Jun Shu1, Yongqiang Tang1, Jingjing Cui2, Ruwu Yang3, Xiaoli Meng3, Zhengting Cai2, Jingsong Zhang1, Wanni Xu4, Didi Wen1, Hong Yin5. 1. Department of Radiology, Xijing Hospital, Fourth Military Medical University, Changle West Road 127#, Xi'an City, 710032, People's Republic of China. 2. Huiying Medical Technology Co., Ltd, Room C103, B2, Dongsheng Science and Technology Park, HaiDian District, Beijing City, 100192, People's Republic of China. 3. Department of Radiology, Xi'an XD Group Hospital, Shaanxi University of Chinese Medicine, FengDeng Road 97#, Xi'an City, 710077, People's Republic of China. 4. Department of Pathology, Xijing Hospital, Fourth Military Medical University, Changle West Road 127#, Xi'an City 710032, People's Republic of China. 5. Department of Radiology, Xijing Hospital, Fourth Military Medical University, Changle West Road 127#, Xi'an City, 710032, People's Republic of China. Electronic address: yinhong@fmmu.edu.cn.
Abstract
OBJECTIVES: To discriminate low grade (Fuhrman I/II) and high grade (Fuhrman III/IV) clear cell renal cell carcinoma (CCRCC) by using CT-based radiomic features. METHODS: 161 and 99 patients diagnosed with low and high grade CCRCCs from January 2011 to May 2018 were enrolled in this study. 1029 radiomic features were extracted from corticomedullary (CMP), and nephrographic phase (NP) CT images of all patients. We used interclass correlation coefficient (ICC) and the least absolute shrinkage and selection operator (LASSO) regression method to select features, then the selected features were constructed three classification models (CMP, NP and with their combination) to discriminate high and low grades CCRCC. These three models were built by logistic regression method using 5-fold cross validation strategy, evaluated with receiver operating characteristics curve (ROC) and compared using DeLong test. RESULTS: We found 11 and 24 CMP and NP features were independently significantly associated with the Fuhrman grades. The model of CMP, NP and Combined model using radiomic feature set showed diagnostic accuracy of 0.719 (AUC [area under the curve], 0.766; 95% CI [confidence interval]: 0.709-0.816; sensitivity, 0.602; specificity, 0.838), 0.738 (AUC, 0.818; 95% CI:0.765-0.838; sensitivity, 0.693; specificity, 0.838), 0.777(AUC, 0.822; 95% CI: 0.769-0.866; sensitivity, 0.677; specificity, 0.839). There were significant differences in AUC between CMP model and Combined model (P = 0.0208), meanwhile, the differences between CMP model and NP model, NP model and Combined model reached no significant (P = 0.0844, 0.7915). CONCLUSIONS: Radiomic features could be used as biomarker for the preoperative evaluation of the CCRCC Fuhrman grades.
OBJECTIVES: To discriminate low grade (Fuhrman I/II) and high grade (Fuhrman III/IV) clear cell renal cell carcinoma (CCRCC) by using CT-based radiomic features. METHODS: 161 and 99 patients diagnosed with low and high grade CCRCCs from January 2011 to May 2018 were enrolled in this study. 1029 radiomic features were extracted from corticomedullary (CMP), and nephrographic phase (NP) CT images of all patients. We used interclass correlation coefficient (ICC) and the least absolute shrinkage and selection operator (LASSO) regression method to select features, then the selected features were constructed three classification models (CMP, NP and with their combination) to discriminate high and low grades CCRCC. These three models were built by logistic regression method using 5-fold cross validation strategy, evaluated with receiver operating characteristics curve (ROC) and compared using DeLong test. RESULTS: We found 11 and 24 CMP and NP features were independently significantly associated with the Fuhrman grades. The model of CMP, NP and Combined model using radiomic feature set showed diagnostic accuracy of 0.719 (AUC [area under the curve], 0.766; 95% CI [confidence interval]: 0.709-0.816; sensitivity, 0.602; specificity, 0.838), 0.738 (AUC, 0.818; 95% CI:0.765-0.838; sensitivity, 0.693; specificity, 0.838), 0.777(AUC, 0.822; 95% CI: 0.769-0.866; sensitivity, 0.677; specificity, 0.839). There were significant differences in AUC between CMP model and Combined model (P = 0.0208), meanwhile, the differences between CMP model and NP model, NP model and Combined model reached no significant (P = 0.0844, 0.7915). CONCLUSIONS: Radiomic features could be used as biomarker for the preoperative evaluation of the CCRCC Fuhrman grades.
Authors: Felix Y Yap; Bino A Varghese; Steven Y Cen; Darryl H Hwang; Xiaomeng Lei; Bhushan Desai; Christopher Lau; Lindsay L Yang; Austin J Fullenkamp; Simin Hajian; Marielena Rivas; Megha Nayyar Gupta; Brian D Quinn; Manju Aron; Mihir M Desai; Monish Aron; Assad A Oberai; Inderbir S Gill; Vinay A Duddalwar Journal: Eur Radiol Date: 2020-08-15 Impact factor: 5.315