Smart Healthcare IoT: Deep Learning-Driven Patient Monitoring and Diagnosis

Full text

2023 9th International Conference on Smart Structures and Systems (ICSSS)
979-8-3503-8420-8/23/$31.00 ©2023 IEEE
Nudurupati Jaya Viswadutt
Department of Electronics and
Communication Engineering
Amrita Vishwa Vidyapeetham, Amritapuri,
India
amenu4eac20049@am.students.amrita.edu
Dileep Kumar Vemula
Department of Electronics and
Communication Engineering
Amrita Vishwa Vidyapeetham, Amritapuri,
India
amenu4eac20075@am.students.amrita.edu
Mamidi Shardunya
Department of Electronics and
Communication Engineering
Amrita Vishwa Vidyapeetham, Amritapuri,
India
amenu4eac20063@am.students.amrita.edu
Narasimha Paleti
Department of Electronics and
Communication Engineering
Amrita Vishwa Vidyapeetham, Amritapuri,
India
amenu4eac20043@am.students.amrita.edu
Kalakunnath Namitha
Department of Computer Science and
Engineering, Amrita School of Computing,
Amrita Vishwa Vidyapeetham, Amritapuri,
India,
namithak@am.amrita.edu
Abstract—In this paper, we introduce an Internet of Things
(IoT)-based framework customized for use in home clinical
settings. This system is designed to remotely monitor patients and
detect potential health issues at an early stage. The system's three
primary sensors are the MAX30102 for oxygen saturation and
heart rate monitoring, the AD8232 ECG sensor module for
capturing ECG signal data, and the Non-contact infrared
thermometry (NCIT) for precise temperature readings.
Individuals' vital health data can be gathered by these devices in
unison. After collection, the Message Queuing Telemetry
Transport protocol is used to safely and effectively send the data
to a centralized server. The categorization of possible illnesses
and health abnormalities is performed on the server side using a
pre-trained deep learning model built on Artificial Neural
Networks (ANN). This model leverages the rich and diverse
dataset collected from the sensors to identify early warning signs
and detect emerging health issues. By combining IoT technology,
advanced sensor capabilities, and artificial intelligence (AI), this
system presents a comprehensive solution for proactive
healthcare management and early intervention. The Proposed
model shines with an impressive accuracy rate of 97.81%,
indicating
its exceptional ability to accurately forecast outcomes for nearly
98% of the cases.
Keywords—Patient Monitoring System.Internet of Things
(IoT), MQTT (Message Queuing Telemetry Transport), Sensors,
Artificial Neural Networks (ANN).
I. INTRODUCTION
There has been a dramatic shift in the healthcare business in
recent years, and it is now a major generator of income and
jobs. In the past, medical practitioners had to conduct invasive
physical examinations of patients to diagnose illness or
abnormality, which sometimes resulted in lengthy hospital
admissions. This method inflated healthcare expenses and put
pressure on healthcare infrastructure, especially in less
populated areas. To diagnose various disorders and monitor
health, nevertheless, the healthcare system has evolved away
from being hospital-centric to being patient-centric due to
technology improvements such as the use of small gadgets like
smartwatches [1]. In addition, an aging population and the rise
of chronic diseases provide formidable obstacles for the
healthcare sector. This has led to the rise of remote medical
surveillance as a potentially advantageous strategy for better
patient care, lower healthcare expenditures, and fewer in-
person hospital visits. Blood oxygen levels, heart rate, body
temperature, and electrocardiogram (ECG) signals are only a
few of the instances of the types of physiological data that
may be acquired remotely for the purpose of health
surveillance. The end goal is to alert medical staff
immediately to any emerging health concerns [2, 3].
The human body, however, cannot be easily connected to
the Internet in the same way that mechanical or digital devices
can. Adding a sensing and networking system allows a digital
device to communicate with other devices and the Internet.
