1. Malaria Disease
Detection
A
Synopsis submitted
in the partial fulfillment of the requirements for the award of the
degree of
Bachelor of Technology
in
Computer Science and Engineering
By
Group 17
Poonam Agrawal - 1805213036(CSE),
Anjali Shivhare - 1805213010(CSE)
Preksha Rai -1805231039(CSE)
Under the guidance of
Dr. Parul Yadav
Dr. Jaswant Kumar
Department of Computer Science and Engineering
Institute of Engineering and Technology, Lucknow
Dr. A.P.J. Abdul Kalam Technical University, Lucknow, Uttar Pradesh.
2. Synopsis
Introduction
Malaria is caused by protozoan parasites of the genus
Plasmodium that are transmitted through the bites of
contaminated female Anopheles mosquitoes and that
infect the red blood cells. Most deaths occur among the
children in Africa, where a child passes away almost
every minute by malaria, and where malaria is a major
cause of childhood neuro-disability. According to the
World Malaria Report 2016, an estimate of 3.2 billion
people in 95 countries and territories are at risk of being
contaminated with malaria and developing disease, and
1.2 billion are at high risk (i.e is >1 in 1000 chance of
getting malaria in a year). There were about 214 million
instances of malaria globally in 2016 and about 438,000
malaria deaths. The burden was heaviest in the African
region, where an estimated 92% of all malaria deaths
occurred, and in children aged under 5 years, who
accounted for more than two thirds of all deaths (see the
malaria death rates from an earlier WHO report in .
Typical symptoms of malaria include fever, fatigue,
headaches, and, in severe instances, seizures and coma,
leading to death.
3. Malaria is usually diagnosed by the microscopic
examination of blood films, and hundreds of millions of
blood films are examined every year for malaria
detection . Although this is the most commonly used
technique, the action of examining films under the
microscope is tedious and susceptible to error.
4. Motivation
Malaria is a life-threatening disease caused by
parasites that are transmitted to people through the
bites of contaminated female Anopheles mosquitoes.
It can be prevented and can be cured if proper
measures are taken.
In 2017, there were 219 million instances of malaria in
90 countries.
In 2017, Malaria deaths reached 435 000.
The WHO African Region carries an unreasonable
high share of the global malaria burden. In 2017, the
region was home to 92% of malaria instances and
93% of malaria deaths.
According to the World Health Organization, In Africa
228 million active instances , 405000 deaths in 2020.
Malaria is very severe in Africa as clearly visible from
the data. Approximately, 94% of these deaths
happened in the African region. It is known to us that
almost 90% of the contaminated malaria instances are
children below 5 years of age. An estimated 3.2 billion
people in 95 countries are at high risk, according to
the 2016 World Malaria report .
Hundreds of millions of blood films are inspected
every year for malaria, which involves manual
counting of parasites and contaminated red blood cells
by a trained microscopist. Accurate parasite counts are
essential not only for malaria detection. They are also
very important for testing drug-resistance, measuring
5. drug-effectiveness, and classifying disease severity.
However, microscopic diagnostics is not systematized
and depends heavily on the experience and skill of the
microscopist. It is very common for microscopists in
low-resource settings to work in isolation, with no
rigorous system in place that can ensure the
maintenance of their skills and thus diagnostic quality.
This leads to incorrect diagnostic decisions in the
field. For false-negative instances, this leads to
unnecessary use of antibiotics, a second consultation,
lost days of work, and in some instances progression
into severe malaria. For false-positive instances, a
misdiagnosis entails unnecessary use of anti-malaria
drugs and suffering from their potential side effects,
such as nausea, abdominal pain, diarrhea, and
sometimes severe complications.
So, these all facts have inspired us to take up this
project to work upon. Malaria Detection using ML
would not only benefit Healthcare but also help in our
academics as Machine Learning is the new boon in the
industry.
6. Previous work
1) AE + SMC BASED MALARIA PARASITE
DETECTION METHOD
The actual autoencoder is a single layer feedforward
network of deterministic approach consisting of an
input layer, one hidden layer, and an output layer. The
input and hidden layer together form an ‘encoder’ part
and the hidden layer and output layer combine to form
a ‘decoder’ part, as shown . The hidden layer is
smaller in size than that of the input layer, as the
autoencoder compresses the information and focuses
on the useful extracted features for an effective
performance of the model .
2) SAE + SMC BASED MALARIA PARASITE
DETECTION METHOD
7. In this model, the input feature learned of SMC classifier
via the use of two hidden layers of stack autoencoder from
an input malaria blood smear microscopic image . In the
SAE + SMC network architecture , the features extracted
by SAE are fed as input to SMC for classifying the
infected and non-infected malaria.
3) TAE + SMC Based Malaria Parasite Detection Method
Three-layer Sparse Autoencoder (TAE) is three hidden layers
based on autoencoder which is also known as ‘deeply stacked auto
encoder’ . In order to handle high dimension input data, a single
hidden layer of actual autoencoder may not be suitable, therefore
stacked based deep third autoencoder is used as shown in
8. 4) RBM + SMC BASED MALARIA PARASITE
DETECTION METHOD
Restricted Boltzmann Machine (RBM) is a variant of the
stochastic recurrent neural network. It is a probability
distribution ‘P’ of the observed input data (visible units) as the
primary layer and the second layer consists of the latent
variables (hidden units). The observed input data units and
hidden units together are linked with symmetric weights, and
the visible and hidden units are pointed by bias weights .
