Machine learning model for predicting epidermal growth factor receptor expression status in breast cancer using ultrasound radiomics

Q: Why is predicting EGFR expression status in breast cancer clinically important?

EGFR expression influences overall and disease-free survival, and EGFR overexpression is linked to accelerated metastasis and recurrence.

Q: How were radiomic features and predictors selected in this study?

Radiomic features were extracted from grayscale and wavelet-transformed ultrasound images, then filtered by reproducibility (ICC), univariate tests (F-test), and LASSO regression, followed by training seven machine-learning models.

Q: Which model performed best and what were the main predictive features?

The random forest model achieved the best performance, and SHAP interpretation highlighted original_ngtdm_Coarseness, original_ngtdm_Strength, and wavelet.LL_glcm_ClusterProminence as top predictors.

Aug 5, 2026

Upload by Clementine

EGFR is a clinically important target because its expression in breast cancer patients affects both overall and disease-free survival. Current EGFR assessment relies on invasive procedures, which increase discomfort, procedural risk, and testing cost. This study develops a machine learning approach that uses ultrasound radiomics to non-invasively predict EGFR expression status, combining feature reproducibility filtering, LASSO-based selection, and model interpretability with SHAP.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Background and Objectives

Methods

Results

Conclusions

Introduction

Key Predictors and Model Interpretation

TYPE Original Research PUBLISHED 17 October 2025 DOI 10.3389/fonc.2025.1683164 OPEN ACCESS EDITED BY Mohamed Shehata, Midway College, United States REVIEWED BY Deepak Nag Ayyala, Takeda Oncology, United States Dimitris Filos, Aristotle University of Thessaloniki, Greece *CORRESPONDENCE Guorong Lyu [lgr_feus@sina.com](lgr_feus@sina.com)[ ](lgr_feus@sina.com)Shanshan Su [susan@fjmu.edu.cn](susan@fjmu.edu.cn) †These authors have contributed equally to this work RECEIVED 10 August 2025 ACCEPTED 06 October 2025 PUBLISHED 17 October 2025 CITATION Xu Z, Ye J, Zhong H, Chen J, Wang H, Zhang X, Lyu G and Su S (2025) Machine learning model for predicting epidermal growth factor receptor expression status in breast cancer using ultrasound radiomics. Front. Oncol. 15:1683164 . doi: 10.3389/fonc.2025.1683164 COPYRIGHT © 2025 Xu, Ye, Zhong, Chen, Wang, Zhang, Lyu and Su. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Machine learning model for predicting epidermal growth factor receptor expression status in breast cancer using ultrasound radiomics Zhirong Xu 1†, Jiayi Ye 2†, Huohu Zhong 1†, Jiemin Chen 1, Han Wang 1, Xiaoqian Zhang 1, Guorong Lyu 1* and Shanshan Su 1* 1 Department of Ultrasound, The Second Afﬁliated Hospital of Fujian Medical University, Quanzhou, China, 2 Department of Nuclear Medicine, The Second Afﬁliated Hospital of Fujian Medical University, Quanzhou, China Background/objectives: The epidermal growth factor receptor (EGFR) is a clinically important target, as its expression in patients with breast cancer inﬂuences both overall and disease-free survival. Current methods for assessing EGFR expression status in a patient are invasive. Therefore, in this study, we developed a machine learning-based approach utilizing ultrasound radiomics to non-invasively predict EGFR expression status in patients with breast cancer. Methods: Radiomic features were extracted from grayscale and wavelettransformed ultrasound images of 321 patients. The dataset was randomly split into training (n = 225) and test (n = 96) sets at a 7:3 ratio with stratiﬁed sampling topreserve the EGFR+/– ratio. Key predictors were identiﬁed using a multi-step procedure—including reproducibility ﬁltering (ICC > 0 . 75), univariate F-test ﬁltering (p < 0 . 05), and L1-regularized selection via LASSO regression. Seven machine-learning models were trained. Model interpretability was assessed using SHAP (Shapley Additive Explanations) . In addition to the hold-out evaluation, we performed stratiﬁed 10-fold cross-validation to reduce selection bias. Results: The random forest model demonstrated the optimal performance, with an area under the receiver operating characteristic curve of 0.86 in the training set and 0 .70 in the test set. It signiﬁcantly outperformed the other models (P < 0. 001) . The Shapley additive explanation method was used to interpret the model, revealing that original_ngtdm_Coarseness, original_ngtdm_Strength, and wavelet. LL_glcm_ClusterProminence were the top predictors. These Frontiers in Oncology 01 [frontiersin.org](frontiersin.org) features reﬂect structural compactness and heterogeneity associated with EGFR overexpression. Conclusions: We present a reliable and interpretable tool for non-invasively assessing EGFR expression status in patients with breast cancer. The most important predictors captured tumor heterogeneity and microstructural uniformity, highlighting the biological relevance of radiomic patterns in EGFRpositive tumors. This model integrates advanced imaging analyses

FAQs

Why is predicting EGFR expression status in breast cancer clinically important?

EGFR expression influences overall and disease-free survival, and EGFR overexpression is linked to accelerated metastasis and recurrence.

How were radiomic features and predictors selected in this study?

Radiomic features were extracted from grayscale and wavelet-transformed ultrasound images, then filtered by reproducibility (ICC), univariate tests (F-test), and LASSO regression, followed by training seven machine-learning models.

Which model performed best and what were the main predictive features?

The random forest model achieved the best performance, and SHAP interpretation highlighted original_ngtdm_Coarseness, original_ngtdm_Strength, and wavelet.LL_glcm_ClusterProminence as top predictors.

Aug 5, 2026

Upload by Clementine