RT - Journal Article T1 - Non-invasive radiomics nomogram model for determining the low and high-grade glioma base on MRI images JF - Int-J-Radiat-Res YR - 2023 JO - Int-J-Radiat-Res VO - 21 IS - 2 UR - http://ijrr.com/article-1-4740-en.html SP - 275 EP - 280 K1 - Glioma K1 - low-grade glioma (LGG) K1 - high-grade glioma (HGG) K1 - radiomics K1 - nomogram. AB - Background: Glioma is the most common type of tumor in the nervous system. Glioma grading remains challenging despite advancements in diagnostic and treatment systems. Preoperative classification is essential to determining optimal treatment and prognosis for gliomas. This study aimed to use magnetic resonance imaging (MRI) to develop accurate nomogram models for glioma grading. Materials and Methods: Eighty-three patients who had undergone a glioma biopsy from June 2017 to November 2021 were retrospectively collected. Two multiparametric MRIs were acquired: T2-weighted and T1-weighted gadolinium contrast-enhanced of 83 glioma patients from one medical institution. Using the open-source python package PyRadiomics, 107 radiomics features were identified for each sequence MRI. We analyzed the probabilities of low-grade gliomas (LGG) and high-grade gliomas (HGG) using logistic regression and the least absolute shrinkage and selection operator regression (LASSO). We identified seven features affecting LGG and HGG differentiated using the lasso algorithm. Next, logistic regression analysis was performed to build a classification model, and five features were obtained. Nomograms were created to predict the incidence of HGG and LLG. To evaluate the prediction performance of the models, receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC), sensitivity, specificity, and accuracy were calculated. Results: For multivariate logistic regression models, according to the best-selected features based on MRI images and clinical data, five parameters were independent predictors of LGG from HGG (P<0.001). The highest prediction performance in terms of AUC, sensitivity, specificity, and accuracy was 0.97, 89.19%, 91.11%, and 90.24%, respectively. Conclusion: The radiomics nomogram models created from quantitative images and clinical data performed well in differentiating LGG from HGG. LA eng UL http://ijrr.com/article-1-4740-en.html M3 10.52547/ijrr.21.2.14 ER -