et

International Journal on Emerging Technologies 10(4): 66-70(2019)

ISSN No. (Print): 0975-8364
ISSN No. (Online): 2249-3255

Image Classification using Deep Neural Networks for Malaria Disease Detection
1 2 3 4 5
E. Laxmi Lydia , G. Jose Moses , N. Sharmili , K. Shankar and Andino Maseleno
1
Professor & Big Data Consultant,
Computer Science and Engineering, Vignan's Institute of Information Technology, India.
2
Professor, Computer Science and Engineering,
Raghu Engineering College (Autonomous), Viskahpatnam (Andhra Pradesh), India.
3
Associate Professor in CSE Department,
Gayatri Vidya Parishad College of Engineering for Women, Viskahpatnam (Andhra Pradesh), India.
4
Department of Computer Applications, Alagappa University, India.
5
Institute of Informatics and Computing Energy, University Tenaga National, Malaysia.
(Corresponding author: E. Laxmi Lydia)
(Received 07 August 2019, Revised 25 October 2019, Accepted 01 November 2019)
(Published by Research Trend, Website: www.researchtrend.net)
th
ABSTRACT: Since the 19 century, Malaria has become a terrifying life-threating disease in most of the
countries. Its been identified that five countries namely Nigeria with 25%, Congo with a ratio of 11%,
Mozambique with ratio of 5%, India with ratio of 4% and Uganda with ratio of 4%. World Health Organization
stated that above 90% of malaria death cases were recorded every year. Most of the Indian states like
Odisha, Madhya Pradesh, Maharastra, northern countries, Chhattisgarh got affected by Malaria. India spotted
death cases of malaria from millions to thousands that have reduced in recent years. Directorate of National
vector Bore disease control program has started malaria control strategies using early case detection and
treatments, vector control, protective measures against mosquito bites and management of Environment.
The major challenge was to identify the disease at early stage. The key contributions avoid malaria disease is
to provide antimalaria drugs, using indoor spray with residual insecticides, mosquito nets. For the treatment,
medical technologies, deep learning architectures related to Convolutional Neural Networks to train and test
performing different combinations for image classification using ResNet34 which helps patient go through
prior examination for microscopic diagnosis. For patients examination, this paper considers Malaria Cell
Images dataset with Parasitized and uninfected images. Thus, this clearly shows that one can easily identify
person’s condition whether he is infected or uninfected by enabling open-source Artificial Intelligence. It
shows the start-of-the-art accuracy by checking individual details.
Keywords: Malaria, ResNet34, Convolution Neural Network, Image Classification, convolutional, neural network.

