Formal Ontologies And Rdf Health Essay

Semantic interoperability of patient informations across information systems has been on the medical information sciences docket for many old ages. Electronic Healthcare Record architectures and theoretical accounts that support interoperability have been proposed and developed, including OpenEHR [ 1 ] , ISO 13606 [ 2 ] , CEM [ 3 ] and HL7 [ 4 ] . Nevertheless the semantic interoperability desideratum is by and big unsated, non merely because of the go oning predomination of textual informations, but besides the inability of most systems to attach intending to informations in a manner that makes them semantically interoperable, despite legion medical nomenclatures and categorization systems have been developed and deployed. The SemanticHEALTH roadmap made a series of recommendations about how semantic interoperability might be achieved [ 5 ] .Formal Ontologies And Rdf Health Essay.  It emphasized that EHR information theoretical accounts are non plenty for stand foring clinical information. Standardized nomenclatures such as SNOMED CT [ 6 ] are every bit required to attach expressed semantics to clinical content. However, there is an convergence between EHR information theoretical accounts and nomenclatures that permits different possible representations for the same clinical information. As an illustration, the SNOMED CT construct Cancer confirmed refers both to a clinical construct ( disease ) and the information about that construct ( diagnosing ) . It is a typical instance for a intercrossed term that blends epistemological and ontological facets.
One major challenge in the context of semantic interoperability consists of observing the information that is semantically tantamount among isosemantic theoretical accounts, i.e. representations of the same content but in different ways. This requires counsel to decently place what has to be represented by EHR information theoretical accounts and what by clinical nomenclatures in order to stand for clinical information in an unambiguous manner.
Such recommendations are being addressed by the two major current interoperability enterprises, viz. SemanticHealthNet [ 7 ] and CIMI [ 8 ] , which are based on the premise that solutions are necessary that would include the coexistence of several EHR criterions. Both enterprises promote EHR architectures in which the semantics of the clinical theoretical accounts can be understood and processed.

  ORDER A PLAGIARISM -FREE PAPER NOW

In this work, our aim will be t o demo how we can uniformly query clinical information coming from heterogenous systems by bring forthing the corresponding RDF datasets and this procedure being guided by a shared model based on formal ontology. CEM and OpenEHR datasets will be transformed into RDF ( Resources Description Framework ) [ 9 ] , and SPARQL questions [ 10 ] will show how heterogenous clinical information beginnings can be queried in a homogenous manner. Formal Ontologies And Rdf Health Essay.
Materials and Methods
Clinical information is normally captured utilizing clinical theoretical accounts. Different clinical theoretical accounts can be created for the same clinical information and such theoretical accounts can be represented utilizing different formalisms, or even utilizing the same 1. This means that clinical information can be captured in many different ways by different theoretical accounts despite the fact that they represent the same clinical content. We call them isosemantic theoretical accounts. In this research, we use CEM and OpenEHR as our usage instance clinical theoretical accounts since most theoretical accounts defined by these two specifications are publically available.
Following, we describe the shared model developed in the context of SemanticHealthNet for stand foring clinical information in a manner that the boundary line between clinical and information entities is clearly delimited by formal-ontological standards. This model will do usage of description logics ( DL ) [ 11 ] ( OWL 2 DL [ 12 ] ) , supported by DL ratiocinators such as HermiT [ 13 ] . Later we will demo how this model can be used to observe semantically tantamount information from isosemantic clinical theoretical accounts.
Shared Formal-Ontological Model
The model proposed consists of the following ontologies which are described following: top degree, sphere, and EHR information entity.
Top-level ontology:
Top-level or high-level ontologies provide a basic, largely domain-independent set of cardinal classs, dealingss and maxims, which are easy reclaimable across specific sphere applications. Without them, if one modeling undertaking is given to different people the ensuing representations will be widely different.Formal Ontologies And Rdf Health Essay.  Therefore, top-level ontologies are meant to turn to this job by standardizing the development procedure and curtailing the picks of ontology applied scientists. Under top-level ontologies, top-domain ontologies further elaborate an upper degree ontology to a specific sphere. We will utilize the top-level ontology BioTopLite [ 14 ] to restrain the manner in which the information and clinical sphere ontologies combine. BioTopLite exhibits ten reciprocally disjoint top categories: Condition, Disposition, ImmaterialObject, MaterialObject, InformationObject, Quality, Role, ValueRegion, Process and Time, under which we will categorise non merely the categories that describe the medical sphere, but besides formal representations of information entities.
