Pneumonia-Classification-Project

Using Python technology and deep learning, I classified pneumonia in the Chest X-ray image.

1.Project Overview

This project focuses on classifying pneumonia cases from chest X-ray images using deep learning techniques. The dataset is sourced from the RSNA Pneumonia Detection Challenge. The main goal is to classify images into pneumonia-positive (1) or pneumonia-negative (0) categories using convolutional neural networks (CNNs).

2.Project Structure

Pneumonia_Classification_Project/
│── notebooks/                      # Jupyter Notebooks
│   ├── pneumonia_classification.ipynb  # Main Jupyter Notebook
│   ├── Pneumonia_Project/               # Project-related data and scripts
│   │   ├── data/                         # Dataset files
│   │   │   ├── stage_2_train_labels.csv  # Training labels
│   │   │   ├── stage_2_train_images/     # X-ray image files
│   │   │   ├── stage_2_test_images/      # Test X-ray images
│   │   │   ├── stage_2_detailed_class_info.csv  # Additional metadata
│   │   │   ├── stage_2_sample_submission.csv  # Sample submission format
│   │   │   ├── rsna-pneumonia-detection-challenge.zip  # Raw dataset archive
│   │   ├── processed_images/             # Preprocessed image files
│── src/                          # Source code for preprocessing and training
│── models/                       # Saved model weights
│── results/                       # Output graphs and metrics
│── README.md          

3.Dataset and Preprocessing

Dataset Used

• stage_2_train_labels.csv: Contains labels for training images.
• stage_2_train_images/: Directory containing X-ray images.

Preprocessing Steps

• Load and visualize DICOM images using pydicom.
• Normalize pixel values for better model performance.
• Augment training data using rotation, flipping, and contrast adjustments.
• Split the dataset into 80% training/validation and 20% testing.

4.Model Architecture

• Implemented a convolutional neural network (CNN) using tensorflow.keras.
• Model Layers:
  • Conv2D layers with ReLU activation for feature extraction.
  • MaxPooling2D layers for dimensionality reduction.
  • Dropout layers to prevent overfitting.
  • Dense layers for final classification.
• Loss Function: Binary cross-entropy.
• Optimizer: Adam.

5.Training and Validation

Training

The model was trained using the following configuration:

history = model.fit(
    train_gen,
    validation_data=val_gen,
    epochs=10,
    steps_per_epoch=len(train_gen),
    validation_steps=len(val_gen)
)

• The dataset was processed using train_gen (training data generator) and val_gen (validation data generator).
• The model was trained for 10 epochs, with batch-wise updates to optimize convergence.
• Each epoch processed all training batches (steps_per_epoch=len(train_gen)).
• Validation was performed after each epoch using validation_steps=len(val_gen).
• Training accuracy improved progressively, and loss steadily decreased.

Validation

• 20% of the training set was used for validation.
• Hyperparameters were adjusted based on validation performance.
• Final validation accuracy: 78.23%

6.Results and Analysis

• Final Test Accuracy: 78.96%
• Evaluated model performance on the test dataset.
• Plotted training and validation loss/accuracy curves.
• Loss and Accuracy Trends:
  • Training accuracy steadily improved over epochs.
  • Validation accuracy reached 78.23%, indicating good generalization.
  • Loss consistently decreased, confirming effective learning.

7.Conclusion

• Successfully classified pneumonia cases from X-ray images using deep learning.
• Future improvements:
  • Experiment with different CNN architectures (e.g., ResNet, EfficientNet).
  • Fine-tune hyperparameters for better generalization.
  • Increase dataset diversity through additional augmentation techniques.

This project demonstrates the application of deep learning in medical imaging and serves as a portfolio that demonstrates exploration in the fields of data science, AI, and medical analytics.