Kempi V, Sutton D G
Department of Radiophysics, Sjukhuset, Ostersund, Sweden.
J Nucl Med. 1995 Jan;36(1):147-52.
Seventy patients with established diagnoses of normal, parenchymally insufficient or acutely obstructed kidneys were subjected to gamma camera renography. Deconvolution was then performed using three main techniques subdivided into six variants. Parameters from time-activity curves as well as retention curves were calculated. Logistic regression analysis was performed to assess the ability of renography and deconvolution methods to differentiate between kidney groups.
Discrimination between the groups was achieved by standard renography using six of 17 tested renogram parameters. Based on a set of six curve parameters, the correct classification rates ranged 86%-100%. Five of the six variants of the deconvolution technique used produced similar results. None, however, produced results which were as robust as those from renography. The sixth deconvolution method was consistently worse than the others.
Standard renography was consistently better than any of the deconvolution techniques used in the separation of the kidney groups. Conceptually, the results of a logistic regression analysis of renogram parameters may raise possibilities in the field of computer-aided diagnosis.
对70例已确诊为正常肾脏、实质功能不全或急性梗阻性肾脏的患者进行γ相机肾动态显像。然后使用三种主要技术（细分为六种变体）进行去卷积。计算时间-活性曲线以及滞留曲线的参数。进行逻辑回归分析以评估肾动态显像和去卷积方法区分不同肾脏组别的能力。
通过标准肾动态显像，利用17个测试肾图参数中的6个实现了组间区分。基于一组6个曲线参数，正确分类率在86% - 100%之间。所使用的去卷积技术的六种变体中有五种产生了相似的结果。然而，没有一种产生的结果像肾动态显像那样可靠。第六种去卷积方法始终比其他方法差。
在区分肾脏组方面，标准肾动态显像始终优于所使用的任何去卷积技术。从概念上讲，肾图参数的逻辑回归分析结果可能会为计算机辅助诊断领域带来新的可能性。
