MRI不同的ROI测量法对卵巢癌表观弥散系数(ADC)值和观察者一致性的影响

目的 通过选取3种不同的感兴趣区域(region of interest,ROI)法测量卵巢癌的最小及平均表观弥散系数(apparent diffusion coefficient,ADC)值,研究对观察者一致性的影响及3种方法的差异性。方法 病理诊断为卵巢癌的45例患者在手术切除前均接受了弥散加权成像磁共振(diffusion weighted imaging-magnetic resonance imaging,DWI-MRI)检查,两个独立的观察者对其进行回顾性分析,分别使用3种不同的ROI选取方法:完全不规则ROI测量法,部分不规则ROI测量法和肿瘤内小ROI取样法。测出ROI的最小和平均ADC值,使用SPSS和MedCalc软件对两个不同观察者所测得的最小和平均ADC值进行比较分析。结果 两个观察者使用肿瘤内小ROI取样法时在最小ADC值时相关系数最大(ICC:0.918 4),而在平均ADC值时相关系数最小(ICC:0.859 5),此法测得的平均ADC值较其他两种方法有显著降低;而部分不规则ROI法在最小(ICC:0.913 8)或平均ADC值(ICC:0.892 5)测定时相关系数及观察者的一致性都很好;完全不规则ROI法在观察者的一致性较差。3种方法在最小ADC值比较时,只有完全和部分不规则ROI法差异无统计学意义(P＝0.12),而其余差异均有统计学意义(P<0.017);3种方法在平均ADC值比较时,两两比较差异均有统计学意义(P<0.001)。结论 ROI测量方法的不同对最小和平均ADC值,观察者一致性的影响有统计学意义,部分不规则ROI法对三者的评估结果均较好,而完全不规则ROI法相对较差。

The effect of different region of interest (ROI) of MRI on apparent diffusion coefficient (ADC) values and interobserver variability in ovarian tumor

Objective To investigate the consistency between observers in the minimium and mean apparent diffusion coefficient (ADC) values measured by 3 kinds of differently shaped regions of interest (ROIs) in ovarian tumors, and to compare the differences of differently shaped ROIs. Methods Forty-five patients recruited with pathology-proven ovarian tumor underwent diffusion weighted imaging-magnetic resonance imaging (DWI-MRI) examinations before the surgical. Two readers measured the minimium and mean ADCs using 3 ROI methods: completely irregular ROI, partial irregular ROI and small solid sample ROI. The interclass correlation coefficient (ICC) and repeated-measures analysis of variance were used to assess their measurement reliability and to compare ADCs for each ROI methods using SPSS and MedCalc software. Results The small solid sample ROI measurement showed the maximum correlation (ICC=0.918 4) in the minimum ADC value of the two readers, while the minimum correlation (ICC=0.859 5) in the mean ADC value. The partial irregular ROI measurement was the best for correlation and interobserver consistency in no matter the minimum ADC (ICC=0.913 8) or the mean ADC values (ICC=0.892 5) of the correlations are the best, while the completely irregular ROI measurement was the worst. In the minimum ADC value, there was no significant difference between the complete and partial irregular ROIs (P=0.12), while the rest showed the significant difference (P<0.017). In the mean ADC value, pairwise comparisons showed significant differences (P<0.001). Conclusions The ROIs had a considerable influence on both the minimum and mean ADCs and the interobserver variation in ovarian cancer. Partial irregular ROI measurement has better evaluation results than the other two methods, and the completely irregular ROI measurement is relatively poor.