I. INTRODUCTION morphologies that detects both annular and disk-like

structures among cells.Microfluidic devices used to track
Millions of people all over the world were affected by the red blood cells in videos to control the concentration
malaria. The major different Parasites that led to malaria levels of oxygen. Due to large number of varying
are p.knowlesi, p.ovale, p.malariae, p.vivax and characteristic features, the video of blood flow of RBC
p.falciparium. These parasites are like small insects that
cells process cell classification for accurate
grow through stages inside the human body and segmentation [19]. In real-time data processing [31],
transforms into harmful bacteria, which leads to malaria. implemented and performed an iterative voting-based
They change in their morphological behavior, size, color detection of cells. To identify the local extreme and
at each level of their stage. The main cause of malaria approximate centroids, multiscale LoG filter implies to
is due to climatic conditions in tropical region areas. know the individual cells, which help in speed up the
Mosquitoes causing malaria, it is a parasite which gets processing and accuracy. This will result in the
transmitted into human blood through mosquito bites. extraction of offline features of color and texture for
Millions of people are getting sick from high fever to
further evaluation of classification and reduction of
deadly states. It is a transmission process from dimensionality. Some of the techniques like normalized
mosquito to the human body, human body to liver, liver red green blue for the extraction of color feature
to red blood cells. The three different modes of malaria information and joint adaptive median binary pattern for
transmission are from mother to child (in womb, who are extracting texture feature information [29]. Classifiers
unborn), one to another person blood transfusions, such as artificial neural networks result in classifying
giving similarly injected needles too other persons. infected and uninfected cells.
Existing approaches for malaria diseases, implemented The aim of the work is to provide the construction of
methodologies such as automatic detection through models using neural networks by detecting the images
RBC count, applying Log filters to identify unique cells, and by decrease the model variance. This will enhance
use of model variance, minimizing the variance provide the robustness and analysis for detection. The efficiency
solutions but not with optimal results. An automated of the CNN classifies the thin-blood smear images [22]
detection system to recognize and distinguish malaria by allowing cross-validation. Some of the performance
parasites in human blood varies from different stages metric calculation on finding classification accuracy is
like trophozoites; schizonts analyze blood slides through receiver-operating characteristic, mean squared error,
a microscope.Detection of RBC count and segmentation precision, and Matthews Correlation Coefficient.
of cell images uses multiscale Laplacian of Gaussian,
another contour-based segmentation for the detection of
Lydia et al., International Journal on Emerging Technologies 10(4): 66-70(2019) 66
Models VGG-19 outruns the classification process on threshold. Oguntimilehin et al., [15] have mentioned all
reaching optimal levels of predictions through multiple the predictive models for imaging and symptomatic
models.This has reached the real-world conditions by approaches for the treatment of malaria detection during
minimizing the variance. Computer-based diagnoses the process of diagnosis [14]. The different models like
such as machine learning algorithms [25] prepared rough set, fuzzy logic, decision trees, analytical
clinical assistance for decision-making. Deep learning hierarchy process, Rule-based approach, linear
methods work in identifying the raw pixel data of programming, Mockler situation analysis through
images, feature extraction, and operating end-to-end knowledge-based systems, expert systems, Neural
extractions and coordination. In this paper, proposed Networks and Support Vector Machines.
Convolution neural networks solve the existing issues in Poostchi et al., (2018) [2] described that they discovered
finding optimal solutions byconsidering patient blood an automated system that identifies the infected blood
sample features to predict the state of the patient cells using Blood Smears. They used the extraction
whether the patient is parasitized and uninfected method to segment the RBC in blood. The algorithm
accurately. used for segmentation is C-EPAC Geodesic Active
Contour Based Segmentation Algorithm. Their
II. LITERATURE SURVEY
experiment analysis was tested [30] on human blood
In 2017, peripheral blood smear images used to detect cells of malaria-infected and mouse blood cells of
malaria parasites through Deep Belief Networks. Bibin malaria-infected. The implemented procedure is they
et al., [4] introduced Boltzmann machines for pre- initially considered microscopy thin smears of blood
training using concatenated color features and texture images, detection of RBC and its segmentation,
features. The different feature extraction techniques that computational evaluation of Cell features like texture
can be applied are color histograms, color coherence features and RBC color, cell classification performance,
vector, Statistical method for gray level co-occurrence and evaluation, finally to achieve parasite
matrix, and textural features positioned on the gray level quantification.Thusthe model has outperformed
run length matrix and local binary pattern. The algorithm automatic counting of parasites. This was already tested
applied is backpropagation with 4 hidden layers and 600 by [3] based on the histogram features. But the
hidden nodes. This has gained 95% specificity. Parveen classification techniques implemented are from Artificial
et al., [5] also proposed an application for malaria Neural Networks and experiment analysis tested using
prediction analysis using Artificial Neural Network which clinical database. Before that Chakrabortya et al., [6]
has given only verbal history and physical appearance have merged algorithms for classification for a thick
of patient without any doctors and laboratory facilities blood smear. Techniques like pixel discrimination and
showing positive and negative results. color based on segmentation. In this approach they
Automatic visual diagnosis of malaria has developed didn’t use training data, implemented using
through a motorized practical microscope [10]. Needed unsupervised approach [32].
samples are collected to perform analysis using patient Later based on mathematical morphology detection
symptoms during curing stage. ANN techniques in [9] systems are introduced for malaria parasites in [7]. They
are applied to evaluate blood smear images. Features have identified certain issues in the diagnosis of malaria
from RBC are collected from database with infected and and performed a practical implementation over spatial
non-infected RBC cells. Situated on the intensity of structures with impressive image processing
features, classifiers detected the images that got techniques. In this, they have selected morphological
infected. A system architecture was designed in [11] mathematical operators that successfully gained image
using two processing systems (system1 and system2), analysis and pre-processing tasks using Non-linear
system1 collects the image samples for testing and they transformation computer-aided architectures suggested
are converted into greyscale using pre-processing for the diagnosis in visual inspection [8]. Image analysis
techniques meanwhile system2 simultaneously collects using deep learning for vision-based has obtained fast
reference samples and converts them into greyscale. identification of malaria disease.
System1 images are compared with the system2 Poostchi et al., (2018) [16] has deep concern over these
images using NCC. If NCC is less than 0.7 then diseases to make use of technology. They have
detection process is continued using morphological implemented an analysis of images and machine
processing and detection pf RBC count. Finally, learning approaches [18]. They concentrated on
diagnosis results will be achieved. automatic cell classification and cell segmentation.
Object detection procedures and computer vision-based Malaria image acquisition techniques like light,
approaches were implemented in [12]. The authors Binocolor, Fluorescent, Polarized, Multi-spectral, multi-
have suggested these approaches based on three modal, image-based cytometer, sub-pixel resolving
reasons. Implemented procedure overstated raw data optofluidic microscopy, quantitative phase imaging,
for color normalization using Reinhard’s method. A quantitative cartridge-scanner, scanning electron, fiber
better comparison of statistical multimodal approaches array-based Raman imaging, serial block-face scanning
using Kalman filtersand illumination issues. Finally, pixel electron microscopy, and sightDx digital imaging
information using a kernel-based approach for localizing scanning.
accurate infected blood samples in image processing. Dharpal and Malviya (2018) [17] suggested the work
Peñas et al., [13] performed analysis-using CNN [21, analyses on image processing for the treatment of
24] for problematic issues identified in rural areas for malaria disease. Parasites are recognized from tainted
blood smears with specified pixel resolutions for and non-contaminated red platelets using local binary
p.falciparum and p.vivax ratio. Laser desorption mass patterns. Here the Plasmodium parasites are detected
spectrometry was implemented long back in the using support vector machine [28].
consequence of detection of malaria parasites [23] Sajana and Narasingarao (2017) [18] worked for the
during the growth of RBC. They consider beam improvement of malaria disease detection using R
parameters with upgraded magnitude. It is a UV laser programming and weka software by establishing a
used to clean up entire blood samples by setting up model-based decision support system on 165 patients
Lydia et al., International Journal on Emerging Technologies 10(4): 66-70(2019) 67
from distinct groups collected from Narasaraopet using ResNet34 Architecture needs residual blocks trained to
Naive Bayesian technique. SMOTE is applied to achieve deeper networks. Initially calculate activations
transform unbalanced dataset to achieve balanced from layer 1 to end of the layers. This reduces the error
dataset. in-network and assumes weight and bias for computing
In 2019, infectious malaria disease has been analyzed layers of predictions.The mathematical form of ResNet
using machine learning approaches. The authors [1] can be evaluated using g(ai) and f(ai).
have selected spatial features through microscopic f(xi) in which the incoming original input ai is combined
images and trained using the Neural network model. to f(ai). where f(ai) is g(ai) - ai (results in residual
Using these images as input, the model needs to learning).
recognize the infected area. Exposure of the malaria i.e, f(ai) = g(ai) - ai (1)
parasite in the images performed testing the transfer- and g(ai) = f(ai) + ai (2)
learning method. This method is an expert learning Eqns. 1 and 2 are the ResNet calculations on network
model that is used as the Unification of VGG network layers to make connections among layers and enable
and SVM. Top layers are trained and the rest of the cross-layer connectivity.
layers are frozen. Images of blood smear samples are
tested using VGG19-SVM. This model has shown its
Dataset Feature Data Missing
potentiality above CNN [20]. Collection Generation Conversion Values
Raut and Marathe (2019) [25] identified the rapid usage
of multimedia websites that transmit videos and retrieve Feature Randomized Data Data
information from images using a classification system. selection/Extraction Data Ordering Normalization Cleaning

