Preliminary Study On Diagnosis Of Cervical Lymph Node Lesions Dased On Deep Learning

Background and purpose:Cervical lymphadenopathy is a common clinical manifestation of many diseases in the neck and even in the whole body.The differentiation of benign and malignant lymphadenopathy is not only helpful for the qualitative diagnosis of the disease,but also an important basis for the grading and staging of the patients with malignant tumors.At the same time,it is of great value for the judgment of the survival,local recurrence and distant metastasis of the patients with tumors.Deep learning(DL)is one of the methods of artificial intelligence(AI).It is a convolutional neural network with neurons as the basic unit.It is one of the most advanced machine learning methods in the medical field.The model has been applied to the identification and differentiation of pulmonary nodules,the identification and location of lymph node metastasis of lung cancer,the classification of liver cancer,the stratification of risk factors of cardiovascular disease and the establishment of prognosis model of tumor patients,and has achieved high diagnostic efficiency.Therefore,the purpose of this study was to investigate whether the deep learning model could make qualitative diagnosis of benign and malignant lymph nodes in a variety of cervical lymph node lesions and predict the source of cervical metastatic lymph nodes.Part 1:Preliminary study on qualitative diagnosis of cervical lymph nodes based on deep learning modelObjiect:To explore the application value of artificial intelligence deep learning model in the qualitative diagnosis of cervical lymph nodes.Methods:Axial CT images of the neck of 115 patients confirmed by histopathology were collected,including 207 malignant lymph nodes and 359 benign lymph nodes.In the order of consultation time,the first 486(169 malignant,317 benign)were used as the training group,and the last 80(malignant 38,benign 42)as the verification group.The training group used DenseNet network for model training.The trained model was used to test the lymph nodes in the verification group.The accuracy,sensitivity,specificity,positive predictive value(PPV),and negative predictive value(NPV)were calculated based on the confusion matrix.The receiver operating characteristic curve(ROC)was drawn.Results:The accuracy,sensitivity,specificity,PPV and NPV of qualitative diagnosis of lymph nodes in the verification group were 83.8%,76.3%,90.5%,87.9%and 80.9%,respectively,and the area under the curve(AUC)was 0.842.Conclusion:The DenseNet model based on deep learning algorithms can be used for the differential diagnosis of benign and malignant cervical lymph nodes.Its diagnostic efficiency is better than the empirical diagnosis of conventional imaging methods,and it can save time.Part 2:Preliminary study on the origin of cervical metastatic lymph nodes based on deep learning modelsObjiect:To explore whether the artificial intelligence deep learning model can make a preliminary judgment on the origin of cervical metastatic lymph nodes in malignant tumors.Methods:The axial images of cervical enhanced CT of 59 patients confirmed by pathology were collected,including 98 metastatic lymph nodes of head and neck cancer and 110 metastatic lymph nodes of other parts of the body.In the order of consultation time,the first 128(54 metastatic lymph nodes of head and neck,74 metastatic lymph nodes of other parts)were used as the training group,and the last 80(44 metastatic lymph nodes of head and neck,36 metastatic lymph nodes of other parts)were used as the verification group.The training group used DenseNet network for model training.The trained model was used to test the lymph nodes in the verification group.The accuracy,sensitivity,specificity,positive predictive value(PPV),and negative predictive value(NPV)were calculated based on the confusion matrix.The receiver operating characteristic curve(ROC)was drawn.Results:The diagnostic accuracy,sensitivity,specificity,PPV,and NPV of the diagnosis of metastatic lymph node origin in the verification group using deep learning algorithms were 67.5%,59.1%,77.8%,76.5%,and 60.9%,respectively,and the area under the curve(AUC)was 0.747.Conclusion:Preliminary research shows that the DenseNet model based on the deep learning algorithm can make a preliminary prediction of the origin of cervical metastatic lymph nodes to a certain extent,and its diagnostic efficiency needs to be further verified on a larger scale data set.