基于人工脉冲神经网络的门诊静脉采血智能预约平台设计-《中国医学物理学杂志》

文章信息/Info

Title:: Design of an intelligent reservation platform for outpatient venous blood collection based on artificial spiking neural network

Author(s):: HU Shiwei; LI Nan; XU Li; Information Center, West China Hospital, Sichuan University, Chengdu 610041, China

Keywords:: Keywords: spiking neural network outpatient venous blood collection multilayer bidirectional LSTM network PDCA model reservation platform

摘要:: 基于人工脉冲神经网络建立新型门诊静脉采血智能预约平台。在数据源整合处理模块中，基于脉冲神经网络设计局部递归的人工脉冲神经网络，实现多类型预约数据源的整合处理。在序列特征挖掘模块中，利用多层双向LSTM网络建立PDCA模型，提取门诊静脉采血预约数据的形态特征和语义特征，通过多层双向LSTM网络融合序列特征实体信息。在预约模块中，门诊静脉采血智能预约主要通过自助设备与互联网来完成，通过JavaScrip编写自助设备与互联网页面的门诊静脉采血预约程序。平台数据库由多种数据表构成，具体包括医生信息表、预约时间点分配表等。平台性能测试结果显示，设计平台的数据库每秒查询率更高，最高可达到54 239次，信息抽取准确率最高为98.60%。

Abstract:: Abstract: A novel intelligent reservation platform for outpatient venous blood collection is established based on artificial spiking neural network. In the data source integration processing module, a local recursive artificial spiking neural network is designed based on the spiking neural network to realize the integration of multiple types of reservation data sources. In the sequence feature mining module, a multilayer bidirectional LSTM network is used to establish a PDCA model for extracting the morphological and semantic features of the outpatient venous blood collection appointment data, and the sequence feature entity information is fused through multilayer bidirectional LSTM network. In the reservation module, the intelligent reservation of outpatient venous blood collection is mainly completed through self-service equipment and the Internet, and the outpatient venous blood collection reservation programs of the self-service equipment and the Internet page are written through JavaScript. The platform database is composed of various data tables, including doctor information table, appointment time point allocation table, etc. The platform performance test results show that the database query rate per second of the designed platform is higher. The highest queries per second reaches 54 239, and the information extraction accuracy is up to 98.60%.