However, even if we were to implant a sensor system in a
human being, we couldn't link it to the web. However, large
measuring equipment has a limitation in that it can only be
utilized in carefully monitored settings for limited amounts of
time [4]. Therefore, using currently available large-scale
sensing and measuring technology, it is difficult to link a
person's body to the Internet anywhere and at all times. IoT's
potential benefits in healthcare and security are severely
limited by the impossibility of connecting the human body to
the internet. The data gathered is sent to a distant server where
it is analyzed by deep learning algorithms for signs of health
problems. Convolutional neural networks (CNNs) and other
deep learning algorithms can automatically learn to assess
enormous volumes of data and extract meaningful
characteristics that are symptomatic of specific health
conditions [5] [6].
The discipline of remote patient monitoring (RPM) is
expanding fast, with the goal of providing doctors with more
Smart Healthcare IoT: Deep Learning-Driven
Patient Monitoring and Diagnosis
2023 9th International Conference on Smart Structures and Systems (ICSSS) | 979-8-3503-8420-8/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICSSS58085.2023.10407108
Authorized licensed use limited to: Amrita School Of Engineering - Kollam. Downloaded on March 06,2024 at 10:12:07 UTC from IEEE Xplore. Restrictions apply.

2023 9thInternational Conference on Smart Structures and Systems (ICSSS)
resources to care for patients in a variety of medical and
surgical wards in general hospitals through the use of flexible
materials for wearable sensors [7]. This is accomplished
through using cutting-edge IoT approaches in healthcare, such
as telehealth apps, wearable devices, and contact-based
sensors. Intelligent medical systems are depicted in Figure 1.
Vital signs and other physiological factors like motion
recognition are often measured using RPM to aid in clinical
decision-making and therapy planning for problems including
movement disorders and mental health issues.
IoT Cloud
Database
Smart Devices
Sleep
Blood Pressure
Heart Rate
Oxygen Level Steps
Medical Store
Ambulance
Doctors
Fig. 1 Smart Healthcare
AI has tremendous potential to improve healthcare delivery,
and researchers and practitioners are investigating its use
across a range of concrete domains, including disease risk
assessment, continuing patient care, and the mitigation or
reduction of problems associated with illness development. AI
is also useful in medicine since it speeds up processes like
genome sequencing and the creation of new therapies and
treatments by allowing researchers to glean insights from
complicated data that were previously impossible to gather or
observe. Machine learning is a branch of AI that has the
potential to speed up the process by which physicians can
analyze complicated data with the use of specialized
algorithms [8] [9]. They can identify the sorts of motion or
activities a patient engages in and help determine the
likelihood of a patient's health deteriorating prematurely.
These artificial intelligence systems are able to handle massive
information, allowing them to learn complicated patterns for
use in decision-making. ANN and deep learning algorithms
that can manage and improve exceedingly complicated
datasets have recently become even more powerful because of
advancements in computer performance. By adopting an IoT
paradigm with a consolidated control hub and user interface,
many mundane chores may be computerized automatically.
Patient safety may be improved if this helps reduce the
likelihood of human mistakes [10].
This study introduces a state-of-the-art IoT system
developed specifically for home clinical contexts, allowing for
remote medical surveillance and early identification of health
concerns. An extensive sensor array built within the system
allows for continuous monitoring of oxygen saturation, heart
rate, and signal data from an electrocardiogram (ECG) sensor
module. A non-contact infrared sensor also allows for accurate
monitoring of core body temperature.
II. RELATED STUDY
The medical industry is the one that will be most affected
by the IoT. Medical infrastructures based on the Internet of
Things will be assessed, expanded, and enhanced through this
study. The essay examines the pros and cons of using the
Internet of Things and edge computing in healthcare. Edge
computing, the authors argue, may enhance healthcare
delivery, affordability, and outcomes [11]. Online health
evaluation by doctors, nurses, and other healthcare
professionals allows for more effective monitoring of patients'
health and the provision of medical treatments from a distance
[12]. It is risky to travel to a medical facility during this
epidemic. Connecting to IoT smart-health infrastructure also
facilitates the use of wearable devices for health monitoring.
