Fisher判别模型在肾透明细胞癌多层螺旋CT征像及病理学WHO/ISUP分级中的价值

Establishment of a Fisher discriminant model for preoperative grading of renal clear cell carcinoma based on MSCT features and WHO/ISUP classification

目的探讨基于术前多层螺旋CT(MSCT)检查结果及术后病理世界卫生组织/国际泌尿病理学会(WHO/ISUP)分级资料建立肾透明细胞癌(CCRCC)术前病理分级Fisher判别模型的临床应用价值。方法收集2017年5月至2020年5月浙江中医药大学附属第二医院行手术治疗的53例CCRCC患者术前MSCT影像,包括肿瘤位置、是否浸润性生长、肿瘤最长径、边缘是否规则、是否发生液化、坏死及钙化等。将具有统计学意义的CT征像建立术前Fisher判别模型,将术前判别结果与术后WHO/ISUP分级结果进行对比,并检验Fisher判别模型的效能。结果53例CCRCC患者中术后病理低级别组(Ⅰ、Ⅱ)28例,高级别组(Ⅲ、Ⅳ)25例。高级别和低级别组肿瘤在肿瘤最长径、边缘是否规则、是否发生液化、坏死及浸润性生长均有统计学意义(均P<0.05),其中边缘规则或有假包膜、肿瘤长径小的CCRCC倾向于低级别组(WHO/ISUPⅠ~Ⅱ级)。长径越大、浸润性生长、发生液化或者坏死的CCRCC倾向于高级别组(WHO/ISUPⅢ~Ⅳ级)。选定具有统计学意义的4项CT征象:肿瘤长径(X1,临界值为5 cm),浸润性生长(X2),发生液化或坏死(X3),肿瘤边缘规则或有假包膜(X4),判别函数为:Z=0.485X1+0.348X2+0.741X3+0.414X4,Zc=1.24;将判别指标代入判别模型,得出低级别组为31例,高级别组为22例,判别模型将病理结果中5例高级别组判为低级别组,2例低级别组判为高级别组,误判率为13.21%,正确率为86.79%,灵敏度为89.66%,特异度为71.43%。结论Fisher判别方法联合MSCT建立的判别模型对CCRCC术前预测分级具有较高的临床实用价值。

Objective To establish a Fisher discriminant model for preoperative pathological grading of renal clear cell carcinoma(CCRCC)based on preoperative enhanced multislice CT(MSC T)findings and postoperative WHO/ISUP classification.Methods The preoperative enhanced MSCT images of 53 patients with CCRCC treated in the Second Affiliated Hospital of Zhejiang Chinese Medical University from May 2017 to May 2020 were analyzed Fisher discriminant model was established based on MSCT findings,the results were compared with those of WHO/ISUP classification,and its validity was tested.Results Among 53 cases of CCRCC,28 cases were in low grade group(Ⅰ,Ⅱ)and 25 cases in high grade group(Ⅲ,Ⅳ).The longest diameter,regular and clear edge,liquefaction and necrosis,and invasive growth of tumors in high-grade and low-grade groups were statistically significant(all P<0.05).The CCRCC with regular edge(pseudocapsule)and shorter tumor diameter tended to be in low-grade group(WHO/ISUP gradeⅠ~Ⅱ).The CCRCC with longer diameter,invasive growth,liquefaction or necrosis tended to be high-grade group(WHO/ISUP gradeⅢ~Ⅳ).Four CT signs with statistical significance:tumor length(X1)(>5 cm),invasive growth(X2),lique fa c tion and necro s is(X3),regular tumor margin(or pseudocapsule)(X4)were selected as discriminant indexes to establish discriminant equation,and the misjudgment rate was calculated by cross validation method.Fisher’s discriminant function was:Z=0.485 X1+0.348 X2+0.741 X3+0.414 X4,Zc=1.24,by substituting the discriminant index into the discriminant model,31 cases of low-grade group and 22 cases of high-grade group were obtained.In the pathological results,5 cases of high-grade group were classified as low-grade group,2 cases of low-grade group were classified as high-grade group,the mis judgment rate was 13.21%,and the correct rate was 86.79%.The sensitivity was 89.66%and the specificity was 71.43%.Conclusion The discriminant model established by Fisher discriminant method combined with MSCT has high clinical value in predicting the CCRCC grades.