Objective: To Identify a subgroup of patients with acute clavicle fractures who would benefit from early surgical intervention

Background: This projects represents an exciting collaboration between the Trauma & Orthopaedic team at University Hospitals Dorset (UHD) and Bournemouth University.

The clavicle is a commonly fractured bone with approximatley 400 patients presenting to A&E at UHD per year. Patients can be treated wither using a simple sling or with surgery, which involves fixing the bone with plates and screws. The long term results of surgery and non-operative treatment are the same as long as the fracture heals.

However 15% of fractures treated with a sling alone do not heal (non-union). Most of these require delayed surgery (which is more difficult) and this leads to patient dissatisfaction.

At the moment there is no accurate way to predict who are the patients likley to develop non-union and would have therefore benefited from early surgery.

Similar studies using ML to predict hip replacement failure suggest that there are relevant features on x-rays that are iinvisible to the human eye. This technology has, as yet, not been applied to clavicle fractures.

Using the UHD fracture database we aim to develop and train a model to assess whether ML can see early x-ray features that predict clavicle healing?

A database of serial x-ray images, patient demographics and patient outcomes will be provided as well as support to provide clinical context to the research project.