The rapid advancement of artificial intelligence (AI) has significantly influenced medical image analysis, allowing automated systems to reach diagnostic accuracy comparable to that of expert clinicians. Among the many AI models, Convolutional Neural Networks (CNNs) have played a central role in this progress, proving highly effective in classifying complex medical images such as MRI, CT, an‎d histopathology slides. Despite their impressive predictive capability, CNNs often operate as “black boxes,” offering little insight into how their decisions are made. This limitation poses a major barrier to their clinical acceptance an‎d raises concerns about trust an‎d accountability. This review examines how explainable artificial intelligence (XAI) can enhance the transparency of CNN-based medical image classification, focusing on two widely used interpretation methods: Gradient-weighted Class Activation Mapping (Grad-CAM) an‎d Local Interpretable Model-Agnostic Explanations (LIME). Grad-CAM generates gradient-based visual heatmaps that show the regions influencing a model’s output, while LIME provides locally interpretable, model-independent explanations by analyzing small input variations. Drawing on literature published between 2020 an‎d 2025, this study compares the theoretical principles, implementation details, strengths, an‎d limitations of both methods across different clinical domains. The findings suggest that Grad-CAM offers faster an‎d more model-faithful results, whereas LIME provides more detailed an‎d intuitive explanations, though at the cost of computational efficiency an‎d consistency. The review concludes that combining these two complementary approaches into a unified explainability framework may improve the interpretability an‎d reliability of medical AI systems, bridging the gap between deep learning performance an‎d real-world clinical decision-making. In addition, the study highlights current research gaps an‎d outlines future directions for developing more robust hybrid explainability models to support transparent an‎d trustworthy clinical applications of AI keywords: Convolutional Neural Networks (CNNs); Medical Image Classification; Explainable Artificial Intelligence (XAI); Grad-CAM; LIME; Hybrid Explainability ; Clinical Decision Support.

10/27/2025 12:00:00 AM

87983

1404/08/08

Review of Medical Image Classification Using CNNs with Explainability Techniques (Grad-CAM, LIME)

Convolutional Neural Networks (CNNs) , LIME , Explainable Artificial Intelligence (XAI) , Grad-CAM , LIME , Clinical Decision Support.