Medical researchers have advanced breast cancer detection with two new artificial intelligence tools. Researchers at AIIMS Delhi developed an AI assistant trained on Indian patient data to flag mammogram anomalies, transferring the technology to BPL Technologies on September 16, 2026. Meanwhile, a New York University study led by Yanqi Xu demonstrated that a deep learning model analyzing longitudinal 3D mammography scans significantly improves five-year breast cancer risk prediction.
DBT-based AI risk prediction model
- ▪The New York University deep learning model integrated up to ten longitudinal digital breast tomosynthesis examinations per patient, patient age, and breast density.
- ▪Researchers at New York University developed a deep learning model using longitudinal digital breast tomosynthesis examinations to predict five-year breast cancer risk.
- ▪The study on the digital breast tomosynthesis risk prediction model, led by Yanqi Xu, was published in the American Journal of Roentgenology.
AIIMS mammography AI tool
- ▪Researchers at the All India Institute of Medical Sciences in Delhi developed an artificial intelligence tool to act as a second reader of mammograms.
- ▪The All India Institute of Medical Sciences mammography artificial intelligence model was trained entirely on Indian patient data over approximately one to one-and-a-half years.
- ▪The development of the All India Institute of Medical Sciences mammography artificial intelligence model received initial funding from the Indian Ministry of Education.
Model performance validation results
- ▪The All India Institute of Medical Sciences mammography artificial intelligence tool demonstrated a 99% negative predictive value during retrospective testing on 3,000 to 4,000 patients.
- ▪The New York University longitudinal risk model achieved a five-year area under the curve of 0.721, outperforming the Mirai and Tyrer-Cuzick models.
Risk stratification by breast density
- ▪The New York University model identified 14.8% of women with fatty breasts as high risk, who exhibited a 2.6% five-year breast cancer incidence.
- ▪The New York University model classified 39.7% of women with extremely dense breasts as average risk, who subsequently showed a 0.8% five-year cancer incidence.
Clinical deployment readiness
- ▪The All India Institute of Medical Sciences mammography artificial intelligence tool has been run in a non-clinical trial mode at the institute for approximately one year.
- ▪The All India Institute of Medical Sciences artificial intelligence tool could potentially be deployed in hospitals within approximately six months from September 2026, pending regulatory approvals.
Technology transfer to BPL
- ▪The All India Institute of Medical Sciences transferred its mammography artificial intelligence technology to BPL Technologies Limited on September 16, 2026.
- ▪BPL Technologies Limited will develop the All India Institute of Medical Sciences model into a finished product, seek regulatory clearances, and manage hospital deployment.
Debatable claims
- ▪Clinical guidelines should use the NYU deep learning model to personalize breast cancer screening
- ▪Indian hospitals should deploy the AIIMS mammography tool to address radiologist shortages
Story comments
Loading comments…