CAPÍTULO 3. EXPERIMENTAL
1. Procedimiento experimental
1 Specify Goals Example - The following information is recorded by this activity for the sample iteration. For each instance we show how the information is modeled and represented in the e-Science Life Cycle Ontology.
– Goal specification name: goalSpecification NIGM
<hasGoalSpecification rdf:resource="#goalSpecification_NIGM"/>
– a scientist username: CADGrid-researcher3
<scientist rdf:about="#CADGrid-researcher3"></scientist>
– a research group he or she is member of: CADGrid-buct.edu.cn
<researchGroup rdf:about="#CADGrid-buct.edu.cn">
<hasMembers rdf:resource="#CADGrid-researcher3"/>
</researchGroup>
– description of objectives, including a name of the description data de-scriptionData gS NIGM
6.3 Concrete Use-Case Scenario 103 – keywords such as The Non-Invasive Blood Glucose Measurement (NIGM)
and NIGM measurement techniques
– a textual description: The Non-Invasive Blood Glucose Measurement (NIGM) method offers a novel promising non-invasive technology for measuring patient’s data. In order to be accepted by general medical com-munity, the quality of provided results will have to be verified in extensive clinical trials following well defined protocol. The main aim of this exper-iment is to specify medical conditions under which values of blood glucose obtained through the NIGM method are as reliable as data provided by the invasive measurement techniques that are currently the only standard used and accepted by western medicine [EHL+08].
<descriptionData rdf:about="#descriptionData_gS_NIGM">
<keywords xml:lang="en">NIGM</keywords>
<Name xml:lang="en">
The Non-Invasive Blood Glucose Measurement (NIGM)
</Name>
<keywords xml:lang="en">blood glucose</keywords>
<description xml:lang="en">
he Non-Invasive Blood Glucose Measurement (NIGM) method
– Additionally, we save a name of the executed life cycle NIGM CADGrid 001.
<scientist rdf:about="#CADGrid-researcher3">
<execute rdf:resource="#NIGM_CADGrid_001"/>
</scientist>
We further save
2 Prepare Data Example - In this experiment there is no need for data integra-tion, thus the final input data set is simply the deployed OGSA-DAI resource.
The following information is recorded by this activity.
– Data preparation name: dataPreparation NIGM
6.3 Concrete Use-Case Scenario 104
<hasDataPreparation rdf:resource="#dataPreparation_NIGM"/>
– Resource name: CADGrid-MMDB1
<resourceFile rdf:about="#CADGrid-MMDB1">
<isResourceFileOf rdf:resource="#dataPreparation_NIGM"/>
</resourceFile>
– OGSA-DAI resource file location: http://.../CADGrid-MMDB1.ResourceFile
<resourceFile rdf:about="#CADGrid-MMDB1">
<resourceFileReference rdf:datatype="&xsd;anyURI">
http://.../CADGrid-MMDB1
</resourceFileReference>
</resourceFile>
– Resource description: Within various experiments on diabetes patients the CADGrid research team has collected large amount of meridian mea-surement data and corresponding blood glucose values. The experiments were applied on 300 diabetic patients, conducted in the Dong Fang hos-pital, Beijing, China. For each patient 50 meridian measurements and equally many measures were applied using the conventional invasive blood glucose method. Meridian measurements are represented as value pairs, the input signal value, and the output signal value. One meridian mea-surement generates around 1000 to 10000 value pairs [EHL+08].
<resourceFile rdf:about="#CADGrid-MMDB1">
<resourceFileDescription rdf:datatype="&xsd;string"
>The experiments were applied on 300 diabetic patients..
</resourceFileDescription>
</resourceFile>
Background data resources participating in this example application are pro-viding information about the patient’s health condition and the meridian on which the measurement was performed. They are considered as background data resources within a dataspace.
3 Select Appropriate Tasks Example - In this example selected tasks are orga-nized within the NIGM workflow, which consists of the following algorithms, deployed as CADGrid services: (1) System Identification, (2) Kalman Filter-ing, (3) Wavelet Transformation, (4) Fast Fourier Transformation, (5) Com-bination Service, and finally (6) Back Propagation Neural Network. The Fol-lowing information is captured.
6.3 Concrete Use-Case Scenario 105 – Task selection name: NIGM Inst23
<hasTaskSelection rdf:resource="#taskSelection_NIGM"/>
– Workflow name: NIGM Inst23 and the above mentioned CADGrid ser-vices.
– Workflow description: The NIGM workflow processes a compute intensive signal phase and followed that sets up a neural network, which can be used for computing humans blood glucose values. Input data sets are on one side meridian measurements, collected using a specially developed instrument and on the other side blood glucose values measured using the conventional method.
>Runs a compute intensive signal processing phase...
<usesService rdf:resource="#CADGrid_service_001"/>
<describes rdf:resource="#NIGM_workflow"/>
</descriptionData>
6.3 Concrete Use-Case Scenario 106 – Workflow reference: http://.../NIGM Inst23.gwa
<workflow rdf:about="#NIGM_workflow">
<workflowReference rdf:datatype="&xsd;anyURI">
http://.../NIGM_Inst23.gwa
</workflowReference>
</workflow>
4 Run Tasks Example - We record the PMML document representing the output of the neural network model set up by the NIGM workflow and its correspond-ing visualization document, all considered as derived data. The information is captured as follows:
– Task execution name: taskExecution NIGM
<hasTaskExecution rdf:resource="#taskExecution_NIGM"/>
– PMML Document: pmmlDocument NIGM
<pmmlDocument rdf:about="#pmmlDocument_NIGM">
<pmmlReference rdf:datatype="&xsd;anyURI"></pmmlReference>
<isPMMLDocumentOf rdf:resource="#taskExecution_NIGM"/>
</pmmlDocument>
– NIGM visualisation: NIGM visualisation
<visualisation rdf:about="#NIGM_visualisation">
<visualisationType rdf:datatype="&xsd;string">
PMLL Visualisation</visualisationType>
<visualisationReference rdf:datatype="&xsd;anyURI">
http://lela.par.univie.ac.at/cadgrid/
CADGrid_OutputModel_NN.svg
</visualisationReference>
<isVisualisationOf rdf:resource="#taskExecution_NIGM"/>
</visualisation>
5 Process and Publish Results Example - A publish mode is dedicated to each output, based on which the outputs are published into a dataspace. Finally a publication mode of the executed Life Cycle is generated, restricting the published results to one particular research group CADGrid-buct.edu.cn. We capture the information as follows:
– Task execution name: resultPublishing NIGM
6.3 Concrete Use-Case Scenario 107
<hasResultPublishing rdf:resource="#resultPublishing_NIGM"/>
– Publication mode: NIGM publicationMode
<publicationMode rdf:about="#NIGM_publicationMode">
<hasPublicationMode rdf:resource="#CADGrid-buct.edu.cn"/>
<isPublicationModeOf rdf:resource="#NIGM_CADGrid_001"/>
</publicationMode>
The information captured by this life cycle represents the relationship among its participating data resources (primary, derived, and background data). It also covers the objectives specified for this experiment. Finally the data is published in the dataspace, thus is available for further data mining studies aiming at further im-provements in diabetic care and meridian theory resulting in higher patient comfort.