18F-FDG PET/CT Texture Analysis In Diagnosis And Prognostic Evaluation Of Renal Cancer And Renal Lymphoma

Purpose:The purpose of this study is to retrospectively analyze the value of ¹⁸F-FDG PET/CT texture analysis parameters,conventional PET/CT parameters,and clinical correlative factors for the identification of renal cancer and renal lymphoma and their impact on the overall survival（OS）of renal cancer.Method:Patients who had underwent ¹⁸F-FDG PET/CT examination at the Second Affiliated Hospital of Zhejiang University School of Medicine from January 27,2014 to December 23,2019,and had been diagnosed with renal cancer or renal lymphoma through pathological examination were enrolled as the subjects.A total of 40 cases were effectively selected,including 22 cases of renal cancer and 18 cases of renal lymphoma.PET texture analysis parameters,CT texture parameteras and conventional PET/CT parameters were extracted from the images using the Life X software package（http://www.lifexsoft.org）.There are 40 PET texture analysis parameters,including 5histogram（HISTO）parameters,3 shape（SHAPE）parameters,7 gray level co-occurrence matrix（GLCM）parameters,11 gray level running length matrix（GLRLM）parameters,3 neighborhood gray level difference matrix（NGLDM）parameters,and 11 gray area length matrix（GLZLM）parameters.There are 40 CT texture parameters mentioned above.There are 9 conventional PET parameters,including the minimum,average,and maximum standardized uptake values（SUVmin,SUVmean,SUVmax）in the target volume,the ratio of the average and the maximum standardized uptake of tumor and renal cortex,（tumor SUVmean/renal cortex SUVmean,tumor SUVmax/renal cortex SUVmax）,the ratio of the average and the maximum of tumor and liver blood pool,（tumor SUVmean/liver blood pool SUVmean,tumor SUVmax/liver blood pool SUVmax）,Tumor Metabolic Active Volume（MTV）,Total Lesion Glycolysis（TLG）,Peak SUV（SUVpeak）.There are 3 CT conventional parameters,including the minimum,average,maximum density（HUmin,HUmean,HUmax）in Hounsfield unit（HU）.The receiver operating characteristic curve（ROC curve）was used to select the most distinguishable parameters in each category.Binary logistic regression analysis was used to convert the most distinguishable conventional PET parameters,PET texture parameters,combined PET parameters,CT combined parameters,PET/CT texture parameters and PET/CT combined parameters into six models.Whether the six models can effectively identify two kinds of kidney tumors is identified by ROC curve analysis.Furthermore,the difference of AUC values between the six models were analyzed by Medcalc software.In survival analysis,the above parameters were divided into high-value and low-value groups by the median of the values.Log-rank test was used for Kaplan-Meier univariate survival analysis.Statistical significance was defined as P<0.05.The significant prognostic parameters from the above univariate analysis results were included in the Cox risk regression model for multivariate analysis.Results:SUVmax,SUVpeak of conventional PET images,13 parameters of PET texture analysis:HISTO＿Entropy＿log10,HISTO＿Entropy＿log2,SHAPE＿Sphericity,GLCM＿Contrast,GLCM＿Entropy＿log10,GLCM＿Entropy＿log2,GLCM＿Dissimilarity,GLRLM＿SRE,GLRLM＿HGRE,NGLDM＿Contrast,GLZLM＿HGZE,GLZLM＿ZP,GLZLM＿ZLNU),1 parameters of CT conventional analysis:HUmin,and 1 parameters of CT texture analysis:HISTO＿Skewness can effectively differentiate between renal cancer and renal lymphoma（AUC:0.689-0.760,P<0.05,respectively）.Binary logistic regression analysis showed that six models of conventional PET parameters,PET texture parameters,combined PET parameters,CT combined parameters,PET/CT texture parameters and PET/CT combined parameters could effectively identify two kinds of kidney tumors（AUC:0.704,0.801,0.843,0.788,0.905,0.902 P=0.028,0.001,<0.001,<0.001,<0.001,<0.001）.The results of Medcalc software analysis showed that only the AUC value difference between the conventional PET parameter model and PET/CT texture parameter model,the conventional PET parameter model and PET/CT combined parameter model were statistically significant（P=0.007,0.008）,and the P values between the remaining models are greater than 0.05.Univariate survival analysis of patients with renal cancer showed that the pathological subtype of was non-clear cell carcinoma（P=0.007）,no surgical treatment（P=0.002）,CONVENTIONAL＿TLG≥206.0（P=0.031）,PET texture parameters HISTO＿Kurtosis≥3.0（P=0.047）,GLRLM＿RLNU≥220.0（P=0.029）,GLZLM＿ZLNU≥3.2（P=0.021）,CT texture parameters GLCM＿Contrast≥125.0（P=0.019）or GLZLM＿LZHGE≥2970.0（P=0.029）were all adverse factors affecting the prognosis of renal cancer（P<0.05,respectively）.In the univariate analysis for renal lymphoma patients,except for the correlation between serum creatinine value and prognosis（P=0.029）,all clinical factors,conventional PET/CT parameters and texture parameters were not factors affecting the prognosis of renal lymphoma patients（P>0.05 respectively）.Multifactorial survival analysis of renal cancer revealed:the overall model consisting of pathological subtype,whether to undergo surgery,TLG,PET texture parameters HISTO＿Kurtosis,GLRLM＿RLNU,GLZLM＿ZLNU,CT texture parameters GLCM＿Contrast and GLZLM＿LZHGE,P=0.045,suggesting that the multifactorial model has an impact on the patients’prognosis,however,but there was no independent predictors in outcome among each of these parameters（P>0.05,respectively）.Conclusion:（1）Both PET/CT texture analysis and conventional PET/CT can effectively differentiate between renal cancer and renal lymphoma and PET/CT texture analysis has more advantages;（2）PET texture analysis parameters HISTO＿Kurtosis,GLRLM＿RLNU,GLZLM＿ZLNU,CT texture analysis parameters GLCM＿Contrast,GLZLM＿LZHGE,the conventional parameter TLG,the clinicopathological subtype of non-clear cell carcinoma,or no surgical treatment are all adverse factors affecting the prognosis of renal cancer.