Automated Disease Detectionand Reporting System

Abstract

Lung diseases such as pneumonia, tuberculosis and lung cancer continue to be the main health challenges worldwide, often resulting in delayed diagnosis and treatment. Tradi tional methods relying on manual interpretation of chest radiographs are time consuming, prone to human error, and may miss early-stage diseases. To address these limitations, this project presents an AI-based system that uses deep learning techniques for auto- mated lung disease detection and report generation. The system employs a hybrid model that combines the DenseNet-121 and EfficientNet-B5 architecture to improve both accu- racy and efficiency in data sets of varying sizes. A React-based front-end, Flask RESTful API, and MongoDB database is integrated to ensure a scalable, secure, and user-friendly platform. Pre-processing techniques such as noise reduction and data augmentation are applied to enhance model robustness. In addition, explainable AI methods such as Grad- CAM are incorporated to provide visual insights into the predictions of the model, in- creasing transparency for clinical users. The system also includes an automated report generation module that creates detailed diagnostic reports to support healthcare decision making. The experimental results demonstrate high diagnostic performance, suggesting that the proposed solution can improve the speed, accuracy, and accessibility of lung disease diagnosis, especially in settings with limited medical expertise.hat