administration, predictive modeling, decision-making, and diagnostics. Artificial intelligence technologies have achieved performance levels akin to humans in several activities; nevertheless, their widespread adoption is impeded by preconceptions of these systems as inscrutable “black boxes”. In response to such requirements, the authors devised “DiaXplain”, an interpretable hybrid artificial intelligence model intended to assist in diagnosing diabetes mellitus. DiaXplain utilizes data from the National Health and Nutrition Examination Survey (NHANES) and integrates a convolutional neural network (CNN) for feature extraction with XGBoost for classification, guaranteeing both elevated accuracy and interpretability. This hybrid methodology allows the model to autonomously discern pertinent properties while the XGBoost element enhances its prediction efficacy. DiaXplain utilizes SHAP (SHapley Additive exPlanations), an explainable artificial intelligence methodology that elucidates each prediction, assisting users in comprehending the reasoning behind individual and aggregate model actions. The model’s performance was assessed using necessary measures, revealing that DiaXplain exceeds current approaches in accuracy, attaining an exceptional 98.24%, with a precision of 95.12% and an F1-score of 97.50. Every prediction is substantiated by SHAP-generated explanations, providing doctors with a transparent understanding of the causes influencing each diagnostic result. Unlike traditional black-box systems, DiaXplain offers local and global interpretability, making its decisions understandable and trustworthy. This dual focus on high diagnostic accuracy and explainability makes DiaXplain a novel contribution to AI-driven healthcare, bridging the critical gap between prediction power and clinical transparency.This clarity is vital for cultivating confidence among healthcare providers, facilitating more informed decision-making, and enhancing the management of diabetes mellitus. DiaXplain signifies a significant progression in artificial intelligence-driven healthcare by merging high diagnostic accuracy with interpretable forecasts, providing healthcare practitioners with a dependable tool for diabetes diagnosis that promotes precision and fosters trust via transparency.