The existing hospitals may now deliver smart-health services.
Wearable devices collect data on a variety of health indicators
and transmit that information to caretakers and other relevant
parties. The potential for information theft or alteration during
such contact raises concerns about patients' safety. Patients
and their anxious guardians may get misleading information,
which may have an adverse effect on their mental health [13].
The proposed method aids in reducing the possibility of such
information theft by the simple implementation of hyper
ledger fabric block chain as well as the accompanying gadgets
aimed at smart health care surveillance. They employ a block-
chain consortium to encrypt communication for the dispersed
and mostly ad hoc deployment of IoT devices. Based on the
findings, we conclude that using block-chain hyper ledger to
address problems associated with the use of constrained IoT
devices in smart healthcare is a promising direction to pursue.
Many recent technical advances can be attributed to the
widespread availability of these tools [14]. The importance of
IoT and how it may be used in futuristic healthcare systems
have been discussed in this chapter. In recent years, smart
healthcare has emerged as a rapidly developing industry that
has great promise for the future. One of the most important
aspects of early diagnosis and treatment is constant monitoring
and prompt reporting of health concerns. To this end, experts
have created IoT-enabled smart healthcare aids. The cloud
makes it possible to remotely view and share sensitive medical
information. The smart Apps analyze the key signals and relay
the results to the smart devices. This completes the picture for
intelligent healthcare delivery. The most consequential result
of this study is a smart App that patients may use on their
smartphones to keep tabs on their health and anticipate its
trajectory in real-time.
III. METHODOLOGY
In the realm of scientific research and technological
advancement, the utilization of sensors is fundamental for data
collection, experimentation, and analysis. In this particular
study, the researchers harnessed the capabilities of three
distinct types of sensors: the Heart Rate Sensor, the ECG
sensor module, as well as the Infrared body Temperature
Sensor. Additionally, they employed a NodeMCU device to
facilitate data acquisition and transmission. This combination
of sensors and a microcontroller represents a comprehensive
approach to monitoring and collecting vital physiological and
Authorized licensed use limited to: Amrita School Of Engineering - Kollam. Downloaded on March 06,2024 at 10:12:07 UTC from IEEE Xplore. Restrictions apply.

2023 9thInternational Conference on Smart Structures and Systems (ICSSS)
environmental data. The details of each component are given
in the following subsections.
A. MAX30102 Sensor:
The MAX30102 sensor is a versatile and valuable
component in medical and health-related research. It is
primarily designed for pulse oximetry and heart-rate
monitoring. By shining a light on the skin and measuring how
much of it is absorbed, pulse oximetry may non-invasively
determine both the blood's oxygen saturation level (SpO2) and
the patient's heart rate. This sensor is commonly placed on a
person's fingertip or earlobe, where it can provide real-time
data about oxygen levels and heart rate.
B. AD8232 ECG Sensor Module
The AD8232 ECG sensor module specializes in
electrocardiogram (ECG) monitoring. Electrocardiograms are
used as a diagnostic tool because of the data they provide
about how the heart functions electrically throughout time. It
provides valuable insights into cardiac health by detecting
irregularities and abnormalities in heart rhythms.
C. Non-contact infrared thermometry (NCIT) Sensor
The NCIT sensor offers a non-invasive means of measuring
body temperature. It relies on the detection of infrared
radiation emitted by the body's surface to provide accurate
temperature readings without the need for direct contact.
D. NodeMCU
The NodeMCU device, based on the ESP8266 WiFi
module, plays a crucial role in collecting, processing, and
transmitting data from the aforementioned sensors. It is a
microcontroller equipped with WiFi capabilities, making it an
excellent choice for Internet of Things (IoT) applications.
