Performance across different versions of an artificial intelligence model for screen-reading of mammograms - Takeaways - MDSpire

Performance across different versions of an artificial intelligence model for screen-reading of mammograms

  • By

  • Marthe Larsen

  • Christoph I. Lee

  • Marie B. Bergan

  • Åsne S. Holen

  • Håkon Lund-Hanssen

  • Solveig R. Hoff

  • Steinar Auensen

  • Jan F. Nygård

  • Kristina Lång

  • Yan Chen

  • Giske Ursin

  • Solveig Hofvind

  • January 13, 2026

  • 0 min

Share

  • 1

    AI models for mammography screening can enhance interpretation, potentially increasing cancer detection rates and reducing false positives.

  • 2

    Version updates of AI models may influence interpretive performance and screening outcomes, necessitating evaluation of each version.

  • 3

    The study analyzed data from 117,709 mammography examinations in BreastScreen Norway to assess performance differences between AI model versions.

  • 4

    Transpara versions 1.7 and 2.1 were compared, with version 2.1 incorporating improved algorithms and broader training data.

  • 5

    Key performance measures included screen-detected cancer rates and interval cancer rates, alongside histopathological tumor characteristics.

Original Source(s)

Performance across different versions of an artificial intelligence model for screen-reading of mammograms

  • by Marthe Larsen, Christoph I. Lee, Marie B. Bergan, Åsne S. Holen, Håkon Lund-Hanssen, Solveig R. Hoff, Steinar Auensen, Jan F. Nygård, Kristina Lång, Yan Chen, Giske Ursin, Solveig Hofvind

  • January 13, 2026

Related Content

Evaluation of SPIO-Enhanced MRI in Distinguishing Metastatic Lymph Nodes from Reactive Hyperplasia in Breast Cancer: Diagnostic Efficacy and Correlation with VEGF-C Levels

Dana-Farber Research News 03.15.2026

This twice-monthly newsletter highlights recently published research where Dana-Farber faculty are listed as first or senior authors. The information is pulled from PubMed and this issue notes papers published from February 16 - 28.

  • by

  • March 15, 2026

  • 21 min
Correction: Evaluation of Large Language Models for Radiologists' Support in Multidisciplinary Breast Cancer Teams: Comparative Study