An Integrated XAI–LLM System for Lung Disease Diagnosis and Report Generation

Abstract

Lung diseases such as pneumonia, tuberculosis, and COVID-19 remain major contributors to global morbidity and mortality. Early and accurate diagnosis is a critical factor in reducing complications and improving patient outcomes. However, the similarity of radiographic features in chest X-ray images often leads to diagnostic ambiguity. This paper presents an intelligent deep learning framework capable of detecting multiple lung diseases from chest X-ray images using Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs). To enhance clinical interpretability, Explainable Artificial Intelligence (XAI) techniques, including Grad-CAM and SHAP, are employed to identify and visualize the regions that most influence the model’s predictions. Furthermore, the framework integrates a Large Language Model (LLM) to automatically generate clear, structured, and human-readable diagnostic reports, thereby reducing clinicians’ documentation burden. The proposed system is implemented as a user-friendly web application that enables real-time disease detection, heatmap visualization, and automated report generation. Experimental evaluation conducted on a public lung disease dataset comprising four classes demonstrates strong classification performance, achieving an accuracy of 92.8%, precision of 94%, recall of 93.2%, and F1-score of 93.1%. Visual analysis confirms accurate localization of disease-affected regions, supporting the reliability of the model’s predictions. Overall, the proposed framework provides a transparent, scalable, and cost-effective AI-based solution for automated lung disease diagnosis, contributing to improved clinical decision-making and facilitating the integration of artificial intelligence into real-world healthcare workflows.