E. MQTT (Message Queuing Telemetry Transport)
MQTT is a lightweight and efficient messaging standard
developed specifically for IoT gadgets. Its major goal is to let
devices with limited resources and bandwidth to communicate
in a reliable and low-latency manner. For the purposes of this
undertaking, sensor readings were gathered by the NodeMCU
device and relayed through the MQTT protocol to a server in
the cloud. As a MQTT client, the NodeMCU device
communicated with the distant server, which acted as the
intermediary between the two parties. Messages containing the
collected sensor data were sent from the NodeMCU device to
the MQTT broker, where they were filed under appropriate
topics. These messages were then picked up by the MQTT
broker and sent on to the customers who had subscribed to its
service. The subscribed client served as the information server
in this case, taking in sensor readings and storing them in
preparation for further processing.
F. Artificial Neural Network
The core concept of ANNs lies in their ability to process
information through layers of artificial neurons. In this
structure, the input layer receives data, which is then
propagated through one or more hidden layers, ultimately
leading to an output layer that produces a result or prediction.
Throughout training, the weights attached to each neuronal
connection are tweaked to improve the network's efficiency.
Every neuron's output is calculated using the activation
function inside it, which takes into account the input's weight.
Figure 2 depicts the overall layout of the proposedmethod.
A
D
C
Controller Cloud
Database
Machine learning
(ANN)
Prediction
Sensors
Oxygen Level
Tempearture
ECG Heart Rate
Fig. 2 The architecture of the ProposedFramework
The ability of ANNs to acquire knowledge from experience
is crucial. Through training on labeled datasets, ANNs adapt
their internal parameters, fine-tuning the weights to minimize
errors or discrepancies between their predictions and the
actual outcomes. This iterative learning process, often
facilitated by algorithms like backpropagation, enables ANNs
to recognize complex patterns and make accurate predictions
in various domains.
G. The Proposed Framework
The proposed framework requires the user to implant three
sensors into their person. Together, these sensors gather
extensive information about the user's health, including
electrocardiogram (ECG) signals, heart rate, blood oxygen
levels, and temperature. This information is sent from the
NodeMCU to an external server by means of the MQTT
protocol so that it may be analyzed. When the information
enters the central server, it is put across the deep learning
model suggested for categorization. In particular, five types of
abnormal cardiac rhythms are identified by carefully
examining the ECG signals as well as heart rate data: normal,
supraventricular premature beats, premature ventricular
contractions, fusion of ventricular beats, as well as
unclassifiable beats. At the same time, a thorough analysis of
the temperature readings is performed to check for fever. The
device also creates a report detailing the patient's temperature,
heart rate, and oxygen levels, and if they are within the ranges
considered normal. Beyond data analysis, the system assumes
the pivotal role of assessing the gravity of the detected
condition. In cases where the situation is deemed alarming, the
Authorized licensed use limited to: Amrita School Of Engineering - Kollam. Downloaded on March 06,2024 at 10:12:07 UTC from IEEE Xplore. Restrictions apply.

2023 9thInternational Conference on Smart Structures and Systems (ICSSS)
model initiates contact with the nearest medical practitioner or
doctor to facilitate further diagnosis and immediate
intervention and treatment. In essence, this holistic framework
amalgamates sensor technology, deep learning algorithms, and
telemedicine capabilities to enable timely and potentially life-
saving responses to health-related concerns.
IV. RESULTS AND DISCUSSIONS
In the context of this study, the combination of these
sensors and the NodeMCU device suggests a comprehensive
approach to data collection and monitoring. The goal of the
study was probably to gather and evaluate physiological
information including heart rate, blood oxygen levels, and
body temperature in a way that did not need intrusive bodily
contact with the subjects. This setup has several potential
applications:
The study focused on monitoring the health status of
individuals, particularly those with specific medical conditions
or in high-risk environments. For instance, the MAX30102
could be used to monitor oxygen saturation and heart rate in
patients with respiratory diseases, while the AD8232 sensor
could track their ECG signals for cardiac health assessment.
The NCIT sensor might have been employed for continuous
body temperature monitoring, which is crucial in detecting
infections or fever. In the era of telemedicine and remote
healthcare, the combination of these sensors and NodeMCU
could enable healthcare providers to remotely monitor
patients' vital signs in real-time, allowing for early
intervention in case of anomalies.
