Sweden’s Mammography Screening with Artificial Intelligence (MASAI) randomised controlled trial recently published its final results (Gommers, et al., 2026, The Lancet), which highlighted that the use of Transpara Detection 1.7.0 contributed to a 12% reduction in the rate of interval cancers with 27% fewer aggressive cancers (non-luminal A subtype). MASAI, which involved over 105,000 women, is a first-of-its kind trial exploring if AI could benefit mammography screening by reducing screen-reading workload and decreasing the rate of interval cancer.
Previous MASAI research from 2025 (Hernström, et al., 2025, The Lancet Digital Health) found that Transpara Detection increased the cancer detection rate for screen-detected cancers by 29% while simultaneously reducing screen-reading workload by 44% compared to the double-reading standard of care. However, as the United States breast screening practice revolves around a single reader model and is contextually different from Europe (FFDM vs DBT, annual screening), what can results from MASAI tell us about the state of breast AI research in the USA?
Key outcome
|
As seen in MASAI |
What could happen in the USA |
|
Reduction of interval cancer |
In aggregate, the MASAI trial observed a large increase in cancer detection (29% increase in the cancer detection rate based on 2025 data) and a non-inferior reduction of interval cancer (a descriptive 12% reduction) without an increase in false positives, resulting in higher sensitivity and similar specificity. |
Research from the University of California, Los Angeles (Yu, et al., 2025, Journal of the National Cancer Institute) finds that Transpara Detection flagged 76% of 3D mammograms that had been originally read as normal but were later linked to an interval breast cancer. It also flagged 90% of missed reading error cases where cancer had been visible on the mammogram but missed or misinterpreted by the radiologist. While this UCLA study is retrospective and cancer-enriched, it is consistent with findings from MASAI and supports the potential to detect more clinically relevant cancers by identifying cases that were previously missed or minimally visible. |
|
Earlier and effective cancer detection |
According to MASAI, there were overall 16% fewer invasive, 21% fewer large (T2+), and 27% fewer non-luminal A interval cancers in the intervention group compared with the control group, suggesting “ a potential clinical benefit of earlier detection of clinically relevant breast cancer.” |
Findings from previous UCLA research (Plimpton, et al., 2024, Journal of Breast Imaging) suggest that Transpara "was able to detect cancers at the time of screening that were missed by the radiologist," and was impactful on flagging false negatives in the cohort, all of which were invasive and a majority of which had minimal signs on the preceding screening mammography, but were identified by Transpara and not by the radiologist. Transpara Detection offers the potential for earlier detection, with cancers identified up to 248 days earlier in a retrospective DBT cohort compared to radiologist interpretation. |
|
Workflow enhancement |
MASAI utilized a screen-reading procedure that used Transpara to triage screening examinations to single or double reading based on their assessed risk (Exams with low and intermediate risk underwent single reading and those with high risk underwent double reading); Transpara was also used as detection support in mammography screening. Together, this resulted in a 44.2% reduction in the screen-reading workload for radiologists with a similarly low false-positive rate. (Hernström, et al., 2025, The Lancet Digital Health) |
Within a single reader environment, the workflow benefits of Transpara would remain limited to a reduction in screen-reading time (e.g. more quickly reviewing unmarked, Low Risk exams) or a reduction in exams requiring human reading at all versus the benefits seen moving from double to single reading. Although research conducted on DBT exams from the USA (van Winkel, et al., 2021, European Radiology;) showed “that a deep learning–based AI system for DBT enables radiologists to increase their breast cancer detection performance in terms of overall accuracy and sensitivity at similar specificity, while reducing reading time" more direct research in the USA breast screening continuum is needed. While not yet FDA-cleared as a use case, a recent study in Nature Medicine (Elías-Cabot, et al., 2026, Nature Medicine) explored a new AI-based reading paradigm. In the AITIC clinical trial from Spain, an AI strategy used Transpara to classify exams (including DBT) as low risk and assess them as normal while the rest were double read with AI support: in this AI strategy, radiologist workload was 63.6% lower without negative clinical implications, suggesting Transpara can “safely reduce workload by excluding low-risk exams from radiologist reading.” Even in a single-reader model, this approach could net significant workload benefits while focusing radiologists’ time on clinically intensive studies. [i] |
|
High negative predictive value |
A high negative predictive value (NPV) enables workflow improvement: if a radiologist is able to confidently understand that a negative test result is extremely likely to mean no cancer is present (in this case a Low Risk exam categorization by Transpara), that radiologist can more quickly review those cases and focus their efforts on higher risk exams. NPV was not explicitly reported in MASAI; however, given the screening prevalence and outcome data, it can be estimated and it would exceed 99.8% for Low Risk exams within this real-world screening population of over 105,000 women. |
Additional research from UCLA (Chen, et al., 2025, American Journal of Roentgenology) finds a 99.8-99.9% NPV for Transpara in both DBT and DM (based on the algorithm’s risk categorization). |
As Scripps Health’s Dr. Eric Topol wrote recently in his newsletter, Ground Truths, “mammography with AI support has emerged as the most rigorously studied of all possible indications. It is a standout for demonstrating superhuman performance—’digital eyes’— that see things which humans can’t. I believe the data we now have is compelling and should set the stage for a new standard of care.”
This rigorous standard extends to the United States in the form of the PRISM randomised controlled trial. Led by UCLA and UC Davis, PRISM will study whether using Transpara Detection can help radiologists detect breast cancer more accurately and reduce unnecessary patient callbacks and anxiety for patients. As results from PRISM are published, its impact on elevating the standard of care will become even clearer even as its positive impact continues to reverberate for women and radiologists across the country.
[i] Disclaimer: Transpara is intended for use as a concurrent reading aid for physicians interpreting screening FFDM and DBT exams, to identify regions suspicious for breast cancer and assess their likelihood of malignancy, in accordance with its FDA clearance and CE certification. Patient management decisions should not be made solely on the basis of analysis by Transpara. The views expressed herein reflect ongoing research, published evidence, and product development considerations. They do not describe or promote use of Transpara beyond its currently cleared and certified intended use. Any discussion of autonomous or rule-out applications represents future or investigational concepts and is not FDA cleared or covered by the current CE certificate.