9
[222] presented a solution for monitoring high-risk Maternal D. Environmental Monitoring
and Fetal Health (MFH) using IoT sensors, data analysis
1) Background of Environmental Monitoring with IoT:
for feature extraction, and a Deep Convolutional Genera-
IoT technologies have been widely used for monitoring envi-
tive Adversarial Network (DCGAN) classifier. It continuously
ronmental conditions such as air temperature, large-scale sea
monitors clinical indicators like heart rate, oxygen saturation,
surface temperature, soil moisture, air quality, etc. According
blood pressure, and uterine tonus, and the proposed system
to the nature of the sensor platform, real-time environmental
effectively classifies MFH status into more than four possible
observationdataisusuallycollectedbyvarioussensorsinclud-
outcomes,showingthatIoT-basedmobilemonitoringofMFH
ing satellite-based sensors, airborne sensors, temporal in situ
for pregnancy care is practical.
sensors,andlong-termsensorsinstalledatmonitoringstations.
AGI’spotentialextendstowearablehealthdevices,enabling The data collected from various environmental monitoring
real-time processing and analysis of data to detect anomalies methodshavetheirownadvantagesanddisadvantagesinterms
in vital signs. The capacity for on-device data processing of spatial resolution, spatial coverage, temporal resolution,
ensures timely alerts and recommendations. Wearable devices and temporal span. For example, accurate and real-time soil
endowedwithAGIcapabilitiesexcelindetectingirregularities moisture (SM) estimates are useful to characterize trends in
in patients’ essential indicators, such as heart rate, blood the global and local climate systems, and for predicting the
pressure, and oxygen saturation, facilitating prompt interven- interactionsbetweenlandandatmosphere[232].SoilMoisture
tions [223]. Kumar et al. [224] developed a Smart Healthcare Active Passive (SMAP)2 is an Earth satellite mission that
System (SHS) by integrating the IoT with AGI. Millions of measures and maps Earth’s soil moisture by the National
devices and sensors capture data for continuous patient health Aeronautics and Space Administration (NASA). SMAP data
monitoring. This data is analyzed using machine learning and canprovideglobal-scalesoilmoistureradiometerdatabutwith
deep learning algorithms to predict disease severity, and the a rather lower spatial resolution (36 km) and lower temporal
insights are wirelessly shared with medical professionals for resolution(every2-3days).Incontrast,highspatial-resolution
appropriate recommendations. soil moisture data can also be collected from airborne sensors
(∼800 m),long-termin situsensors(3- 5km),andtemporal
Telemedicine and telesurgery platforms equipped with AGI
in situ sensors (∼ 5 − 10 cm) such as soil moisture data
offer personalized, real-time feedback and support to patients.
measured during the joint NASA-United States Department
Virtual health assistants powered by AGI can interact with
ofAgriculture(USDA)soilmoisturevalidationcampaignsfor
patients in natural language, addressing queries, guiding them
SMAP airborne scale (∼ 800 m), sub-pixel scale (3 - 5 km),
through treatment plans, and providing information and rec-
and point scale (∼ 5 − 10 cm) such as soil moisture data
ommendations [225]. By integrating AGI, telemedicine plat-
measured during the joint NASA-United States Department
forms can also improve patient engagement and adherence
of Agriculture (USDA) soil moisture validation campaigns
to treatment plans [226]. Meanwhile, LLMs, a rising trend
for SMAP [233], [234]. However, these observations can
in AGI, will revolutionize how patients and clinicians access
only be collected by request with small spatial coverage
and obtain information. It is crucial for telehealth clinicians
(e.g., airborne data, temporal in situ sensor data), or have
to understand LLMs and recognize their potential and lim-
rather sparse spatial distribution (e.g., data from long-term in
itations [227]. With a telesurgery system powered by AGI
situ monitoring stations). In order to provide real-time high-
medical robotic systems, like the da Vinci Surgical System
resolutionsoilmoistureobservationsoveralargespatialscale,
from Intuitive Surgical (Sunnyvale, CA, USA), more complex
the best approach is to integrate observation data collected
surgeries can be performed remotely to reduce the imbalance
from various sensors.
in medical resources across geographical areas. AGI methods
2) Foundation Models for Earth and Environmental Moni-
like GANs are also pivotal in filling knowledge voids and
toring: Recently,significanteffortshavebeenmadetodevelop
speeding up the incorporation of telemedicine and telesurgery
FMs for climate and weather forecasting based on various
into clinical practice [228], especially for sim2real transfer
environmental observation data. For example, ClimaX [235]
learning to bridge the domain gap between simulated and real
is a recently developed FM for weather and climate science
data in the development of data-driven models for medical
whicharetrainedusingheterogeneousclimate,environmental,
segmentation and detection tasks that require human labeling
and earth observation data including 6 atmospheric variables
[229], [230].
at 7 vertical levels, 3 surface variables, and 3 constant fields.
AGI-integrated smart pill dispensers manage medication ClimaX shows promising performance on various weather
for patients effectively. By learning from patient medication global/regional forecasting, sub-seasonal to seasonal predic-
adherencepatterns,AGIcanalertpatientsortheircaregiversin tion,climateprojection,andclimatemodeldownscalingtasks.
caseofmisseddosesandpredictpotentialhealthissuesdueto In addition, IBM recently released its newest geospatial
non-compliance. Johnson et al. [231] explored the application foundation model on the open-source AI platform Hugging
of AGI in medication management, highlighting its potential Face 3. This geospatial FM was first pre-trained on NASA’s
to improve medication adherence. Harmonized Landsat Sentinel-2 satellite data (HLS) over one
WithanyfutureadvancementsinAGI,thehealthcaresector
2https://smap.jpl.nasa.gov/data/
is poised for significant transformations that will benefit both
3https://newsroom.ibm.com/2023-08-03-IBM-and-NASA-Open-Source-
patients and healthcare providers. Largest-Geospatial-AI-Foundation-Model-on-Hugging-Face