Domain ontology:
As an ontology for stand foring the health care sphere, parts of SNOMED CT will be used. SNOMED CT is a nomenclature system partly built on formal-ontological rules, utilizing the description logics EL++ , for formalising necessary and sufficient conditions for its about 311,000 constructs, which are embedded in a comprehensive multihierarchical taxonomy. The selected SNOMED CT content will be placed under classs provided by the BioTopLite ( e.g. Procedure, Quality, Condition, etc. ) .
EHR Information entity ontology:
The EHR information entity ontology represents clinical information entities, which are results of clinical actions like observations, probes, or ratings. They have in common that they refer to clinical entities or to intensional category definitions and are farther qualified by properties that represent the epistemic and contextual facets of the information ( e.g. past history, confirmed or suspected diagnosing, etc. ) . All categories of this ontology will be represented as subclasses of the top-level class information object of BioTopLite. This category is defined as a piece of information that exists independently of any possible stuff bearer. Information entities will mention to clinical entities by agencies of the relation isAboutSituation defined as a specialisation of the relation isAbout that relates an information artifact to an entity ( e.g. clinical entity ) . As subclasses of information object we have placed the categories
shn_clinical_information_item for clinical information entities that refer to some clinical state of affairs of a patient in which he/she might hold some clinical status ( s ) . Formal Ontologies And Rdf Health Essay.
shn_clinical_information_set for clinical information entities that aggregate other clinical information points that together stipulate some patient clinical state of affairs.
shn_clinical_composition for aggregations of clinical information entities that refer to some patient healthcare state of affairs.
Additionally, information_object_attribute has been created to stand for epistemological properties of information entities like Suspected, Probable, etc. There are other information entities such as clinical datatypes, that are non described here and that will be included in subsequent versions. Figure 1 depicts the ontological model. The discontinuous pointers depict how categories from the different ontologies relate each other. Merely some of them have been included in the figure, as they will assist to understand the attack presented. For case, it can be observed how an information point is an information entity that refers to a clinical construct from SNOMED CT, more specifically a clinical finding/Situation. IHTSDO, in order to transport out the harmonisation between SNOMED CT and the new ICD-11 disease categorization, is analyzing the reading of SNOMED CT upsets as clinical state of affairss. We propose to construe SNOMED happening constructs as clinical state of affairss, e.g. as stages of a patient ‘s life during which a certain status ( organic structure construction, temperament, procedure of clinical involvement ) exists during all clip [ 15 ] . Information entities refer to ( types of ) patient state of affairss and are the consequence of ( outcomeOf ) executing some clinical action or process. These processs will be farther described by agencies of a clinical procedure ontology, which will stand for the different health care procedures, activities, etc. together with their interactions that provide the clinical information. Since the usage instance questions in this research do non necessitate health care procedures, the execution of this ontology and its harmonisation with SNOMED CT is beyond of the range of this paper.
Figure 1 – Shared Formal-Ontological Model
Use of the ontological model for observing semantically tantamount clinical information from isosemantic theoretical accounts
In order to utilize the model presented to observe semantically tantamount clinical information from different clinical theoretical accounts we propose to footnote the information entities of the clinical theoretical account cases by utilizing OWL DL looks conforming to the ontologies presented, which normally represent complex types of patient state of affairss. For case a diagnosing of a patient with suspected disease ( X ) would be expressed as:
shn_SuspectedDiagnosis and isAboutSituation merely X
where shn_SuspectedDiagnosis is a defined category:
shn_SuspectedDiagnosis equivalentTo
shn_clinical_information_item
and outcomeOf some shn_DiagnosingAction
and hasInformationObjectAttribute some Suspected
We do non utilize OWL DL notes to stand for the whole clinical theoretical account informations case. Clinical information in an EHR normally answer the inquiries WHERE ( e.g. healthcare installation ) , WHEN ( e.g. day of the month clip ) , WHO ( e.g. clinician ) , HOW ( e.g. clinical procedure ) and WHAT ( e.g. clinical status ) .Formal Ontologies And Rdf Health Essay. OWL DL notes will merely reply the WHAT inquiry, which may go randomly complex, harmonizing to the grade of post-coordination needed. In this manner description logics concluding and questioning would be restricted to the WHAT clause, whereas database questioning would be addressed by the HOW, WHO, WHEN, and WHERE clauses. Even if utilizing DL sentence structure for stand foring all clause replies, the logical thinking would be restricted to the filler of the isAboutSituation relation. All extra averments, e.g. who made the diagnosing, when it was made etc. could so be made at A-box ( informations ) degree. This would suitably turn to the fact that DL representations follow an open-world semantics, whereas information theoretical accounts – at least implicitly – follow a closed-world semantics.
