基于PACS测量实体瘤最长径的变异性分析

Variability in longest diameter measurement of solid tumors in a PACS environment

目的以非小细胞肺癌(NSCLC)的肺内靶病灶为例对当前临床实践中基于影像存档与通信系统(PACS)测量实体瘤最长径的变异性进行分析。方法收集2017年9月至2018年2月78个NSCLC肺内靶病灶。两名放射科医师通过PACS工作站分别独立测量肿瘤最长径,其中1名放射科医师间隔4周后重复测量1次。两名肿瘤科医师共同对测量图像进行复查和比较,记录测量结果并判定不同次测量之间是否选择了明显不一致的测量径线。应用Bland-Altman法计算观察者内及观察者间变异性,应用组内相关系数(ICC)评价相应两组测量结果的一致性。结果分别对肺窗和纵隔窗的病灶最长径进行3次测量,均值的差异均无统计学意义(P>0. 05)。对同一病灶测量时,不同观察者之间或同一观察者不同时间选择的最长径可能不同,不同观察者选择测量径线的变异性高于同一观察者。当最长径选择相同时,Bland-Altman散点图显示,肺窗测量的观察者内变异性为-7. 1%至7. 4%,观察者间变异性为-7. 4%至6. 8%;纵隔窗测量的观察者内变异性为-7. 6%至5. 8%,观察者间变异性为-5. 3%至7. 1%。测量绝对值的变异在±3 mm之间。无论是肺窗还是纵隔窗,每两组测量值的ICC均在0. 9至1. 0之间。结论由同一观察者对NSCLC肺内病灶最长径进行测量时,选择测量径线的变异性较低,肺窗最长径变化在-7. 1%至7. 4%之间,纵隔窗最长径变化在-7. 6%至5. 8%之间,同时测量值绝对值差异大于3 mm时可以认为是肿瘤的客观变化,而不是测量误差。

Objective To detect variability in longest diameter measurement of solid tumors in a picture archiving and communication systems(PACS)environment,and pulmonary lesions of non-small cell lung cancer(NSCLC)was taken as a paradigm.Methods From September2017to February2018,78pulmonary lesions from NSCLC patients chest images were pre-selected.Two radiologists measured the target lesions independently at PACS workstation.Repeat measurement was performed4weeks later by one of the radiologists.The measured images were reviewed in pairs whether an approximately same longest diameter was selected for the same lesion by two oncologists in a side-by-side manner.Bland-Altman methods were used to assess intra-and interobserver variability.Intraclass correlation coefficients were used to assess intra-and inter-observer agreement.Results There was no statistical difference in the mean of the longest diameter on lung window or mediastinal window measurements.Longest diameter selection for a same lesion may be different between different observers or the same observer in defferent times.Interobserver variability was higher than intraobserver in measurement diameter selection.Bland-Altman methods showed that intra-and inter-oberver variability were-7.1%to7.4%,-7.4%to6.8%for lung window,and-7.6%to5.8%,-5.3%to7.1%for mediastinal window respectively.In addition to the relative variability,the absolute difference variability was close to a cut-off value of3mm.The intraclass correlation coefficients for intra-and inter-observer agreement rates were between0.9and1.0for lung window or mediastinal window.Conclusion Variability can be reduced if the same reader performs the diameter selection for NSCLC pulmonary lesions.Longest diameter changes of-7.1%to7.4%in lung window and changes of-7.6%to5.8%in mediastinal window with an absolute difference of at least3mm at CT can be considered true changes rather than measurement variation.