NHS Grampian, which provides health and social care services to more than 500,000 people in Scotland, is trialling Kheiron Medical Technologies’ artificial intelligence to help medical staff detect cancer, cut the backlog of mammograms to be read and deliver results to patients faster.

More than two million women have a breast cancer screening each year in the UK. As a result of the COVID-19 pandemic, staff pressures and long waiting lists, there is a backlog of mammograms to be read.

The detection of breast cancer can be difficult with an incidence of 6 in 1000 women which is why in the UK and in Europe, two specialists read every mammogram. This ensures the majority of cancers are detected at screening. Occasionally, cancers are missed at or occur after screening. These are known as interval cancers.

Under an Innovate UK funded programme, the Industrial Centre for Artificial Intelligence Research in Digital Diagnostics (iCAIRD), Gerald Lip, Lead Radiologist at NHS Grampian, and Professor Lesley Anderson, Chair in Health Data Science, University of Aberdeen led a research team reviewing Mia’s performance on historical mammograms. Partnering with Kheiron in iCAIRD, the team ran Mia on deidentified scans with known outcomes from a previous 3-year screening cycle in Grampian. Over 55,000 screening events comprising more than 220,000 mammogram images were read by Mia and compared with the radiologists opinions.

Looking at a test dataset of mammograms over a 3 year period, the standard process involving two specialists would have detected 303 cancers. A single specialist reading alone, would have picked up 261 cancers and Mia reading alone would have picked up 271.  Mia also picked up a further 47 (33.8%) additional interval cancers, which would not have been detected by the two specialists

Dr Lip said: “We have seen significant results using Mia. Not just in the detection of difficult cancers but also in the time it takes to get results to patients. Radiologists can’t work 24 hours a day, but the AI can. By considering replacing one human review of a screening with AI we project a 30%-40% reduction in reporting workload on our team. This is time that can be spent with patients. But it’s important to remember that the AI isn’t being used in isolation. It is great at detecting breast cancer, but it is unable to review the findings in the context of past scans and medical history. This is why we are using a combination of a highly trained radiologist with the AI in all screenings.

Full article published 25 November and available at Microsoft New Centre UK here.