Once information entities from clinical theoretical accounts cases have been annotated, a DL ratiocinator can calculate the equality between the different DL looks used in the theoretical accounts ( see Figure 2 ) .
Figure 2 – Owl DL note and logical thinking
The ratiocinator detects tantamount clinical information, which is represented by utilizing heterogenous informations constructions in their several theoretical accounts. For case, in one theoretical account the location of a disease can be represented together with the disease and in another one by different informations constructions.
Coevals of the RDF EHR information depositories and note of informations with the RDF inferred theoretical account
We use RDF for stand foring the patient informations. An RDF statement consists of a ternary subject-predicate-object and a aggregation of RDF statements represents a graph. RDF graphs can be easy connected by associating their related nodes. In our usage instance, an RDF graph is created for each clinical informations depository. In these graphs the nodes represent the categories of all ( single ) information entities and the borders the relationship among them. As discussed in the old subdivisions, we have added OWL DL notes to the different information entities of the clinical theoretical account informations cases and computed the equality of the notes by utilizing the DL ratiocinator. Formal Ontologies And Rdf Health Essay. As a consequence we have obtained an inferred ontology which has been transformed into RDF to be used together with the graphs generated for each clinical informations depository.
Figure 3 shows how the information entities from two clinical theoretical account informations cases relate by agencies of the linkedTo relation with the RDF graph theoretical account that contains the note cases.
Figure 3 – Relation between the RDF clinical informations cases and the shared notes
Since the RDF graph with the notes cases has been created from the inferred theoretical account, each note case includes as rdf: type all the illations obtained.
Exploitation of the depository
Once the three RDF graphs have been created and defined the relation between their nodes, we will utilize these dealingss to do SPARQL questions to the different RDF clinical informations depositories in a homogenous manner.
Figure 4 describes the question procedure. A question is executed over the graph of the shared ontological model. By agencies of the relation linkedTo it is possible to acquire informations from heterogenous beginnings, in this instance the theoretical accounts A and B.
Figure 4 – Description of how heterogenous clinical informations are queried and how consequences from both datasets can be obtained
Consequences
We have evaluated the presented attack by following these stairss: ( 1 ) coevals of EHR informations in RDF ; ( 2 ) note of clinical informations and compute equalities ; and ( 3 ) SPARQL question edifice.
Coevals of EHR informations in RDF
The starting point is one set of patient informations but rendered in two different ways, one utilizing CEM theoretical accounts and another one utilizing OpenEHR originals and both represented in RDF. We have used CEM theoretical accounts and OpenEHR originals for stand foring patient demographic, diagnosing and medicine information. For each patient we represent the EHR informations by utilizing an RDF graph. In order to bring forth the CEM RDF data the CEM-OWL ontology has been used. CEM-OWL represents elaborate CEM theoretical accounts utilizing OWL to offer a formal definition of these theoretical accounts [ 16 ] . The CEM OWL ontologies are publically available and with regard to these ontologies, a set of fake patient informations has been generated and stored in an RDF three-base hit shop [ 17 ] . For bring forthing the OpenEHR dataset we used an archetype-based function tool to bring forth valid OpenEHR XML instances that follow OpenEHR mention theoretical account from available CEM XML cases. Formal Ontologies And Rdf Health Essay. These OpenEHR cases include both clinical informations and OpenEHR model context information. In order to bring forth the OpenEHR RDF data an ontology for OpenEHR has been used [ 18 ] .
Note of clinical informations and compute equalities
The diagnosing and medicine informations of patients from the CEM and the OpenEHR RDF depositories have been annotated with OWL DL looks harmonizing to the shared model antecedently introduced. Tables 1 and 2 describe the representation for the diagnosing instance for both CEM and OpenEHR. In CEM most of the diagnosing related information is encoded by the SNOMED CT codification 5969009 for Diabetes mellitus associated with familial syndrome while in OpenEHR two information entities are used. The codification 73211009 bases for Diabetes mellitus and 290028006 bases for Genetic syndrome. The relation between both can non be deduced since it is non explicitly represented in the clinical theoretical account. The present part enables acknowledgment of the term-term relationship in deciding isosemantic renditions manifest in an information theoretical account.
