IoT in Provenance Management of Medical Data
353
for the single sensor measurement. As soon as necessary samples from the sensor
are collected, they can be regarded as an instance for learning, inference, and rules
generation.
Annotations for incoming data are derived from the level of ontologies. According
to that fact that system is supposed to be heterogeneity-agnostic, ontology layer
is combined of multiple ontologies that may be connected with each other either
via intermediate ontologies or independent ones responsible for separate field of
knowledge.
Regarding the problem of data provenance, most notions will be established at the
start of the system with minimal corrections during its work. Provenance notions can
be placed into a single ontology and exploited when initially annotated RDF-tuples
are stored to the database.
Abovementioned semantic rule engine is represented by SPARQL-queries. The
queries incorporate application logic and form a core of the system from the point
of view of application processing. Specific query is triggered whenever data related
to its parameter arrive to the storage. For instance, let us consider an example where
the patient reports that blood pressure test is performed. Storage engine writes this
parameter to the RDF-store with time parameter. Then, corresponding rule is triggered that checks if it is performed in the recommended time interval. Violation
of the recommended interval boundaries results that customer receives notification
about this event.
As has been mentioned earlier, rules can be modified and added by an expert. In our
case, a doctor plays a role of the expert. He/she accesses system’s data via dedicated
interface. While, in most cases, the system runs in autonomous mode, observation
from the doctor’s side is fundamental and control from the expert prevents situations when inference engine might cause malfunction or other severe consequences.
Therefore, the doctor should approve each generated rule first.
The schematic presentation of semantic medical system with all mentioned
components included is depicted in Fig. 1.
In this system, provenance of medical data is possible to establish on different
system levels for different cases involving actors of the system. It has to provide
a service that indicates wrong conclusions from correct data and prevents usage of
incorrect data to make decisions. It is especially critical for the healthcare system
responsible for the state of patients.
3.2 Medical Data Formats
Considering provenance of medical data, it is important to take into consideration
presentation of medical data.
Obviously, that text-based formats are the preferred choice for usage in semantic
system. As semantic origins imply that text information is prevalent in the system and
semantic annotations themselves are performed in text format, it is easy to combine
such representation with annotations in a meaningful. However, addition of the new
353
for the single sensor measurement. As soon as necessary samples from the sensor
are collected, they can be regarded as an instance for learning, inference, and rules
generation.
Annotations for incoming data are derived from the level of ontologies. According
to that fact that system is supposed to be heterogeneity-agnostic, ontology layer
is combined of multiple ontologies that may be connected with each other either
via intermediate ontologies or independent ones responsible for separate field of
knowledge.
Regarding the problem of data provenance, most notions will be established at the
start of the system with minimal corrections during its work. Provenance notions can
be placed into a single ontology and exploited when initially annotated RDF-tuples
are stored to the database.
Abovementioned semantic rule engine is represented by SPARQL-queries. The
queries incorporate application logic and form a core of the system from the point
of view of application processing. Specific query is triggered whenever data related
to its parameter arrive to the storage. For instance, let us consider an example where
the patient reports that blood pressure test is performed. Storage engine writes this
parameter to the RDF-store with time parameter. Then, corresponding rule is triggered that checks if it is performed in the recommended time interval. Violation
of the recommended interval boundaries results that customer receives notification
about this event.
As has been mentioned earlier, rules can be modified and added by an expert. In our
case, a doctor plays a role of the expert. He/she accesses system’s data via dedicated
interface. While, in most cases, the system runs in autonomous mode, observation
from the doctor’s side is fundamental and control from the expert prevents situations when inference engine might cause malfunction or other severe consequences.
Therefore, the doctor should approve each generated rule first.
The schematic presentation of semantic medical system with all mentioned
components included is depicted in Fig. 1.
In this system, provenance of medical data is possible to establish on different
system levels for different cases involving actors of the system. It has to provide
a service that indicates wrong conclusions from correct data and prevents usage of
incorrect data to make decisions. It is especially critical for the healthcare system
responsible for the state of patients.
3.2 Medical Data Formats
Considering provenance of medical data, it is important to take into consideration
presentation of medical data.
Obviously, that text-based formats are the preferred choice for usage in semantic
system. As semantic origins imply that text information is prevalent in the system and
semantic annotations themselves are performed in text format, it is easy to combine
such representation with annotations in a meaningful. However, addition of the new