Skip to content

Latest commit

 

History

History
156 lines (106 loc) · 4.16 KB

README.MD

File metadata and controls

156 lines (106 loc) · 4.16 KB
Raw

Multi-Label Chest X-Ray Classifier

This project implements a robust pipeline for training a multi-label image classification model on chest X-ray data, using deep learning and transfer learning techniques. The model predicts the presence of multiple conditions in a single X-ray image.

Table of Contents

Features

  • Handles multi-label classification tasks for medical images.
  • Utilizes the EfficientNetB0 architecture with transfer learning.
  • Provides detailed evaluation metrics including classification report, Hamming loss, and Jaccard scores.
  • Implements advanced callbacks for training optimization:
    • ReduceLROnPlateau
    • EarlyStopping
  • Extensive error handling and logging for image loading and preprocessing.

Technologies Used

  • Python
  • TensorFlow/Keras for deep learning model implementation.
  • Pandas for data manipulation.
  • NumPy for numerical computations.
  • scikit-learn for data preprocessing and evaluation metrics.
  • EfficientNetB0 as the base model for transfer learning.

Project Structure

.
|-- script.py                # Main implementation file
|-- merged_CXLSeg_data.csv   # Input dataset (CSV format)
|-- images/                  # Directory containing chest X-ray images
|-- README.md                # Documentation
|-- requirements.txt         # Python dependencies

Dataset

The project uses a dataset containing chest X-ray images annotated with multiple conditions. Each X-ray image is associated with one or more labels indicating the medical conditions present. The dataset should:

  • Contain a CSV file with image paths and labels.
  • Include a column named DicomPath_y specifying the file paths to the X-ray images.

Example CSV Format:

DicomPath_y Atelectasis Cardiomegaly Pneumonia ...
path/to/image1.dcm 1 0 0 ...
path/to/image2.dcm 0 1 1 ...

Installation

  1. Clone the repository:

    git clone https://github.com/your-repo/multi-label-xray-classifier.git
cd multi-label-xray-classifier

  2. Install required dependencies:

    pip install -r requirements.txt

  3. Ensure the dataset is available in the correct structure.

Usage

  1. Initialize the classifier:

    from script import ConsistentMultiLabelXRayClassifier

classifier = ConsistentMultiLabelXRayClassifier(data_path='merged_CXLSeg_data.csv')

  2. Train and evaluate the model:

    model, history, mlb = classifier.train_and_evaluate(epochs=15)

  3. View evaluation metrics in the console output.

Model Training and Evaluation

Training

The training pipeline:

  • Loads and preprocesses images.
  • Uses transfer learning with the EfficientNetB0 model.
  • Fine-tunes the model with a fully connected head for multi-label classification.

Evaluation

The evaluation includes:

  • Classification report with precision, recall, and F1-score for each label.
  • Hamming loss.
  • Jaccard scores (micro and macro averages).

Results

  • Accuracy: Achieved XX% on validation data.
  • Precision/Recall/F1: Comprehensive metrics for each condition.
  • Hamming Loss: YY%.
  • Jaccard Score: ZZ (micro), AA (macro).

Future Improvements

  • Add support for additional datasets.
  • Implement data augmentation to improve model generalization.
  • Experiment with other architectures like EfficientNetV2 or Vision Transformers.
  • Optimize preprocessing for large datasets using parallelization.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

  • TensorFlow/Keras team for providing pre-trained EfficientNet models.
  • Dataset contributors for their invaluable work in medical imaging.