Wednesday, November 28 | 10:40 a.m.-10:50 a.m. | SSK01-02 | Room E451A
The performance of synthesized 2D mammography images created from digital breast tomosynthesis scans in a screening program is mixed, according to findings being presented on Wednesday morning.In a study conducted by University of Pennsylvania investigators, synthesized 2D had comparable cancer detection rates and decreased recall rates compared with digital mammography, but it also had higher false-negative rates.
"Synthesized 2D imaging is replacing 2D digital mammography in digital breast tomosynthesis [DBT] screening to reduce radiation dose," wrote the team led by Dr. Samantha Zuckerman. "However, there have been reports of decreased rates of detection of in situ carcinomas with synthesized 2D/DBT screening. [We sought] to compare screening outcomes as well as false-negative rates and lesion types in digital mammography/DBT versus synthetic 2D/DBT screening."
Zuckerman and colleagues included 37,184 women who underwent digital mammography/DBT screening between October 2011 and January 2015 and 37,996 women who had synthetic 2D/DBT screening between January 2015 and January 2018. They calculated recall, cancer detection, and false-negative rates per 1,000 women screened.
The researchers found the following:
- Synthetic 2D/DBT decreased the recall rate compared with digital mammography/DBT, at 6.9% versus 8.9% (p < 0.001).
- The difference in cancer detection rate between the two modalities was not statistically significant, at 6 per 1,000 for digital mammography/DBT and 5 per 1,000 for synthetic 2D/DBT (p = 0.37).
- However, the false-negative rate more than doubled for synthetic 2D/DBT compared with digital mammography/DBT, at 0.84 per 1,000 versus 0.40 per 1000 (p = 0.02).
Why did synthesized DBT produce such a high false-negative rate? It could be due to the learning required when synthesized 2D technology is incorporated into a radiology practice, according to the researchers.
"The increase in false-negative rates ... may be intrinsically related to the new technology and/or a learning curve in implementing the synthetic 2D technology," the group concluded.












![A normal mammogram confirmed by three-year radiologic follow-up illustrates reader-marked regions of interest (ROIs) during (A) unaided (round 1) and (B) artificial intelligence (AI)–assisted (round 2) reading. Each colored dot represents an ROI for recall by a human reader. Readers could mark more than one ROI per case, represented by multiple dots of the same color. During AI-assisted reading, the AI system displayed three visible prompts: two with suspicion of malignancy scores of 35% (left mediolateral oblique [L MLO] and craniocaudal [L CC]) and one with a suspicion of malignancy score of 10% (right craniocaudal [R CC]), shown as polygonal overlays. Without AI, six of 10 readers (60%) marked a false-positive ROI. With AI assistance, this fell to two of 10 (20%). R MLO = right mediolateral oblique.](https://img.auntminnie.com/mindful/smg/workspaces/default/uploads/2026/07/2026-07-14-radiology-mammogram-ai-auto-bias.H0bYO8QlWs.jpg?auto=format%2Ccompress&fit=crop&h=112&q=70&w=112)






