直肠癌术前扩散加权成像层面及感兴趣区的选择对ADC值一致性及重复性的影响

背景与目的：扩散加权成像(diffusion weighted imaging，DWI)定量参数即表观扩散系数(apparent diffusion coefficient，ADC)的一致性及可重复性是众多研究关注的重点。既往研究结果显示b值、感兴趣区(region of interest，ROI)的选择等因素对ADC值一致性及可重复性均有一定影响。本研究将探讨直肠癌术前DWI层面及ROI的选择对ADC值一致性及重复性的影响。方法：回顾性分析80例病理证实为直肠腺癌的患者资料，此组患者均行直肠MR及DWI检查。2名不同年资的医师分别在ADC图像上采用肿瘤中心法及轮廓法勾画ROI，层面选择了肿瘤最大层、肿瘤最大层及其上下两层及肿瘤全层三种方法。其中肿瘤轮廓法采用手动勾画模式，尽量沿肿瘤轮廓勾画ROI。肿瘤中心法使用圆形ROI工具，肿瘤实性部分尽可能圈入而避开病变内坏死或囊变区。两名医师用三种层面和两种ROI放置法间隔2周进行重复测量。使用ANOVA法比较不同层面之间ADC值差异，配对t检验比较两种ROI之间ADC值差异。计算组内相关系数(interclass correlation coefficient，ICC)，并分别比较同一医师前后两次及两名医师之间测量值的可重复性(Bland-Altman法)。结果：将三种层面法测得ADC值进行两两比较，差异均无统计学意义(P>0.05)。相同层面的情况下，肿瘤轮廓法测得ADC值均大于肿瘤中心法，且差异有统计学意义(P<0.05)。选择肿瘤全层-肿瘤中心法及轮廓法测量ADC值，组内及组间一致性可达到0.931、0.803和0.913、0.822。医师1前后两次分别采用肿瘤中心法及肿瘤轮廓法测定ADC值，测得的ADC值95%一致性区间为(-0.089 3~0.083 1)×10^-3mm²/s(P:Mean=0,P>0.05)，(-0.066 8~0.096 3)×10^-3 mm²/s(P:Mean=0，P<0.05)；医师1与医师2使用采用肿瘤中心法及肿瘤轮廓法测定ADC值，测得的ADC值95%一致性区间为(-0.127 5~0.141 6)×10^-3mm²/s(P:Mean=0，P>0.05)，(-0.112 0~0.150 2)×10^-3mm²/s(P:Mean=0，P<0.05)。Bland-Altman散点图显示，肿瘤中心法较轮廓法在同一医师前后两次间及两名医师间重复测量一致性区间小，重复性好。结论：不同层面选择对直肠癌ADC值影响不大；而轮廓法测得ADC值大于中心法。肿瘤全层-肿瘤轮廓法及中心法测得ADC值一致性均较高。但肿瘤全层-中心法重复性稍好且操作简便；而肿瘤全层-轮廓法重复性较中心法稍差，但能反映肿瘤整体情况。

ADC value of rectal cancer at DWI: effect of slice protocol and region of interest on measurement consistency of parameters

Background and purpose: The apparent diffusion coefficient (ADC) values are the quantitative parameters of diffusion weighted imaging (DWI) which are influenced by many factors, such as b values and region of interest (ROI), and have become the concerns of research. Our study aimed to determine the measurement consistencyof ADC in rectal cancer based on different ROIs. Methods: Eighty patients with histologically proven rectal cancer were examined using echo-planar DW MRI values (b value=800 s/mm2). ADC values were measured on ADC map. Mean tumor ADCs were measured according to three distinct slice protocols: ‘whole-slices’,‘single-slice of largest tumor’ and ‘tumor parts of largest independent tumor-containing slices and two adjacent slices’. The freehand ROI was drawn along the border of the high signal of the tumor on the b=800 images to cover the entire tumor area of each slice. Another ROI was drawn using circle tool to cover the entire tumor solid parts as large as possible. After 2 weeks of repeating the measurement between the two physicians. Differences in ADC values were measured using one-way ANOVA. The two protocols were compared for differences in ADC values using paired t-test. We calculated intra- and inter-observer variability using intraclass correlation coefficient (ICC) and Bland-Altman plot. Results: No significant difference was observed by ROIs for three distinct slice protocols (P>0.05). The ADC values obtained by ROIs for the outlined were higher than the ADC values obtained by ROIs for the center analysis (P<0.05). Intraand inter-observer ICC were 0.931, 0.803; 0.913, 0.822 by whole-slices-center and outline ROI respectively. The intraobserver 95% limits of consistency of ADC values were (-0.089 3 to 0.083 1)×10^-3mm²/ s (P: Mean=0, P>0.05), (-0.066 8 to 0.096 3)×10^-3mm²/s (P<0.05) with center and outline ROI, respectively. The inter-observer 95% limits of consistency of perfusion parameters were (-0.127 5 to 0.141 6)×10^-3mm²/s (P>0.05), (-0.112 0 to 0.150 2)×10^-3mm²/s (P<0.05) with center and outline ROI, respectively. Bland-Altman plot showed that whole-slices-center ROI could provide more reproducible and stable results. Conclusion: There was no statistically significant difference among three distinct slice protocols analysis of ADC values in rectal cancer. ADC values by whole-slices-center ROI protocol appropriately improved intra- and inter-observer consistency and could provide more reproducible and stable results with less time. ADC values by whole-slices-outlined ROI protocol appropriately indicated the overall situation, but needed more time.