New Delhi: Artificial intelligence (AI) can help detect breast cancer earlier and reduce the number of women diagnosed with cancer years after screening, a large randomised trial in Sweden has shown.
The study, published in The Lancet Friday, followed more than 100,000 women in Sweden and tested the use of AI-supported mammography—a breast X-ray used to look for early signs of cancer—in a national screening programme.
The researchers, from Sweden, Norway, Netherlands and Denmark, found that AI-supported mammography detected more cancers during routine screening and reduced breast cancer diagnoses in the next two years by 12 percent.
Mammography is an X-ray test that takes pictures of each breast from different angles, called cranio-caudal (CC), which is from top to bottom, and medio-lateral-oblique (MLO), which is at a slant from the side. It is the main way to check for breast cancer.
Results of the scan are reported using Breast Imaging-Reporting and Data System (BI-RADS), a method that helps doctors describe what they see and how likely it is to be cancer. While mammography is the gold standard, screening programs face challenges like unnecessary biopsies, misdiagnoses, and delayed detection.
To improve accuracy, AI-based tools have been developed since the 1990s. Using deep learning, these models analyse breast tissue through advanced algorithms, helping doctors interpret scans faster and more accurately, supporting early cancer detection.
Breast cancer remains a major global health burden, causing an estimated 6,70,000 deaths globally in 2022, according to the World Health Organization. In India, it is the most commonly diagnosed cancer among women, with 1,92,020 new cases reported in 2022.
Screening programmes aim to detect cancer early, when treatment works better. In Europe, mammograms are usually read independently by two radiologists. Even then, some cancers are missed and are diagnosed later, between screening rounds. These are known as interval cancers and are often harder to treat.
The new findings come from the Mammography Screening with Artificial Intelligence (MASAI) trial, the first randomised controlled trial to test AI use in breast cancer screening at this scale.
“Our study is the first randomised controlled trial investigating the use of AI in breast cancer screening and the largest to date looking at AI use in cancer screening in general. AI-supported screening improves the early detection of clinically relevant breast cancers, which leads to fewer aggressive or advanced cancers diagnosed in between screenings,” said Dr Kristina Lång from Sweden’s Lund University, lead author of the study, in a communication.
Earlier results from the MASAI trial had already suggested benefits. Interim safety findings published in The Lancet Oncology in 2023 showed that AI reduced radiologists’ screen-reading workload by 44 percent. Another analysis, published in The Lancet Digital Health in 2025, reported a 29 percent increase in cancer detection without a rise in false positives.
The full results now show that these gains also translate into fewer cancer diagnoses in the years following a breast cancer screening.
Also Read: Cancer registry covers only 10% of Indians. Call to make it notifiable disease is gathering steam
A look at the study
In the MASAI trial, held between April 2021 and December 2022, women attending routine screening at four centres in Sweden were randomly assigned to either AI-supported screening or standard screening without AI. The median age was 54 years.
In the AI-supported screening group, a computer system analysed each mammogram before doctors reviewed it. The AI checked the breast X-ray for signs that could point to cancer and gave each case a risk score. Mammograms marked as low risk were read by one radiologist, while higher-risk cases were read by two radiologists, which is the standard process.
The AI screening also highlighted areas on the image that looked suspicious, helping doctors decide whether a woman should be called back for further tests. Radiologists made the final decisions in all cases.
Over two years of follow-up, the AI group recorded 1.55 interval cancers per 1,000 women, compared with 1.76 per 1,000 in the standard screening group. This represents a 12 percent reduction.
The AI group also had fewer serious cases. There were fewer cancers that had spread, fewer large tumours, and fewer aggressive types compared to women who went through standard screening.
AI also helped doctors find more cancers during the screening itself. About 81 percent of cancers in the AI group were caught during screening, compared to 74 percent in the standard group. At the same time, the number of false alarms stayed almost the same in both groups.
First author of the study Jessie Gommers, a PhD student at Radboud University Medical Centre in the Netherlands, said AI is meant to support, not replace, doctors.
“Our study does not support replacing healthcare professionals with AI,” she said. “The AI-supported mammography screening still requires at least one human radiologist, but with support from AI.”
The authors said the findings suggest AI could help reduce the heavy workload faced by radiologists, allowing them to spend more time on other clinical work and possibly cut waiting times for patients.
The researchers also noted that the study was conducted in one country using one AI system and one type of mammography machine, and that data on race and ethnicity were not collected.
“Introducing AI in healthcare must be done cautiously,” Dr Lång said. “Further studies and cost-effectiveness analyses will help clarify the long-term benefits and risks, especially as many countries face shortages of trained radiologists.”
(Edited by Nida Fatima Siddiqui)
Also Read: 1 in 7 women aged 50-74 who undergo mammogram ‘overdiagnosed’ for breast cancer, finds US study