Pneumonia Detection

This repository contains a project for detecting pneumonia from chest X-ray images using a deep learning model. The dataset consists of labeled JPEG images (Pneumonia/Normal) from a retrospective cohort of pediatric patients aged 1-5 years from Guangzhou Women and Children Medical Center in Guangzhou.

Author

Patryk Klytta

Project Structure

The repository is organized as follows:

• chest_xray/: Directory containing the X-ray images categorized into train, val, and test subdirectories.
• train/: Training dataset.
• val/: Validation dataset.
• test/: Test dataset.
• pneumonia_detection.ipynb: Jupyter notebook containing the code for training and evaluating the model.

Setup

1. Clone the repository:

   git clone https://github.com/yourusername/pneumonia_detection.git
   cd pneumonia_detection

2. Install the required dependencies (libraries and packages):

3. Ensure TensorFlow is configured to use the GPU if available.

Dataset

The dataset used in this project is organized into three subsets:

• Training Set: Used to train the model.
• Validation Set: Used to tune the model hyperparameters.
• Test Set: Used to evaluate the model's performance on unseen data.

Data Preprocessing

The images are preprocessed using the ImageDataGenerator class from Keras, which includes the following transformations:

• Rescaling
• Shearing
• Zooming
• Horizontal and vertical flipping
• Brightness adjustments
• Width shifting
• Rotation

Model Architecture

The model is built using the ResNet50V2 architecture with the following modifications:

• Global average pooling layer
• Dense layer with ReLU activation
• Output layer with sigmoid activation for binary classification

The ResNet50V2 layers are frozen to leverage transfer learning, only training the added layers.

Training

The model is trained using the following configurations:

• Optimizer: Adam
• Loss function: Binary Cross-Entropy
• Metrics: Accuracy
• Number of epochs: 30
• Batch size: 32

Evaluation

The model's performance is evaluated on the test set, with metrics including accuracy, precision, recall, and F1-score. Additionally, confusion matrices are generated for a detailed performance analysis.

Example Results

• Training Accuracy: 95.91%
• Validation Accuracy: 91.66%
• Test Accuracy: 91%

Visualizations

The training and validation loss and accuracy are plotted to visualize the model's learning process. Confusion matrices are also plotted for the training, validation, and test sets.

Model Saving

The trained model is saved in the resnet_pneumonia_model_PKlytta.keras file for future inference.

How to Run

To train and evaluate the model, run the provided Jupyter notebook pneumonia_detection.ipynb or execute the script in your Python environment.

Contributing

Contributions are welcome! If you encounter any issues or have suggestions for improvements, please feel free to open an issue or submit a pull request.

License

This project is licensed under the MIT License.

art. 74 ust. 1 Ustawa o prawie autorskim i prawach pokrewnych, [Zakres ochrony programów komputerowych](https://lexlege.pl/ustawa-o-prawie-autorskim-i-prawach-pokrewnych/a