Fast and Efficient Lung Abnormality Identification With Explainable AI: A Comprehensive Framework for Chest CT Scan and X-Ray Images

Fast and Efficient Lung Abnormality Identification With Explainable AI: A Comprehensive Framework for Chest CT Scan and X-Ray Images

Radiologists face mounting workloads interpreting thousands of chest scans, creating demand for fast, trustworthy AI screening tools. This project delivers a comprehensive explainable AI (XAI) framework for detecting lung abnormalities — including pneumonia, tuberculosis, nodules, and COVID-19 patterns — from chest X-rays and CT scans. The system employs a high-performance transformer-CNN hybrid architecture that combines the local feature strength of CNNs with the global context modeling of vision transformers, achieving both speed and accuracy. Crucially, every prediction is paired with Grad-CAM and attention heatmaps that visually justify the model's decision, addressing the black-box problem that limits clinical adoption of AI. The framework includes a triage module that prioritizes urgent cases, real-time inference optimized for standard hardware, and a radiologist-friendly dashboard displaying anomaly localization, confidence scores, and comparative analytics. Trained and validated on public chest imaging datasets, the system reports competitive accuracy, sensitivity, and specificity metrics. Developed in Python with PyTorch, OpenCV, and Streamlit, this is a flagship explainable AI healthcare project for medical imaging students and hospital AI initiatives.

Components

Python 3.8+
PyTorch
Transformer-CNN hybrid architecture
Grad-CAM
OpenCV
Streamlit radiologist dashboard

Key Features

• Transformer-CNN hybrid combining local feature strength with global context modeling
• Detection of pneumonia, tuberculosis, nodules and COVID-19 patterns from X-rays and CT scans
• Grad-CAM and attention heatmaps justify every prediction
• Triage module that prioritizes urgent cases
• Real-time inference optimized for standard hardware

Applications

Hospitals, radiology labs, TB and pneumonia screening programs, and developers building clinical decision support AI.