Fracture Spotter in Musculoskeletal Radiographs

I need an expert in medical imaging and AI technology who can develop a system for detecting fractures in musculoskeletal radiographs. Here's what you need to know: - Specialization: The project primarily requires detecting fractures. However, due to the complexities and variety of fractures, the system should be open-ended enough to detect any kind depending on the input. - Time Frame: This project needs to be completed as soon as possible. The sooner we can get this up and running, the better. - Skills & Experience: Ideal candidates would be those with experience in radiology, medical imaging, machine learning, and AI. Prior work in fracture detection would be a definite advantage. The proposed methodology involves training individual deep learning models for the humerus, elbow, and wrist on their respective parts of the MURA dataset. Features are extracted from the intermediate layers of each model during training and concatenated to form a unified feature representation for the upper extremity. A new neural network architecture is designed to take these concatenated features as input, incorporating additional layers for fine-tuning and feature combinations. Modifications are made to enable individual models to output both class probabilities and relevant localization information. The hybrid model is then trained on a comprehensive dataset encompassing abnormalities in the elbow, shoulder, and wrist. Evaluation metrics include classification (accuracy, precision, recall) and localization (intersection over union, dice coefficient). I look forward to working with a professional who can deliver quality work swiftly.