Data Mining & Cyberinfrastructures in Biomedical InformaticsMain ConceptsNature of Biomedical DataReviewSlide 5Slide 6Slide 7Slide 8Slide 9Data Mining in BMISlide 11Figure: Clinical Data RepositorySlide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Cyberinfrastructures in BMISlide 25Slide 26Slide 27Slide 28Slide 29Slide 30Slide 31Slide 32Slide 33Slide 34Slide 35Slide 36Slide 37Slide 38Slide 39Slide 40Slide 41Slide 42Slide 43Slide 44Slide 45Data Mining and Cyberinfrastructures in Biomedical Informatics - 1CSE 300Data Mining & Cyberinfrastructures in Data Mining & Cyberinfrastructures in Biomedical InformaticsBiomedical InformaticsRyan McGivernRyan McGivernCSE5095CSE5095May 1, 2011May 1, 2011Data Mining and Cyberinfrastructures in Biomedical Informatics - 2CSE 300Main ConceptsMain ConceptsData MiningData MiningKnowledge DiscoveryCyberinfrastructuresCyberinfrastructuresCollaborative ResearchData Mining and Cyberinfrastructures in Biomedical Informatics - 3CSE 300Nature of Biomedical DataNature of Biomedical DataHealth care is more than numbers and readingsHealth care is more than numbers and readingsCan’t replace the subjective sense of disease severity that a physician has in momentsCapture data in a way that best captures observationCapture data in a way that best captures observationData representationPrecisionData Mining and Cyberinfrastructures in Biomedical Informatics - 4CSE 300ReviewReviewMedical datumMedical datumAny single observation of a patientKnowledgeKnowledgeDerived through formal/informal analysis of dataInformationInformationCombine knowledge with data for new informationHeuristics and research modelsBMI Data-Knowledge SpectrumBMI Data-Knowledge SpectrumWhat information constitutes the substance of medicineData Mining and Cyberinfrastructures in Biomedical Informatics - 5CSE 300Nature of Biomedical DataNature of Biomedical DataKnowledge at one level of abstraction might be Knowledge at one level of abstraction might be considered data at anotherconsidered data at anotherMedical Database is a Collection of individual patient Medical Database is a Collection of individual patient observations observations EHR is in some sense simply a databaseUsing historical patient data from the EHR system can Using historical patient data from the EHR system can facilitate the deduction of new knowledge related to facilitate the deduction of new knowledge related to health care strategieshealth care strategiesData Mining and Cyberinfrastructures in Biomedical Informatics - 6CSE 300Nature of Biomedical DataNature of Biomedical DataHumans can intuitively decompose information from Humans can intuitively decompose information from unitary view of dataunitary view of dataBut nothing is intuitive to computational systemsExampleExampleClinical settingBP of 120/80 may suffice to indicate a normal readingAnalytical settingSystolic BP = 120 mm HgDiastolic BP = 80 mm HgData Mining and Cyberinfrastructures in Biomedical Informatics - 7CSE 300Nature of Biomedical DataNature of Biomedical DataData mining in health is mainly related to Clinical Data mining in health is mainly related to Clinical Research SupportResearch SupportClinical Data Repositories (CDRs)New knowledge learned through aggregated info from New knowledge learned through aggregated info from a large number of patientsa large number of patientsCan be facilitated by EHRsUnfortunatelyUnfortunatelyCDRs generally limited to admin data sourcesRarely store patient chartsData Mining and Cyberinfrastructures in Biomedical Informatics - 8CSE 300Nature of Biomedical DataNature of Biomedical DataCDRs support Clinical Research StudiesCDRs support Clinical Research StudiesRetrospective studiesRetrospective studiesInvestigate a hypothesis that was not a subject of the study at the time the data were collectedProspective studiesProspective studiesClinical hypothesis known in advanceResearch protocol designed to collect future dataData Mining and Cyberinfrastructures in Biomedical Informatics - 9CSE 300Nature of Biomedical DataNature of Biomedical DataKnowledge baseKnowledge baseFactsHeuristicsComplex modelsSemantic linkingConduct case based problem solvingMedical data is intrinsically heterogeneousMedical data is intrinsically heterogeneousIllusory to conceive ‘complete medical dataset’Data selective based on treatmentData Mining and Cyberinfrastructures in Biomedical Informatics - 10CSE 300Data Mining in BMIData Mining in BMIData miningData miningKnowledge discovery techniqueSophisticated statistical methodsIdentify trend patterns hidden amongst the sheer size of the dataset Data warehouseData warehouseMultiple heterogeneous data sources Organized under a unified schema Single site Facilitate management and decision makingData Mining and Cyberinfrastructures in Biomedical Informatics - 11CSE 300Data Mining in BMIData Mining in BMICDR is essentially a data warehouseCDR is essentially a data warehouseArchitecture consists of four tiersArchitecture consists of four tiersExternal data sourcesOperational databases, flat files, etc.Data storage layerUnified schema, metadata, data martsOLAP LayerData mining enginePresentation layerGUIUsually web-basedData Mining and Cyberinfrastructures in Biomedical Informatics - 12CSE 300Figure: Figure: Clinical Data RepositoryClinical Data RepositoryData Mining in BMIData Mining in BMIData Mining and Cyberinfrastructures in Biomedical Informatics - 13CSE 300Data Mining in BMIData Mining in BMIData integration mechanismData integration mechanismExtractionTransformationRefreshScrubbingData martsData martsSubsets of data tailored to a user groupCache resultant datasetsData Mining and Cyberinfrastructures in Biomedical Informatics - 14CSE 300Data Mining in BMIData Mining in BMIData integrationData integrationHeterogeneous data under a unified schemaOntologiesLink primary data expressions to structured vocabulariesData now available to search and algorithmic processing at different levels of abstractionClinical domainNotorious for overwhelming presence of natural language textNatural language processingData Mining and Cyberinfrastructures in Biomedical Informatics - 15CSE 300Data Mining in BMIData Mining in