Specht H D, Brown P H, Haines J E, McNeill M
Dept. of Clinical Pathology, Oregon Health Sciences University, Portland 97201.
J Nucl Med. 1987 Dec;28(12):1826-30.
To improve ease of use, precision, and reduce interobserver variability of the 67Ga lung index, we developed and tested a computer method that yields similar numeric values and uses previous indexing principles, except that the computer matrix unit is the fractional area assessed. Patients were referred for suspected interstitial lung disease. Fifty-three image studies were available for both manual and computer indexing. Linear regression analysis gave a correlation of 0.884. Decision matrix analysis of 58 different nonimmunosuppressed patients resulted in 93% sensitivity and 76% specificity. Receiver operating characteristic curve analysis showed that the ideal index cutoff was 50. Because there is tighter control over 67Ga uptake intensity and spatial distribution assessment, the computerized 67Ga index appears to perform better than the manual analysis.
为提高镓-67肺指数的易用性、精确性并减少观察者间的变异性，我们开发并测试了一种计算机方法，该方法产生相似的数值并采用先前的索引原则，不同之处在于计算机矩阵单元是评估的分数面积。患者因疑似间质性肺病前来就诊。有53项影像研究可用于手动和计算机索引。线性回归分析得出的相关性为0.884。对58名不同的非免疫抑制患者进行决策矩阵分析，结果显示敏感性为93%，特异性为76%。受试者工作特征曲线分析表明，理想的指数截断值为50。由于对镓-67摄取强度和空间分布评估的控制更为严格，计算机化的镓-67指数似乎比手动分析表现更好。
