🦴 Bone Fracture Classifier using Deep Learning

This project focuses on building a deep learning model to classify bone fractures from X-ray images using a convolutional neural network (CNN). The dataset used contains multi-region X-ray images categorized into fractured and not_fractured.

📂 Dataset

• Source: Fracture Multi-Region X-Ray Data
• The dataset is structured into three splits:

  dataset_bone_fracture/
  ├── train/
  │   ├── fractured/
  │   └── not_fractured/
  ├── val/
  │   ├── fractured/
  │   └── not_fractured/
  └── test/
      ├── fractured/
      └── not_fractured/
  

Note: The dataset is large and is not included in this repository. Please download it from Kaggle and place the unzipped dataset_bone_fracture folder in the same directory as the notebook.

📊 Exploratory Data Analysis

• Directory structure and file counts explored using a custom explore_dataset() function.
• Grid visualization of sample images from each class.
• File integrity checks to remove any corrupted images.
• Distribution plots showing image counts across train, val, and test splits.

🧪 Preprocessing and Augmentation

• All images resized to 224x224.
• Image data generators apply random transformations:
  • Rotation, width/height shift, zoom, horizontal/vertical flip.
• Pixel values normalized to the [0, 1] range.

🧠 Model Architecture

• Sequential CNN with:
  • Convolutional layers + MaxPooling
  • Batch Normalization and Dropout
  • Dense layers with ReLU and softmax activation
• Compiled with:
  • Loss: Categorical Crossentropy
  • Optimizer: Adam
  • Metrics: Accuracy

📈 Training and Evaluation

• Trained using:

  history = model.fit(
      train_data,
      validation_data=val_data,
      epochs=10,
      steps_per_epoch=len(train_data),
      validation_steps=len(val_data)
  )

• Visualizations:
  • Accuracy and Loss vs. Epoch
  • Confusion Matrix
  • Classification Report

✅ Results

• Evaluation on the test set shows classification performance using precision, recall, and F1-score.
• The confusion matrix helps visualize the model's performance in classifying both fractured and not_fractured images.

🚀 How to Run

1. Download the dataset from Kaggle.
2. Unzip it and place the dataset_bone_fracture folder in the same directory as the notebook.
3. Install required packages:

   pip install tensorflow matplotlib seaborn scikit-learn

4. Run all cells in the notebook.