Year: 2025
Type: article
Abstract: : Pneumonia remains a global health challenge, especially in developing countries where diagnostic tools are limited or unavailable. Early diagnosis is vital in preventing severe complications, reducing mortality, and ensuring effective patient management. In this paper, we propose a deep learning-based pneumonia detection system utilizing Vision Transformer (ViT) architecture for interpreting chest X-ray images. The ViT model, with its powerful attention-based mechanism, captures both local and global dependencies within X-ray scans, providing highly accurate classification results. To enhance transparency and foster trust in AI-based diagnostics, Grad-CAM is integrated to visually highlight regions influencing the model’s decisions. The model was trained and tested on the publicly available Kaggle Chest X-ray dataset, achieving significant improvements over traditional CNN approaches. The complete pipeline is deployed using Streamlit and FastAPI, enabling real-time, browser-accessible diagnostics. Our system demonstrates the potential of combining explainable AI with advanced transformer-based models to augment clinical decision-making and support radiologists in diagnosing pneumonia more reliably. Keywords: Pneumonia Detection, Vision Transformer, Chest X-ray, Explainable AI, Grad-CAM, Deep Learning, Streamlit, FastAPI, Medical Imaging, AI in Healthcare
DOI: 10.55041/ijsrem48224
Author: R. Ameen
Language: English
Topic: COVID-19 diagnosis using AI
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W4410475537