Radiogenomics in HER2-Positive Breast Cancer
- Dhanvin Raj Puppala
- 5 hours ago
- 6 min read
Introduction
HER2-positive breast cancer is defined by increased HER2 protein expression and/or ERBB2 gene amplification, and HER2 testing is clinically important because it identifies patients who may benefit from HER2-directed therapy.¹‚²
The American Cancer Society notes that HER2-positive breast cancers tend to grow and spread faster than HER2-negative cancers but are more likely to respond to drugs targeting HER2.¹
The 2023 ASCO-College of American Pathologists guideline update reinforces that HER2 testing focuses on detecting HER2 protein overexpression or ERBB2 amplification to guide therapy that disrupts HER2 signalling.²
Despite major treatment advances, response to therapy remains variable; for example, HER2-directed combinations such as pertuzumab, trastuzumab and docetaxel improved survival in metastatic HER2-positive breast cancer, but not every patient achieves the same depth or duration of response.³‚⁴
What radiogenomics means
Radiogenomics is the study of associations between imaging features and genomic, transcriptomic, molecular or histopathological tumour characteristics.⁵⁻⁸
In breast cancer, radiogenomics usually begins with imaging data such as magnetic resonance imaging (MRI), mammography, ultrasound, computed tomography (CT), positron emission tomography/computed tomography (PET/CT), or contrast-enhanced mammography, then relates quantitative imaging patterns to tumour biology.⁵⁻⁹
Radiomics is a key component of radiogenomics because it extracts quantitative imaging features, including shape, intensity, texture and spatial heterogeneity, from standard medical images.⁵‚⁶
Radiogenomics differs from routine visual image interpretation because it attempts to convert imaging patterns into measurable biomarkers that can be tested against molecular data and clinical outcomes.⁵⁻⁹
Why HER2-positive breast cancer is a relevant setting
HER2-positive breast cancer is biologically and therapeutically distinct because HER2 status directly influences systemic treatment selection.¹⁻⁴
HER2 testing is performed on tumour tissue using methods such as immunohistochemistry and in situ hybridisation, but tissue sampling may not fully represent spatial heterogeneity across the whole tumour or metastatic disease burden.¹‚²‚⁷‚⁸
Medical imaging can assess the whole visible tumour and, in metastatic disease, multiple tumour sites; therefore, radiogenomics is attractive because it may provide non-invasive information about tumour heterogeneity and treatment response.⁵⁻⁹
This does not mean imaging replaces biopsy or pathology: current HER2 classification and treatment eligibility still depend on validated pathology testing and established clinical guidelines.¹‚²
Biological rationale
Tumours contain differences in cellular density, vascularity, necrosis, extracellular matrix composition and immune microenvironment, and these features can influence how lesions appear on imaging.⁵⁻⁸
MRI features may reflect vascular permeability, diffusion restriction or enhancement patterns, while PET/CT features may reflect metabolic activity or tracer uptake.⁵‚⁶‚⁸
Radiogenomic research asks whether these imaging phenotypes correlate with molecular phenotypes such as receptor status, gene-expression patterns, ERBB2 amplification, proliferation signatures, or immune-related biology.⁷⁻⁹
In breast cancer, machine-learning radiogenomic studies have shown that MRI-derived image features can be analysed against molecular and clinical characteristics, although such models require external validation before clinical use.⁹⁻¹²
Potential clinical applications
1) Prediction of treatment response
A major goal is to predict response to neoadjuvant systemic therapy before surgery, including whether a patient is likely to achieve pathological complete response.⁵⁻⁹
If validated, radiogenomic tools could help identify patients who may benefit from treatment escalation, de-escalation, or closer response monitoring, but such use must be proven prospectively before routine implementation.¹⁰⁻¹³
2) Monitoring tumour heterogeneity
HER2 expression and other biological features may vary within a tumour or between tumour sites, while imaging can repeatedly evaluate the whole lesion and metastatic burden over time.²‚⁵⁻⁸Radiogenomics may therefore help identify imaging patterns that suggest biologically aggressive or heterogeneous disease, although the interpretation of these patterns remains investigational.⁷⁻¹⁰
3) Prognosis and recurrence risk
Radiomic and radiogenomic signatures are being studied as prognostic biomarkers for outcomes such as recurrence, disease-free survival, progression-free survival and overall survival.⁵⁻⁹‚¹²Prediction-model reporting guidance such as TRIPOD+AI is relevant because many radiogenomic models use regression or machine-learning approaches and must report model development, validation, performance and transparency clearly.¹²
4) Trial enrichment and precision oncology
Radiogenomics may support precision oncology research by helping stratify patients according to imaging-based tumour phenotypes, molecular features and predicted response patterns.⁷⁻⁹‚¹²‚¹³
However, radiogenomic biomarkers should only be used to guide treatment after they demonstrate reproducibility, external validity, clinical usefulness and acceptable safety in relevant patient populations.⁶‚¹⁰⁻¹³
How radiogenomic models are usually developed
Step | Description | Why it matters |
Image acquisition | MRI, CT, PET/CT, ultrasound, mammography or other imaging is collected using defined protocols.⁵‚⁶ | Scanner settings and protocols can affect extracted features.⁵‚⁶ |
Segmentation | The tumour or region of interest is outlined manually, semi-automatically or automatically.⁵‚⁶ | Different segmentations can produce different radiomic features.⁵‚⁶ |
