Jiva.ai, in collaboration with Aberdeen’s Robert Gordon University (RGU), wins the recent SBRI Bone Fracture Detection call, which focuses on machine learning solutions to streamline fracture identification from scans. This builds upon the existing scan-based detection vertical that Jiva.ai has been working on for a number of years.

Background

The SBRI Bone Fracture Detection call highlights the requirement in UK A&E departments to improve detection statistics. Sensitivity amongst doctors in A&E can be as low as 65%, which results in both long term discomfort for misdiagnosed patients and ongoing costs to the NHS in the form of downstream complications.  The cost of delayed and compromised healing in terms of incapacity places a huge burden on both the taxpayer and the patient.

The SBRI have released 1000s of pre-labelled images to encourage the creation of new AI models to automatically detect fractures.

Jiva.ai and RGU’s proposed solution takes two fundamental steps in changing the clinical pathway. Firstly, creating an active data pipeline that transports newly captured scans from the hospital EMR/PACS system to a suite of algorithmic tools that analyse the data. Secondly (the focus of the SBRI project), a machine learning-generated kernel that will be able to recognise fractures within scans better and faster than a human, and alert the relevant staff via a customer front end. The partnership between Jiva.ai and RGU is a significant one, not least because RGU is a leading team of data science and health tech experts in Scotland;  they will provide the basis of a strong validation of a breadth of models that could be applied to this problem domain. Moreover, the partnership is intended to be a long term one in which either party can continue to draw upon each another’s resources and skills for future projects.

The project is due to begin in 2019 Q4.