摘要
针对人工镜检分类白细胞准确率和效率低的问题,基于深度学习和机器学习算法,提出了一种基于迁移学习和支持向量机的白细胞分类方法。首先对迁移模型进行微调训练,其次用微调训练后的迁移模型进行特征提取,然后将特征输入至神经网络和支持向量机中进行训练,最后通过神经网络和支持向量机的组合分类器对白细胞进行分类。实验结果表明,白细胞分类准确率由最初微调训练的83.26%,随着迁移模型的优化提升为90.43%,最后通过组合分类器再次提升为93.52%,可以在临床实践中帮助医生提高诊断的准确率和效率。
Aiming at the low accuracy and inefficiency of manual microscopic examination of white blood cell classification,based on deep learning and machine learning algorithms,a white blood cell classification method based on transfer learning and support vector machine was proposed.Firstly,the transfer model was fine-tuned,then the features were extracted from the fine-tuned transfer model and the features were input into the neural network and support vector machine for training.Finally,the white blood cells were classified by the combined classifier of neural network and support vector machine.The experimental result showed that the accuracy of white blood cell classification increased from 83.26%in the initial fine-tuning training to 90.43%with the optimization of the transfer model,and finally to 93.52%through the combined classifier,which can help doctors improve the accuracy and efficiency of diagnosis in clinical practice.
作者
张剑飞
郭笑颜
王波
崔文升
ZHANG Jian-fei;GUO Xiao-yan;WANG Bo;CUI Wen-sheng(College of Computer and Control Engineering, Qiqihar University, Qiqihar 161006, China;College of Computer Science and Information Technology, Daqing Normal University, Daqing 163712, China)
出处
《科学技术与工程》
北大核心
2021年第19期8113-8119,共7页
Science Technology and Engineering
基金
国家社会科学基金(19BGL241)
黑龙江省省属高校基本科研业务费科研项目(135509402)。
关键词
白细胞分类
迁移学习
神经网络
支持向量机
white blood cell classification
transfer learning
neural network
support vector machine
