An interpretable machine learning model based on intratumoral and peritumoral-3 mm ultrasound radiomic features can effectively evaluate the Ki-67 expression level in breast cancer
PENG Qian
HONG Qianyao
LU Jiao
LIU Zhao
ZHUO Xiaoying
WANG Xingtian
Abstract：Objective To investigate the value of an interpretable machine learning model integrating intratumoral and peritumoral ultrasound radiomics with clinical sonographic features in evaluating Ki-67 expression levels in breast cancer,and to analyze the impact of different peritumoral region widths on predictive performance.Methods A retrospective analysis was performed on 185 breast cancer patients who underwent surgical treatment at our institution.Based on postoperative pathology,patients were stratified into high and low Ki-67 expression groups.Tumor regions of interest(ROIs)were delineated,with peritumoral regions automatically expanded outward(1,3,5 mm).Radiomics features were extracted and screened to construct four radiomics models:intratumoral,peritumoral-1mm,peritumoral-3mm,and peritumoral-5mm.The optimal peritumoral model(3 mm)was combined with the intratumoral model to establish the optimal radiomics signature,and a radiomics score was calculated.Clinically significant ultrasound features were selected to develop a clinical model.A nomogram was then constructed by integrating the optimal radiomics model and clinical model.Model performance was evaluated using ROC curves,decision curve analysis,and calibration curves.To enhance interpretability,SHapley Additive exPlanations analysis was employed to assess feature importance.Results The peritumoral-3mm model demonstrated superior predictive performance compared to other peritumoral models.Maximum lesion diameter,presence of microcalcifications,and sonographically abnormal axillary lymph nodes were incorporated into the clinical model.The nomogram combining the optimal radiomics model and clinical model exhibited strong predictive performance in both training(AUC=0.923)and validation(AUC=0.883)cohorts,with favorable clinical utility and calibration.Conclusion The combination of intratumoral and peritumoral-3 mm radiomics features with clinical ultrasound characteristics provides reliable assessment of Ki-67 expression in breast cancer,offering potential clinical decision-making support.
Keywords：breast neoplasmsultrasonographyradiomicsKi-67 antigen
Publication Date：2025-12-20
Online Publishing Date：2026-01-07(First online date of this platform, not the publication date of the document)
Pages：9( 1506-1514 )