Triple-negative and non-triple-negative breast cancer prediction by mammographic radiomics features
10.3760/cma.j.issn.1005-1201.2018.11.006
- VernacularTitle:基于乳腺X线影像组学特征的预测模型在鉴别三阴型与非三阴型乳腺癌中的价值
- Author:
Wenjuan MA
1
;
Yumei ZHAO
;
Yu JI
;
Yujuan HAO
;
Junjun LIU
;
Peifang LIU
Author Information
1. 300060 天津医科大学肿瘤医院乳腺影像诊断科,国家肿瘤临床医学研究中心,天津市"肿瘤防治"重点实验室,天津市恶性肿瘤临床医学研究中心,乳腺癌防治教育部重点实验室
- Keywords:
Breast neoplasms;
Triple-negative breast cancer;
Molecular subtypes;
Radiomics;
Machine learning;
Mammogram
- From:
Chinese Journal of Radiology
2018;52(11):842-846
- CountryChina
- Language:Chinese
-
Abstract:
Objective To develop and validate a radiomics predictive model based on mammogram for preoperative predicting triple-negative breast cancer (TNBC) or non-triple-negative breast cancer (NTNBC). Methods We retrospectively analyzed 459 Chinese women who were diagnosed with invasive breast cancer (confirmed by pathology) during August 2015 to November 2015. Our cohort included 34 TNBC and random selected 102 NTNBC cases. Regions of interest (ROIs) were manually selected from craniocaudal and mediolateral oblique mammograms by radiologists through manual lesion segmentation, and 43 radiomics features were evaluated. Craniocaudal (CC) single-view, mediolateral oblique (MLO) single-view and CC and MLO double-view classification model were constructed respectively. Classification performance was evaluated by the area under receiver operating characteristic curve (AUC), accuracy, sensitivity and specificity. Kruskal-Walls U test and t test were used to compare the radiomics features between TNBC and UTNBC. Results The model that used the combination of both the CC and MLO view images achieved the overall best performance than using either of the two views alone, yielding an AUC of 0.791, accuracy of 0.798, sensitivity of 0.776 and specificity of 0.806 for TNBC comparing with NTNBC. Three features were selected by the model (gray scale span and inverse different moment for CC, roundness for MLO) showed a statistical significance (P<0.05) and AUC>0.6 in the subtype classification. Conclusion This research constructed model based on mammograms classification model can effectively distinguish between TNBC and NTNBC. This model has potential value for breast cancer molecular subtype classification and clinical treatment.
