Prediction model of neck lymphoma based on gray-scale ultrasonography radiomics
张旭,黄品同
DOI: https://doi.org/10.3760/cma.j.cn131148-20201222-00960
2021-01-01
Abstract:Objective:To evaluate the value of ultrasound imaging in cervical lymphoma by constructing a prediction model of cervical lymphoma based on gray-scale ultrasonography radiomics.Methods:Retrospective analysis of ultrasonic sonographic data of 294 patients with cervical lymphadenopathy from June 2016 to June 2020 in the Affiliated Hangzhou Chest Hospital of Zhejiang University School of Medicine. The image features were extracted by imaging method, and the computer-generated random numbers were assigned to the training queue and the verification queue according to 7∶3 ratio. The minimum absolute contraction and selection algorithm (LASSO) were used to screen the optimal features, and support vector machine (SVM) was used to construct the imaging prediction model. The area under the ROC curve (AUC) was calculated to evaluate the performance of the prediction model.Results:Of the 294 cervical lymph nodes, 95 were lymphoma (32 Hodgkin′s lymphoma, 63 Non-Hodgkin′s lymphoma ), 199 cases were other lymph node diseases(34 metastatic lymph nodes, 67 lymph node tuberculosis, 98 reactive proliferative lymph nodes). A total of 107 features were extracted from gray scale ultrasound images, and after pretreatment and screening, 5 features were significantly associated with lymphoma. AUC, sensitivity, specificity and accuracy in the training and validation cohorts were 0.835, 85.1%, 85.7%, 85.5%, 0.793, 82.1%, 83.1%, 82.8%, respectively. The AUC, sensitivity, specificity and accuracy for a senior ultrasund doctor were 0.714, 65.3%, 79.9%, 75.2% in the idagnosis of cervical lymphoma.The diagnostic efficacy of the imaging prediction model was better than that of senior ultrasound doctor (AUC: 0.793 vs 0.714, Z=0.434), and the difference was statistically significant ( P<0.05). Conclusions:The imaging features based on gray scale ultrasound can provide more information on the diagnosis of lymphoma, and have good efficacy in the diagnosis of cervical lymphoma. Its diagnostic efficacy is better than that of the senior ultrasound attending physician.