X-Ray Insights: Comprehensive Dataset for Bone Fracture Detection Across Diverse Anatomical Regions

Effective diagnosis and treatment in medical imaging depend on exact identification of bone fractures. This paper presents a fresh approach for bone fracture diagnosis with MobileNetV2, a lightweight and effective convolutional neural network derived from deep learning. Leveraging transfer learning, the model fine-tunes a pre-trained MobileNetV2 architecture using an X-ray picture dataset. The approach consists in several phases: data collecting, pre-processing, model construction, training, evaluation, and application. Resizing X-ray images to $224 \times 224$ pixels, standardizing pixel values, and using data augmentation methods to improve model resilience are among pre-processing tasks. Custom classification layers-Dense and Dropout layers-along with some layer freezing to preserve pre-trained features-extend the MobileNetV2 model to solve the classification job. Built on binary cross-entropy loss function and Adam optimizer, the model is trained on a well-split dataset including test, validation, and training sets. Evaluation shows that the model is efficient; a classification report reveals exceptional performance criteria. The model achieves precision, recall, and F1-score of 0.95 for both fractured and non-fractured classes, therefore suggesting great accuracy in bone fracture diagnosis. The model’s general accuracy is 0.95; macro and weighted averages regularly show performance over all classes. With its great accuracy and efficiency, which improve diagnosis capability and patient outcomes, this MobileNetV2-based model presents a dependable tool for medical professionals that fits very well for clinical.