直肠癌患者MRI体素内不相干运动成像中感兴趣区选择方法对参数一致性的影响

Intra-voxel incoherent motion parameters of rectal cancer at 3.0 T MRI： effect of region of interest on measurement consistency of parameters

目的探讨直肠癌患者体素内不相干运动（IVIM）成像中采用不同ROI选择方法对参数一致性的影响。方法回顾性分析经纤维结肠镜检查病理证实为直肠腺癌的43例患者资料，患者均行盆腔常规MRI和IVIM检查。2名医师分别在轴面T2WI上肿瘤最大层面采用肿瘤轮廓法和肿瘤中心法勾画ROI。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；肿瘤中心法采用固定小ROI勾画法，在近肿瘤中心区放置大小约40个像素的类圆形ROI。测量IVIM参数值，包括真实扩散系数（D值）、灌注分数（f值）、灌注相关假扩散系数（D^＊值）。其中医师1间隔3个月后重复上述测量。采用配对样本t检验比较医师1前后两次之间、医师1与医师2间采用不同ROI勾画方法所得IVIM参数的差异，并计算组内相关系数（ICC），采用Bland-Altman法对同一医师2次测量的结果以及不同医师间的测量结果进行对比。结果医师1第1次采用肿瘤轮廓法测得的D值、f值、D^＊值分别为（1．08±0．24）×10^-3mm^2／s、（0．16±0．06）和（26．59±19．54）×10^-3mm^2／s，采用肿瘤中心法测得的D值、f值、D^＊值分别为（1．06±0．27）×10^-3mm^2／s、（0．17±0．07）和（30．79±20．85）×10^-3mm^2／s，差异均无统计学意义（t值分别为1．113、-0．259、-1．660，P值分别为0．272、0．797、0．104），ICC值分别为0．863、0．469、0．663。同一医师两次采用肿瘤轮廓法勾画ROI，测得的D值、f值、D^＊值差值的95％一致性区间分别为（-0．012～0．038）×10^-3mm^2／s、（-0．003～0．007）、（-0．923～1．166）×10^-3mm^2／s；采用肿瘤中心法勾画ROI，测得的D值、f值、D^＊值差值的95％一致性区间分别为（-0．024—0．044）×10^-3mm^2／s、（-0．005～0．015）、（-1．670—4．195）×10^-3mm^2／s。Bland．Altman散点图显示肿瘤轮廓法测定的f值分布更集中，提示肿瘤轮廓法在同一观察者重复测量中一致性区间小，准确性好。2名医师采用肿瘤轮廓法勾画ROI，测得的D值、f值、D^＊值差值的95％一致性区间分别为（-0．047～0．009）×10^-3mm^2／s、（-0．015—0．009）、（-7．206—3．190）×10^-3mm^2／s；采用肿瘤轮廓法勾画ROI，测得的D值、f值、D^＊值差值的95％一致性区间分另0为（-0．068～0．048）×10^-3mm^2／s、（-0．005—0．041）、（-17．657～0．779）×10^-3mm^2／s。Bland—Altman散点图显示肿瘤轮廓法测定的f值分布相对更集中，提示肿瘤轮廓法在不同医师的重复测量中一致性区间变异度小，准确性好。结论直肠癌患者采用肿瘤轮廓法与肿瘤中心法勾画选择ROI所得IVIM参数无差异，采用肿瘤轮廓法重复测量一致性评估变异度小，可重复性好。

Objective To determine the measurement consistency of diffusion coefficient D, perfusion fraction f and pseudodiffusion coefficient D＊ in rectal cancers based on different ROIs. Methods Forty-three patients with histologically proven reetal cancers were examined using echo-planar DW-MRI with eight b values （0 to 1 000 s/mm^2）. Intravoxel incoherent motion parameters were measured on intravoxel incoherent motion map that contained the largest tumor cross-section, according to two distinct ROI protocols： freehand outline ROI and semi-automatic tumor center ROI. The two protocols were compared for differences in IVIM parameters and the interclass correlation coefficient （ICC） were also calculated. intra-and inter-observer variability using paired t test and Bland-Ahman plot. Results The IVIM parameters（D, f and D＊） obtained by ROIs for outlined and center analysis were （1.08±0.24）×10^-3mm^2/s, （0.16±0.06）, （26.59±19.54）×10^-3mm^2/s and （1.06±0.27）×10^-3mm^2/s, （0.17±0.07）, （30.79±20.85）×10^-3mm^2/s, respectively. No significant differences were observed between the means of the IVIM parameters （D, f, D＊） calculated by the two methods （t=1. 113,-0.259,-1.660; P=0.272, 0.797,0.104, respectively）,and the relative ICC were 0.863, 0.469, 0.663, respectively. The intra-observer 95% limits of consistency of IVIM parameters were （-0.012-0.038）×10^-3mm^2/s, （-0.003-0.007）, （-0.923-1.166）×10^-3mm^2/s with ROI outline tumor, respectively; （-0.024-0.044）×10^-3mm^2/s, （-0.005-0.015）, （-1.670-4.195）×10^-3mm^2/s with center ROI, respectively. The inter-observer 95% limits of consistency of perfusion parameters were （-0.047-0.009） ×10^-3mm^2/s, （-0.015-0.009）, （-7.206-3.190）×10^-3mm^2/s with ROI outlined tumor, respectively; （-0.068-0.048）×10^-3mm^2/s, （-0.005-0.041）, （- 17.657-0.779）×10^-3mm^2/s with center ROI, respectively. Conclusions There was no statistically significant difference between the outlined ROI and tumor center ROI analysis of rectal cancers＇ IVIM parameters. The tumor analysis by outlined ROI protocol appropriately improves intra- and inter-observer consistency and can provide more reproducible and stable results.