Evaluation of an artificial intelligence (AI)-based system for assessing diagnostic image quality in mammography

Evaluation of an artificial intelligence (AI)-based system for assessing diagnostic image quality in mammography - AI-DBQ

Status

Recruiting

Phases

Unknown

Study type

Observational

Source

DRKS

Registry ID

DRKS00037958

Enrollment

1500

Registered

2025-10-06

Start date

2025-10-15

Completion date

Unknown

Last updated

2026-06-22

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

C50 D05

Interventions

Group 1: AI-BRAYZ: The b-rayZ software is used to check diagnostic image quality.

Eligibility

Sex/Gender

Female

Age

50 Years to 75 Years

Inclusion criteria

Inclusion criteria: Participants in the mammography screening program, four standard views of mammography

Exclusion criteria

Exclusion criteria: Men, postoperative mammograms, incomplete and technically unusable image data

Design outcomes

Primary

Measure	Time frame
Retrospective AI performance evaluation The AI's automated diagnostic image quality analysis is compared with the assessments of three independent experts who are unaware of either the AI's results or the other experts' results. Performance is measured both for overall image quality (e.g., grade 1/2/3) and for individual technical quality attributes (e.g., inframammary fold, pectoralis muscle, nipple in profile).	—

Measure

Time frame

Retrospective AI performance evaluation The AI's automated diagnostic image quality analysis is compared with the assessments of three independent experts who are unaware of either the AI's results or the other experts' results. Performance is measured both for overall image quality (e.g., grade 1/2/3) and for individual technical quality attributes (e.g., inframammary fold, pectoralis muscle, nipple in profile).

—

Secondary

Measure	Time frame
Simulation of future workflow integration In this adaptive phase, the practical utility and efficiency of the AI ??tool in the clinical setting are investigated. The specific methodology will be determined by the results of Phase 1. 1. Scenario A: High AI Agreement If Phase 1 demonstrates high agreement between the AI ??and the expert consensus, the study will focus on quantifying the efficiency gains of a fully automated workflow for assessing diagnostic image quality. This will be achieved by comparing the time required for the current manual process (baseline) with an automated process with no time required for the reader. 2. Scenario B: Moderate AI Agreement If Phase 1 demonstrates that the AI ??is effective but requires human supervision, the study will simulate an AI-assisted workflow. The readers are presented with the pre-analyzed AI results via a graphical user interface (GUI). The primary endpoint is the time required to confirm a correct AI assessment or correct an incorrect one. This “confirmation/correction time” is compared with the manual baseline to quantify the reduction in assessment time.	—

Measure

Time frame

Simulation of future workflow integration In this adaptive phase, the practical utility and efficiency of the AI ??tool in the clinical setting are investigated. The specific methodology will be determined by the results of Phase 1. 1. Scenario A: High AI Agreement If Phase 1 demonstrates high agreement between the AI ??and the expert consensus, the study will focus on quantifying the efficiency gains of a fully automated workflow for assessing diagnostic image quality. This will be achieved by comparing the time required for the current manual process (baseline) with an automated process with no time required for the reader. 2. Scenario B: Moderate AI Agreement If Phase 1 demonstrates that the AI ??is effective but requires human supervision, the study will simulate an AI-assisted workflow. The readers are presented with the pre-analyzed AI results via a graphical user interface (GUI). The primary endpoint is the time required to confirm a correct AI assessment or correct an incorrect one. This “confirmation/correction time” is compared with the manual baseline to quantify the reduction in assessment time.

—

Countries

Germany

Contacts

Public ContactSusanne Wienbeck

Referenzzentrum Mammographie Nord

wienbeck@referenzzentrum-nord.de+4941130930800

Outcome results

None listed

Source: DRKS (via WHO ICTRP) · Data processed: Jun 29, 2026