Fraxpert: AI Powered Fracture Detection System

Client

NHS Grampian launched their first SBRI competition in collaboration with the Chief Scientists Office, Opportunity North East (ONE), Canon Medical Research Europe, the University of Aberdeen and ICAIRD to develop artificial intelligence (AI) solutions to support fracture diagnosis in hospitals.


Background

NHS Grampian identified a big challenge that was affecting how patients get care: people living in areas with smaller hospitals or clinics had to make long trips to big city hospitals for detailed checks on fractures, especially for injuries to wrists and ankles. By making these diagnostic services available closer to where patients live, it wouldn't just save them from unnecessary travel; it would also lighten the workload for the radiographers and make the whole healthcare system more efficient.


Additionally, there's been a noticeable slowdown in getting accurate diagnoses for complex health issues, creating a bottleneck. Often, this meant patients had to come back for more detailed, time-consuming treatments that could have been minimised or even avoided with faster, earlier diagnosis. Improving how quickly and accurately we can assess patients right from the start is crucial for better patient care and outcomes.


Fractures can be challenging to diagnose, particularly in emergency situations. Radiologists and A&E clinicians are exhausted and overworked. Fraxpert aims to bridge the gap between supply and demand for image diagnostic interpretation, so that patients receive their correct diagnosis at the first point of consultation and radiologists never have to work outside their contracting hours.



Solution


Seeai worked in collaboration with NHS Grampian and Canon Medical Research Europe to develop an AI-supported fracture diagnostic system to deliver fast and accurate plain film interpretation.

The project used innovative data technology to screen radiographs of wrist and ankle fractures to enable radiologists to focus on more complex cases.


Seeai collaborated radiologists and doctors to create one of the largest hand-labelled datasets for fracture detection. Our AI algorithms were trained on the development dataset, created from 7,242 patients consisting of 12,926 abnormal images and 8,286 normal images. The dataset contained 48,770 annotations of localised fractures. Moreover, we prepared a separate group of 1,050 patient data for evaluation.


Our AI algorithm achieved specialist-like performance with an accuracy of 95.67%, a recall of 95.05% and a precision of 97.24% on the test dataset. 

Ready to get started?

Seeai LTD.

All rights reserved.

Find us

Platform, New Station St, LS1 4JB, Leeds, UK info@seeai.co.uk

Seeai LTD.

All rights reserved.

Find us

Platform, New Station St, LS1 4JB, Leeds, UK info@seeai.co.uk