New Delhi: A landmark study based on the world’s first randomised clinical trial to assess Artificial Intelligence efficacy in detecting breast cancers has found that AI-supported screen-reading procedures resulted in a significant 29 percent increase in identifying early malignancies.
Published this week in The Lancet Digital Health, the study Mammography Screening with Artificial Intelligence trial (MASAI) has shown that AI use in mammography/breast cancer screening reduced screen-reading workload without increasing false positives.
The results are based on randomised clinical trials as a part of Sweden’s national screening programme. Involving over 1 lakh women, the trials were held between April 2021 and 7 Dec 2022. Half of the women underwent AI-supported screening, and the other underwent the standard screening procedure.
Initial data analysis from the trial came out in 2023, reporting that the integration of AI in breast cancer screening programmes was safe.
The latest findings suggested that cancer detection rates were 6·4 per 1,000 screened in the intervention group subjected to mammography with AI use compared to five per 1,000 in the control group.
Also, AI-supported screening increased the detection of in-situ cancer, or stage 0 cancer—when the abnormal cells are still where they first formed.
The advanced screening method also led to increased detection of invasive cancers—mainly small lymph node-negative cancers or cancers that had not spread to the lymph nodes and were detected early.
The results, said scientists, indicated that an AI-supported screen-reading procedure could contribute to the early detection of breast cancer, likely to be clinically progressive.
“Use of AI in mammography screening has the potential to reduce the screen-reading workload and increase cancer detection, which could affect patient outcomes,” the scientists associated with several Swedish institutions noted.
According to the GLOBOCAN 2022, released by the World Health Organisation’s International Agency for Research in Cancer (WHO-IARC), last year, breast cancer accounted for the second-highest number of malignancies—2.3 million or 11.6 percent of total cancer cases.
Worldwide, lung cancer is the leading form of cancer.
The report also showed that breast cancer continued to be the leading form of cancer in India, affecting 1,92,020 women and killing 98,337 in the country in 2022.
According to experts, delays in detecting cancer, apart from lack of awareness to undergo the screening regularly, is a key reason why breast cancer outcomes in a majority of women in low-income countries such as India are poor.
According to estimates, over 50 percent of breast cancer cases get diagnosed in late stages—3 or 4—in India, and the five-year survival rate for women with breast cancer is just 61 percent for Indian women compared to more than 80 percent in the US.
Experts hope the situation will significantly change with the extensive use of advanced AI tools.
A senior researcher in oncology attached with the Indian Council of Medical Research—who did not wish to be named—underlined that the findings might be relevant for mammography screening services worldwide, be that for evaluating, planning or beginning to implement AI for mammography.
“They can also be important for population screening programmes,” the ICMR scientist added.
Some others agreed, too.
“AI-powered imaging solutions, trained on vast datasets of mammograms and ultrasound scans, can identify subtle patterns and anomalies that may not be visible to the human eye,” Kabir Mahajan, chief operating officer with Delhi-based Mahajan Imaging & Labs, told ThePrint.
AI systems assist radiologists in detecting cancer at its earliest, most treatable stages while reducing false positives and unnecessary biopsies, minimising patient anxiety and healthcare costs, he explained.
Also Read: House panel urges covering medicine costs under Ayushman Bharat, especially for diseases like cancer
How AI was used to identify malignancy
Mammography screening—a medical imaging technique that uses low-dose X-rays to examine the breast for cancer and can detect malignancy—has been effective in reducing breast cancer mortality.
In most developed countries, standard screening examination involves the double reading of a mammogram. As a part of that, two radiologists independently review a mammogram to increase the accuracy of cancer detection. A second opinion on the images increases the chances of catching subtle abnormalities that a single reader might have missed.
In the MASAI trial, the inclusion criterion was women eligible to participate in population-based mammography screening, including general screening for women aged 40–74 years at 1·5- to 2-year screening intervals and annual screening for those with a moderate hereditary risk of breast cancer or a history of breast cancer.
The project, which involved the AI system Transpara version 1.7.0 made by Netherlands-based ScreenPoint Medical, was used to triage screening examinations to single- or double-reading and as detection support, highlighting suspicious findings.
To train AI to read mammograms, technicians input information from hundreds of thousands to millions of mammograms, and the AI software then creates a mathematical representation of what a normal mammogram looks like and what a mammogram with cancer looks like.
Transpara provided an examination-based, malignancy risk score on a scale of 1 to 10. It was pre-configured and calibrated to place approximately a tenth of the screening examinations in each risk score group. The categories were low risk (1–7), intermediate risk (8–9), and high risk (10).
Low- and intermediate-risk underwent single reading, and those with high risk underwent double reading. The screening exams of the control group were standard double reading.
AI-supported screening resulted in increased detection of invasive cancers, precancerous lesions or in-situ cancers, with about half of the increased detection being of high-grade, in-situ cancers.
The scientists noted that the significant increase in detected small, lymph node-negative, invasive cancers suggested that downstaging cancerous tumours after earlier detection with AI use was possible. It could be of clinical benefit since stage majorly influenced breast cancer treatment and prognosis, they added.
“Altogether, the use of AI has the potential to increase the early detection of clinically relevant breast cancer without unduly increasing the harm of false positives and overdiagnosis of low-grade in situ cancer,” they pointed out.
Also Read: Cancer registry covers only 10% of Indians. Call to make it notifiable disease is gathering steam
Far-reaching implications
Buoyed by the findings, the scientists highlighted that the MASAI trial results indicated that an AI-supported screen-reading procedure was safe to reduce the screen-reading workload and that the significant increase in cancer detection probably contributed to the early detection of clinically relevant breast cancer.
Experts back home told ThePrint that it was possible to integrate AI tools into X-ray mammography machines inside mobile vans. So, AI tools could play a pivotal role in promoting breast cancer screening programmes across the length and breadth of countries, such as India.
“Also, AI-driven risk assessment models analyse imaging data, lab medicine data, genomics data, along with patient history, genetic predisposition, and lifestyle factors, to predict susceptibility, enabling proactive screening strategies,” Mahajan said.
He added that integrating AI in telemedicine further ensured that women in remote and underserved areas gained access to timely and expert-driven diagnosis, bridging critical gaps in healthcare.
As AI continues to evolve, its role in breast cancer detection will extend beyond diagnosis to treatment planning, personalised therapies, and continuous patient monitoring, he stressed.
However, specialists also insisted that while AI was a powerful augmentation tool, its success lay in collaboration with healthcare professionals, ensuring that human expertise and ethical considerations guide its application.
Dr Rupa Ananthasivan, consultant of radiology with Manipal Hospitals in Bengaluru, underlined that with enhanced capacity of speed and accuracy of analysing MRIs, ultrasounds, and mammograms, AI is becoming a powerful tool in breast cancer screening.
“With its quick identification of suspicious lesions, AI is now empowering radiologist decisions in prioritising the most urgent cases,” she said.
Dr Ananthasivan also added that despite this, it may be key to remember it lacked legal authority to make decisions alone.
“Diagnostic centres should ensure AI’s use as a useful tool, not a substitute. AI must be properly trained in indigenous and diverse communities to be effective. It can help radiologists increase productivity and results, but it may never fully replace their knowledge and discretion,” she pointed out.
(Edited by Madhurita Goswami)
Also Read: How cancer daycare centres promised in Budget 2025 aim to ease treatment costs, access