Differential Diagnosis Value Of Benign And Malignant Lymph Nodes In Head And Neck Based On Radiomics Model

Background: In clinical diagnosis and treatment activities,radiologists use existing imaging tools and make subjective interpretations from the location,shape and size,density or signal,enhancement mode and other characteristics of lesions according to their own experience,which inevitably has some subjective biases sex.In order to make up for the lack of qualitative diagnosis of head and neck lymph nodes by conventional imaging tools（CT/MRI/PET,etc.）,especially the qualitative diagnosis of small-sized lymph nodes in deep soft tissue spaces such as the posterior pharyngeal wall The smaller it is,the less effective the clinician will be when performing puncture or lymph node dissection.The level of medical technology is accompanied by the continuous progress of science and technology.Traditional imaging tools and technical methods have exposed many shortcomings in the increasingly developing digital precision medical model.Therefore,this study will use a new image analysis technology-radiomics,which aims to extract a large number of radiomics features that cannot be recognized and differentiated by the human eye from CT plain and enhanced images,and further combine lymph node size,morphology,patient age,Gender,CT value and other medical data were screened by multi-factor logistic regression to establish a combined diagnosis and prediction model,and a prediction nomogram was made according to the proportion of each diagnostic factor,and the benign and malignant lymph nodes to be examined were numerically calculated.Through the research of this project,the combination of imaging diagnosis and artificial intelligence can avoid the experiential and subjective bias of radiologists in the interpretation of imaging features,and apply the emerging technology of radiomics in the diagnosis and treatment of benign and malignant head and neck lymph nodes.Objective: To investigate the application value of CT-based radiomics model in the differential diagnosis of benign and malignant lymph nodes in the head and neck regions;Combine a variety of clinical diagnostic factors to jointly establish an imaging diagnostic prediction nomogram,output the diagnostic results in the form of probability,and further explore its clinical application value.Methods: The clinical and CT images of 200 patients with benign or malignant lymph nodes confirmed by pathology or puncture biopsy in our hospital from 2010 to 2019 were retrospectively analyzed.There were 105 patients in the benign group and 95 patients in the malignant group.All patients were randomly divided into training（n=133）and test（n=67）cohort.Mazda software was used to extract the radiomics features of lymph nodes based on CT images,and LASSO method was used to reduce the dimension and establish the radiomics signature.After multivariate logistic regression analysis,radiomics and non-radiomics（Size+Z）diagnostic models were established from relevant variables shown statistical difference between the benign or malignant lymph nodes,including the clinical index,short diameter of（Size）,CT values in enhanced arterial phase（Z:arterial phase Z1,venous phase Z2）and radiomics signature（score）of lymph nodes,and use the calibration curve to observe the fit of the two models.Results:After feature reduction,three radiomics features S（2,-2）Correlat 、 S（0,3）Inv Df Mom 、 S（4,0）Contrast were selected for establishing the radiomics signature.The AUC in the training cohort and the test cohort was 0.884 and 0.749,respectively;and the radiomics diagnostic model showed better discrimination efficacy than that of non-radiomics model in training cohort（AUC: 0.958 vs.0.908,P<0.05）and validation cohort（AUC: 0.847 vs.0.806,P<0.05）,the calibration curve is in good agreement with the ideal curve（P<0.05）.In the established nomogram prediction tool,this study proposed 78 points as the critical value for distinguishing benign and malignant lymph nodes.Conclusion: CT-based radiomics model is valuable in the differential diagnosis of benign and malignant lymph nodes in head and neck regions;The nomogram prediction tool established by combining various clinical diagnostic information has great clinical practical value.