Ai-powered chest x-ray diagnostic web application with multi-colored heatmap visualization and structured disease outcome analysis

Abstract

An AI-powered chest X-ray diagnostic system utilizing a pretrained DenseNet-121 convolutional neural network model, trained on over 100,000 publicly available NIH chest X-ray images, is presented. The system processes uploaded chest X-ray images in standard formats (JPEG, JPG, PNG) and generates an 8×8 probabilistic heatmap grid with smooth transitions using a Plasma colormap for enhanced clarity, and an optional Cividis colormap for accessibility. Diagnostic outcomes are displayed above the heatmap, while a structured summary of 14 predefined diseases is presented below, mapped through disease-specific bell curve models into five confidence ranges: Highly Unlikely, Unlikely, Possible, Likely, and Highly Likely, each color-coded for intuitive interpretation. The system supports dynamic disease switching, automatic saving of analysis results, and dual deployment options for both offline installations and cloud-based web applications, ensuring flexibility, accuracy, and seamless integration into clinical workflows.