| Research BackgroundColorectal cancer(CRC)is a common malignancy of the digestive system.The presence of regional lymph node metastasis in patients with CRC is closely related to TNM stage,treatment options and prognosis.At present,the diagnosis of regional lymph node metastasis in patients with CRC is difficult,the diagnostic sensitivity and accuracy is low.Therefore,to explore more effective methods to improve the diagnostic efficacy of regional lymph node metastasis of colorectal cancer has important clinical value.Texture analysis is an emerging imaging image analysis method in recent years,which can be used to quantitatively describe the heterogeneity of the spatial distribution of radioactive tracer in the lesion.Currently PET image texture analysis in colorectal cancer applications mainly focused on the prediction of tumor gene mutations,with or without distant metastasis,prognosis and treatment effect evaluation.If we can analyze the clinical information and PET texture parameters of patients with CRC to predict the presence or absence of regional lymph node metastasis,the patient can be accurately clinical staging,which has important clinical value.Research purposes1.To explore the relationship between CRC regional lymph node metastasis and the clinical indicators,PET image texture parameters as well as PET/CT image data.2.To establish a regional lymph node metastasis diagnostic model of CRC based on clinical information and texture parameters,and evaluate its predictive performance.Materials and methodsAll patients with CRC who underwent 18 F-FDG PET/CT at PET Center of NanFang Hospital from May 2014 to September 2017 were retrospectively collected and screened for compliance.Clinical data like demographic data,the tumor location,with or without ileus and postoperative pathology were collected.CRC lesions were found in syngo.via image workstation under the guidance of nuclear medicine physician,three-dimensional auto-outline of the ROI on the PET image with SUVmax = 2.5.Then,extract texture parameters from CGITA.Observed lymph nodes in the drainage area of CRC,measure the size and the concentration of radioactivity of the lymph nodes,clustered lymph nodes near the CRC and distant metastasis of CRC patients were observed as well.Chi-square test was used to analyze the difference of demographic data,the location of colorectal cancer tumor,with or without ileus between two groups of patients.Difference of texture parameters and traditional parameters of two groups of patients were measure by sample t test or Mann-Whitney U test.Binary logistic regression model was established to predict regional lymph node metastasis in CRC.We define SUVmax>2.5 and regional lymph node short diameter>1.0 of lymph node as PET diagnostic criteria and CT diagnostic criteria,the two combined as a PET/CT diagnostic criteria.The diagnostic efficacy of each method was evaluated by ROC curve analysis.Heat map analysis of CRC texture parameters using Morpheus.ResultAccording to inclusion and exclusion criteria included 83 cases of regional lymph node metastasis-positive CRC patients,97 cases of negative patients.Univariate analysis showed that the TNM staging,ileus,depth of tumor invasion,presence of regional lymph nodes with SUVmax≥2.5,maximum short axis of regional lymph nodes≥1.0,SUV Variance extracted by gray-level histogram,Large number emphasis extracted by Neighborhood gray level dependence matrix,Coarseness and Mean convergence extracted by texture feature coding method were statistically significant different in two groups of patients(P<0.05).Multivariate analysis showed that ileus,regional lymph node SUVmax,regional lymph node short diameter,texture parameter Coarseness and Mean convergence were independent predictors of regional lymph node metastasis in CRC,then we build a lymph node metastasis diagnostic model.The AUC(95%CI)of the diagnostic model,PET diagnostic criteria,CT diagnostic criteria and PET/CT diagnostic criteria were 0.859(0.805-0.912),0.698(0.620-0.776),0.689(0.610-0.769)and 0.714 0.790),of which the AUC of the prediction model was the highest(all P<0.05).The sensitivities were 78.3%,67.5%,55.4%and 74.7%,respectively.The sensitivity of the diagnostic model was significantly higher than the latter two(x2=6.2 and 17.2,P = 0.022 and<0.001,respectively).But there was no significant difference in sensitivity(x2=1.3,P = 0.453)between diagnostic model and PET/CT diagnostic criteria.The specificity was 76.3%,72.2%,82.2%and 68.0%respectively.There was no significant difference between the diagnostic model and the latter three(x2=0.7,1.5 and 2.5,P=0.523,0.307 and 0.169,respectively).ConclusionThis study established a regional lymph node metastasis prediction model based on 18F-FDG PET/CT and texture parameters.The efficacy of this model used to predict CRC regional lymph node metastasis is higher than CT,PET and PET/CT diagnostic criteria... |
PDF Full Text Request