Skrypniuk John V, Bailey Dale, Cosgriff Philip S, Fleming John S, Houston Alex S, Jarritt Peter H, Whalley David R
IPEM Nuclear Medicine Software Working Party, York, UK.
Nucl Med Commun. 2005 Mar;26(3):205-15. doi: 10.1097/00006231-200503000-00005.
To examine the variability of results obtained from computer analysis of left ventricular gated blood pool (LVGBP) images by nuclear medicine centres in the UK.
Twelve data sets of LVGBP images were distributed via commercial software suppliers to nuclear medicine centres in the UK. Two of the data sets were duplicates and three were acquired from the same patient with different total counts in the images. The quality of the images was also variable and two images had poorly defined left ventricular walls. A questionnaire was used to identify the parameters used during the analysis and to give an indication of the number of LVGBP scans per year routinely carried out by each centre as well as report the results obtained from the analysis.
Results were received from 63 nuclear medicine centres using 77 computer systems. The vast majority of participants (57) carried out fewer than 10 scans per month. Only two centres performed more than 30 scans per month. Sixteen centres did not quote a minimum normal value for left ventricular ejection fraction (LVEF) and 36 did not record a maximum value. The remainder recorded between 0.40 and 0.60 for the minimum of normal range and 0.60-0.90 for the maximum of normal range. Analysis of returns showed that LVEF estimates for the data sets were highly variable between centres and computer systems. The overall standard deviation of results compared to the mean for each study was 0.076. Approximately half this variation was due to systematic variation between centres. The overall precision taking into consideration this systematic variation, was 0.055. Lower variability was found between studies with higher overall counts and this was highly significant.
研究英国核医学中心对左心室门控血池（LVGBP）图像进行计算机分析所获结果的变异性。
通过商业软件供应商向英国核医学中心分发了12套LVGBP图像数据集。其中两套数据集是重复的，三套是从同一患者获取的，但图像中的总计数不同。图像质量也各不相同，有两张图像的左心室壁定义不清。通过问卷调查来确定分析过程中使用的参数，并了解每个中心每年常规进行的LVGBP扫描数量，同时报告分析所得结果。
收到了来自63个核医学中心使用77台计算机系统得出的结果。绝大多数参与者（57个）每月进行的扫描少于10次。只有两个中心每月进行的扫描超过30次。16个中心未给出左心室射血分数（LVEF）的正常最小值，36个中心未记录最大值。其余中心记录的正常范围最小值在0.40至0.60之间，最大值在0.60至0.90之间。对返回结果的分析表明，各中心和计算机系统对数据集的LVEF估计值差异很大。与每项研究的平均值相比，结果的总体标准差为0.076。约一半的这种差异是由于各中心之间的系统差异。考虑到这种系统差异后的总体精度为0.055。在总计数较高的研究之间发现变异性较低，且这一差异非常显著。