The data collected from these sensors can be valuable for
various research purposes. In particular patient populations,
investigators may utilize the information to examine the
relationship between core body temperature, heart rate, and
oxygen saturation, or to assess the efficacy of treatments.
Beyond healthcare, this sensor setup could have applications
in smart homes, where individuals can monitor their health
and wellness using connected devices.
Evaluation Metrics
When assessing the system's performance, we utilized a set
of standard evaluation metrics to measure its effectiveness.
The F1-score, along with accuracy, precision, and recall, are
all useful measures. The accuracy of a classification system is
measured as the fraction of a dataset's samples that were
properly labeled. It's determined by dividing the total number
of correct diagnoses (TP + TN) by the sum of the incorrect
diagnoses (FP + TN + FN). On the other hand, precision is an
indicator of how well a model predicts true positives relative
to the overall amount of samples. Comparatively, recall
calculates how many TP samples there are relative to how
many positive samples there are in the dataset as a whole. The
F1-score is calculated to evaluate the model's overall efficacy;
it takes into account both accuracy and reliability.
TABLE 1. PERFORMANCE EVALUATION OF VARIOUS CATEGORIES
Types Accuracy Precision Recall
Normal Beat 0.95 0.94 0.96
Premature Beat 0.98 0.98 0.97
Ventricular Beat 0.96 0.94 0.92
Unclassifiable Beat 0.97 0.96 0.97
Table 1 provides an in-depth analysis of the efficacy of a
system of categorization in identifying four distinct types of
heartbeats (normal, premature, ventricular, and unclassifiable)
utilizing the most important assessment criteria. These
measures are essential for determining how well the model can
identify and categorize different types of heartbeats.
In the case of Normal Beat, the model exhibits an
impressive accuracy of 95%, signifying that 95% of the
instances of normal heartbeats were correctly identified.
Additionally, the precision and recall values of 0.94 and 0.96,
correspondingly, indicate that the method has a high degree of
precision in classifying normal heartbeats, minimizing false
positives, and a commendable recall, effectively capturing
most of the actual normal heartbeats.
Moving to the Premature Beat category, the model's
performance is even more impressive, with an accuracy of
98%. What this means is that the model was able to detect
98% of arrhythmias before they became life-threatening. A
recall of 0.97 indicates that the algorithm used is able to
efficiently catch the vast majority of true premature heartbeats,
suggesting that the algorithm's prediction of a premature beat
is accurate 98% of the time.
In the Ventricular Beat category, the model exhibits a
slightly lower accuracy of 96%, indicating that it correctly
identified 96% of ventricular beats. The precision and recall
values of 0.94 and 0.92, respectively, reflect a balanced
performance, with a reasonably high precision in classifying
ventricular beats while maintaining a satisfactory recall
rate.Finally, a 97% accuracy in the Unclassifiable Beat
category demonstrates that the program can reliably recognize
a large subset of beats that defy easy categorization. The
model's 0.96 accuracy in categorizing unclassifiable beats and
0.97 recall make it ideal for assuring a few false positives. The
evaluation metrics for each heartbeat category provide
valuable insights into the model's effectiveness. The high
accuracy, precision, and recall values indicate that the model
is proficient in distinguishing between different heartbeat
types, thus holding significant promise for applications in
Authorized licensed use limited to: Amrita School Of Engineering - Kollam. Downloaded on March 06,2024 at 10:12:07 UTC from IEEE Xplore. Restrictions apply.

2023 9thInternational Conference on Smart Structures and Systems (ICSSS)
cardiac health monitoring and diagnosis.The confusion matrix
is depicted in Figure 3.
