Abstract

In this project, we focus on the classification of chest X-ray images into Normal and Pneumonia categories. Using Convolutional Neural Networks (CNNs), we train a model to classify chest x-ray images. The dataset used for training consists of labeled chest X-ray images, containing both Normal and Pneumonia cases. The model is trained on a subset, validated and tested on a different set to measure its performance. Image augmentation and preprocessing techniques, along with transfer learning was used to enhance the model's performance.

Methodology

1. Dataset:

• The dataset used for this project is a publicly available Chest X-ray dataset which is the Chest X-ray Images (Pneumonia) dataset. It contains labeled chest X-ray images categorized into Normal and Pneumonia classes.
• The images are resized and preprocessed to ensure they are consistent in dimensions and format for the deep learning model.

2. Data Preprocessing:

• Image Resizing: The original X-ray images are resized to a standard resolution (e.g., 224x224 pixels) to fit the input requirements of the Convolutional Neural Network.
• Normalization: The pixel values of the images are normalized to a range of [0, 1] to improve the convergence of the model during training.
• Data Augmentation: To improve the model's generalization and reduce overfitting, various augmentation techniques such as random rotations, flips, color shifts are applied to artificially increase the size of the training dataset.

3. Model Architecture:

• Convolutional Neural Network (CNN): A CNN model is implemented using PyTorch.
• Transfer Learning: ResNet50, ViT, Swin models are fine-tuned on the chest X-ray dataset out of which ResNet50 gives the best results.

4. Model Training:
The model is trained using a binary cross-entropy loss function, which is appropriate for binary classification tasks.

• Optimizer: The Adam optimizer is used for optimization, adjusting weights during the backpropagation process.
• Batch Size and Epochs: The model is trained with a batch size of 32 and for 10 epochs.

5. Model Evaluation:
The model is evaluated using accuracy score to measure its performance.

Results

After training and evaluating the CNN-based model for chest X-ray classification, the following results were obtained:

• Accuracy: The model achieved an overall classification accuracy of 94% on the training set, indicating that the model correctly classified the X-ray images into Normal and Pneumonia categories most of the time.

• Validation and Test Performance:
  The model performed similarly on the validation and test sets with an accuracy score of 87%, demonstrating good generalization and minimal overfitting.

Deployment

The fine-tuned model is deployed on streamlit for an easy-to-use interface.

Try out the app here.