Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jun 30, 2026

Clinical Imaging of Microwave Mammography
05:28

Clinical Imaging of Microwave Mammography

Published on: November 14, 2025

Cross-Device Adaptation of Mirai for Mammography-Based Breast Cancer Risk Prediction.

Adriana Sistig, Joseph H Rothstein, Tejomay Gadgil

    Medrxiv : the Preprint Server for Health Sciences
    |June 29, 2026
    PubMed
    Summary

    Related Concept Videos

    You might also read

    Related Articles

    Articles linked to this work by shared authors, journal, and citation graph.

    Sort by
    Same author

    Genetic survey of biomarkers at early and mid-pregnancy identifies pregnancy-specialized immune regulation.

    PLoS genetics·2026
    Same author

    Comparative 10-year performance of mammography artificial intelligence, polygenic, and clinical breast cancer risk models in the Kaiser Permanente Research Bank.

    Journal of the National Cancer Institute·2026
    Same author

    Transcriptome-wide association studies implicate RCC1 and PHACTR4 in prostate cancer survival.

    European journal of human genetics : EJHG·2026
    Same author

    Breast Arterial Calcification on Mammography: Act Now Please.

    Journal of breast imaging·2026
    Same author

    Inherited genetic risk factors in young-onset lung cancer.

    medRxiv : the preprint server for health sciences·2026
    Same author

    AlphaGenome Enhances Personal Gene Expression Prediction but Retains Key Limitations.

    bioRxiv : the preprint server for biology·2026

    Adapting breast cancer risk models like Mirai to new mammography equipment can harm performance on older systems. A new device-invariant method improves accuracy across different imaging platforms, enhancing robustness for diverse clinical settings.

    Area of Science:

    • Artificial Intelligence
    • Medical Imaging
    • Biomedical Engineering

    Background:

    • Deep learning models for medical imaging require adaptation to new datasets.
    • Device-specific domain shifts can limit the generalizability of these models.
    • Breast cancer risk prediction models face challenges with diverse mammography equipment.

    Purpose of the Study:

    • To evaluate the performance of the Mirai mammography-based deep learning model across different digital mammography systems (Hologic, GE Premium View, GE Tissue Equalization).
    • To investigate the impact of device-specific fine-tuning on model performance and generalizability.
    • To develop and validate a device-invariant model to improve robustness across heterogeneous imaging platforms.

    Main Methods:

    • Evaluated the native Mirai model on a large screening cohort with Hologic and GE mammography systems.

    More Related Videos

    Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
    15:48

    Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging

    Published on: December 15, 2014

    Related Experiment Videos

    Last Updated: Jun 30, 2026

    Clinical Imaging of Microwave Mammography
    05:28

    Clinical Imaging of Microwave Mammography

    Published on: November 14, 2025

    Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
    15:48

    Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging

    Published on: December 15, 2014

  • Assessed the effects of fine-tuning on GE Tissue Equalization (TE) images, noting performance changes on Hologic images.
  • Developed a device-invariant model using interleaved multi-device sampling and conditional adversarial training to mitigate domain shift issues.

    • Main Results:

    • Native Mirai showed reduced performance on GE TE images compared to Hologic or GE Premium View (PV) images.
    • Fine-tuning on TE images improved TE performance but caused significant performance degradation on Hologic images (catastrophic forgetting).
    • The developed device-invariant model largely restored Hologic performance while maintaining improved TE performance, demonstrating enhanced cross-platform robustness.

    Conclusions:

    • Device-specific fine-tuning of mammography risk models offers benefits but carries risks of reduced generalizability.
    • A balanced domain-adaptation strategy using interleaved sampling and adversarial training can improve model robustness across diverse imaging equipment.
    • Performance gains are most significant for short- and intermediate-term breast cancer risk predictions.