Proceedings of the Fifth International Conference on Inventive Computation Technologies (ICICT 2022)

IEEE Xplore Part Number: CFP22F70-ART; ISBN:978-1-6654-0837-0

IoT Sensors Empowered with Deep Learning

for Brain Depletion Recognition
Dr.G.Revathya , Dr. Durga Karthikb , Dr.B.Sreedevic,
2022 International Conference on Inventive Computation Technologies (ICICT) | 978-1-6654-0837-0/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICICT54344.2022.9850582

1 Assistant Professor, School of Computing, SASTRA Deemed University, Thirumalaisamudram.

b Assistant Professor, Department of Computer Science and Engineering, SRC, SASTRA Deemed University, Kumbakonam.
C Senior Assistant Professor, Department of Computer Science and Engineering, SRC, SASTRA Deemed University, Kumbakonam.
revathyjayabaskar@gmail.com , durgakarthik@src.sastra.edu, sreedevi@src.sastra.edu

ABS TRACT and intra-cerebral hemorrhages are all terms

used to describe blood clots in the brain.
In recent years, Internet of Things enabling Approximately 13% of all strokes are caused by
applications, which have provided excellent answers them. Brain swelling develops when blood from
to a variety of challenges. This fast-growing industry
a severe injury irritates brain tissues. Cerebral
is led by wireless sensor networks. S mart medical
devices and wearables, for example, play an edema is the medical term for this. A hematoma
important part in the Internet of Things, as they may is a blood clot that has accumulated in a mass.
collect a variety of longitudinal patient-generated These situations increase the strain on
health data while also presenting preliminary adjacent brain tissue by limiting the critical blood
diagnosis options. As part of their efforts to serve flow and resulting in the death of brain cells.
patients with IoT-based solutions, experts apply ml to Internal bleeding, bleeding between the brain and
give effective resolutions in bleeding detection. This
its surrounding membranes, bleeding between
work describes a smart IoT-based solution for human
the layers of the brain's covering, and bleeding
brain hemorrhage diagnostics that uses deep learning
algorithms to reduce death rates and provide correct between the skull and the brain's covering are all
treatment recommendations. The S VM and possibilities. Trauma to the scalp, skull, or brain
Recurrent Neural Network were used to classify the results in a head injury. Traumatic brain injury,
images from the computed tomography scans for the or TBI, occurs when the brain is harmed.
intracranial dataset. When compared to prior A clot that forms under the skull in the
techniques such as naive bayes, KNN, and K-medoids, brain is called an intracranial hematoma (ICH).
the classification results for the S VM and Recurrent Hematomas in the brain are defined by where
neural network are high. According to the findings,
they start and range in intensity from moderate
the recurrent neural network beats other methods for
identifying intracranial images. The output of the to severe. This work looks at how IoT sensors
classification tool offers information on the type of and deep learning algorithms are being used to
brain hemorrhage, which helps to validate an expert's predict ICH in humans and save lives. Many
diagnosis and is utilized as a learning tool for trainee people are hesitant to go to the hospital because
radiologists to eliminate errors in existing systems. they are terrified of catching one of these
diseases as a result of the high pandemic
Keywords: IoT, wireless sensor networks, machine condition that occurred during Covid'19. Based
learning algorithms, support vector machines, on naive bayes classification and recurrent
recurrent neural networks
neural networks, this research provides an IoT-
INTRODUCTION based model for brain hemorrhage diagnosis.

