11 Healthcare IoT
527
medical data, vital signs, and bio-signals across a wide geographical area in
the presence of various environmental conditions and store and process them
in different data centers or the cloud. Instead, moving intelligence to the edge is
a promising approach to eliminate latency and evolve IoT e-Health solutions. In
this approach, an edge node with localized processing capability enables us to
respond more quickly than the cloud by making time-sensitive decisions closer
to the source of data. Thereby, this solution results in a more efficient solution
that can better handle low-latency demands of e-Health applications.
• Traffic reduction on overburdened networks: Considering the limited network
bandwidth, it is not practical and in certain use cases even not necessary to
transfer enormous volume of big raw data from millions of e-Health devices
to the cloud. Edge computing reduces the data transport costs, which can be
significant for data-intensive applications, such as genomic-association analysis,
generating several GBs of raw time-series data within a day. In this regard, edge
nodes can process, filter, and compact the medical data before delivering it to the
cloud to dramatically minimize bandwidth requirements.
Other important tasks of a fog node are explained below:
• Two-way connectivity: Fog nodes establish a secure reliable bidirectional data
flowing between e-Health devices and the cloud platform. An edge node gathers
feeds in real time from health devices using an appropriate protocol and,
after processing, sends the corresponding summary periodically to the cloud to
facilitate the long-term data sets aggregation, exploration, analysis, and globally
intelligent decision making. On the other hand, it might also need to receive
commands, configuration data, etc., from the cloud. Note that the edge node is
also dealing with the compliance challenges associated with connectivity such
as protocol translation, security, switching, routing, and networking analytics.
For example, nodes might not be assigned with a public IP address. Therefore,
to enable reachability from the cloud, an edge node can rely on different
mechanisms such as WebSocket, MQTT (message queue telemetry transport),
and IP tunneling.
• Time-series data capture: Edge nodes can either use an interrupt or a polling
mechanism for data acquisition. Depending on the application, time precision
might also be required to be able to extract the trend over time. In such a case,
accuracy is increased if the time stamp is generated in close proximity to the
e-Health device generating the data. In this regard, edge nodes time-stamp the
incoming data and store it in a historian database.
• Transient data storage: Fog nodes are required to provide short-time historical
storage for e-Health device data. For example, filtering outliers of data (i.e., in
the case of deviation from normal) depend on previous samples of the data.
• Device management: This includes device discovery, device registration, and
device control.
• Edge processing: A rich set of applications can be executed on the edge node. For
example, edge nodes are capable of on-demand data cleaning, data normalization, filtering, data reducing, compressing, integrity check, and formatting, data
527
medical data, vital signs, and bio-signals across a wide geographical area in
the presence of various environmental conditions and store and process them
in different data centers or the cloud. Instead, moving intelligence to the edge is
a promising approach to eliminate latency and evolve IoT e-Health solutions. In
this approach, an edge node with localized processing capability enables us to
respond more quickly than the cloud by making time-sensitive decisions closer
to the source of data. Thereby, this solution results in a more efficient solution
that can better handle low-latency demands of e-Health applications.
• Traffic reduction on overburdened networks: Considering the limited network
bandwidth, it is not practical and in certain use cases even not necessary to
transfer enormous volume of big raw data from millions of e-Health devices
to the cloud. Edge computing reduces the data transport costs, which can be
significant for data-intensive applications, such as genomic-association analysis,
generating several GBs of raw time-series data within a day. In this regard, edge
nodes can process, filter, and compact the medical data before delivering it to the
cloud to dramatically minimize bandwidth requirements.
Other important tasks of a fog node are explained below:
• Two-way connectivity: Fog nodes establish a secure reliable bidirectional data
flowing between e-Health devices and the cloud platform. An edge node gathers
feeds in real time from health devices using an appropriate protocol and,
after processing, sends the corresponding summary periodically to the cloud to
facilitate the long-term data sets aggregation, exploration, analysis, and globally
intelligent decision making. On the other hand, it might also need to receive
commands, configuration data, etc., from the cloud. Note that the edge node is
also dealing with the compliance challenges associated with connectivity such
as protocol translation, security, switching, routing, and networking analytics.
For example, nodes might not be assigned with a public IP address. Therefore,
to enable reachability from the cloud, an edge node can rely on different
mechanisms such as WebSocket, MQTT (message queue telemetry transport),
and IP tunneling.
• Time-series data capture: Edge nodes can either use an interrupt or a polling
mechanism for data acquisition. Depending on the application, time precision
might also be required to be able to extract the trend over time. In such a case,
accuracy is increased if the time stamp is generated in close proximity to the
e-Health device generating the data. In this regard, edge nodes time-stamp the
incoming data and store it in a historian database.
• Transient data storage: Fog nodes are required to provide short-time historical
storage for e-Health device data. For example, filtering outliers of data (i.e., in
the case of deviation from normal) depend on previous samples of the data.
• Device management: This includes device discovery, device registration, and
device control.
• Edge processing: A rich set of applications can be executed on the edge node. For
example, edge nodes are capable of on-demand data cleaning, data normalization, filtering, data reducing, compressing, integrity check, and formatting, data