Medical Image Classification: Binary Classification of X-Ray Images to Classify Fractures in Bones

This paper addresses the classification of X-ray images to determine whether bone fractures are present or not. Using carefully selected Bone Fracture Dataset and Convolutional Neural Networks (CNNs), this research investigates the use of deep learning for diagnosis from X-ray images. The dataset is subjected to extensive pre-processing and augmentation, which improves the model's ability to generalize over a variety of complex medical image types. In the considered CNN architecture, there is a flattening layer, a dense layer with rectified linear units as activation function, and three convolutional layers with increasing filter sizes and matching max-pooling operations. A dropout layer is added to reduce overfitting, and a final layer with a sigmoid activation function for binary classification completes the system. The model is trained using the Adam optimizer and binary cross entropy loss. The training process involves iterative updates of model parameters over ten epochs, with batches of augmented images. The model is subsequently evaluated on a distinct testing dataset. With a test accuracy of 79.67%, precision of 75%, recall of 85.7143%, and F1-score of 80%, the model demonstrated its ability to discriminate between X-ray images that show bone fractures and those that don't.