Table 1 – Datas harmonizing to CEM
CEM
AdministrativeDiagnosis
Information entity
Value
AttribRecordedTime
2001-09-27T07:19:37
hasCode 1
5969009
hasCodeSystem
SNOMED CT
Status 2
Active
Table 2 – Datas harmonizing to OpenEHR
OpenEHR
openEHR-EHR-EVALUATION.problem-diagnosis.v1
Information entity
Value
Date clinically recognised
2001-09-27T07:19:37
Diagnosis 1
SNOMED-CT: :73211009
Aetiology 2
SNOMED-CT: :290028006
Status 3
Active
In order to observe that both theoretical accounts are isosemantic, i.e. different constructions but the same semantic content, we can footnote the above CEM and OpenEHR information entities with OWL DL looks shown in Tables 3 and 4 severally. Each case of the diagnosing of a patient has to be interpreted as the concurrence of these DL looks. Formal Ontologies And Rdf Health Essay.
In Table 3 the reading for the SNOMED CT construct Diabetes mellitus associated with familial syndrome is provided. We have interpreted it as a state of affairs of a patient with both diabetes mellitus and familial syndrome findings. In SNOMED CT the relation associated with represents the interaction between two constructs beyond simple accompaniment in the patient, without either asseverating or excepting a causal or consecutive relationship between the two.
Table 3 – Owl DL look for CEM informations cases
CEM
AdministrativeDiagnosis
OWL DL look
1
shn_Diagnosis equivalentTo
shn_information_item and
outcomeOf some shn_diagnosingAction
shn_Diagnosis_diabetes_assocWith_ genetic_syndrome
equivalentTo
shn_Diagnosis and
isAboutSituation merely
shn_SituationWithDiabetesAndGeneticSyndrome
2
shn_Diagnosis_active equivalentTo
shn_Diagnosis and
shn_hasInformationObjectAttribute some shn_Active
Once the OWL DL notes are created, we classify the ontology to obtain the illations by utilizing a DL ratiocinator and we transform them into RDF in order to let subsequently constructing questions to both RDF depositories. Then, we will associate the information entities from two clinical theoretical account informations cases with the RDF notes graph by agencies of the linkedTo relation as shown in Figure 5. There, we can detect the dealingss defined between the CEM information entity AdministrativeDiagnosis and the corresponding OpenEHR information entities for stand foring the diagnosing, its position and the cause with respects to the shared notes.Formal Ontologies And Rdf Health Essay.  These last 1s allow recovering patient diagnosing informations represented harmonizing to CEM or OpenEHR in a homogenous manner.
Table 4 – Owl DL look for OpenEHR informations cases
OpenEHR
openEHR-EHR-EVALUATION.problem-diagnosis.v1
OWL DL look
1
shn_Diagnosis equivalentTo
shn_information_item and
outcomeOf some shn_diagnosingAction
shn_Diagnosis_diabetes equivalentTo
shn_Diagnosis and
isAboutSituation merely shn_SituationWithDiabetes
2
shn_Diagnosis_disease_assocWith_genetic_syndrome equivalentTo
shn_Diagnosis and
isAboutSituation merely
shn_SituationWithDiseaseAndGeneticSyndrome
3
shn_Diagnosis_active equivalentTo
shn_Diagnosis and
shn_hasInformationObjectAttribute some shn_Active
Figure 5 – Linkss between OpenEHR and CEM diagnosing information entities with respects to the shared notes
SPARQL question edifice
In order to construct questions that can recover patient informations from both CEM and RDF repositories we use the illations ensuing from calculating the equality between the information entity notes. Harmonizing to the notes we can inquire for all patients diagnosed with diabetes mellitus through the undermentioned SPARQL question:
PREFIX SHN: & lt ; hypertext transfer protocol: //semantichealthnet/SharedOnt.owl # & gt ;
SELECT? src1? s? src2? ten
FROM NAMED & lt ; CEM-Model-Data & gt ;
FROM NAMED & lt ; Shared-Annotations & gt ;
FROM NAMED & lt ; OpenEHR-Model-Data & gt ;
WHERE {
GRAPH? src1 {
? s rdf: type SHN: shn_Diagnosis_diabetes_mellitus.
