13 Smart Learning Using Big and Small Data for Mobile and IOT e-Health
609
to conduct ecological momentary assessment and further provide tailored feedback
from clinicians to patients, based on each person’s ongoing performance [23–26].
For example, by using mobile and wearable devices to record a user’s exercise
patterns, caloric intake and health status, the user is then provided with personalized
advices that help make sense of his own data and reflect on the impact of exercise
or food on himself. As another example, mobile and IOT e-health data are used
to personalize messages from clinicians [27, 28] and to support the tracking and
communication of outcomes between clinical visits for conditions such as obesity
[29], chronic obstructive pulmonary disease [30] and post-traumatic stress disorder
[31].
Beyond self-monitoring and clinician-patient communication, mobile and IOT
e-health technologies have also been studied for its benefits on delivering behavioral interventions to sustain healthy behavior change in a variety of self-care
settings. The applications widely range from behavior change theory-driven lifestyle
interventions for chronic disease management [32, 33] to self-regulation-based
psychotherapeutic interventions for the management of mental health [34], stress
[35] and substance use disorders [36].
13.1.1 Key Challenges in Smart Learning for Mobile and IOT
e-Health
The increasing accumulation of mobile and IOT e-health data has enabled an
immense potential of applying AI for smart learning. The recent emerging AI
technologies (including but not limited to the neural modeling and reinforcement
learning approaches) have offered promises to capture the patient-specific, dynamic
and intricate relationships through an end-to-end learning framework. The framework is flexible enough to allow for the incorporation of additional exogenous
patient data sources, ranging from patient-reported outcomes and contexts to sensor
measurements such as heart rate variability (HRV) and electroencephalogram (EEG)
signals.
Meanwhile, the detection of context-aware physiological and behavioral patterns
over time enables healthcare professionals to understand how a treatment plan
or care management program is working for a particular patient. More frequent
checkpoints between clinical visits allow for more pertinent analysis of how a
patient responds to a certain medication, treatment plan or engagement strategy. In
the cases where constant monitoring is feasible, real-time analysis can be provided
to care teams at any given moment. It will also allow for just-in-time intervention at
the time of adverse events.
One great domain wherein smart learning for mobile and IOT e-health can make
great impacts is to generate feedback and recommend behavioral interventions
based on both the small (personal-level) and big (population-level) data collected.
Two key challenges emerge. One is regarding how to generate adaptive strategies
609
to conduct ecological momentary assessment and further provide tailored feedback
from clinicians to patients, based on each person’s ongoing performance [23–26].
For example, by using mobile and wearable devices to record a user’s exercise
patterns, caloric intake and health status, the user is then provided with personalized
advices that help make sense of his own data and reflect on the impact of exercise
or food on himself. As another example, mobile and IOT e-health data are used
to personalize messages from clinicians [27, 28] and to support the tracking and
communication of outcomes between clinical visits for conditions such as obesity
[29], chronic obstructive pulmonary disease [30] and post-traumatic stress disorder
[31].
Beyond self-monitoring and clinician-patient communication, mobile and IOT
e-health technologies have also been studied for its benefits on delivering behavioral interventions to sustain healthy behavior change in a variety of self-care
settings. The applications widely range from behavior change theory-driven lifestyle
interventions for chronic disease management [32, 33] to self-regulation-based
psychotherapeutic interventions for the management of mental health [34], stress
[35] and substance use disorders [36].
13.1.1 Key Challenges in Smart Learning for Mobile and IOT
e-Health
The increasing accumulation of mobile and IOT e-health data has enabled an
immense potential of applying AI for smart learning. The recent emerging AI
technologies (including but not limited to the neural modeling and reinforcement
learning approaches) have offered promises to capture the patient-specific, dynamic
and intricate relationships through an end-to-end learning framework. The framework is flexible enough to allow for the incorporation of additional exogenous
patient data sources, ranging from patient-reported outcomes and contexts to sensor
measurements such as heart rate variability (HRV) and electroencephalogram (EEG)
signals.
Meanwhile, the detection of context-aware physiological and behavioral patterns
over time enables healthcare professionals to understand how a treatment plan
or care management program is working for a particular patient. More frequent
checkpoints between clinical visits allow for more pertinent analysis of how a
patient responds to a certain medication, treatment plan or engagement strategy. In
the cases where constant monitoring is feasible, real-time analysis can be provided
to care teams at any given moment. It will also allow for just-in-time intervention at
the time of adverse events.
One great domain wherein smart learning for mobile and IOT e-health can make
great impacts is to generate feedback and recommend behavioral interventions
based on both the small (personal-level) and big (population-level) data collected.
Two key challenges emerge. One is regarding how to generate adaptive strategies