Using artificial intelligence to screen for breast cancer poses a “significant risk of overdiagnosis”, a leading tech body has said.
The technology is often seen as “like magic” in healthcare but there is little scientific evidence to support its use in screening, BCS, The Chartered institute for IT said.
Responding to a recent consultation by the UK National Screening Committee on the use of AI in breast cancer screening, the body said it was keen to support the technology once “robust research” has been carried out to ensure safety.
Dr Philip Scott, chair of the BCS health and care executive, said: “Unfortunately, there is so much hype about AI that some people treat it like magic. Most AI in healthcare is early stage and not shown to work clinically.
“If AI were adopted now in the screening of breast cancer, there is significant risk of overdiagnosis with all the anxiety that would cause. We need to educate and inform the public to maintain trust, and that includes being honest about the immaturity of most AI tools.
“AI has the potential to be of huge benefit or of huge harm to society, and standards for the design, development, and adoption of AI systems must be regulated to ensure we get the very best out of them.”
The BCS agreed with the NSC proposal that the use of AI for image analysis in breast cancer screening should not be endorsed in the UK at present.
The consultation on AI in breast cancer screening, which closed last month, said there was “insufficient evidence in quality and quantity to recommend implementation of AI into clinical practice of the NHS breast screening programme”.
It noted that, overall, studies to test algorithms used to detect the disease were “sparse” and “lacked applicability” to the UK as none of the datasets used in studies were from the UK.
It recommended that until limitations in research and evidence had been addressed a rapid review of clinical effectiveness of the technology could not take place.
‘Unclear benefits’
A recent evaluation of AI studies found it is not yet accurate enough to make judgement calls in breast cancer screening.
The analysis of 12 studies into the use of artificial intelligence (AI) for diagnosing breast cancer concluded it was “unclear” where the technology may be beneficial on the clinical pathway.
Some 94% of 36 AI systems examined in studies were “less accurate than a single radiologist” and all were less accurate than two radiologists, experts from Warwick Medical School found.
In three of the studies AI used for triage screened out 10%, 4% and 0% of cancers that were detected by radiologists – meaning in some cases it was missing 100% of cases spotted by radiologists.
The analysis, published in the BMJ, also found many of the studies lacked test accuracy of AI and had poor methodical quality.
“Current evidence for AI does not yet allow judgement of its accuracy in breast cancer screening programmes, and it is unclear where on the clinical pathway AI might be of most benefit,” experts concluded.
“AI systems are not sufficiently specific to replace radiologist double reading in screening programmes. Promising results in smaller studies are not replicated in larger studies.”