Prediction of deep learning-based radiomic features for neoadjuvant radiochemotherapy response in locally advanced rectal cancer
10.3760/cma.j.cn113030-20191112-00472
- VernacularTitle:基于深度学习的放射影像组学特征预测局部晚期直肠癌新辅助放化疗反应研究
- Author:
Ning LI
1
;
Qi SHARON
;
Lingling FENG
;
Yuan TANG
;
Yexiong LI
;
Ye REN
;
Hui FANG
;
Yu TANG
;
Bo CHEN
;
Ningning LU
;
Hao JING
;
Shunan QI
;
Shulian WANG
;
Yueping LIU
;
Yongwen SONG
;
Jing JIN
Author Information
1. 国家癌症中心/国家肿瘤临床医学研究中心/中国医学科学院北京协和医学院肿瘤医院放疗科 100021
- From:
Chinese Journal of Radiation Oncology
2020;29(6):441-445
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To evaluate the effectiveness of deep learning (DL)-based radiomic features extracted from pre-treatment diffusion-weighted magnetic resonance images (DWI) for predicting neoadjuvant chemoradiation treatment (nCRT) response in patients with locally advanced rectal cancer (LARC).Methods:Forty-three patients receiving nCRT from 2016 to 2017 were included. All patients received DWI before nCRT and total mesorectal excision surgery 6-12 weeks after completion of nCRT. The patient-cohort was split into the responder group ( n=22) and the non-responder group ( n=21) based on the post-nCRT response assessed by postoperative pathology, MRI or colonoscopy. DL-based radiomic features were extracted from the apparent diffusion coefficient map of the DWI using a pre-trained convolution neural network, respectively. Least absolute shrinkage and selection operator-Logistic regression models were constructed using extracted radiomic features for predicting treatment response. The model performance was evaluated with repeated 20 times stratified 4-fold cross-validation using receiver operating characteristic (ROC) curves. Results:The model established with DL-based radiomic features achieved the mean area under the ROC curve of 0.73(SE, 0.58-0.80).Conclusion:DL-based radiomic features extracted from pre-treatment DWI achieve high accuracy for predicting nCRT response in patients with LARC.
