The tests, involving historic patient records in Leeds and the East Midlands, are the latest to probe the use of AI in the healthcare sector
Kheiron Medical is set this month to launch trials with the NHS of an artificial intelligence system aimed at assisting in breast cancer diagnoses, in the latest tests of AI in the healthcare sector.
Kheiron, which announced UK government funding for the trials late last November, said its Mia tool is aimed at tackling an acute shortage of the staff needed to accurately interpret mammograms.
Each mammogram currently requires two consultant radiologists to analyse it, but the service is threatened by staff shortages and impending retirements, with the NHS facing soaring costs for outsourcing analyses.
Mia, which is intended to act as an independent reader alongside human radiologists, is the first deep learning image analysis system to be tested within the NHS at this scale, Kheiron said.
AI training
The firm said it is working with the East Midlands Radiology Consortium (EMRAD) on the project, the largest collaborative network of hospitals in the UK to share a single image-sharing platform.
Funding for the trials comes as part of NHS England’s Test Bed Programme.
The trial is to use historic scans at an NHS trust in Leeds, as well as tens of thousands of historic scans from the East Midlands, Kheiron told the Financial Times.
Firms developing similar technology include the Netherlands’ ScreenPoint Medicaland Google’s DeepMind, which began a trial with the NHS last year.
Kheiron’s technology has already been trained on about 500,000 scans from hospitals in Hungary, and the firm found Mia beat the average performance of a human radiologist when tested against 3,500 scans, clinical director Hugh Harvey told the FT.
Harvey said the analysis of historic data was a first step, which it intended tofollow with tests using current patients once the algorithm was performing “optimally”.
Workforce shortage
“There are considerable workforce issues in breast cancer screening with unfilled posts and many centres in the UK struggling to meet targets,” Dr Jonathan James, a breast radiology consultant at the Nottingham Breast Institute, said on the occasion of the trials’ initial announcement last year. “AI has the potential to provide a solution to the workforce crisis by supplementing at least one of these human mammography readers.”
DeepMind began testing its mammogram AI tech with the Cancer Research UK Imperial Centre at Imperial College, London in April of last year, and has also signed a five-year deal with the Jikei University Hospital in Japan to analyse historic mammograms of 30,000 women.
Healthcare is considered a major potential market for AI, but has faced initial hurdles establishing itself in the field due in part to the necessity of delivering large amounts of patient data into the hands of private companies such as Google for analysis.
A 2016 arrangement between DeepMind and the NHS, which saw the analysis of the records of 1.6 million patients by DeepMind, was ruled illegal the following year by the Information Commissioner’s Office (ICO), since patients had not been informed.
Last year Google abruptly took direct control of DeepMind Health and its flagship NHS app, Streams, in a move that added to concerns over the relationship between Google and the NHS.
DeepMind had said earlier in the year that it was planning clinical trials of an AI diagnostic tool for eye diseases after tests found it was more accurate than human specialists.
