Reader study comparing AI-based HER2 scoring versus Standard of Care
Research type
Research Study
Full title
Study comparing pathologists' HER2 scoring of breast IHC slides when supported by the Galen™ Breast HER2 AI-powered solution versus the standard of care
IRAS ID
328938
Contact name
Dana Mevorach
Contact email
Sponsor organisation
Ibex Medical Analytics
Duration of Study in the UK
0 years, 5 months, 31 days
Research summary
This performance reader study aims to validate an AI-based HER2 scoring application which is meant to assist qualified pathologist in the measurement of images from microscope slides of breast cancer specimens stained for the presence of HER-2/neu receptor protein.
The proposed study is a two-arm observational reader study with retrospectively, anonymized collected IHC stained slides (4B5 antibody) from breast specimen.
The study will compare pathologists' HER2 scoring when assisted with the AI-based application versus when not assisted with it (working as per Standard of Care – using a microscope or digital viewer only).
The study will involve two sites - the Betsi Cadwaladr University Health Board, and the Cambridge University Hospitals NHS Foundation Trust.
2 reader pathologists from each site will review the retrospectively collected IHC stained slides in both arms:
Arm A (Manuel Read) – review using a digital viewer and manually score HER2 according to 2018 ASCO/CAP guidelines as in routine lab practice.Arm B (AI-assisted Read) - review and score HER2 IHC slides using the Galen™ Breast HER2 application.
Each reader pathologist will review all study slides in both arms, with a washout period of two weeks between arms.
The reader pathologists’ HER2 scoring in both arms will be compared to a ground truth set by additional blinded 2 breast pathologists.
The GT pathologists will review the cases as in their routine workflow, whether by a microscope or a digital viewer (without AI). In case of discordance between the 2 breast specialists, a third specialist will review the cases for adjudication. The final GT will be set as the concordance between all the GT pathologists.
Accuracy of HER2 score reporting with the Galen™ Breast HER2 application (Arm B) vs. score reporting without AI (Arm A) will be expressed as the percentage of cases that were in concordance between the readers and the GT in both study arms.
REC name
N/A
REC reference
N/A