TY - JOUR T1 - <strong>PET/MRI Radiomics Analysis in Predicting the Ascending and Descending Nasopharyngeal Carcinoma</strong> JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 3099 LP - 3099 VL - 63 IS - supplement 2 AU - Jiangtao Liang Y1 - 2022/06/01 UR - http://jnm.snmjournals.org/content/63/supplement_2/3099.abstract N2 - 3099 Introduction: To evaluate the value of PET/MRI radiomics model in predicting the ascending and descending types of nasopharyngeal carcinoma (NPC).Methods: From June 2017 to October 2021, the clinical data and PET/MR imaging data of 327 NPC patients (72 ascending type and 255 descending type) with undifferentiated carcinoma from a medical center were collected. The patients were randomly divided into training group (n=229) and verification group (n=98). We measured the metabolic parameters (SUVmax, MTV, TLG) of the NPC primary focus with GE post-processing workstation, and used AK software to extract the most relevant radiomics features to NPC types, then constructed the corresponding radiomics signature. Multivariable logistic regression analysis was performed with radiomics signature and clinical variables for developing the prediction model. The receiver operating characteristic(ROC) analysis was used to evaluate the prediction model.Results: Eight features were selected by the LASSO from 2600 radiomics features. The established radiomics signature has better prediction efficiency for identifying the ascending and descending NPC. The AUCs were 0.899 (95% CI: 0.847 - 0.996) and 0.886 (95% CI: 0.735 – 1.000) in the training group and the verification group, respectively. On multivariable logistic regression, the radiomics signature and total lesion glycolysis (TLG) were considered to be independent and significant risk factors for NPC type (The ascending type vs. The descending type) of NPC. The combined prediction model showed good discrimination in both training group (AUC=0.954, 95%CI:0.801 to 0.998; sensitivity=0.935, specificity=0.734, positive predictive value=0.998, negative predictive value=0.890) and verification group (AUC=0.968, 95% CI: 0.856 to 0.998, sensitivity=0.925, specificity=1.000, positive predictive value=0.913, negative predictive value=0.900).Conclusions: The radiomics predictive model which integrated with the radiomics signature and metabolic parameters (TLG) can be used as a promising and applicable adjunct approach for predicting the ascending and descending NPC. ER -