Outcomes of Single-View Asymmetries Recalled From Screening Mammography and Exploratory Analysis of AI Detection.

Authors

Huijuan Wang, Shrouq Solimanie, Fatma Eldehimi, Amit Katyan, Jean M Seely

Published in

Canadian Association of Radiologists journal = Journal l'Association canadienne des radiologistes. Pages 8465371261475941. Aug 20, 2026. Epub Aug 20, 2026.

Abstract

To evaluate the diagnostic outcomes of single-view asymmetries (SVAs) recalled from screening mammography and to explore the association between artificial intelligence (AI) detection and malignancy.
This REB-approved retrospective single-center study included women recalled for SVAs from screening mammography performed at The Ottawa Hospital between March 1, 2024 and March 31, 2025. SVAs were defined as asymmetries identified on a single mammographic projection. Screening performance metrics were obtained from the EPIC Mammography Quality Standards Act module, and imaging outcomes were reviewed in PACS. Primary outcomes included cancer detection rate (CDR), positive predictive value for recall (PPV1), and positive predictive value for biopsy (PPV3). Following implementation of AI software (Transpara) in September/October 2024, exploratory analyses evaluated associations between AI flag status and malignancy.
Among 32 412 screening examinations, 2455 patients (7.6%) were recalled, including 410 patients with 464 SVAs. Five cancers were identified, corresponding to a CDR of 0.16 per 1000 screened patients, significantly lower than the overall screening CDR of 6.08 per 1000 (P < .0001). PPV1 for SVAs was 1.22% (5/410), significantly lower than the overall PPV1 of 8.02% (197/2455) (P < .0001). Twenty-five patients (5.4%) underwent biopsy, yielding a PPV3 of 20%. Overall, 98.8% of recalled SVAs were non-malignant. Following AI implementation, malignancy rates were 0.6% (1/154) among non-flagged SVAs and 2.9% (2/68) among AI-flagged SVAs (P = .17). All malignancies were Stage I at diagnosis.
SVAs were a common but very low-yield cause of recall in screening mammography. Although exploratory, the low malignancy yield among non-AI-flagged SVAs suggests a potential role for AI as a supportive adjunct in risk stratification and recall decision-making.

PMID:
42624507
Bibliographic data and abstract were imported from PubMed on 21 Aug 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.