The radiomics model based on dual-sequence MRI can effectively predict the Ki-67 expression level in patients with breast cancer

LIANG Xiaohan; QIAO Jiaye; CHEN Yan; MA Yichuan

doi:10.12122/j.issn.1674-4500.2025.08.10

Abstract：Objective To explore the value of dual-sequence MRI radiomics model for prediction of Ki-67 expression level in breast cancer. Methods A retrospective analysis of MRI and clinical data of 177 patients with breast cancer confirmed by postoperative pathology at the First Affiliated Hospital of Bengbu Medical University from January 2023 to August 2024. They were divided into low expression group and high expression group according to immunohistochemical results, 3D Slicer software was used to manually extract the radiomics features from dynamic enhanced phase 2 and DWI images and screen the best combination features. Univariate and multivariate logistic regression analyses were used to screen the independent risk factors of clinical and imaging data, clinical model, single-sequence radiomics model, dual-sequence radiomics model, and a combined model were established, model performance was assessed using ROC curves, while calibration curves and decision curve analysis were used to evaluate clinical utility. Results Compared with the clinical and single-sequence radiomics models, the diagnostic efficacy of the dual-sequence radiomics model was better, and the AUC values in the training group and the validation group were 0.83 and 0.74, respectively. The diagnostic efficacy of the combined model was further improved, and the AUC values in the training group and the validation group were 0.85 and 0.83, respectively. Calibration and decision curves analysis indicated good agreement and favorable clinical benefit. Conclusion The dual-sequence MRI radiomics model has good diagnostic efficiency in predicting the expression level of Ki-67 in breast cancer, which is better than the single-sequence radiomics model and clinical model, these findings indicate that it is expected to become a non-invasive tool and provide help for clinical individualized treatment decision-making.