The Internet of Things (IoT) is a Introduction to Internet of Things

network of physical devices or people that use
software, electronics, networks, and sensors to Health monitoring is one of the most
collect and exchange data. IoT promises to common scientific applications of wearable
connect ordinary activities such as toasters to the electronics. It can be used to monitor and alert
internet, in addition to traditional devices such as the practitioner remotely in the event of an
computers, smartphones, and tablets. Wearable emergency. There are a variety of wearable
diabetes monitoring examples of IOT. A monitoring devices available in the market.
hemorrhage in the brain is known as a stroke. SenseWear, SleePic, Heally, M1, Zeo, iBrain,
When a cerebral artery bursts, the surrounding and other wearable sleep monitoring devices are
tissues hemorrhage. Brain cells die as a result of the recent examples. Brainwaves, muscular
the hemorrhage. "Blood" is a Greek term. "Blood activity, and eye movement can all be tracked
rushing forth" is what hemorrhage means. with a wearable gadget called Zeo, which looks
Cerebral hemorrhages, intracranial hemorrhages, like an elastic band. A modest EEG brainwave

978-1-6654-0837-0/22/$31.00 ©2022 IEEE 1309

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on November 12,2022 at 06:02:27 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fifth International Conference on Inventive Computation Technologies (ICICT 2022)
IEEE Xplore Part Number: CFP22F70-ART; ISBN:978-1-6654-0837-0

monitoring forms the basis of the proposed conditions are routinely evaluated, and values
device. are checked on a regular basis.

Introduction to SVM

Support Vector Machines (SVM) are

algorithms that map engrossments to complex
dimensional feature spaces. SVM is a machine
learning technique that uses a hyperplane to
eliminate the characteristic space, allowing it to
utilize the margin between the occurrences of
distinct classes. SVM classifier creates a model
to predict different classes for new samples
based on a collection of features and labels. It
gives one of the classes with fresh
specimen/facts points. A binary SVM classifier
is one that has only two classes. SVM classifiers
are divided into two categories: Non-linear SVM
classifier and linear SVM classifier

RECURRENT NEURAL NETWORKS

RNN can now act in a time- dependent

manner. RNNs are based on feed forward neural
networks, which employ their internal state to
deal with input sequences of varying length
(memory). As a result, activities like
unsegmented, linked handwriting recognition
and speech recognition can be accomplished. To
simply describe, SVM claims the existence of
one feature in a class, which has no influence on
the presence of other characteristics in the same
class. The SVM model is simple to construct and
performs well on huge datasets. SVM is
recognized to outperform even the most complex
classification methods due to its simplicity.

DATASET

PSG polysomnography (ECG& EEG),

blood pressure, and air flow are all recorded
together with the date and time.

PROPOSED SYSTEM

Blood flow, blood pressure, and air

flow detecting sensors are all connected to an
Arduino board with a Wi-Fi module, and the
information is fed into a SVM model. Then the
SVM divides data obtained from the IoT sensor
into two groups: safe and non-safe. Individuals
who are in the Non-Safe zone, intracranial
images are obtained and relayed to recurrent
neural networks. Patients, who are in the danger
of bleeding are identified by using recurrent
neural networks, and a notification will be sent
to the doctor or family members, who take
appropriate action. Non-hemorrhage patients'

978-1-6654-0837-0/22/$31.00 ©2022 IEEE 1310

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on November 12,2022 at 06:02:27 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fifth International Conference on Inventive Computation Technologies (ICICT 2022)
IEEE Xplore Part Number: CFP22F70-ART; ISBN:978-1-6654-0837-0

BLOCK DIAGRAM

IOT DEVICES

The values are recorded and passed

SUPPORT VECTOR
MACHINES

Non safe zone values are passed.

RECURRENT NEURAL NETWORKS

Alert to Family and Doctor for Further

Treatment

Figure 1 Block diagram of the proposed system

IMPLEMENTATION devices. Below is a diagram of the circuit with

sample sensors. The sensors are wired together,
The Arduino board is used in and a sample output is momentarily captured.
conjunction with a GSM module to develop IoT

Figure 2 Circuit Diagram

978-1-6654-0837-0/22/$31.00 ©2022 IEEE 1311

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on November 12,2022 at 06:02:27 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fifth International Conference on Inventive Computation Technologies (ICICT 2022)
IEEE Xplore Part Number: CFP22F70-ART; ISBN:978-1-6654-0837-0

The output is passed to Support Vector Machine (SVM) and the execution is performed using Orange tool.

