自动血细胞形态预分类系统的性能验证方法的探索

The Exploration of the Methods of Performance Verification of Automatic Blood Cell Morphology and Pre-classification System

目的按ICSH2014年版血细胞分析仪评价指南，对CelldiffBCM-2自动血细胞形态分析系统的白细胞分类计数进行验证及临床应用评估。方法Celldiff分析仪对401例需显微镜复检的住院患者血涂片、101例健康体检者血涂片进行白细胞分类。（1）通过对Celldiff预分类结果与人工纠正后结果符合性分析。（2）人工纠正后结果与人工镜检结果相关性分析。结果住院患者标本，绝对偏差从低到高分别为中性中幼粒细胞（3．3％）、中性分叶核粒细胞（11．8％），单核细胞（20．5％）、嗜酸性粒细胞（28．4％）、中性杆状核粒细胞（31.6%）、淋巴细胞（42．0％）、中性晚幼粒细胞（68．0％）、嗜碱性粒细胞（77．1％）、异型淋巴细胞（100．0％）、幼稚淋巴细胞（100．0％）、原始细胞（236．5％）、涂抹细胞（306．7％）、早幼粒细胞（413．3％）和有核红细胞（1321．2％）。健康体检者标本，绝对偏差从低到高分别为单核细胞（0．4％）、中性分叶核粒细胞（2．1％）、中性杆状核粒细胞（14．1％）、淋巴细胞（15．5oA）、嗜酸性粒细胞（21．8％）、嗜碱性粒细胞（69．2％）和涂抹细胞（728．6％）。镜检分类与人工纠正后的相关性除了嗜碱性粒细胞的线性有明显偏差（P〈0．01），其余中性杆状核粒细胞、中性分叶核粒细胞、淋巴细胞、单核细胞、嗜酸性粒细胞及幼稚粒细胞的线性均无明显偏差。结论Celldiff虽然预分类的结果不是非常理想，但经过人工审核后同人工镜检结果的相关性较好还是具有较好的临床应用以及教学的价值，再结合全自动外周血制片机利于实现白细胞分类的自动化与标准化。

Objective According to ICSH guidelines for the evaluation of blood cell analyzers,2014 edition to evaluate the clinical performance of celldif BCM-2 automated hematology analysis system for leukocyte differentiation. Methods 401 blood smears of hospitalized patients which complied with the rules of microscope re-examination and 101 blood smears of healthy controls were analyzed by celldiff automatic hematology analyzer and manual microscope of leukocyte classification. 1. We evaluated accordance rate： pre-classification results carried out by celldiff vs the results by manual correction. 2. Correlation analysis between the results by manual correction and the results by manual microscope. Results the absolute deviation of the hospitalized patients specimens from low to high was respectively ： neutrophilic myelocyte（3.3 % ）, neutrophilic segmented granulocyte（ 11.8 % ）, monocyte （20.5 % ）, eo-sinophils （28.4% ）, band form neutrophilic granulocyte （ 31. 6 % ）, lymphocyte （ 42% ）, netrophilic metamyelocyte （ 68 % ）, basophlis （77.1%） ,atypical lymphocyte（ 100%）, prolymphocyte（ 100%）, blasts （236.50/00）, smear cell （306.7%）, premyelocyte（413.3%） and NRBC （1321.2 % ）. The absolute deviation of healthy controls from low to high was respectively monocyte（0.4 %）, netrophilic segmented granulocyte（2.1% ）, band form neutrophilic granulocyte（ 14.1% ）, lymphocyte （ 15.5% ）, eosinophils （ 21.8% ）, basophils （69.2%） and smear cell（728. 6%）. There was a significant difference correlation between the microscopic examination and the manual correction in the linearity of the basophils（P^0.01）. There was no significant difference in the linearity of the other neutro- philic segmented granulocyte,band form netrophilic granulocyte,lymphocyte, monocyte, eosinophils and immature granulocyte. Conclusion even though the results of pre-classification by celldiff is not very ideal,the better correlation between the results by manu al review and the results by manual microscope shows good clinical application and teaching value. And with the full automatic pe- ripheral blood production machine can improve the automation and standardization of leukocyte differentiation.