352
G. Chuiko et al.
establishes a basic framework to annotate data regardless of the source. However, the
pitfall of the generic ontologies is that they lack tools for the specific domain. In this
situation, it may cause not fully annotated data that notably limits possibilities for
inference engine as multiple parameters are omitted and not provided to the system
[19].
To provide further overview of the state-of-the-art technologies, first, we consider
dataflow of medical data in general healthcare system, then, discuss types of medical
data formats and, finally, provide our vision and possible solution to the problem of
medical data provenance.
3.1 Dataflow of Medical Data
Let us first describe the general semantic-enabled information system with medical
data. Perception layer is represented by devices that can capture physical parameters. They enlist devices that perform measurements in specific interval as well as
devices that constantly observe state of the human. We also differentiate mass market
wearable devices (e.g. smart watches with enriched functionality to monitor physical state) and specific medical sensors that is responsible for precise information
capturing. This factor is directly concerned with provenance of the medical data.
Medical data are submitted from multiple IoT-agents in concurrent manner. It makes
possibly to aggregate data in a single storage and investigate them to devise complex
dependencies between them.
At the edge level, devices support preliminary data processing and preparation
for further transmissions to the top layers. Edge devices also facilitate storage capabilities and can even be used for the initial data annotation. However, this point of
interest should be meticulously devised because it notably affects further flow of the
data. To be precise, this situation is concerned with situation when decision about
physical state is made on the early processing stage. While simple inference from the
sensor measurements might demonstrate no problems, top level processing engine
can generate the recommendation that contradicts to the previous one. It will result
in ambiguous state of the system.
On the other side, edge layer is suitable for diagnosis of sensor devices. Edgelocated computers are supposed to exchange diagnosis messages with sensors to
control their state.
In opposite to streaming applications where decisions are generated exclusively
regarding stream state with analysis of data on the interval of time, presence of the
storage layer is obligatory for the semantic-enabled system. Availability of the storage
layer provides an opportunity to improve semantic rule engine on the whole space
of available data. As this paper deals with semantic data, we assume that storage
is represented by sets of RDF-tuples with annotations added immediately before
the write operation. Apparently, it entails that only basic information is available
G. Chuiko et al.
establishes a basic framework to annotate data regardless of the source. However, the
pitfall of the generic ontologies is that they lack tools for the specific domain. In this
situation, it may cause not fully annotated data that notably limits possibilities for
inference engine as multiple parameters are omitted and not provided to the system
[19].
To provide further overview of the state-of-the-art technologies, first, we consider
dataflow of medical data in general healthcare system, then, discuss types of medical
data formats and, finally, provide our vision and possible solution to the problem of
medical data provenance.
3.1 Dataflow of Medical Data
Let us first describe the general semantic-enabled information system with medical
data. Perception layer is represented by devices that can capture physical parameters. They enlist devices that perform measurements in specific interval as well as
devices that constantly observe state of the human. We also differentiate mass market
wearable devices (e.g. smart watches with enriched functionality to monitor physical state) and specific medical sensors that is responsible for precise information
capturing. This factor is directly concerned with provenance of the medical data.
Medical data are submitted from multiple IoT-agents in concurrent manner. It makes
possibly to aggregate data in a single storage and investigate them to devise complex
dependencies between them.
At the edge level, devices support preliminary data processing and preparation
for further transmissions to the top layers. Edge devices also facilitate storage capabilities and can even be used for the initial data annotation. However, this point of
interest should be meticulously devised because it notably affects further flow of the
data. To be precise, this situation is concerned with situation when decision about
physical state is made on the early processing stage. While simple inference from the
sensor measurements might demonstrate no problems, top level processing engine
can generate the recommendation that contradicts to the previous one. It will result
in ambiguous state of the system.
On the other side, edge layer is suitable for diagnosis of sensor devices. Edgelocated computers are supposed to exchange diagnosis messages with sensors to
control their state.
In opposite to streaming applications where decisions are generated exclusively
regarding stream state with analysis of data on the interval of time, presence of the
storage layer is obligatory for the semantic-enabled system. Availability of the storage
layer provides an opportunity to improve semantic rule engine on the whole space
of available data. As this paper deals with semantic data, we assume that storage
is represented by sets of RDF-tuples with annotations added immediately before
the write operation. Apparently, it entails that only basic information is available