Texture features of pre-operative contrast enhanced MRI early phase images in predicting complete response for breast cancer after neoadjuvant chemotherapy
10.3760/cma.j.issn.1005?1201.2018.07.007
- VernacularTitle:早期增强MRI纹理特征分析对乳腺癌新辅助化疗后病理完全缓解的判断能力
- Author:
Kun CAO
1
;
Hui LIU
;
Bo ZHAO
;
Yanling LI
;
Yuhong QU
;
Yingshi SUN
Author Information
1. 100142,北京大学肿瘤医院暨北京市肿瘤防治研究所医学影像科恶性肿瘤发病机制及转化研究教育部重点实验室
- Keywords:
Breast neoplasm;
Magnetic Resonance Imaging;
Pathological complete response;
Texture analysis
- From:
Chinese Journal of Radiology
2018;52(7):523-527
- CountryChina
- Language:Chinese
-
Abstract:
Objective To evaluate the ability of texture analysis in early phase of enhanced MRI in predicting pathological complete response(pCR) after neoadjuvant chemotherapy(NAC) for breast cancer. Methods This retrospective study enrolled 64 breast cancers samples from 64 patients that were diagnosed by core-needle biopsy and received NAC before operation in Peking University Cancer Hospital between July and Dec 2015. MRI were conducted after NAC. Regions of interest were drawn to cover the whole enhanced areas on subtraction images of early phase to pre-enhanced phase on MRI, and were sent to an in-house developed texture-analyzing software to achieve parameters including average signal intensity (SIav), mean signal intensity (SIm), signal intensity range(SIr), skewness, kurtosis, energy and entropy. Groups of pCR (no invasive tumor) and non-pCR were separated based on pathology results. Differences of MRI parameters were compared by independent-sample t test (normal distribution) or Mann-Whitney U test (abnormal distribution) and ROC curve were drawn to evaluate the diagnostic abilities. Results Post-operation pathology found 28 pCR and 36 non-pCR. ROIs of 13 samples were not drawn because no residual enhanced areas could be found on subtraction images of post-NAC MRI. For 51 lesions (17 pCR and 34 non-pCR) that still had residual enhancement, tumor volume, SIav, SIr, energy and entropy of pCR group were all significantly lower than that of non-pCR group (P＜0.05). ROC curves were drawn, yielding AUC=0.669 for non-enhancement criterion, and the accuracy, sensitivity and specificity were 70.3%, 39.3% and 94.4%. AUCs for volume, SIav, SIr, Energy and Entropy were 0.870, 0.772, 0.810, 0.883 and 0.881 respectively. Conclusion Texture analysis on early-enhanced phase of breast MRI is able to help to improve the diagnostic ability in predicting complete response on in breast cancer after NAC.
