Predicting the response to neoadjuvant chemotherapy for breast cancer: wavelet transforming radiomics in MRI

BMC Cancer

Jinghui Lu¹  Wenli Cai¹  Maosheng Xu²  Changyu Zhou²  Xiaobo Lai²  Chen Gao²  Jiali Zhou²  Jingjing Zeng³ 

[1] Department of Radiology, Massachusetts General Hospital and Harvard Medical School;Department of Radiology, The First Affiliated Hospital of Zhejiang Chinese Medical University;The Third Affiliated Hospital of Zhejiang Chinese Medical University;

关键词: Radiomics;    Breast cancer;    Neoadjuvant chemotherapy;    Pathological complete response;   

DOI  :  10.1186/s12885-020-6523-2

来源: DOAJ

【 摘 要 】

Abstract Background The purpose of this study was to investigate the value of wavelet-transformed radiomic MRI in predicting the pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) for patients with locally advanced breast cancer (LABC). Methods Fifty-five female patients with LABC who underwent contrast-enhanced MRI (CE-MRI) examination prior to NAC were collected for the retrospective study. According to the pathological assessment after NAC, patient responses to NAC were categorized into pCR and non-pCR. Three groups of radiomic textures were calculated in the segmented lesions, including (1) volumetric textures, (2) peripheral textures, and (3) wavelet-transformed textures. Six models for the prediction of pCR were Model I: group (1), Model II: group (1) + (2), Model III: group (3), Model IV: group (1) + (3), Model V: group (2) + (3), and Model VI: group (1) + (2) + (3). The performance of predicting models was compared using the area under the receiver operating characteristic (ROC) curves (AUC). Results The AUCs of the six models for the prediction of pCR were 0.816 ± 0.033 (Model I), 0.823 ± 0.020 (Model II), 0.888 ± 0.025 (Model III), 0.876 ± 0.015 (Model IV), 0.885 ± 0.030 (Model V), and 0.874 ± 0.019 (Model VI). The performance of four models with wavelet-transformed textures (Models III, IV, V, and VI) was significantly better than those without wavelet-transformed textures (Model I and II). In addition, the inclusion of volumetric textures or peripheral textures or both did not result in any improvements in performance. Conclusions Wavelet-transformed textures outperformed volumetric and/or peripheral textures in the radiomic MRI prediction of pCR to NAC for patients with LABC, which can potentially serve as a surrogate biomarker for the prediction of the response of LABC to NAC.

【 授权许可】

Unknown