Fig. 3 Confusion Matrix of Proposed Model
The suggested system relies heavily on heart rate as a vital
metric for evaluating the cardiovascular health of the
monitored individual. Heart rate data is segmented by age,
allowing for consideration of age-specific normal heart rate
ranges. For instance, a newborn's heart rate can be anywhere
from 70 to 190 beats per minute (bpm) during their first
month, and it can be anywhere from 80 to 160 bpm from their
second month onwards. Heart rates tend to drop with age, with
adults and the elderly often keeping theirs between 60 and 100
beats per minute (bpm). Our method can identify abnormal
heart rates by comparing them to the usual range for a certain
age group, and then immediately alerting medical staff so that
the patient may be evaluated and, if necessary, treated.
We also use a simple thresholding approach with a
predetermined threshold value of 100.4°F to classify instances
as fever or non-fever. The method entails sorting patients into
two groups, those with fever and those without. Using
predetermined temperature cutoffs as a diagnostic tool is
standard practice in the medical field. Although basic
thresholding has its uses, it may not be suitable in all cases;
more sophisticated approaches may be necessary.
TABLE 2. PERFORMANCE COMPARISON OF VARIOUS MODELS
Models Accuracy Precision Sensitivity Specificity
CNN 85.45 86.59 85.51 87.65
MobileNet 88.59 84.56 90.56 89.56
GoogleNet 89.96 90.78 91.36 91.82
Inception 90.56 97.56 88.15 89.56
ResNet 88.62 89.65 90.56 90.89
Proposed 97.81 96.59 97.28 97.06
In general, jobs where false positives are expensive to
create are better suited to models with higher precision, which
includes the "Inception" model's 97.56%. Sensitivity, often
called recall, is a measure of how many correct positive
predictions there were. The "GoogleNet" model has a high
sensitivity of 91.36%, indicating that it is good at detecting
true positives. The level of specificity measures how many
times accurate negative predictions were made. When it comes
to successfully recognizing negative occurrences, an indicator
having high specificity, like "GoogleNet" with 91.82%,
shines. Overall, the "Proposed" model performs the best when
categorizing information, as seen by its high levels of
accuracy and sensitivity. Nevertheless, as various models tend
to place varying emphasis on various elements of
effectiveness, selecting the best one for a specific project is
very context-dependent. Our system also maintains a vigilant
monitoring of patients' oxygen saturation levels, which
typically fall within the range of 95 to 100 percent. Conditions
such as lung illness, heart disease, anemia, and carbon
monoxide poisoning can all be indicated by readings outside
of this range. However, excessive amounts of oxygen can have
the opposite effect and damage cells, causing inflammation.
Since keeping an eye on oxygen levels is so important, our
device gives continuous feedback to doctors in real-time.
V. CONCLUSION AND FUTURE SCOPE
In conclusion, the proposed system presents a
comprehensive and innovative approach to healthcare
monitoring, leveraging a combination of sensor data, age-
specific metrics, and advanced artificial intelligence
techniques. By meticulously tracking key physiological
parameters, including heart rate, body temperature, and
oxygen levels, our system provides a holistic assessment of an
individual's health status in real-time. The utilization of age-
based normal ranges for heart rate offers a personalized and
nuanced perspective on cardiovascular health, allowing us to
swiftly detect any deviations from the expected values.
Additionally, our fever detection method, while
straightforward, provides a reliable means of identifying
elevated body temperatures, a common indicator of illness.
The monitoring of oxygen saturation levels further enhances
our system's diagnostic capabilities, enabling the early
detection of conditions that affect respiratory and circulatory
health. Real-time updates and alerts to healthcare providers
ensure timely intervention when abnormal conditions arise.
The integration of advanced machine learning techniques,
includingANN with attention layers, for ECG analysis allows
Authorized licensed use limited to: Amrita School Of Engineering - Kollam. Downloaded on March 06,2024 at 10:12:07 UTC from IEEE Xplore. Restrictions apply.

2023 9thInternational Conference on Smart Structures and Systems (ICSSS)
us to accurately classify complex heart conditions, enhancing
our diagnostic precision.
The exploration of this domain offers numerous promising
research directions that can advance the capabilities of
healthcare monitoring systems like the one proposed.
