AI-supported mammography screening saves time & detects more cancer
A new study has found that a single radiologist screening mammograms picked up more incidents of breast cancer and was more efficient when supported by AI. The researchers say their approach would be a safe alternative to having two radiologists ‘double read’ the scans.
For many women, having regular mammograms is the best way of detecting breast cancer early, when it’s easier to treat and before it’s big enough to feel or cause symptoms.
Australia and many European countries employ ‘double reading’ of mammograms, meaning that the scans are reviewed by two radiologists, each giving an independent opinion. The rationale is that using two sets of eyes increases the likelihood that cancer will be detected. In the US, a single radiologist plus computer-aided detection (CAD), a computer program that scans the mammogram and marks potentially abnormal areas, is more common.
A new study by researchers at Lund University, Sweden, has examined whether replacing one of the breast screening radiologists with AI is safe and feasible compared to the standard practice of double reading.
“In our trial, we used AI to identify screening examinations with a high risk of breast cancer, which underwent double reading by radiologists,” said Kristina Lång, lead author of the study. “The remaining examination were classified as low risk and were read by only one radiologist. In the screen reading, radiologists used AI as detection support, in which it highlighted suspicious findings on the images.”
The researchers included 80,033 women in the study, divided into two groups: one that underwent AI-supported breast screening and a control group that underwent standard double reading without AI support.
“We found that using AI resulted in the detection of 20% more cancers compared with standard screening, without affecting false positives,” Lång said. “A false positive in screening occurs when a woman is recalled but cleared of suspicion of cancer after workup.”
The largest disadvantage in using CAD software, as opposed to modern AI, is its high rate of false positives, creating unnecessary anxiety and the requirement for further testing. Unlike modern AI, traditional CAD uses more limited techniques that can only be trained on small datasets, which can lead to inaccuracy. And, by virtue of human variability, despite the input of two radiologists, false positives can still arise with double reading.
In addition to being accurate, the current study found that the support provided by AI also reduced the radiologists’ workload by 44%. The number of AI-supported screenings was 46,345 compared to 83,231 standard screenings. With AI assistance, the researchers estimated it took about five months less for a radiologist to read approximately 40,000 screens.
Mindful that the study was conducted at only one site, the researchers plan to conduct further studies to see if the results can be replicated.
“The study was conducted on a single site in a Swedish setting,” said Lång. “We need to see whether these promising results hold up under other conditions, for example, with other radiologists or other AI algorithms. There may be other ways to use AI in mammography screening, but these should preferably also need to be investigated in a prospective setting.”
The study is part of the Mammography Screening with Artificial Intelligence (MASAI) trial, the first randomized controlled trial evaluating the effect of AI-supported screening. A total of 100,000 women have been enrolled in the trial; the next step is to investigate which cancer types were detected with and without AI support to determine the interval-cancer rate. Interval cancer is diagnosed between screenings and usually has a poorer prognosis than screen-detected cancer. The interval-cancer rate will be assessed after all the women in the trial have had at least a two-year follow-up.
“What’s important is to find a method that can identify clinically significant cancers at an early stage,” Lång said. “However, this has to be balanced with the harm of false positives and the overdiagnosis of indolent cancers. The results from our first analysis shows that AI-supported screening is safe since the cancer detection rate did not decline despite a substantial reduction int. he screen-reading workload.”
The study was published in the journal The Lancet Oncology.
Source: Lund University