Figure 3 Connections of SVM for Prediction

Figure 4 Test Result of SVM

SVM shows 99% accuracy.

Figure 5 The values are classified into Safe and Non-Safe zone

SVM splits the complete dataset into two verified for intracranial images and the same are
categories one is safe zone dataset which is passed to recurrent neural networks for further
indicated blue in color and another one is non safe investigation. For the other values the process is
zone dataset, which is indicted yellow in color. continuously repeated.
The values that contain non safe zones are

978-1-6654-0837-0/22/$31.00 ©2022 IEEE 1312

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on November 12,2022 at 06:02:27 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fifth International Conference on Inventive Computation Technologies (ICICT 2022)
IEEE Xplore Part Number: CFP22F70-ART; ISBN:978-1-6654-0837-0

Figure 6 Neural Network implication with intracranial images

Figure 7 Neural net implication with hidden layer

Recurrent Neural networks predicts the Doctor. The overall accuracy of recurrent neural
hemorrhage patient and through Wi-Fi module networks in 99%.
Arduino board it informs to the Patient family and

Figure 8 SVM and RNN comparison with all other Algorithms

SVM combined with RNN shows higher values when compared each time with all other deep

978-1-6654-0837-0/22/$31.00 ©2022 IEEE 1313

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on November 12,2022 at 06:02:27 UTC from IEEE Xplore. Restrictions apply.
 Proceedings of the Fifth International Conference on Inventive Computation Technologies (ICICT 2022)
IEEE Xplore Part Number: CFP22F70-ART; ISBN:978-1-6654-0837-0

learning algorithms. Hence the result will be patients with COVID-19. Kidney Int. 2020
May;97(5):829–38.
more accurate than any other algorithms.
[2] Klok FA, Kruip MJHA, van der Meer NJM, Arbous MS,
Gommers D, Kant KM, et al. Confirmation of high
CONCLUSION cumulative incidence of thrombotic complications in
critically ill ICU patients with COVID-19: an updated
analysis. T hromb Res. 2020;191:148–50.
It is incredibly impossible to admit every patient
[3] Martínez-Martínez, José M., et al. "Prediction of the
to a hospital during a pandemic situation like hemoglobin level in hemodialysis patients using machine
Covid-19 to examine their medical condition. learning techniques." Computer methods and programs in
Patients who require treatment or in an biomedicine 117.2 (2014): 208-217.
[4]Mrs.G.Revathy and Dr.K.Selvakumar,”Increasing quality of
emergency situation might be admitted and services in wireless mesh networks”, International journal
treated at the hospital. Furthermore, the total cost of advanced research in computer engineering and
of the components is relatively modest (about technology, vol 7, issue 3,march 2018. ISSN 22781323
350 rupees) as they may quickly adapt to the [15]Mrs.G.Revathy and Dr.K.Selvakumar, “Escalating
quality of services with channel assignment andtraffic
proposed method and save money on MRI and scheduling in wireless mesh networks”, Cluster computing,
CT scans. They can resume their therapies when Jan 2018. ISSN no 13867857.
necessary. The proposed approach has achieved a [5]Mrs G.Revathy and Dr.K.Selvakumar, “Route maintenance
98% accuracy rate, ensuring the survival of a using tabu search and priority scheduling in wireless mesh
networks”, Journal of advanced research in dynamical and
patient's life. control systems, vol 9,sp-6, 2017. ISSN 1943023X
[17] https://orange3.readthedocs.io
REFERENCES
[1] Heng Y, Luo R, Wang K, Zhang M, Wang Z, Dong L, et al.
Kidney disease is associated with in- hospital death of

978-1-6654-0837-0/22/$31.00 ©2022 IEEE 1314

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on November 12,2022 at 06:02:27 UTC from IEEE Xplore. Restrictions apply.