Eilertsen H, Henriksson C E, Hagve T-A
Department of Multidisciplinary Laboratory Medicine and Medical Biochemistry, Akershus University Hospital, Lørenskog, Norway.
Faculty of Health Sciences, Oslo and Akershus University College of Applied Sciences, Oslo, Norway.
Int J Lab Hematol. 2017 Aug;39(4):423-428. doi: 10.1111/ijlh.12648. Epub 2017 Mar 23.
The CellaVision™DM96 is a digital pattern recognition system that classifies white blood cells. The aim of this study was to evaluate whether the CellaVision preclassification feature, without a subsequent re-classification, was a sufficient approach to follow up flags reported by Sysmex XE-5000.
Pairs of blood smears from 400 samples reported with suspect flags were examined using conventional microscopy and the CellaVision features. The effect of pre- vs. re-classification, and intersmear and between-technologist variation, on blast counts was assessed using generalized linear mixed models (GLMM).
The GLMM analysis showed a significant difference between the blast counts of preclassification vs. re-classification (P = 0.009). The analysis showed no significant difference between duplicate smears (P = 0.621) or between technologists (P = 0.542). Preclassification showed blasts in 105 samples (26%), where the re-classification feature did not detect any blasts. Not a single sample that was re-classified as positive for blasts was preclassified as negative. Compared to manual microscopy, the sensitivity and specificity of the preclassification feature were 0.83 and 0.66, respectively.
The preclassification feature alone is sufficient to verify the absence of blasts in flagged samples. When the preclassification feature detects blasts, the finding has to be verified or reclassified by an experienced technologist. However, the use of CellaVision™DM96 in the follow-up of blast reports has to be questioned due to the finding of false-negative samples in the preclassification feature, but also after re-classification, compared to manual slide review.
CellaVision™ DM96是一种用于对白细胞进行分类的数字模式识别系统。本研究的目的是评估CellaVision预分类功能（无需后续重新分类）是否足以跟进Sysmex XE - 5000报告的警示标记。
使用传统显微镜检查和CellaVision功能，对400份报告有可疑警示标记的样本的成对血涂片进行检查。使用广义线性混合模型（GLMM）评估预分类与重新分类以及涂片间和技术人员间差异对原始细胞计数的影响。
GLMM分析显示预分类与重新分类的原始细胞计数之间存在显著差异（P = 0.009）。分析显示重复涂片之间（P = 0.621）或技术人员之间（P = 0.542）无显著差异。预分类在105个样本（26%）中检测到原始细胞，而重新分类功能未检测到任何原始细胞。没有一个重新分类为原始细胞阳性的样本被预分类为阴性。与手工显微镜检查相比，预分类功能的敏感性和特异性分别为0.83和0.66。
仅预分类功能就足以验证警示标记样本中是否不存在原始细胞。当预分类功能检测到原始细胞时，该结果必须由经验丰富的技术人员进行验证或重新分类。然而，由于在预分类功能中发现了假阴性样本，而且与手工玻片复查相比，重新分类后也存在假阴性样本，因此在原始细胞报告的跟进中使用CellaVision™ DM96值得质疑。
