Artificial intelligence found more breast cancer cases than doctors with years of training and experience and cut doctors’ mammogram readings almost in half, a new early-stage study found.
This does not mean that your hospital will let a computer determine if you have cancer at some point. There is still much more research to be done, but the study was published Tuesday in the journal The Lancet Oncologyshows that artificial intelligence is safe to use in breast cancer detection and can make doctors even more efficient at finding cancer than they are now.
Other studies have shown that artificial intelligence can be useful in predicting breast cancer risk, but they use models or have focused on retrospective data. The new research is believed to be the first randomized management trial to compare AI-assisted detection of breast cancer with detection performed by well-trained humans alone.
The researchers looked at scans from more than 80,000 women in Sweden who underwent a mammogram between April 2021 and July 2022. Half of the women were assigned to a group where AI read the mammogram before it was analyzed by a radiologist. The other group’s mammograms were read by two radiologists without the use of AI. All the radiologists in the study were considered highly experienced.
The group whose scans were read by a radiologist together with AI had 20% more cancers detected than the group whose mammograms were read by two radiologists without the additional technical assistance.
Overall, the screenings supported by AI resulted in a cancer detection rate of 6 per 1,000 women screened compared to 5 per 1,000 with the standard method.
But the researchers say they didn’t realize the AI was too sensitive. It did not increase the number of false positives when a mammogram is diagnosed as abnormal even though no cancer is present.
The group that used AI had an additional benefit: a reduced reading workload of 44%. The trial did not measure the specific amount of time saved by AI, but the researchers calculated that if radiologists read about 50 mammograms per hour, it would have taken a single radiologist four to six months less to read about 40,000 screening exams with the help of AI than it would take two radiologists alone.
“The greatest potential of AI right now is that it can enable radiologists to be less burdened by the excessive amount of reading,” said study co-author Dr. Kristina Lång, lecturer in radiology diagnostics from Lund University in Sweden.
In Europe, guidelines recommend that two radiologists screen a mammogram. The US does not have the same standard, so the workload may be different in different countries.
However, Europe and the United States both have a shortage of radiologists, according to Radiological Society of North America. If further research shows that this technology really works, it could help ease some of these staffing issues as well as make radiologists even better at their jobs.
Demand for radiologists is expected to increase as the global population ages and requires even more imaging.
Many radiologists see the opportunities as welcome news rather than threats to their job security.
“With mammography, our goal is to detect breast cancer as early as possible to give each patient the best prognosis, so anything that will make us more accurate is a wonderful thing,” said Dr. Stamatia Destounis, a radiologist specializing in breast imaging at Elizabeth Wende Breast Care in Rochester, New York, who were not involved in this study.
Any kind of technology that can help with breast screening can make a big difference. The incidence of breast cancer has increased by 0.5% per year, according to American Cancer Society, although there has not been a corresponding increase in the number of deaths. While breast cancer is still the No. 2 killer of women dying from cancer, behind only lung cancer, more women are surviving than decades ago, largely because of effective screening. When breast cancer is caught earlya person’s chance of survival increases significantly.
But mammography isn’t perfect, experts say. It is a very subjective skill. Overall, screening mammograms miss about 20% of breast cancers, according to the National Cancer Institute.
Detecting the complex pattern that is breast cancer is extremely difficult, even with years of specialized training. Basically, a radiologist needs to detect a tumor that is white in the middle of a white background. AI may one day be able to help with that pattern detection, but a radiologist’s job is much more than pattern recognition, said Dr. Laura Heacock, a breast radiologist at NYU Langone Perlmutter Cancer Center, who was not involved in the new study.
“If you spend a day with a radiologist, you’ll see that how an artificial intelligence looks at screening a mammogram is really just part of how radiologists practice medicine, even in breast imaging,” she said. “These tools work best when paired with highly trained radiologists who make the final call on your mammogram. Think of it as a tool like a stethoscope to a cardiologist.”
Heacock said that with more research, her colleagues may use AI in this way in the future. Radiologists already use a relatively crude form of computer image analysis called CAD, developed in the 1990s, which can recognize patterns in mammograms.
“AI algorithms are more flexible and trained with much more cutting-edge deep neural networks that allow advanced feature recognition and application, and they’re cross-trained on all the commercial models and the research models are externally validated,” Heacock said. An AI model looks at an image differently than a human eye would, is trained on different material, and can make different predictions based on what it can and can’t see, she said.
Although artificial intelligence is still a new technology, artificial intelligence is beginning to capture the imagination of scientists. It is used in drug discovery and development, and it has helped doctors communicate better with patients. AI even passed the practice exam doctors use to get their licenses, so it’s being used to help write better test questions.
Several AI programs are also being developed to help doctors detect cancer. A program at MIT has been created to detect high risk of future breast cancer based on current mammograms, something doctors are unable to do right now.
Many of these programs are showing real promise, Heacock said.
“I think of AI as more validation. It doesn’t sleep. AI doesn’t get tired. The AI doesn’t get tired, and it’s been shown that it can tremendously augment our less experienced doctors, which if you see something rare, AI “One might be more likely to flag it if you haven’t seen it before,” she said.
She will also welcome the day when research moves forward.
“You wouldn’t turn down a stethoscope if it was offered to you, you know?” she added.