Radiomics features on ultrasound imaging for the prediction of disease-free survival in triple negative breast cancer: A multi-institutional study
10.3760/cma.j.cn131148-20210104-00014
- VernacularTitle:基于多中心超声影像组学特征预测三阴性乳腺癌无病生存期的应用价值
- Author:
Feihong YU
1
;
Jianxiang WANG
;
Jing DENG
;
Jing HANG
;
Ao LI
;
Chun ZHAO
;
Bin YANG
;
Xinhua YE
Author Information
1. 南京医科大学第一附属医院超声科 210029
- Keywords:
Ultrasonography;
Triple negative breast cancer;
Radiomics;
Prognosis
- From:
Chinese Journal of Ultrasonography
2021;30(6):519-525
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate the effectiveness of radiomics model based on preoperative ultrasound in predicting disease-free survival (DFS) in patients with triple negative breast cancer (TNBC) from multicenter data.Methods:A total of 418 patients with TNBC between July 2012 and December 2016 were consecutively recruited for this study from three different institutions including the First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital of Chinese Medicine and General Hospital of Eastern Theater Command. In the training cohort ( n=271) from institution 1(the First Affiliated Hospital of Nanjing Medical University), least absolute shrinkage and selection operator (LASSO) logistic regression algorithm was employed to select recurrence-related radiomics features and build a signature derived from the grayscale US images. The relationship between the radiomics score (Rad-score) and DFS was evaluated. Univariate and multivariate cox regression were utilized to identify the significant radiomics features and clinical-pathologic variables, which were integrated into a radiomics nomogram. An independent external cohort ( n=147) from the other two institutions was validated for evaluating the calibration and discrimination of the predictive nomogram. Results:Higher Rad-score was an independent risk predictor of worse DFS in two cohorts (both P<0.05). The radiomics model, comprising axillary lymph node status, Ki-67 index and radiomics signature, showed better prognostic performance ( P<0.01) than those of the clinical-pathologic model or tumor node metastasis (TNM) staging system with the concordance index (C-index) of 0.75 (95% CI=0.72-0.78) and 0.73(95% CI=0.71-0.75) in the training and validation cohorts respectively. Furthermore, the calibration curves achieved satisfactory agreement and the decision curves further confirmed the clinical utility of the radiomics nomogram. Conclusions:The US-based radiomics signature is a powerful predictor for the assessment of DFS in patients with TNBC. Moreover, the proposed radiomics model integrating the optimal radiomics signature and clinical-pathologic data could improve personalized DFS estimation.
