Breast cancer is the most common cancer among adults, with more than 2.3 million cases diagnosed each year, according to the World Health Organization. Early detection and treatment can save lives, but the current screening methods are not perfect and can miss some cancers or generate false positives. To improve the accuracy and efficiency of breast cancer screening, researchers are exploring the potential of artificial intelligence (AI) to assist radiologists in reading mammograms and identifying high-risk lesions.
AI-Supported Screening Increases Cancer Detection by 20%
A recent study published in the journal The Lancet Oncology is the first randomized control trial to compare AI-assisted breast cancer detection with detection done by well-trained humans alone. The study involved 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 in which 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 results showed that the group whose scans were read by a radiologist along 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 screened women, compared with 5 per 1,000 with the standard approach. The AI did not increase the number of false positives, when a mammogram is diagnosed as abnormal even though no cancer is present.
The study also found that the AI reduced the reading workload of the radiologists by 44%. The researchers estimated that if radiologists read about 50 mammograms an 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 could allow radiologists to be less burdened by the excessive amount of reading,” said study co-author Dr. Kristina Lång, an associate professor of radiology diagnostics from Lund University in Sweden.
AI Performs Similarly to Radiologists in Reading Mammograms
Another smaller study, published earlier this week in the journal Radiology, found that radiologists and AI came to similar conclusions after reviewing the same mammograms. The study involved 240 women who had mammograms at a hospital in Aberdeen, Scotland, between January and March 2020. The mammograms were read by two radiologists and by an AI tool called Mia, which was developed by a company called Kheiron Medical Technologies.
The study found that Mia agreed with the radiologists’ decisions in 90% of the cases. Mia also detected 13% more cancers than the radiologists did, but also had more false positives. The researchers said that Mia could be used as a second reader to support the radiologists, or as a triage tool to prioritize the cases that need urgent attention.
The study also reported that Mia helped detect early-stage breast cancer for one of the participants, who was a healthcare assistant and volunteered for the trial. She had a mammogram that was initially read as normal by the radiologists, but Mia flagged it as suspicious. A subsequent biopsy confirmed that she had cancer and she is now set to undergo surgery.
“I’m very grateful that I took part in the trial and that the AI picked up the cancer. It was a complete shock, but I’m glad it was detected early,” she said.
AI Still Has Limitations and Challenges
Despite the promising findings, the researchers cautioned that AI still has limitations and challenges to overcome before it can be widely used in breast cancer screening programs. They said that AI needs to be validated in larger and more diverse populations, and that it needs to be integrated with the existing workflows and systems of the radiology departments. They also said that AI needs to be transparent and explainable, so that the radiologists can understand how it reaches its decisions and trust its recommendations.
Moreover, the researchers said that AI is not meant to replace human radiologists, but to augment their skills and expertise. They said that radiologists still have the final say in the diagnosis and treatment of the patients, and that they need to be trained and updated on the use and performance of AI.
“AI is not a magic bullet, but it can be a valuable tool to help radiologists improve the quality and efficiency of breast cancer screening and detection. We need to continue to evaluate and monitor its impact and potential, and to ensure that it is used in a safe and ethical way,” said Dr. Stephen Duffy, a professor of cancer screening at Queen Mary University of London and a co-author of the Lancet Oncology study.
