Correlation Analysis Between CT And Pathology In Risk Classification Of Gastrointestinal Stromal Tumor

OBJECTIVE: To investigate the correlation between CT features and pathological findings in the risk classification or mitotic figures of gastrointestinal stromal tumors(GISTs)according to the computer tomography(CT)imaging findings and pathological immunohistological features of GISTs.METHODS: The CT imaging findings and clinical data of 112 patients with primary GISTs confirmed by pathology and immunohistochemistry in two hospitals were retrospectively analyzed from June 1,2014 to January 1,2019.Record qualitative data of CT imaging findings of GISTs such as location,shape(regular or irregular),growth pattern(endoluminal,exophytic or mixed),necrosis,calcification,enhancement pattern(homogeneous or heterogeneous),enhanced peak period(arterial phase,venous phase or delayed phase),surrounding lymph nodes(short diameter ≥ 1.0cm),metastasis,and quantitative data such as long diameter(LD),short diameter(SD),LD/SD,long diameter and short diameter corresponding to the maximum angle(BA)and its diagonal(SA),BA/SA,plain scan and multi-phase enhanced absolute CT values(Ap,Vp,Dp);record the pathological and immunohistological data of the lesion such as mitotic figures(number of mitotic figures per 50 high power fields)),Ki-67,CD117,CD34,DOG-1,SAM,S-100,etc.All observations were statistically analyzed using SPSS version 22.0 software:(1)The GISTs risk classification was classified into three groups—very low and low risk,intermediate risk,and high risk.The mitotic figures was classified into three groups—≤5,>5and≤10,and>10/50HPFs(2)When comparing the significance and value of each indicator,the continuous variables were analyzed by one-way ANOVA or Kruskal-Wallis test,and the categorical variables were analyzed by Kruskal-Wallis test or chi-square test.And then screen out statistically significant indicators.(3)Calculate meaningful factors using the ROC curve,which include the area under the ROC curve,the optimal threshold,and the sensitivity and specificity.(4)Regression analysis was used to analyze the meaningful research factors,and to find out the influencing factors and independent influencing factors of the risk grading or mitotic figures of GISTs in CT imaging findings and pathological or immunohistological data.RESULTS:(1)After statistical analysis,differences were significant between risk classification of GISTs and shape,growth pattern,cystic or necrosis,calcification,enhancement pattern,enhanced peak period,plain scan,Ap,Vp,BA,BA/SA,and Ki-67(P<0.05),but grade-regression logistic analysis showed that only BA/SA and Ki-67 had significant effects on the classification of GISTs,with P values of 0.001 and 0.024,respectively.(2)LD and BA or BA/SA were analyzed by bivariate correlation analysis.The Spearman correlation coefficients were 0.414 and 0.416,respectively,with P<0.001,respectively,which suggested that LD was positively correlated with BA or BA/SA,but the correlation was not close(P<0.05).According to the ROC curve analysis,the P values of LD,BA and BA/SA were all less than 0.001,and the corresponding AUC values were 9.37,8.77 and 8.44,respectively.The corresponding optimal threshold values were 5.05 cm(Sen=0.977,Spe= 0.879),89.5°(Sen=0.932,Spe=0.773)and 1.11(Sen=0.909,Spe=0.773).Combining the Kruskal-Wallis test of the two-way comparison between the three groups,the difference among BA or BA/SA in different risk grading of GISTs was significant,while the difference between LD in intermediate risk group and high risk group was not significant(P=0.106).LD,One of the determinants of the risk classification of GISTs,cannot make accurate differential diagnosis of GISTs in intermediate risk grading and high risk grading.(3)There were significant differences in the location,shape,peak period of enhancement,Ki-67,necrosis,LD,BA,BA/SA,Ap between different mitotic figures and lesions.Combined with grade-regression logistic analysis,Ki-67 and Ap can be used as an independent imaging factor.The ROC curve analysis shows that the AUC corresponding to Ap was 0.713(P=0.004),and the optimal threshold was 24.5HU(Sen=0.83,Spe=0.65).Conclusion: The different risk grading of GISTs is determined by a variety of factors.The BA/SA of the lesion has a similar predictive power as the lesion size,especially in the intermediate-risk and high-risk GISTs.Ki-67 and BA/SA are independent influencing factors for predicting the classification of GIST risk.In addition,Ki-67 and Ap are independent influencing factors for predicting mitotic figures.