摘要
群体医疗数据的分析技术对于提高区域医疗资源利用率有着重要的意义,特别是对于高职院校等集体生活人数较多的特殊群体,智能化的数据分析可以有效保障群体的健康水平。文中针对群体医疗数据中的影像数据,提出了一种卷积自动编码器深度学习框架,并建立了医疗数据分析系统。该方案通过未标记的数据实现了对肺结节的无监督图像特征学习,且该过程仅需要少量的标记数据即可进行有效的特征学习。综合实验测试数据结果表明,该方案的性能优于其他方案,有效解决了人工图像标注过程中固有的劳动密集问题。此外,还验证了所提出的卷积自动编码器方法可以扩展为肺结节图像的相似性测量,与传统方法相比具有显著的优越性。
The analysis technology of group medical data is of great significance to improve the utilization rate of regional medical resources,especially for the special groups with more collective life such as higher vocational colleges,intelligent data analysis can effectively guarantee the health level of the group.In this paper,a deep learning framework of convolutional automatic encoder is proposed for the image data of group medical data,and the medical data analysis system is established.This scheme realizes unsupervised image feature learning of pulmonary nodules through unlabeled data,and only a small amount of labeled data is needed for effective feature learning.The experimental results show that the performance of this scheme is better than other schemes,and the inherent labor-intensive problem in the process of artificial image annotation is effectively solved.In addition,it is verified that the proposed convolution automatic encoder method can be extended to the similarity measurement of pulmonary nodule images,and has significant advantages over the traditional methods.
作者
邢文娜
宁睿
XING Wenna;NING Rui(Xi’an Vocational and Technical College of Aeronautics and Astronautics,Xi’an 710089,China)
出处
《电子设计工程》
2021年第14期75-79,共5页
Electronic Design Engineering
基金
2019年陕西高校辅导员工作研究课题(2019FKT35)。
关键词
群体医疗数据分析
深度学习
神经网络
特征学习
group medical data analysis
deep learning
neural network
feature learning
