Enhancing HER2 testing in breast cancer - Original Article - predicting fluorescence in situ hybridization (FISH) scores from immunohistochemistry images via deep learning

Sep 23, 2026

Uploaded by Ezra

Breast cancer incidence worldwide requires precise HER2 biomarker testing to support effective therapy, especially emerging antibody-drug conjugates for tumors with low HER2 expression. Conventional immunohistochemistry (IHC) and fluorescence in situ hybridization (FISH) have important tradeoffs: IHC is semi-quantitative and vulnerable to interobserver variability, while FISH offers higher precision but is costlier and workflow-intensive despite faster turnaround than some molecular methods. A deep learning approach was developed to predict HER2 and FISH scores from IHC imagery, reducing reliance on reflex FISH testing where it is unavailable or prohibitive.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Abstract

Introduction

HER2 relevance and clinical rationale

Current HER2 testing methods and limitations

The Journal of Pathology: Clinical Research J Pathol Clin Res March 2025; 11: e70024 Published online in Wiley Online Library (wi[leyonlinelibrary.com](leyonlinelibrary.com). DOI: 10.1002/2056-4538.70024)[).](leyonlinelibrary.com). DOI: 10.1002/2056-4538.70024)[ DOI: 10.1002/2056-4538.70024](leyonlinelibrary.com). DOI: 10.1002/2056-4538.70024) ORIGINAL ARTICLE Enhancing HER2 testing in breast cancer: predicting ﬂuorescencein situ hybridization (FISH) scores from immunohistochemistry images via deep learning Daniel O Macaulay* , Wenchao Han, Mark D Zarella, Chris A Garcia and Thomas E Tavolara Division of Computational Pathology and AI, Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, MN, USA *Correspondence to: Daniel O Macaulay, Division of Computational Pathology and AI, Department of Laboratory Medicine and Pathology, Mayo Clinic, 200 1st Street SW, Rochester, MN 55905, [USA. E-mail:](USA. E-mail: macaulay.daniel@mayo.edu)[ macaulay.daniel@mayo.edu](USA. E-mail: macaulay.daniel@mayo.edu) Abstract Breast cancer affects millions globally, necessitating precise biomarker testing for effective treatment. HER2 testing is crucial for guiding therapy, particularly with novel antibody-drug conjugates (ADCs) like trastuzumab deruxtecan, which shows promise for breast cancers with low HER2 expression. Current HER2 testing methods, including immunohistochemistry (IHC) and in situ hybridization (ISH), have limitations. IHC, a semi-quantitative assay, is prone to interobserver variability. While ISH provides higher precision than IHC, it remains more resourceintensive in terms of cost and workﬂow. However, turnaround time is typically faster than that of other advanced molecular methods such as next-generation sequencing. We adapted the clustering-constrained-attention multipleinstance deep learning model to improve IHC testing and reduce dependence on reﬂex ﬂuorescence ISH (FISH) tests. Using 5,731 HER2 IHC images, including 592 cases with FISH testing, we trained two models: one for predicting HER2 scores from IHC images and another for predicting FISH scores from equivocal cases. The HER2 IHC score prediction model achieved 91% ± 0.01 overall accuracy and a receiver operating characteristic (ROC) area under the curve (AUC) of 0.98 ± 0.01. The FISH score prediction model had an ROC AUC of 0.84 ± 0.07, with sensitivity at 0.37 ± 0.13 and speciﬁcity at 0.96 ± 0.03. External validation on cases from 203 institutions showed similar performance. The HER2 IHC model maintained a 91% ± 0.01 accuracy and an ROC AUC of 0.98 ± 0.01, while the FISH model had an ROC AUC of 0.75 ± 0.03, with sensitivity at 0.28 ± 0.04 and speciﬁcity at 0.93 ± 0.01. Our model advances HER2 scoring by reducing subjectivity and variability in current scoring methods. Despite lower accuracy and sensitivity in the FISH prediction model, it may be a beneﬁcial option for settings where reﬂex FISH testing is unavailable or prohibitive. With high speciﬁcity, our model can serve as an effective screening tool, enhancing breast cancer diagnosis and treatment selection. Keywords: HER2; deep learning; breast cancer; immunohistochemistry; biomarkers; in situ hybridization; ﬂuorescence Received 23 July 2024; Revised 8 February 2025; Accepted 18 February 2025 No conﬂicts of interest were declared. Introduction Human epidermal growth factor receptor 2 (HER2) is a tyrosine-kinase transmembrane receptor that has been identiﬁed as a role player in the aggressive growth of breast cancer cells [1] . About 20% of breast cancer cases exhibit HER2 positivity, and testing is recommended for all primary invasive breast cancers, as HER2 status is critical for guiding targeted therapies [1,2] . Thanks to the discovery of new anti-HER2 therapies, the prognosis for patients with HER2-positive breast cancer has signiﬁcantly improved [3] . Antibody-drug conjugates (ADCs) like trastuzumab deruxtecan have changed the landscape signiﬁcantly [2,4] . A

FAQs

Why is accurate HER2 testing important in breast cancer?

HER2 status guides targeted therapies, and emerging treatments such as antibody-drug conjugates increase the need for reliable biomarker measurement.

What limitations affect current HER2 testing methods?

IHC is semi-quantitative and prone to interobserver variability, while FISH is more precise but more costly and workflow-intensive.

How does the proposed deep learning model improve HER2 and FISH scoring?

The model predicts HER2 scores from IHC images and predicts FISH scores for equivocal cases, aiming to reduce subjectivity and lessen dependence on reflex FISH testing.

Sep 23, 2026

Uploaded by Ezra