RBM is used for fine-tuning the deep neural network to
minimize the error. For each example, let’s define, ‘V’ is the
visible units and ‘H’ is the hidden unit which together is
9. restricted to form a bipartite graph allowing implementation
of more efficient training algorithms.
10. Methodology
The image will be resized as a consequence to the
model’s input. Based on the input size of the image,
various data augmentation techniques like horizontal
flipping, shear zooming, etc will be applied to the images
which make the images more model friendly. After the
data-augmentation the images will be sent to the models
for training and the results will be calculated.
11. Proposed Model
Artificial Neural Networks
An artificial neuron is a computational unit which will
make a computation based on
other units it is connected to. In the case of a single
artificial neuron, it will be connected to the input
description of an object you want to extract information
from. Artificial neurons will read the information from
the input and perform a particular computation and
compute its value. Based on the value, a neuron will
decide whether some characteristics are present in
the input or not.
This computation step is divided into two parts:
a) Neuron Pre-Activation (Input
activation) In scalar form,
In vector form,
where,
12. a: Pre-Activation Function
x: Input Vector
w: vector of the connection weights. It represents the
strength between connections.
b: bias. If we have no inputs, b will be the input for the
neuron.
b) Neuron Activation (Output)
We use values from the pre-activation function
to compute the activation function.
Here,
h(x):: is the result of neuron
g(x):: activation function.
There are various types activation functions which can
be used to apply the nonlinearity
to the inputs.
Some of them are discussed below.
a): Linear activation function
In this case, g takes the pre-activation value and then
returns the same pre-activation value without
manipulating it. It doesn’t perform any input squashing.
It is not an upper
bounded and lower bounded and neither introduces any
non-linearity.
b): Sigmoid Activation Function
It takes pre-activation and then computes
activation value using the formula below:
13. It squashes the activation function between 0 and 1. It is
always positive and is bounded between 0 and 1. It is
strictly increasing. The bigger the pre-activation higher
the activation
will be.
c):Hyperbolic tangent (tanh) Activation Function
It is one of the most commonly used
activation function.It squashes the pre-activation
between -1 and 1 and is strictly
increasing..The pre-activation values are used as an
input and used for computation of values of the neuron
using following function:
d):Rectified linear (Relu) Activation Function
It is the most famous activation function.It is lower
bounded by zero and is monotonically
14. increasing.It gives neurons sparse activity. It often gives
neuron which have
values zero as whenever the input is negative or zero it
gives value zero.It is maximum
between zero and the given input:
A neural network is a set of layers(a layer has a set of
neurons) stacked together sequentially.
The output of one layer, would be an input of the next
sheet.
Three layers that are :
Input layer: A set of input neurons, where each neuron
represents the feature in our dataset. It takes the inputs and then
passes them to the next layer.
Hidden layer: A set of neurons where each neuron has a
weight assigned to it. It
takes the input from previous layer and does the dot
product of inputs and weights,
applies activation function ,produces the result and
passes the data to the next layer.
Output layer: it is the same as hidden layer except it
gives the final result.
For the given input, the networks calculates the results
based on the inputs by passing
the inputs in the first layer of network and passing the
values to the subsequent
layers. This process is known as ‘forward propagation’,
and based on the values generated
15. by the network , errors are calculated and then the
weights are adjusted to minimized
the error by the process of the ‘back propagation’.
16. Merits of Artificial Neural Network (ANN):
1. Parallel processing capability
2. Storing data on the entire network
3. Capability to work with incomplete knowledge
4. Having a memory distribution
5. Having fault tolerance
Demerits of Artificial Neural Network (ANN):
1. Assurance of proper network structure
2. Unrecognized behavior of the network
3. Hardware dependence
4. Difficulty of showing the issue to the network
5. The duration of the network is unknown
18. Conclusion
Malaria is the deadliest disease on the earth and big
hectic work for the health department. The traditional
way of diagnosing malaria is by the schematic
examination of blood smears of the human beings for the
checking of parasite-infected red blood cells under the
microscope by lab or qualified technicians.This action is
inefficient and the detection depends on the experience
and well knowledgeable person needed for the
examination. However, practical performance has not
been efficient so far. We will use an approach and an
algorithm to detect Malaria using Deep Learning. We
will implement Artificial Neural Network and
Convolution Neural Network for the classification of the
infected and uninfected images of blood samples. We
have gone through various research papers published in
this domain to get an idea as to what extent the work is
done. We'll try our best to get an accurate result.
19. References
● Ministry of Health and Family Welfare,WHO World Malaria Report
2020: India continues to make Impressive Gains in reduction of
Malaria Burden, 02 DEC 2020.
● Priyadarshini Adyasha Pattanaik, Mohit Mittal, Mohammad
Zubair Khan, Unsupervised Deep Learning CAD Scheme for
the Detection of Malaria in Blood Smear Microscopic
Images, pp 94936 – 94946, 20 May 2020.
● K. M. Faizullah Fuhad, Jannat Ferdousey Tuba, Md. Rabiul Ali
Sarker, Sifat Momen, Nabeel Mohammed, and Tanzilur Rahman*
,
Deep Learning Based Automatic Malaria Parasite Detection from
Blood Smear and Its Smartphone Based Application, 2020 May 20.
● Mahdieh Poostchi, Kamolrat Silamut, Stefan Jaeger, George Thoma,
Image analysis and machine learning for detecting malaria, 12
January 2018.