A new machine learning process designed to identify and classify hip fractures has proven to be more effective in human clinics.

Two convulsive neural networks (CNN) developed at the University of Bath were able to identify and classify hip fractures from X-rays with a 19% higher level of accuracy and reliability than clients. hospital, in results published this week in Nature Scientific Reports.

A research team from Bath's Center for Therapeutic Innovation and the Institute for Mathematical Innovation, as well as colleagues from the Royal United Hospitals Trust Bath, North Bristol NHS Trust and Bristol Medical School, are committed to developing a new process to help clients who 'cause a hip fracture. more effective care and promote better patient outcomes.

They used a total of 3,659 hip X-rays classified by at least two experts to train and test neural networks, achieving overall accuracy of 92% and 19%. more precisely than hospital-based clinics.

Hip fractures are a major cause of illness and death in the elderly, leading to high health and social care costs. The classification of fracture before surgery is important for assistance in selecting the right interventions for fracture processing and reconciliation.

The ability to easily classify fracture, accurate and reliably, key is: delay operation than 48 hours can increase the risk of poor consequences.

Fractures are divided into three classes - non-returns, orders or subtroples that occur with joints.

Important as a higher patient date: People who strengthen the hip fracture next year than the age of the quantity. This, the team, the development of ways to improve hip fracture and their morumbvokation, health care costs are high priority.

The critical question affecting the use of diagena images is the mismatch between the need and source: including X-rays) increased by 2596 to 2014. The new required departments means that they do not report results over time.

Professor Richie Gill, lead author and co-director of the Center for Therapeutic Innovation, said: “Machine learning techniques and neural networks offer a new and powerful way to automate diagnostics and predict results, so this new technique we share has great very strongly determines the surgical treatment, and therefore the outcome of the patient, there is currently no standardized process in the UK who determines this classification - whether it is performed by orthopedists or radiologists specializing in musculoskeletal disorders.