Soft tissue traume imaging and diagnosis using AI

Imaging is a fundamental diagnostic tool for patients suffering soft-tissue trauma. It is used to grade the severity of a suspected tear and thus inform whether physiotherapy, or surgery is required. MRI is the only widely used method as it has >90% specificity and sensitivity.

But it is expensive. A machine costs (up to)~£1m, requires specialist rooms in a hospital and has high running costs. Further, the imaging technique has broad utility across healthcare and hence is in high demand.

As a result, patients are waiting 18 weeks for their scan, which delays access to necessary treatment, leaving patients in pain and disabled and at risk of degeneration.

Ultrasound scans are faster, two-orders of magnitude cheaper, and machines are more space efficient and readily available. The technology could even be suitable for using in a GP or accident and emergency setting. However, the images are near impossible to interpret for all but the most experienced sonographers, and even for musculoskeletal specialists, it is hard to transfer their ability in imaging one anatomical site, to another. Hence, despite its promise, ultrasound has remained an underutilised resource for decades.

The aim of this PhD project is to leverage recent advances in signal processing and deep learning to deskill use of ultrasound machines for the diagnosis of soft-tissue trauma. The rotator cuff will be focused on as an exemplar application given its increasing prevalence and impact on the UK population.