M.H. Gluck Equine Research Center, Department of Veterinary Science, University of Kentucky, Lexington, KY, USA.
College of Veterinary Medicine, Lincoln Memorial University, Harrogate, TN, USA.
Vet Parasitol. 2021 Dec;300:109623. doi: 10.1016/j.vetpar.2021.109623. Epub 2021 Nov 22.
Fecal egg counts are essential monitoring tools in veterinary parasite control. In recent years, several groups have developed automated egg counting systems based on image analysis and deep learning algorithms. Work in our laboratory demonstrated that an automated system performed with significantly better precision than traditional egg counting techniques. However, while the counting process is no longer operator dependent, the pre-analytical homogenization steps still are. This study aimed at evaluating the influence of sample homogenization on diagnostic performance on an automated equine strongylid egg counting system. Samples were collected from 12 horses and assigned to three egg count categories (four samples per category): Low (0-500 eggs per gram (EPG)), Moderate (501-1000 EPG), and High (1001-2000 EPG). Within each category, all samples were divided into four portions and each was analyzed with the automated system using the following four homogenizing procedures using a homogenizing device supplied with the system: 1) pressing the plunger five times and pouring directly into the counting chamber, 2) pressing the plunger five times and shaking the bottle prior to pouring, 3) pressing the plunger ten times with direct pouring, and 4) pressing the plunger ten times with shaking the bottle before pouring. There were no differences in precision expressed as coefficient of variation between these four procedures but shaking of the bottle prior to pouring was significantly associated with higher counts (p = 0.0068). These results demonstrate that the homogenization process can affect the diagnostic performance of an automated egg counting system and suggest that more efforts should be invested in standardizing and optimizing homogenization procedures.
粪便虫卵计数是兽医寄生虫控制中重要的监测工具。近年来，已有多个团队基于图像分析和深度学习算法开发了自动化虫卵计数系统。我们实验室的工作表明，自动化系统的精确度明显优于传统的虫卵计数技术。然而，虽然计数过程不再依赖操作人员，但预分析的均化步骤仍然如此。本研究旨在评估样本均化对自动化马属强旋尾虫卵计数系统诊断性能的影响。从 12 匹马中采集样本，并将其分为三个虫卵计数类别（每个类别四个样本）：低（0-500 个卵/克（EPG））、中（501-1000 EPG）和高（1001-2000 EPG）。在每个类别中，所有样本均分为四部分，并使用以下四种匀浆程序（由系统提供的匀浆设备进行）通过自动化系统进行分析：1）按压柱塞五次并直接倒入计数室，2）按压柱塞五次并在倾倒前摇动瓶子，3）直接按压柱塞十次，4）按压柱塞十次并在倾倒前摇动瓶子。这四个程序之间的精密度（表示为变异系数）没有差异，但在倾倒前摇动瓶子与更高的计数显著相关（p = 0.0068）。这些结果表明，匀浆过程会影响自动化虫卵计数系统的诊断性能，并提示应更加努力标准化和优化匀浆程序。
