کنفرانس بین المللی فناوری اطلاعات و دانش

English

صفحه اصلی / شانزدهمین کنفرانس بین المللی فناوری اطلاعات و دانش

Explainable AI for Medical Image Diagnosis Using Hybrid Attention-CAM Mechanisms

نویسندگان :

Negin Amirzadeh¹

1- دانشگاه ازاد قزوین

کلمات کلیدی :

(Explainable Artificial Intelligence (XAI،Medical Image Diagnosis،Hybrid Attention-CAM،Attention Mechanism،Gradient-weighted Class Activation Mapping (Grad-CAM)،Chest X-ray

چکیده :

Deep learning models have shown remarkable success in medical image diagnosis; however, their black-box nature limits clinical trust and adoption. To address this challenge, explainable artificial intelligence (XAI) methods such as Gradient-weighted Class Activation Mapping (Grad-CAM) and attention mechanisms have been explored. Nevertheless, existing approaches often suffer from noisy visual explanations or insufficient localization of disease-related regions. In this work, we propose a novel Hybrid Attention-CAM framework that integrates Grad-CAM with self-attention maps to generate more accurate and interpretable heatmaps for medical image diagnosis. The framework employs DenseNet-121 as a backbone for feature extraction, followed by an attention module to highlight clinically relevant regions. The heatmaps from Grad-CAM and attention are normalized and fused to provide refined visual explanations with reduced noise. We evaluate our method on the NIH ChestX-ray14 dataset, covering 14 thoracic diseases. Experimental results demonstrate that our approach improves diagnostic accuracy and AUC compared to baseline methods, while significantly enhancing explanation fidelity. The proposed framework enables physicians not only to receive reliable predictions but also to visualize the precise suspicious regions, bridging the gap between deep learning models and clinical interpretability.