Wide-Angle Tomosynthesis and AI in Diagnostic Mammography

Official title

Evaluation of Wide-Angle Tomosynthesis and AI in Diagnostic Mammography at The Ottawa Hospital

Overview

Breast cancer remains the most commonly diagnosed cancer and a leading cause of cancer-related mortality among women globally. Timely and accurate detection is crucial for improving prognosis and survival outcomes. While digital mammography has long served as the gold standard for screening, it is limited by overlapping tissue structures, particularly in women with dense breasts, which can obscure malignancies or create false positives. To address these limitations, digital breast tomosynthesis (DBT), especially wide-angle DBT, has been developed to offer three-dimensional imaging and reduce tissue overlap. Siemens' MAMMOMAT B.brilliant system, which incorporates wide-angle DBT, enhances spatial resolution and improves lesion conspicuity. This technology may offer significant benefits in diagnostic populations, where accuracy and confidence in imaging interpretation are crucial. In parallel, artificial intelligence (AI) tools such as the Transpara system have been introduced to further improve mammographic interpretation. Previously the evaluation of Transpara in a sample of 310 Japanese women and found that while human readers outperformed AI in overall diagnostic performance, the system showed promising sensitivity levels, highlighting the potential of AI as a decision-support tool rather than a standalone reader. More robust evidence is provided by the Mammography Screening with Artificial Intelligence (MASAI) trial, which assessed AI-supported screen reading in a controlled study of over 80,000 women. The trial found that AI-supported reading led to a comparable cancer detection rate as standard double reading (6.1 vs. 5.1 per 1000 participants) but reduced reading workload by 44.3% without increasing false positives or recall rates. A related analysis by the same team emphasized the capability of AI to triage exams effectively and highlighted that AI-flagged "extra high risk" mammograms accounted for a substantial portion (over 55%) of all screen-detected cancers, with a high positive predictive value. Despite these encouraging findings, most studies have been limited to screening-based settings. There remains a lack of prospective evidence on the real-world diagnostic application of wide-angle DBT and AI in populations at higher risk, such as symptomatic patients or those recalled from screening. This represents a critical knowledge gap, especially given increasing concerns about radiologist workload and diagnostic delays. The purpose of this prospective observational study is to evaluate the integration and diagnostic value of wide-angle tomosynthesis and AI (Transpara) in a clinical diagnostic setting. Specifically, it aims to assess their influence on radiologist confidence, diagnostic accuracy and the need for supplementary imaging. By addressing these questions, the study seeks to inform future implementation strategies that balance accuracy, efficiency, and clinical utility.

Detailed description

This prospective observational study evaluates the use of wide-angle digital breast tomosynthesis (DBT) and an artificial intelligence (AI) decision-support tool (Transpara) during diagnostic mammography at The Ottawa Hospital. All imaging performed in the study is part of routine clinical care and uses the Siemens MAMMOMAT B.brilliant system. The first 700 patients will have images interpreted without AI, and the next 700 with AI available to the radiologist. No additional imaging or procedures are required beyond standard care. The study will compare diagnostic confidence, need for supplementary imaging, biopsy outcomes, and overall workflow efficiency between the AI-supported and non-AI groups. Clinical follow-up for up to two years will be used to assess diagnostic accuracy and cancer outcomes.

Primary outcome measures

• Diagnostic Confidence and Diagnostic Accuracy With and Without AI Support [Time frame: 1- Day 1: Assessments at the diagnostic imaging visit (scan with or without AI). Biopsy collected. Radiologist reader confidence (BI-RADS). 2- Day 1 up to 6 months: Positive Predictive Value of Biopsy (PPV3). 3- 2 year follow-up: Diagnostic accuracy.]

Eligibility criteria

Inclusion criteria

• Provides verbal consent to participate.
• Referred for diagnostic breast imaging at The Ottawa Hospital due to:
• Recall from a screening mammogram for a soft-tissue lesion, or
• Breast symptoms (e.g., palpable mass, nipple discharge) with last screening mammogram >6 months prior.
• Able to undergo wide-angle DBT and Insight 2D views on the Siemens MAMMOMAT B.brilliant system.

Exclusion criteria

• Presence of breast implants.
• History of breast surgery on the breast being evaluated.
• Required imaging views not obtained (wide-angle DBT + Insight 2D views).
• Unable or unwilling to complete the imaging procedure per standard protocol.
• Declines the use of AI on the mammography unit (patients who decline are imaged on another machine and not included).

Criteria are shown verbatim from the registry (in English). Final eligibility is always assessed by the study center.

Healthy volunteers: No

Study design

Observational model

Cohort

Study locations

Center list to be confirmed — check the primary protocol.

Publications

• Lang K, Josefsson V, Larsson AM, Larsson S, Hogberg C, Sartor H, Hofvind S, Andersson I, Rosso A. Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): a clinical safety analysis of a randomised, controlled, non-inferiority, single-blinded, screening accuracy study. Lancet Oncol. 2023 Aug;24(8):936-9 PMID 37541274
• Sasaki M, Tozaki M, Rodriguez-Ruiz A, Yotsumoto D, Ichiki Y, Terawaki A, Oosako S, Sagara Y, Sagara Y. Artificial intelligence for breast cancer detection in mammography: experience of use of the ScreenPoint Medical Transpara system in 310 Japanese women. Breast Cancer. 2020 Jul;27(4):642-651. doi: 10.1007/s12282-020-01061-8. Epub 2020 Feb 12. PMID 32052311

Identifiers

NCT: NCT07491055 · 20250599-01H

Primary sources (government registries)

• ClinicalTrials.gov (US NIH)

View this study on ClinicalTrials.gov ↗