Can clinical radiomics nomogram based on 3D multiparametric MRI features and clinical characteristics estimate early recurrence of pelvic chondrosarcoma?

Background Chondrosarcoma (CS) is the second most common primary malignant bone tumor, with a relatively high recurrence rate. However, an effective method that estimates whether pelvic CS will recur after surgery, which influences the formulation of a clinical treatment plan, remains lacking. Purpose To develop and validate a clinical radiomics nomograms based on 3D multiparametric magnetic resonance imaging (mpMRI) features and clinical characteristics that could estimate early recurrence (ER) (<= 1 year) of pelvic CS. Study Type Retrospective. Population In all, 103 patients (ER = 41, non-ER = 62) with histologically proven CS were retrospectively analyzed and divided into a training set (n = 72) and a validation set (n = 31). Field Strength/Sequence 3.0T axial T-1-weighted (T-1-w), T-2-weighted (T-2-w), diffusion weighted imaging (DWI), contrast-enhanced T-1-weighted (CET1-w). Assessment Risk factors (sex, age, type, grade, resection margins, etc.) associated with ER were evaluated. Five individual models based on T-1-w, T-2-w, DWI, CET1-w, and clinical data were built. Then we compared the performance of models based on T-1-w, T-2-w, CET1-w and their combination. Lastly, two nomograms based on the best model + clinical data and DWI + clinical data were built. Statistical Tests The area under the receiver operating characteristic curve (AUC) and accuracy (ACC) were used to evaluate different models. Results Grade was the most important univariate clinical predictor of ER of pelvic CS patients (odds ratio [OR]1 = 4.616, OR2 = 8.939, P < 0.05). T-1-w + T-2-w + CET1-w had a significantly higher performance than CET1-w in the training set (P = 0.01). Radiomics features are more important than clinical characteristics in clinical radiomics nomograms, especially for multisequence combined features (OR = 3.208, P < 0.01). Clinical radiomics nomogram based on combined features (T-1-w + T-2-w + CET1-w) + clinical data achieved an AUC of 0.891 and ACC of 0.857, followed by DWI + clinical data (AUC = 0.882, ACC = 0.760) in the validation set. Data Conclusion The clinical radiomics nomogram had good performance in estimating ER of pelvic CS patients, which would be helpful in clinical decision-making. Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2020;51:435-445.