EXPLAINABLE ARTIFICIAL INTELLIGENCE IN MEDICAL DIAGNOSIS: A COMPARATIVE REVIEW OF MODERN APPROACHES

Nowadays, machine learning and deep learning models are widely used for analyzing medical images, electronic health records, and clinical data. Despite their high predictive performance, many AI models operate as "black-box" systems, making their decision-making process difficult to understand. This lack of transparency limits clinicians' trust and slows the adoption of AI in clinical practice. Explainable Artificial Intelligence (XAI) has emerged as a promising solution to improve the transparency and interpretability of AI systems. XAI methods help clinicians understand how AI models generate predictions, thereby increasing confidence in automated medical diagnosis. This paper presents a comparative review of modern XAI techniques used in healthcare, including LIME, SHAP, Grad-CAM, Integrated Gradients, Attention Mechanisms, and Counterfactual Explanations. A systematic literature review was conducted to analyze recent studies published in reputable scientific journals. The reviewed approaches were compared based on interpretability, computational complexity, clinical applicability, and advantages and limitations. The analysis indicates that no single XAI method is universally suitable for all medical applications. Instead, selecting an appropriate explainability technique depends on the AI model, data type, and clinical objective. The findings highlight the importance of XAI in developing transparent, trustworthy, and reliable AI-assisted diagnostic systems. The study also discusses current challenges and future research directions for explainable medical AI.