Clinical-Radiomics Nomograms for Preoperative Differentiation of Sacral Chordoma and Sacral Giant Cell Tumor Based on 3D Computed Tomography and Multiparametric Magnetic Resonance Imaging
Ping Yin1, Sicong Wang2, and Nan Hong1

1Department of Radiology, Peking University People's Hospital, Beijing, China, 2Pharmaceutical Diagnostic team, GE Healthcare, Life Sciences, Beijing, China


The purpose of our study was to build and evaluate clinical-radiomics nomograms for the preoperative differentiation of SC and SGCT. We compared individual features and mixed features based on CT and MRI respectively. And we also added clinical data to compare these models in terms of their performance of distinguishing SC and SGCT. Our results reveal that the model based on DWI features had the highest performance among individual scans. Mixed CT features performed better than individual scan, while mixed MR features achieved a low performance. Radiomics model can perform better when combined with clinical data.


To develop and validate clinical-radiomics nomograms based on 3D computed tomography (CT) and multiparametric magnetic resonance imaging (mpMRI) for preoperative differentiation of sacral chordoma (SC) and sacral giant cell tumor (SGCT).

Materials and Methods

A total of 83 SC and 54 SGCT patients diagnosed through surgical pathology were retrospectively analyzed. We built six models based on CT, CT enhancement (CTE), T1-weighted, T2-weighted, diffusion weighted imaging (DWI), and contrast-enhanced T1-weighted features, two radiomics nomograms based on mixed CT or MRI features, and two clinical-radiomics nomograms combined radiomics mixed features with clinical data. The area under the receiver operating characteristic curve (AUC) and accuracy (ACC) analysis were used to assess the performance of the models.


SC and SGCT presented significant differences in terms of age, sex, and tumor location (tage=9.00, χ2sex=10.86, χ2location=26.20; P<0.01). For individual scan, the radiomics model based on DWI features yielded the highest AUC of 0.889 and ACC of 0.885, followed by CT (AUC=0.857; ACC=0.846) and CTE (AUC=0.833; ACC=0.769). For the combined features, the radiomics model based on mixed CT features exhibited a better AUC of 0.942 and ACC of 0.880, whereas mixed MRI features achieved a lower performance than the individual scan. The clinical-radiomics nomogram based on combined CT features achieved the highest AUC of 0.948 and ACC of 0.920.


The radiomics model based on CT and mpMRI present a certain predictive value in distinguishing SC and SGCT, but combined CT features are more recommended. The performance of radiomics nomograms increases when combined with clinical data.


No acknowledgement found.


No reference found.


Figure 1 The workflow of this study.

The ROC curve of radiomics models. Mixed CT features + clinical data achieved the best performance among the four models based on mixed CT (A-B) and mpMRI (C-D) features.

CT based (A-D) and mpMRI based (E-H) clinical-radiomics nomograms showed radscore was the most important feature for predicting SC and SGCT (the first column). The final total points were calculated by summing the score of each point. Calibration curves of the clinical-radiomics models in training set and validating set (the second and third column). The actual probability was represented on the y axis and the nomogram-predicted probabilities was shown in x axis. A closer distance between the two dotted line indicated a better prediction. Radscore of each patient was shown in the validating set (the fourth column).

DCA for the different models. The y axis measures the net benefit. The dark blue line represents the radiomics model. The light blue line represents the clinical-radiomics model. The gray line represents the assumption that all patients were diagnosed correctly. The black line represents the assumption that no patients were diagnosed correctly. The clinical-radiomics nomograms achieved more clinical utility with almost the whole threshold probabilities than radiomics nomograms based on both CT (A) and mpMRI (B).

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)