}
GRAPH? src2 { ? x SHN: linkedTo? s. }
}
In this question the FROM NAMED ” clause specifies the three graphs we are utilizing. Each graph is identified by agencies of the GRAPH ” keyword. In this manner, the question consequence will give the name of the graphs where the information was found. In the WHERE ” clause we are inquiring for all the CEM or OpenEHR clinical informations cases that are linkedTo cases of the type shn_Diagnosis_diabetes_mellitus. This information was explicitly coded in the OpenEHR informations, but non in the CEM informations, which was represented by the individual SNOMED CT codification diabetes mellitus associated with familial syndrome. Since we have reinterpreted SNOMED CT findings as state of affairss ( see Table 3 ) , this would mention to Clinical state of affairs with diabetes mellitus and with some familial syndrome, classified as a subclass of Situation with diabetes and as a subclass of Situation with familial syndrome. Therefore, the question will recover the patient diagnosing information entities from both the CEM and the OpenEHR information depositories. Formal Ontologies And Rdf Health Essay.
The usage of the shared ontological model for making notes facilitates complex questions, such as obtaining all patients with some disease associated with a familial syndrome:
PREFIX SHN: & lt ; hypertext transfer protocol: //semantichealthnet/SharedOnt.owl # & gt ;
SELECT? src1? s? src2? ten
FROM NAMED & lt ; CEM-Model-Data & gt ;
FROM NAMED & lt ; Shared-Annotations & gt ;
FROM NAMED & lt ; OpenEHR-Model-Data & gt ;
WHERE {
GRAPH? src1 {
? s rdf: type
SHN: shn_Diagnosis_disease_assocWith_genetic_syndrome.
}
GRAPH? src2 {
? ten SHN: linkedTo? s.
}
}
It is non easy to recover this information from the CEM or OpenEHR patient cases without utilizing the semantic logical thinking offered by our attack. In the first instance, the SNOMED CT happening Diabetes mellitus associated with familial syndrome is defined as a subclasss of Diabetes mellitus but non of Genetic syndrome with which is associated by agencies of the SNOMED CT concept theoretical account relation associated with. In the OpenEHR instance it is represented in the information entity textually described as Aetiology ” which is further refined by agencies of the three information entities ‘Agent ‘ , ‘complication of ‘ and description.

ORDER A PLAGIARISM -FREE PAPER NOW

For the running illustration, a state of affairs with both diabetes mellitus and a familial syndrome is classified as a subclass of the category that represents all the diseases ( see Table 4 ) associated with some familial syndrome and hence this question retrieves both patient diagnosing information entities. Therefore the information could be retrieved by a simple question shown above.
Decisions and Future Work
Semantic interoperability of clinical information requires decently identifying and sharing the significance of the information to be exchanged. Due to the deficiency of standardisation of what can be represented by information and clinical entities, there is more than one representation of the same clinical content. Therefore one major challenge is the sensing of information that is semantically tantamount among isosemantic theoretical accounts.
Targeting to this challenge, we envision a shared model developed by the SemanticHealthNet web, based on formal-ontological standards. In this paper, we discussed how this model can be used to observe semantically tantamount information from two heterogenous datasets based on the Clinical Element Model and the OpenEHR specifications. Formal Ontologies And Rdf Health Essay. The attack is based on the add-on of OWL DL semantic notes which used the illations obtained by DL concluding for constructing questions that retrieve patient informations from both datasets.
Several hereafter waies need to be pursued. First, we would wish to implement automatic attacks to bring forth proper semantic notes of clinical entities. Second, in this work the informations that corresponds to the WHERE, WHEN and WHO clauses have non been included in the ontological model as unfastened universe description logics concluding is limited to the WHAT clause. In order to recover this information it is hence necessary to question the specific CEM or OpenEHR theoretical account. However, as it has been already mentioned, this could be added to the ontology at A-box degree.
Recognitions
This work has been funded by the SemanticHealthNet Network of Excellence within the EU 7th Framework Program, Call: FP7-ICT-2011-7, understanding no. : 288408, hypertext transfer protocol: //www.semantichealthnet.eu/ , by the Spanish Ministry of Science and Innovation through grant TIN2010-21388-C02-02 and co-funded by FEDER. MC Legaz-GarciI?a is supported by the FundacioI?n SeI?neca ( 15555/FPI/2010 ) . Partial support besides from US HHS/ONC Cooperative Agreement ( 90TR000201-2 ) SHARPn Area 4: Secondary Use of EHR Data.

Formal Ontologies And Rdf Health Essay