基于对比学习的急性早幼粒细胞白血病智能检测算法模型在全血细胞分析中的建立与验证

Establishment and validation of intelligent detection model for acute promyelocytic leukemia based on contrastive learning in complete blood cell analysis

目的利用学习统计软件建立基于对比大模型的急性早幼粒细胞白血病(M3)智能检测算法模型(简称M3模型),并验证其有效性。方法通过实验室信息系统(LIS)和医院信息系统(HIS)检索并统计北京协和医院8256例行全血细胞分析的门诊及住院患者数据,建立M3筛查模型。采用2023年7—10月本院行全血细胞分析的门诊及住院患者数据对M3模型进行验证。结果M3模型对全血细胞分析中M3的筛查具有一定应用价值,在筛查中性粒细胞毒性变化方面有一定作用,成功筛出2例中性粒细胞蓝绿色包涵体。结论M3模型对M3诊断的特异性有待提高。后续研究将增加M3阳性病例以优化模型,在保证高敏感性的同时提高特异性,为全血细胞分析智能审核提供帮助。

Objective To establish an intelligent detection algorithm model for acute promyelocytic leukemia(M3 model)based on a contrast large model using machine learning statistical software and validate its effectiveness.Methods The data from 8256 outpatients and inpatients who underwent complete blood cell analysis at Peking Union Medical College Hospital were retrieved and analyzed using the laboratory information system(LIS)and hospital information system(HIS).A M3 screening model was established and validated using the data from outpatients and inpatients who underwent complete blood cell analysis at our hospital from July to October 2023.Results The M3 model demonstrated potential application value in screening for M3 disease in complete blood cell analysis,which showed certain efficacy in screening for neutrophil toxicity changes,particularly in identifying two cases of blue-green inclusion bodies in neutrophils.Conclusion The M3 model exhibited low specificity for M3 diagnosis.Future research should focus on increasing the number of M3-positive cases to optimize the model,ensuring high sensitivity while improving specificity.This model will provide assistance for the intelligent review of complete blood cell analysis.