MRI-based radiomic features in histological grading of breast invasive ductal carcinoma
10.16571/j.cnki.1008-8199.2018.09.008
- VernacularTitle: 基于MRI的影像组学特征在鉴别乳腺浸润性导管癌病理分级中的价值
- Author:
Pei-qi WU
1
;
Zai-yi LIU
1
;
Chang-hong LIANG
1
Author Information
1. Department of Radiology, Guangdong General Hospital (Guangdong Academy of Medical Sciences), Guangzhou 510080, Guangdong, China
- Publication Type:Journal Article
- Keywords:
breast invasive ductal carcinoma;
histological grade;
magnetic resonance imaging;
radiomics
- From:
Journal of Medical Postgraduates
2018;31(9):938-942
- CountryChina
- Language:Chinese
-
Abstract:
Objective The histological grade of breast cancer is closely related with the treatment and prognosis of the malignancy, and radiomics plays a valuable role in the identification of its grade. This article aimed to investigate the values of the conventional parameters of breast MRI and breast MRI-based imaging features in the histological grading of breast invasive ductal carcinoma (IDC).Methods This retrospective study included 71 cases of breast cancer treated in our hospital from June 2015 to June 2016. We obtained the traditional quantitative parameters of MRI, including the apparent diffusion coefficient (ADC) and initial enhancement rate (IER), performed manual segmentation of the ADC and DCE maps, extracted the radiomic features and analyzed the differences in the radiomic signatures between low- and high-grade IDC. Using logistic regression analysis, we assessed the values of ADC and IER and the radiomic signatures of the ADC and DCE images in differentiating low-grade from high-grade IDC.Results The values of ADC, B_sum_variance, L_SRE and R_RP were significantly lower (P<0.05) while those of B_uniform, B_GLN, L_energy, R_homogeneity 2 and R_IDN remarkably higher in the high-grade than in the low-grade IDC patients (P<0.05), but no statistically significant difference was observed in the IER value between the two groups (P>0.05). In differentiating high-grade from low-grade IDC, the ADC image-based radiomic signature model achieved a significantly higher AUC (0.858 [0.774-0.924]) than the ADC (0.709 [0.588-0.830]) and DCE model (0.691 [0.565-0.818]), and the former also manifested markedly higher accuracy, specificity, and rates of positive and negative prediction than the latter two.Conclusion ADC- and MRI-based radiomic features play a valuable role in differentiating high-grade from low-grade IDC, particularly the former, which could provide even more clinical information, while IER is of little value in this aspect.
