【摘要】 目的利用粗糙集、决策树及二元logistic回归法等三种方法根据常规MRI分级诊断胶质瘤,比较三种方法的诊断性能。方法 275例确诊胶质瘤病例(低级别胶质瘤151例,高级别胶质瘤124例),术前常规MRI平扫及增强检查,提取的MRI征象包括病灶数目、形态、边缘、水肿、坏死、占位效应、钙化、出血、T1WI、T2WI及增强特点。粗糙集基于Rosetta软件使用遗传算法进行属性约简并产生诊断规则,决策树使用CRT算法建立胶质瘤分级诊断规则,回归法使用二元logistic回归法建立胶质瘤诊断模型。结果粗糙集、决策树树及回归法的诊断准确性分别为84.4%、83.3%、83.6%;敏感度分别为75%、74.2%、79.8%;特异度分别为92.1%、91.3%、86.8%,三种方法的ROC曲线下面积分别为0.92、0.907和0.902,ROC曲线下面积之间无明显差异性。相比其他两种方法,粗糙集可以得到更多的确定性诊断规则。结论粗糙集具有与其他两种方法一样的诊断性能,却可以得出明确及清晰的诊断规则,具有更好的临床应用价值。更多还原

【Abstract】 Objective:In this study we were to compare rough set with Decision tree and two-binary-logistic-regression for the grade diagnosis of glioma on MR images and to evaluate the effect of Rough set to the others’ diagnostic performance.Methods:MR images of 275 patients with gliomas (151 low-grade gliomas,124 high-grade gliomas) examined before surgery were collected.The features of MRI with gliomas included numbers,shape,margin,edema,necrosis,mass effect,calcification,hemorrhage,intensity of T1WI and T2WI,enhancement style after administration of contrast agent.The attributes of glioma was reduced by genetic algorithm in Rosetta software with Rough set and then the diagnostic rules about the grade diagnosis came from reduced attributes.The diagnostic rules were calculated by CR&T algorithm in decision tree and an equation was constructed to predict the grade of glioma in two-binary-Logistic-Regression.Results:The accuracy,sensitivity and Specificity of Rough set,Decision tree and two binary Logistic Regression were 84.4%,83.3%,83.6%;75%,74.2%,79.8%;92.1%,91.3%,86.8% respectively.The averaged area under the ROC curve of the three methods above were 0.92,0.907,0.902,and there were no significant differences among the AUROCs above.Compared to the other methods,more definitive rules could get from Rough set.Conclusion:Rough set had good diagnosis performance as well as that of decision tree and regression methods.With the explicit and definitive rules,Rough set was more suitable for practice in diagnosis by experts.更多还原