Radiolytica: AI – Enhanced Radiology Workflow Revolution

Abstract

This project presents an AI-based diagnostic system that will help radiologists identify pneumonia, tuberculosis, and cardiomegaly at early stages using the analysis of chest X-rays. It has a secure web-based interface where radiologists can sign in, enter patient information, and upload the X-ray images. The uploaded images are subjected to backend preprocessing in the form of resizing, normalization, and enhancement, followed by lung segmentation based on the U-Net architecture to remove background noise and extract meaningful lung regions. The segmented images are then examined through DenseNet201-based deep learning model for detecting pneumonia and tuberculosis, whereas cardiomegaly is determined through Cardiothoracic Ratio (CTR) calculation. The system provides outputs in the form of confidence scores, severity levels, percentages of infection, and Grad-CAM heatmaps that outline infected regions for visual interpretability. Results are also presented on a userfriendly user interface and stored automatically with patient information on an Excel sheet to maintain systematic record-keeping. The system also allows radiologists to authenticate and validate predictions in real-time for ensuring clinical reliability.

Ähnliche Arbeiten