Sunday, November 27 | 11:05 a.m.-11:15 a.m. | SSA01-03 | Arie Crown Theater
In this scientific paper presentation, University of Pittsburgh researchers will discuss how interpretation errors in screening ultrasound are similar in frequency to interpretation errors in mammography and MRI.Dr. Wendie Berg, PhD, and colleagues will present results from a study they conducted using data from the American College of Radiology Imaging Network (ACRIN) 6666 trial to determine sources of false-negative mammography, sonography, and MRI -- in particular, whether these errors were caused by bad image quality or errors in interpretation.
Berg's team asked three breast imaging radiologists to review breast imaging performed the year before diagnosis and the year of diagnosis for each of 130 malignant lesions in 110 women in the ACRIN 6666 trial. The reviewing radiologists recorded various factors that affected lesion detection, including technical, detection, and interpretation issues.
The group found that errors in interpretation of screening ultrasound were similar in prevalence (21% of misses) to errors in mammographic (28% misses) and MRI interpretation (20% misses).
![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=100&q=70&w=100)
![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&dpr=2&fit=crop&h=167&q=70&w=250)
