Li-Da Chen1, Jin-Yu Liang1, Hui Wu2, Zhu Wang1, Shu-Rong Li3, Wei Li1, Xin-Hua Zhang2, Jian-Hui Chen2, Jin-Ning Ye2, Xin Li4, Xiao-Yan Xie1, Ming-De Lu5, Ming Kuang5, Jian-Bo Xu6, Wei Wang7. 1. Department of Medical Ultrasonics, Institute of Diagnostic and Interventional Ultrasound, The First Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China. 2. Department of Gastrointestinal Surgery, The First Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China. 3. Department of Radiology, The First Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China. 4. GE Healthcare, Shanghai, China. 5. Department of Medical Ultrasonics, Institute of Diagnostic and Interventional Ultrasound, The First Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China; Department of Hepatobiliary Surgery, The First Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China. 6. Department of Gastrointestinal Surgery, The First Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China. Electronic address: xjianb@mail.sysu.edu.cn. 7. Department of Medical Ultrasonics, Institute of Diagnostic and Interventional Ultrasound, The First Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China. Electronic address: wangw73@mail.sysu.edu.cn.
Abstract
AIMS: To establish multiparametric radiomics of rectal tumor for the preoperative prediction of lymph node (LN) metastasis. MATERIALS AND METHODS: This prospective study consisted of 115 consecutive patients with rectal carcinoma between April 2015 and April 2017. The multiparametric radiomics scores were extracted from the endorectal ultrasound (ERUS), computed tomography (CT) and shear-wave elastography (SWE) features of the rectal tumor, LN, and peripheral tissues. The three radiomics scores were generated. Further validation as an independent predictor was performed using multivariate logistic regression together with clinical data, and a nomogram was subsequently developed. The predictive performance of the multiparametric radiomics nomogram was compared with that of conventional radiomics. KEY FINDINGS: All three scores (ERUS, CT, and SWE) were significantly higher in patients with LN metastasis than in patients with negative LN metastasis (all P < 0.05) in both training and validation set. Multivariate analysis indicated that CT and SWE scores were independent risk variables (odds ratio, OR = 6.764 and 5.482, respectively). In validation cohort, the multiparametric radiomics nomogram showed the highest predictive accuracy for LN metastasis, with a concordance index (C-index) of 0.857 compared with the conventional radiomics nomogram (C-index, 0.703, P = 0.100), resulting in a significantly improved net reclassification index (NRI) (P < 0.05) and integrated discriminatory improvement (IDI) (P = 0.002). Decision curve analysis showed that the multiparametric radiomics nomogram had a higher overall net benefit. SIGNIFICANCE: Multiparametric radiomics of rectal cancer, which captures blood supply and stiffness phenotypes, is a useful tool for predicting LN metastasis preoperatively and has marked discrimination accuracy compared to conventional radiomics.
AIMS: To establish multiparametric radiomics of rectal tumor for the preoperative prediction of lymph node (LN) metastasis. MATERIALS AND METHODS: This prospective study consisted of 115 consecutive patients with rectal carcinoma between April 2015 and April 2017. The multiparametric radiomics scores were extracted from the endorectal ultrasound (ERUS), computed tomography (CT) and shear-wave elastography (SWE) features of the rectal tumor, LN, and peripheral tissues. The three radiomics scores were generated. Further validation as an independent predictor was performed using multivariate logistic regression together with clinical data, and a nomogram was subsequently developed. The predictive performance of the multiparametric radiomics nomogram was compared with that of conventional radiomics. KEY FINDINGS: All three scores (ERUS, CT, and SWE) were significantly higher in patients with LN metastasis than in patients with negative LN metastasis (all P < 0.05) in both training and validation set. Multivariate analysis indicated that CT and SWE scores were independent risk variables (odds ratio, OR = 6.764 and 5.482, respectively). In validation cohort, the multiparametric radiomics nomogram showed the highest predictive accuracy for LN metastasis, with a concordance index (C-index) of 0.857 compared with the conventional radiomics nomogram (C-index, 0.703, P = 0.100), resulting in a significantly improved net reclassification index (NRI) (P < 0.05) and integrated discriminatory improvement (IDI) (P = 0.002). Decision curve analysis showed that the multiparametric radiomics nomogram had a higher overall net benefit. SIGNIFICANCE: Multiparametric radiomics of rectal cancer, which captures blood supply and stiffness phenotypes, is a useful tool for predicting LN metastasis preoperatively and has marked discrimination accuracy compared to conventional radiomics.