XAI-Driven Early Detection and Prevention of Brain Tumors: Integrating Grad-CAM and MedAlpaca-۷b for Interpretable Medical Imaging Reports

In this study, we present a novel explainable AI (XAI) pipeline for brain tumor diagnosis from medical images. Our approach integrates a convolutional neural network (CNN)-based classifier with Gradient-weighted Class Activation Mapping (Grad-CAM) to localize suspicious regions in brain MRI scans. We extract critical information such as the tumor type and probability score and convert it into structured textual prompts. These prompts are then fed into MedAlpaca-۷B, a biomedical language model, to generate natural language explanations of the tumor diagnosis. The proposed system enhances interpretability by bridging visual model outputs with human-understandable medical narratives, offering support for radiologists in clinical decision-making and improving transparency in AI-assisted diagnostics. By offering early and interpretable insights, our system contributes not only to accurate diagnosis but also to timely intervention and secondary prevention of cancer progression, helping clinicians monitor and manage high-risk cases before they worsen. This positions our method as a valuable tool in AI-powered early response strategies for brain tumor management.

کلیدواژه ها:

Brain Tumor Diagnosis ، Explainable AI (XAI) ، Grad-CAM ، MedAlpaca-۷B ، Textual Report Generation