Bone Fracture Detection Using Resnet50 Algorithm

Abstract

Bone fractures are a serious medical issue that need to be diagnosed rapidly and accurately to ensure timely treatment and getting back to normal. Conventional X-ray-based diagnosis is a laborious and error-prone, human skill-intensive process. This study employs a state-of-the-art deep learning-based system using Convolutional Neural Networks (CNNs) for automatic identification of fractures in medical imagery. To increase classification performance, ResNet50 models were trained on the MURA dataset. The system framework has two primary steps (1) Determine which part of the bone is fractured (2) Identification of the type of fracture. Providing accurate and quick diagnostics have always been a strong workload to radiologist and other medical experts, the model's robustness can be greatly decreased with the strategies of transfer learning and data augmentation.