Pretreatment prediction of cervical lymph node metastasis in salivary gland carcinoma based on 18F-FDG PET/CT radiomics / 中华放射医学与防护杂志
Chinese Journal of Radiological Medicine and Protection
; (12): 361-366, 2022.
Article
Dans Zh
| WPRIM
| ID: wpr-932611
Responsable en Bibliothèque :
WPRO
ABSTRACT
Objective:To explore the value of 18F-FDG PET/CT radiomics in predicting the cervical lymph node metastasis in salivary gland cancer. Methods:Sixty-eight patients with salivary gland carcinoma treated in the Peking University School and Hospital of Stomatology were retrospectively studied. They were randomly divided into training group ( n=40), validation group ( n=14), and test group ( n=14). The primary tumor lesions were semi-automatically delineated on PET images as regions of interest (ROIs) and the radiomic features were extracted from ROIs. After feature selection and dimension reduction, an artificial neural network (ANN) prediction model was constructed. The prediction performance of the model was assessed using receiver operating characteristic (ROC) curves, the area under ROC curves (AUC), accuracy, sensitivity, and specificity. Moreover, the performance of various models was compared using the Delong test. Results:The radiomic model yielded an AUC of 0.88 (95% CI: 0.78-0.95), a sensitivity of 75%, specificity of 92.3%, and accuracy of 88.2%. By contrast, the combined model constructed based on the clinical node status (cN) reported by PET/CT and radiomic features yielded an AUC of 0.97 (95% CI: 0.89-0.99), a sensitivity of 87.5%, specificity of 100%, and accuracy of 97.1%. The Delong test showed that there was a statistically significant difference between the combined model and cN ( Z=2.27, P<0.05), but there was no statistically significant difference between the radiomic model and cN ( P>0.05). Conclusions:The ANN model based on 18F-FDG PET/CT radiomics combined with cN reported by PET/CT can more accurately predict cervical lymph node metastasis in patients with salivary gland carcinoma.
Texte intégral:
1
Indice:
WPRIM
Type d'étude:
Prognostic_studies
langue:
Zh
Texte intégral:
Chinese Journal of Radiological Medicine and Protection
Année:
2022
Type:
Article