Today Dr Nisha Sharma and Dr Jonathan James – both NHS breast cancer radiologists and principal investigators for Kheiron Medical Technologies – presented the first internal results of the impact of Mia, Kheiron’s breast screening AI solution, on the UK’s National Health Service breast screening programme. The results were presented at the Radiological Society of North America’s (RSNA) annual conference and prove that Mia is a feasible second reader in breast screening.
In the UK every woman’s screening mammograms are read by two independent human radiologists who each decide independently if she needs to come back (be recalled) for further assessment or not. This is called double reading screening and is considered the gold standard when it comes to cancer detection and minimising the number of women who are called back unnecessarily. The study looked at replacing one of these readers with Mia.
The early results on over 40,000 mammograms showed that if Mia is introduced as a second radiologist, the overall double reading recall rate (the percentage of women who are called back for further investigation) is 4-5% and the cancer detection rate is 8.4 per 1000.
When compared to the results for two humans working together (standard historical double reading) on the same case sample, the recall rate remained at 4-5% and the cancer detection rate was 8.5 per 1000 women. This indicated that it is safe to pair Mia with a human radiologist as there is no significant impact on the performance of the screening service. All these results are well within the UK’s acceptable limits.
Additionally, Mia and the independent human readers agreed whether to recall women or not in 81.9% of the cases. This means that reading time for the second human is reduced by 81.9% compared to a traditional double reading programme. This shows Mia together with just one human reader can provide a meaningful solution capable of addressing the workforce crisis in breast screening while ensuring each woman’s mammogram is still seen by a human reader.
In other markets, such as the USA, only one radiologist will view a mammogram. This single reader model results in more women being unnecessarily recalled and more missed cancers compared to the double reader models used in the UK and Europe. Our study suggests Mia could be introduced to US screening services in a way that offers gold standard double reader screening performance, but with a single human reader. Mia could also be a solution for safely dealing with the backlog of cases building up due to the impact of COVID-19 halting screening services. There are already more than 1 million women in the UK who have not been screened due to COVID3.
The study included 40,588 Mammograms from 3 NHS Breast Screening Centres in the UK between 2012-2019. Of these cases, 40,230 had a normal outcome and 358 were biopsy proven cancers. This internal study is the first step in proving Mia’s efficacy in worldwide breast screening services. The results of our much larger clinical study of Mia which is being led by an independent Contract Research Organisation, consisting of over 250k mammograms from 7 sites in two countries will be published shortly. It represents the largest clinical study of its kind to date.
Peter Kecsekmethy, co-founder and CEO of Kheiron Medical said:
“These internal results demonstrate a promising first step in using safe AI for addressing the workforce crisis in breast screening. When Mia is used as a second radiologist there’s hardly any change in the performance of the screening service and there are reasons to believe that the net result will be an improvement to the quality of service. So we are really showing how Mia helps solve the needs in breast screening by acting as an independent reader. It is important that Mia is not beating a reader in our study, Mia is supporting, being combined with the radiologists to create a solution that maintains the standard of care for women who are being screened. We are very excited to share the results from the wider study as soon as we can. They will build on the story we are sharing today and offer even more assurance that Mia is a safe,effective and a thoroughly validated AI solution for breast screening services anywhere.”
Dr Nisha Sharma, Director of Breast Screening and Clinical Lead for breast imaging, Leeds Teaching Hospital NHS Trust said:
“What we have presented today represents the next step in what we need to see for AI – how it will actually impact the real world. AI holds a lot of promise but when it’s being used to intervene in people’s lives in healthcare we have to be robust in our clinical research and studies. That’s exactly what we have done in this study and the larger study with Kheiron. We’ve worked hard to produce results that are a meaningful outcome for radiologists like me and to also create a study that represents the real population of breast screening as much as possible.”
Dr Jonathan James, consultant breast radiologist, Nottingham NHS Trust said:
“We know that double reading breast screening means better outcomes for women. But we also have a growing workforce crisis. So we need to find a way to solve that problem and keep offering women the best service possible when it comes to screening for breast cancer. The results we’ve presented today show that Mia could be that solution. It’s a privilege to share them today with our peers at RSNA and give them the first piece of evidence that Mia could be a safe and effective tool for their screening service.”