An end-to-end AI-powered medical imaging application that analyzes X-ray images to detect bone fractures, provides confidence-based predictions, and generates professional, downloadable PDF medical reports through a clean, patient-friendly web interface.
- live url here you can visit and test my project
- Active link
Bone fractures are one of the most common injuries worldwide. In many regions:
- Access to expert radiologists is limited
- Manual X-ray interpretation is time-consuming
- Early-stage or subtle fractures are often misdiagnosed
This creates delays in treatment and increases the risk of complications.
👉 Goal: Build an AI-assisted screening tool that can help detect bone fractures quickly, consistently, and accessibly.
- Automatically detect bone fractures from X-ray images
- Provide clear Normal / Fractured predictions
- Display prediction confidence to improve transparency
- Generate a professional medical-style PDF report
- Deliver a clean, easy-to-use patient-facing interface
- Deploy the system as a live web application
This project uses a Convolutional Neural Network (CNN) trained on labeled X-ray images to classify fractures. The trained model is integrated into a Streamlit web application, allowing users to upload X-ray images and receive instant results.
The system is designed for:
- Educational use
- Research demonstrations
- AI-assisted screening (not a medical diagnosis replacement)
Bone_Fracture_Detection_CNN/
│
├── 01_Data/ # Dataset (train / test)
├── 02_Notebooks/ # Phase-wise development notebooks
├── 03_Models/ # Trained CNN model (.h5)
├── 04_Inference/ # Inference & prediction logic
├── 05_UI/ # Streamlit web application (app.py)
├── 06_Reports/ # Generated PDF reports
├── README.md # Project documentation
- Architecture: Convolutional Neural Network (CNN)
- Input Size: 224 × 224 RGB X-ray images
- Output: Binary classification (Fractured / Normal)
- Activation: Sigmoid
- Loss Function: Binary Crossentropy
- Framework: TensorFlow / Keras
The model outputs a probability score which is converted into a confidence percentage for better interpretability.
- Programming Language: Python
- Deep Learning: TensorFlow, Keras
- Data Processing: NumPy, Pillow (PIL)
- Web Framework: Streamlit
- Report Generation: ReportLab
- Deployment: Streamlit Cloud
- Version Control: Git & GitHub
- Upload X-ray image (JPG / PNG)
- Instant prediction result
- Confidence score display
- Clean and patient-friendly layout
Each report includes:
- Prediction result
- Confidence percentage
- Timestamp
- AI disclaimer
- Project branding
Reports are downloadable directly to the user’s device.
-
Footer and reports include:
© SHUBHAM MADDHESIYA
- Upload an X-ray image
- Model analyzes the image
- Prediction displayed (Normal / Fractured)
- Confidence score shown
- Download detailed PDF report
This system is intended for educational and research purposes only. It is not a substitute for professional medical diagnosis or treatment. Always consult a qualified healthcare professional.
- Built a complete AI pipeline from training to deployment
- Applied CNNs to real-world medical imaging problems
- Learned end-to-end ML product development
- Gained experience in Streamlit deployment
- Implemented automated PDF report generation
Shubham Maddhesiya AI & Deep Learning Enthusiast
Give it a ⭐ on GitHub and feel free to connect!