基于中医四诊、西医检验和生物信息的多类型传感器数据归一化分层处理架构的研究

Based on the Four Diagnostic Methods of TCM,Western Medicine Examination and Biological Information of Multi-Types Sensors Data Normalization Hierarchical Processing Architecture Research

在传感器技术、电子信息技术、5G网络、机器学习等技术高速发展的时代背景下,中医四诊数据,西医临床检验数据和生物信息数据的积累也随之呈指数式增长。本文针对3类异构数据的高效处理问题,围绕各类传感器的采集原理进行深入分析,探索建立一种多类型医用传感器数据归一化分层处理架构。此架构分为3层:基于时间轴的医用传感器集群层,多模态数据协议转换、安全传输层,医用传感器数据湖、专题库及API接口层。通过此架构可完成数据采集后的时间序列机器标注、协议转换、组网通信及安全传输、数据粒度的降维升阶、分层特征提取、实体关系建模等任务,旨在实现3类异构数据的归一化处理,为后续实现疾病画像、活态知识图谱、疾病全方位智能诊疗算法及建立中医、西医、生物信息3类数据的分层特征关联网络提供前期基础。

Under the background of the rapid development of sensor technology,electronic information technology,5G network,machine learning and other technologies,the accumulation of TCM four diagnosis data,Western clinical test data and biological information data also showed exponential growth.Aiming at the problem of efficient processing of three types of heterogeneous data,this paper makes an in-depth analysis around the collection principle of various sensors,and explores the establishment of a normalized hierarchical processing architecture for multi-type medical sensor data.This architecture is divided into three layers:medical sensor cluster layer based on the time axis,multi-mode data protocol transformation,secure transmission layer,medical sensor data lake,thematic library and API interface layer.Through this architecture,tasks such as time series machine annotation,protocol conversion,networking communication and secure transmission,dimensionality reduction and upgrading of data granularity,hierarchical feature extraction,entity relationship modeling can be completed after data collection,aiming at achieving normalization processing of three types of heterogeneous data.It provides a preliminary basis for the subsequent realization of disease portrait,active knowledge map,comprehensive intelligent diagnosis and treatment algorithm of disease,and the establishment of hierarchical feature association network of TCM,Western medicine and biological information data.