AI Detects Nearly One-Third of Breast Cancers Previously Undetected by Radiologists on DBT Imaging
The AI program also correctly identified 90% of true positive cases, according to new data presented at the Society for Breast Imaging's annual symposium.
An artificial intelligence (AI) program used to read 3D mammography scans correctly identified about 90% of breast cancers initially detected by radiologists and 32% of cancers that were initially missed, according to a new study conducted at Massachusetts General Hospital (MGH).1
The AI, Hologic’s Genius AI Detection 2.0, performed better at spotting invasive ductal carcinomas, grade 3 invasive cancers, and lymph node-positive cancers, although the technology was less effective at detecting invasive lobular carcinomas, grade I cancers, and tumors presenting as subtle asymmetries, according to the study, published in The Journal of Breast Imaging.1 The findings were presented at the Society of Breast Imaging (SBI) Annual Symposium, April 24-27, in Colorado Springs, CO.
Researchers retrospectively analyzed 5,000 digital breast tomosynthesis exams performed between 2016 and 2019. The data set included 500 cancers initially detected by radiologists and 100 cancers that had been missed during initial readings but later confirmed to be malignant.1
The AI program correctly identified approximately 90% of the cancers originally found by radiologists and detected 32% of the cancers that were initially missed. Program performance varied depending on tumor type and presentation: it was more accurate in identifying invasive ductal carcinomas and cancers presenting as masses or in dense breast tissue, although some findings were not statistically significant. It was also more successful at detecting grade 3 invasive carcinomas and lymph node-positive cancers, but less effective with grade I invasive carcinomas, and cancers that appeared as subtle breast asymmetries.1
Notably, the AI was more likely to correctly detect previously missed cancers when radiologists had already noted subtle mammographic findings, particularly breast calcifications, which can serve as early indicators of cancer.1 In terms of specificity, among the 4,400 exams deemed cancer-free by radiologists, the AI correctly categorized 55% as negative, the study revealed.1
“The purpose of our (new) study was to fill that gap in knowledge or understanding by providing an analysis of the types of breast cancers that are missed versus detected by a commercial AI based CAD algorithm for tomosynthesis,” Manisha Bahl, MD, MPH, director of MGH’s breast imaging fellowship program and associate professor of radiology at Harvard Medical School, remarked in an interview with Diagnostic Imaging.2 She presented the findings at the SBI meeting.
“As AI continues to evolve, I believe it will become an increasingly vital tool for radiologists, helping to transform breast cancer detection and ultimately reduce the burden of this disease for patients,” Bahl said in a Hologic statement.3
The Genius AI Detection 2.0 program was launched in 2023. Hologic said it reduced false-positive markings per case by more than 70% compared to its previous ImageChecker CAD solution.
According to the company, its updated Genius AI Detection PRO workflow platform has been shown to reduce radiologists’ total reading time by 24%.1
“It’s an exciting time for breast health innovation, and we’re proud to showcase our latest advancements at this year’s SBI event,” Mark Horvath, president of breast and skeletal health solutions at Hologic, said in the statement.3
The MGH study provides important insights into how AI tools can complement radiologists, particularly in complex cases, but also underscores the need for further refinement to maximize cancer detection without increasing false positives, Bahl commented.3
References
1. Pacile S, Germaine P, Sclafert C, Bertinotti T, Fillard P Long SS. Evaluation of a multi-instant multimodal artificial intelligence system supporting interpretive and noninterpretive functions. J Breast Imaging. 2025;7(2):155–164. doi:10.1093/jbi/wbae062
2. Groundbreaking new data on Hologic’s AI-powered mammography technology to be presented at SBI. News release. Hologic. April 24, 2025. Accessed April 28, 2025. https://investors.hologic.com/press-releases/press-release-details/2025/Groundbreaking-New-Data-on-Hologics-AI-Powered-Mammography-Technology-to-Be-Presented-at-SBI/default.aspx
3. Hall J. What new research reveals about the impact of AI and DBT screening: an interview with Manisha Bahl., MD. Diagnostic Imaging. April 25, 2025. https://www.diagnosticimaging.com/view/new-research-ai-and-dbt-screening-interview-manisha-bahl-md
