Design And Implementation Of Multi Dimension Association Mining Based Clinical Decision Support System

Electronic Medical Records(EMR),medical data that record patient information by electronic means,provide services beyond paper medical records as the main source of information in the diagnosis and treatment process.With the development of medical services,especially the wide use of mobile devices and Internet technology,health data are keeping accumulating,and take on the characteristics of big data and heterogeneous distribution.This data structure is complex,strong independent but related to each other.The EMR usually uses a local system in a single hospital as the operational environment,and is difficult to be utilized since the weak semantic.The main reasons for the weak semantic include: there are heterogeneous among different hospital systems;instance data contains a large number of(semi-)structured data;hospital systems are enclosed and knowledge sharing is difficult.This paper intends to use semantic annotation and external open linked data fusion to enrich the semantic information of data in EMR,so as to enhance the machine understandability and interoperability of data.Using these as a basis,further associations mining and data utilizing can then be done.Main works of this paper include:(1)Designing a semantic data dictionary generating method.At first,this paper designed an automatic meta-data annotating method according to the characteristics of the meta-data of medical data.Based on word segmentation and connected words replacement,the method can generate semantic data dictionary of the meta-data.(2)Designing a medical instance correlation model generating method based on semantic data dictionary.Utilizing the semantic data dictionary,this paper annotates the(semi-)structured data,such as diagnosis result,medical history and symptoms,and then extracts the medical instance correlation model of every instances.By means of building these correlation model,the study can associate medical data to open linked data.(3)Building the multi-dimensional similarity calculation model based on medical instance correlation model.To provide a foothold for this study,this paper design a similarity calculation model on multi-dimensional information of medical data,and calculate the weights of the dimensions by using Multidimensional linear regression analysis.(4)Developing the prototype system based on multi-dimensional similarity calculation model.As the last step,this paper develops a clinic design support system based on the foregoing studies.This system can acquire the similar instances to provide design supports to medical workers.The annotating methods on meta-data and instance data proposed by this paper can achieve good efficiency and accuracy when handling heterogeneous mass data.By using the semantic data dictionaries and the medical instance correlation models,the medical data can be connected to open linked data and be integrated as a rich experience base.On this basis a multi-dimensional similarity calculation model in which the weights have been optimized can then mines associations among instances well and so as to provide a better support for decision supporting.