About the Project
This project explores how artificial intelligence can help identify common quality issues in X-ray images. Instead of focusing on disease detection, the model focuses on the technical quality of an image and determines whether it is suitable for further evaluation.
Dataset & Preparation
The dataset was personally curated and organized from multiple publicly available X-ray image sources. Images were reviewed and categorized according to the quality characteristics being studied, followed by preprocessing and preparation for model training.
How It Works
The X-ray image is first prepared using standard image preprocessing techniques and resized to 224 × 224 pixels. A Vision Transformer then analyzes the image and learns visual patterns associated with different quality categories. The trained model produces a predicted quality class along with its confidence.
Results
The model achieved strong performance on the held-out test set, showing that deep learning can be useful for automated assessment of X-ray image quality.
Validation Accuracy
96.08%Test Accuracy
92.31%Interactive Application
Try the X-ray Quality Control Classifier
The project was also developed as an interactive Streamlit application. Users can upload an X-ray image and receive an automated quality classification with the model's confidence.
Open Live App →Limitations
This project is intended as a learning and research exploration rather than a clinically validated diagnostic system. The model evaluates image quality only and does not identify diseases or clinical findings. Performance may also vary across different hospitals, imaging systems, acquisition protocols, and patient populations.
Tools & Technologies
Python • PyTorch • Torchvision • TIMM • Pillow • Streamlit • Hugging Face • GitHub