Machine learning-based ultrasound radiomics for predicting risk of recurrence in breast cancer - Original Research

Q: What is the document’s primary goal?

To develop an ultrasound radiomics-based machine learning model that predicts the risk of recurrence in breast cancer patients.

Q: How were patients categorized for recurrence risk?

Patients were classified into low-medium-risk and high-risk groups according to St. Gallen recurrence risk criteria.

Q: Which features and algorithms were used to build the models?

Radiomics features were extracted using Pyradiomics, selected with mRMR and LASSO, and then used to construct nomogram models trained with eight machine learning algorithms.

Aug 5, 2026

Upload by River Wang

A retrospective study develops and validates a machine learning ultrasound radiomics nomogram to predict recurrence risk in breast cancer. Data from 420 pathologically confirmed patients were split into training and test sets, with an independent external validation cohort of 90 patients, and risk group assignment based on St. Gallen criteria. Radiomics features extracted via Pyradiomics were selected using mRMR and LASSO, then modeled with eight machine-learning algorithms. Three nomogram variants (Clin-US, Clin-Rad, Clin-US-Rad) were assessed using ROC, calibration, and decision curve analysis. Results identified 12 informative features and showed strong discriminative and clinical utility performance, supporting a non-invasive approach for guiding personalized care.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Purpose

Methods

Study design and cohorts

Feature extraction and model building

Nomogram models and evaluation

Results

Conclusion

TYPE Original Research PUBLISHED 12 May 2025 DOI 10.3389/fonc.2025.1542643 OPEN ACCESS EDITED BY Domenico Pomarico, University of Bari Aldo Moro, Italy REVIEWED BY Yan Zheng, The First Afﬁliated Hospital of Soochow University, China Yu Du, Shanghai General Hospital, China Alessandro Fania, University of Bari Aldo Moro, Italy *CORRESPONDENCE Lei Zhang [zhanglei6@hrbmu.edu.cn](zhanglei6@hrbmu.edu.cn)[ ](zhanglei6@hrbmu.edu.cn)Jiawei Tian [jwtian2004@163.com](jwtian2004@163.com)[ ](jwtian2004@163.com)Xiuhua Yang [yangxiuhua@hrbmu.edu.cn](yangxiuhua@hrbmu.edu.cn) †These authors have contributed equally to this work and share ﬁrst authorship RECEIVED 10 December 2024 ACCEPTED 17 April 2025 PUBLISHED 12 May 2025 CITATION Fan W, Cui H, Liu X, Zhang X, Fang X, Wang J, Qin Z, Yang X, Tian J and Zhang L (2025) Machine learning-based ultrasound radiomics for predicting risk of recurrence in breast cancer. Front. Oncol. 15:1542643 . doi: 10.3389/fonc.2025.1542643 COPYRIGHT © 2025 Fan, Cui, Liu, Zhang, Fang, Wang, Qin, Yang, Tian and Zhang. 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-based ultrasound radiomics for predicting risk of recurrence in breast cancer Wei Fan 1†, Hao Cui 2†, Xiaoxue Liu 2, Xudong Zhang 1, Xinran Fang 2, Junjia Wang 2, Zihao Qin 2, Xiuhua Yang 1*, Jiawei Tian 2* and Lei Zhang 1* 1 Department of Ultrasound Medicine, the First Afﬁliated Hospital of Harbin Medical University, Heilongjiang, China, 2 Department of Ultrasound Medicine, the Second Afﬁliated Hospital of Harbin Medical University, Heilongjiang, China Purpose: To develop a radiomics model based on ultrasound images for predicting risk of recurrence in breast cancer patients. Methods: In this retrospective study, 420 patients with pathologically conﬁrmed breast cancer were included, randomly divided into training (70%) and test (30%) sets, with an independent external validation cohort of 90 patients. According to St. Gallen recurrence risk criteria, patients were categorized into two groups, low-medium-risk and high-risk. Radiomics features were extracted from aradiomics analysis set using Pyradiomics. The informative radiomics features were screened using the minimum redundancy maximum relevance (mRMR) and the least absolute shrinkage and selection operator (LASSO) algorithms. Subsequently, radiomics models were constructed with eight machine learning algorithms. Three distinct nomogram models were created using the features selected through multivariate logistic regression, including the Clinic-Ultrasound (Clin-US), Clinic-Radiomics (Clin-Rad), and Clinic-Ultrasound-Radiomics (ClinUS-Rad) models. The receiver operating characteristic (ROC), calibration, and decision curve analysis (DCA) curves were used to evaluate the model ’s clinical applicability and predictive performance. Results: A total of 12 ultrasound radiomics features were screened, of which wavelet. LHL ﬁrst order Mean features weighed more and tended to have a high risk of recurrence. The higher the risk of recurrence, the higher the radiomicsscore (Rad-score) in all three sets (training, test, and external validation set, all p < 0. 05) . Rad-score is equally applicable in four different subtypes of breast cancer. In the test set and external validation set, the Clin-US-Rad model achieved the highest AUC values (AUC = 0 . 817 and 0 . 851, respectively) . The calibration and DCA curves also demonstrated the good clinical utility of the combined model. Frontiers in Oncology 01 [frontiersin.org](frontiersin.

FAQs

What is the document’s primary goal?

To develop an ultrasound radiomics-based machine learning model that predicts the risk of recurrence in breast cancer patients.

How were patients categorized for recurrence risk?

Patients were classified into low-medium-risk and high-risk groups according to St. Gallen recurrence risk criteria.

Which features and algorithms were used to build the models?

Radiomics features were extracted using Pyradiomics, selected with mRMR and LASSO, and then used to construct nomogram models trained with eight machine learning algorithms.

Aug 5, 2026

Upload by River Wang