Authors: Sayyed Yasmin Sultana Begum, Associate Professor M.Jyothi
Abstract: Accurate and transparent clinical evaluation of Metabolic Syndrome using radiographic imaging remains a significant challenge due to the subjective nature of manual interpretation and the limited explainability of conventional deep learning models. Although recent advances in artificial intelligence have substantially improved automated disease classification, the lack of model interpretability restricts their adoption in real-world clinical environments where transparency and trust are essential. This paper proposes an Explainable and Interpretable Smart Diagnostic Framework that combines deep convolutional feature learning with Explainable Artificial Intelligence (XAI) techniques to provide both accurate predictions and clinically meaningful explanations. The framework incorporates comprehensive image preprocessing, hierarchical feature extraction, multi-scale severity classification, and explanation-driven decision support through visualization and feature attribution mechanisms. Attention-based localization and feature contribution analysis enable clinicians to identify the anatomical regions and predictive factors influencing the diagnostic outcome, thereby enhancing model transparency and clinical confidence. Furthermore, training on heterogeneous radiographic datasets improves robustness, generalization, and adaptability across diverse patient populations. Experimental evaluation demonstrates that the proposed framework achieves high diagnostic performance while providing reliable interpretability, making it an effective decision-support system for intelligent clinical assessment and early disease severity evaluation.
