AUTHOR=Luo Zhendong , Li Jing , Liao YuTing , Liu RengYi , Shen Xinping , Chen Weiguo 

TITLE=Radiomics Analysis of Multiparametric MRI for Prediction of Synchronous Lung Metastases in Osteosarcoma

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.802234

DOI=10.3389/fonc.2022.802234

ISSN=2234-943X

ABSTRACT=Purpose: 
To develop and validate predictive model based on multiparameter MRI and clinical features for predicting synchronous lung metastases (SLM) in osteosarcoma.
MATERIALS AND METHODS:
Seventy-eight consecutive patients with osteosarcoma (training dataset, n = 54; validation dataset, n = 24) were enrolled in our study. MRI features were extracted from the T1WI, T2WI, and CE-T1WI sequences of each patient. Least absolute shrinkage and selection operator (LASSO) regression and multifactor logistic regression were performed to select key features and build radiomics models in conjunction with logistic regression (LR) and support vector machine (SVM) classifiers. Eight individual models based on T1WI, T2WI, CE-T1WI, T1WI+T2WI, T1WI+CE-T1WI, T2WI+CE-T1WI, T1WI+T2WI+CE-T1WI, and clinical features, as well as two combined models, were built. Two combined models were constructed. The area under the receiver operating characteristic curve (AUC), sensitivity and specificity were used to evaluate the different models.
RESULTS:
Tumor size was the most important univariate clinical predictor. (1) The AUC values of the LR predictive model based on T1WI, T2WI, CE-T1WI, T1WI+T2WI, T1WI+CE-T1WI, T2WI+CE-T1WI, and T1WI+T2WI+CE-T1WI were 0.686, 0.85, 0.87, 0.879, 0.736, 0.85, and 0.914, respectively. (2) The AUC values of the SVM predictive model based on T1WI, T2WI, CE-T1WI, T1WI+T2WI, T1WI +CE-T1WI, T2WI +CE-T1WI, and T1WI+T2WI+CE-T1WI were 0.629, 0.829, 0.771, 0.879, 0.643, 0.829, and 0.929, respectively. (3) The AUC values of the clinical, combined 1 (clinical and LR-radiomics) and combined 2 (clinical and SVM-radiomics) predictive models were 0.779, 0.957, and 0.943, respectively.
CONCLUSION:
The combined model had good performance in predicting SLM of osteosarcoma, and may be helpful in clinical decision-making.