Feature extraction | Quantitative features such as shape, intensity and texture are extracted from the image.⁵‚⁶ | IBSI standardisation aims to reduce inconsistency between software tools.⁶ |
Molecular linkage | Imaging features are compared with pathology, HER2/ERBB2 status, gene expression or other tumour biology data.⁷⁻⁹ | This step creates the radiogenomic association.⁷⁻⁹ |
Model building | Statistical or machine-learning methods are used to predict biology, response or prognosis.⁹‚¹² | Transparent reporting and appropriate validation are essential.¹⁰⁻¹² |
Validation | Performance is tested in internal and ideally external datasets.¹⁰⁻¹² | External validation is needed before clinical translation.¹⁰⁻¹³ |
Key limitations
1) Reproducibility
Radiomic features can vary according to scanner type, acquisition protocol, reconstruction method, segmentation approach, preprocessing and feature-extraction software.⁵‚⁶
The Image Biomarker Standardisation Initiative was created to standardise radiomic feature definitions and improve comparability between radiomics software implementations.⁶
2) Small datasets and overfitting
Many radiogenomic studies use relatively small single-centre datasets, which increases the risk that a model performs well in the development cohort but poorly in new patients.⁵‚⁷⁻⁹‚¹²
External validation, calibration assessment, confidence intervals and transparent reporting are therefore necessary before a model can be considered clinically reliable.¹⁰⁻¹³
3) Clinical interpretability
Complex machine-learning models may identify image patterns that are statistically predictive but difficult for clinicians to interpret biologically.¹⁰⁻¹²
Guidelines such as CLAIM and TRIPOD+AI emphasise transparent reporting of data sources, model development, validation and intended clinical use.¹⁰⁻¹²
4) Not a replacement for pathology
Radiogenomics is best understood as a complementary approach rather than a substitute for established pathology-based HER2 testing.¹‚²‚⁷‚⁸
Treatment decisions for HER2-positive breast cancer should continue to rely on validated clinical, pathological and guideline-based evidence until radiogenomic tools are prospectively validated.¹⁻⁴‚¹⁰⁻¹³
Current position
Radiogenomics in HER2-positive breast cancer is promising but remains primarily a research field rather than a routine clinical tool.⁷⁻¹³
The strongest near-term role is likely to be in research settings, where radiogenomic models can be tested for response prediction, biological stratification, recurrence risk and trial enrichment.⁷⁻¹³
For clinical translation, future studies should use standardised radiomics workflows, clearly defined HER2-positive cohorts, robust outcome definitions, external validation, calibration assessment, decision-curve analysis and transparent reporting according to AI and prediction-model guidelines.⁶‚¹⁰⁻¹³
Conclusion
Radiogenomics combines medical imaging and tumour biology to better characterise HER2-positive breast cancer.⁵⁻⁹
Its potential value lies in non-invasive assessment of tumour heterogeneity, prediction of treatment response and support for precision oncology research.⁷⁻¹³
At present, the field requires stronger validation, standardisation and prospective testing before radiogenomic tools can be used to guide treatment decisions in routine HER2-positive breast cancer care.⁶‚¹⁰⁻¹³
References
1. American Cancer Society. What is HER2 status? Breast Cancer. Updated 2025. Available from: https://www.cancer.org/cancer/types/breast-cancer/understanding-a-breast-cancer-diagnosis/breast-cancer-her2-status.html
2. Wolff AC, Somerfield MR, Dowsett M, et al. Human epidermal growth factor receptor 2 testing in breast cancer: ASCO-College of American Pathologists Guideline Update. J Clin Oncol. 2023. PMID: 37284804. Available from: https://pubmed.ncbi.nlm.nih.gov/37284804/
3. National Cancer Institute. Two drugs that hit one target improve survival in women with metastatic breast cancer. 2015. Available from: https://www.cancer.gov/types/breast/research/two-drugs-one-target
4. Swain SM, Baselga J, Kim SB, et al. Pertuzumab, trastuzumab, and docetaxel in HER2-positive metastatic breast cancer. N Engl J Med. 2015;372:724-734. doi:10.1056/NEJMoa1413513.
5. Mayerhoefer ME, Materka A, Langs G, et al. Introduction to radiomics. J Nucl Med. 2020;61(4):488-495. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC9374044/
6. Zwanenburg A, Vallières M, Abdalah MA, et al. The Image Biomarker Standardisation Initiative: standardized quantitative radiomics for high-throughput image-based phenotyping. Radiology. 2020;295(2):328-338. PMID: 32154773.
7. Gallivanone F, Panzeri MM, Canevari C, Losio C, Gianolli L, De Cobelli F. Radiogenomics, breast cancer diagnosis and characterization. Cancers (Basel). 2022;14(22):5523. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC9611546/
8. Demetriou D, et al. Advancing breast cancer diagnosis with radiogenomics. Cancers (Basel). 2024. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC10930980/
9. Saha A, Harowicz MR, Grimm LJ, et al. A machine learning approach to radiogenomics of breast cancer. Br J Cancer. 2018;119:508-516. PMID: 30033447.
10. Tejani AS, Klontzas ME, Gatti AA, et al. Checklist for Artificial Intelligence in Medical Imaging (CLAIM): 2024 update. Radiol Artif Intell. 2024. Available from: https://pubs.rsna.org/doi/10.1148/ryai.240300
11. Mongan J, Moy L, Kahn CE Jr. Checklist for Artificial Intelligence in Medical Imaging (CLAIM): a guide for authors and reviewers. Radiol Artif Intell. 2020;2(2):e200029.
12. Collins GS, Dhiman P, Ma J, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378. PMID: 38626948.
13. Liu X, Rivera SC, Moher D, Calvert MJ, Denniston AK. Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension. BMJ. 2020;370:m3164. PMID: 32909959.
Image: Ong YH, Zheng W, Khong PL, Ni Q. Application of radiogenomics in head and neck cancer: A new tool toward diagnosis and therapy. iRADIOLOGY. 2024;2(2):113–127. doi:10.1002/ird3.61.
Assessed and Endorsed by the MedReport Medical Review Board