Expanding the range of integrated sensors can significantly
enhance the depth and accuracy of health monitoring. A
person's health state can be supplemented with useful
information from integrated sensors that assess factors like
blood pressure, glucose levels, and respiration rates. We can
have a better understanding of an individual's health as a
whole if we smoothly incorporate these sensors into the
existing framework.
REFERENCES
[1] R. A. Dar, et al., "IOT Based Smart Wheelchair for Elderly Healthcare
Monitoring," 2023.
[2] M. Hooda et al., "Need Assessment and Design of an IoT based Healthcare
Solution through Participatory Approaches for a Rural Village in Bihar,
India," 2021 IEEE R10-HTC, Bangalore, India, 2021, pp. 1-6, doi:
10.1109/R10-HTC53172.2021.9641697.
[3] K. Lokesh, S. Srivastava, M. P. Kumar, S. Arun, S. Padmapriya and R.
Krishnamoorthy, "Detection of Stomach Cancer Using Deep Neural
Network in Healthcare Sector," 2021 3rd International Conference on
Advances in Computing, Communication Control and Networking
(ICAC3N), 2021, pp. 521-526, doi:
10.1109/ICAC3N53548.2021.9725656.
[4] V. Bhatt, et al., "Real-time healthcare monitoring using smart systems: A
step towards healthcare service orchestration Smart systems for futuristic
healthcare," 2021.
[5] A. K. Maurya, K. Lokesh, Sandeep, R. Kumar, S. Arun and R.
Krishnamoorthy, "Deep Neuro-Fuzzy Logic Technique for Brain
Meningiomasa Prediction," 2022 7th International Conference on
Communication and Electronics Systems (ICCES), 2022, pp. 1244-1248,
doi: 10.1109/ICCES54183.2022.9836008.
[6] S. Vishnu et al., Chapter 11 - Simple-Link Sensor Network-Based Remote
Monitoring of Multiple Patients, Editor(s): D. Jude Hemanth, Deepak
Gupta, Valentina Emilia Balas, In Intelligent Data-Centric Systems,
Intelligent Data Analysis for Biomedical Applications, Academic Press,
2019, Pages 237-252, ISBN 9780128155530
[7] A. Juneja, et al., "Smart Healthcare Ecosystems backed by IoT and
Connected Biomedical Technologies," 2022.
[8] R. Hegde, et al., "Survey on Development of Smart Healthcare Monitoring
System in IoT Environment," 2021.
[9] G. N. Vivekananda, A. R. H. Ali, S. Arun, P. Mishra, R. Sengar and R.
Krishnamoorthy, "Cloud Based Effective Health Care Management
System With Artificial Intelligence," 2022 IEEE 7th International
conference for Convergence in Technology (I2CT), 2022, pp. 1-6, doi:
10.1109/I2CT54291.2022.9825457.
[10] S. Nayar et al., "Deep Learning Based Model for Multi-class Classification
of Cervical Cells Using Pap Smear Images," 2022 IEEE I2CT, Mumbai,
India, 2022, pp. 1-6, doi: 10.1109/I2CT54291.2022.9825187.
[11] R. Punugoti, et al., "Exploring the Impact of Edge Intelligence and IoT on
Healthcare: A Comprehensive Survey," 2023.
[12] A. K, R. Anand et al., S., "IoT based Smart Healthcare System to Detect
and Alert Covid Symptom," 2021 ICCES, Coimbatre, India, 2021, pp.
685-692, doi: 10.1109/ICCES51350.2021.9488952.
[13] N. Banerji, et al., "Block-chain enabled secure smart IoT healthcare
monitoring for COVID-19 patients," 2022.
[14] V. Nalla et al., "Influence of Convolutional Neural Network Depth on the
Efficacy of Automated Breast Cancer Screening Systems," 2023.
[15] M. V. S. Girish, et al., "IoT Enabled Smart Healthcare Assistance for Early
Prediction of Health Abnormality," 2021.
Authorized licensed use limited to: Amrita School Of Engineering - Kollam. Downloaded on March 06,2024 at 10:12:07 UTC from IEEE Xplore. Restrictions apply.