Applied modules to retrieve image information such as Model Parameter Evaluate

Training
word-level textual cue encoding, visual cue encoding, selection Optimization & Testing
finally classification. Singh and Singh (2019) [26]
analyzed image classification through SVM, and Analysis
Convolution based PSO. They progressed these
techniques for the detection of diseases that occur on Fig. 1. Block diagram of Convolution Neural Networks.
leaves. Rao et al., (2019) [27] determined the count of Convolution Neural Networks process in different layers
blood cell images from the proposed convolution neural of execution. For the model selection, the Convolutional
network models using color split channel, binary, layer involvesa group of convolutional kernels that
mathematical morphology for segmentation. associates with small image regions (receptive field) by
Experimental results were counted based on the dividing the image and convolving those values withset
hierarchical multiclass Support vector machine. of weights.Represented as
A. Methodology ‫ܨ‬௟௞ = ‫ܫ‬௫,௬ ∗ ‫ܭ‬௟௞ (3)
Convolution neural networks designed based on some Eqn. (3) represents the operation of the convolution.
samples through periodic downsampling with the The input of the image to the network is represented by
th
increase of feature maps. Many computer vision Ix, y , ‫ܭ‬௟௞ shows the kernel l and the k layer of the
problems have identified through machine learning convolution and spatial locality are given by x, y. This
image processing progress through image classification, implementation additionally performs categorization with
object detection, and image segmentation. Various respect to various filters, padding, and direction. Pooling
Deep Convolution Neural Networks are Spatial layer downsamples the features pointing to a specific
exploitation based Convolution Neural Networks, Depth local region which is represented as below
௟
based Convolution Neural Networks, Multi-path ܼ௟ = ݂௣(‫ܨ‬௫,௬ ) (4)
basedConvolution Neural Networks, width based multi- Equation (4) represents the operation of the pooling by
connection Convolution Neural Networks, feature map determining the output feature map which reduces the
(channelfmap) exploitation based Convolution Neural overfitting problem.
Networks, channel(Input Channels) exploitation based Activation Function helps to work on complex patterns
Convolution Neural Networks, Attention-based and provide decision function. Such as
Convolution Neural Networks. ܶ௟௞ = ݂஺ (‫ܨ‬௟௞ ) (5)
ResNet34 architecture followed deep residual networks, Eqn. (5) represents the activation function, activation
which are adequate with a large number of complex function can use different functions like maxout, ReLU
functionalities and input descriptions. A large number of (ELU, PReLU), tanh, sigmoid.
layers indicate mapping and a deeper model that Fig. 1 describes the procedure for image classification
generates no further training error than empty its using a convolution neural network at three different
counterpart. Residual connections initiate network stages. Initially at first stage data is collected
architectures by enabling all the deeper structures. andperformed feature generation, at second stage data
Image recognition through deep neural networks is hard will be pre-processed and feature selection will be
to train but using ResNet 34 training networks is easy by performed, at third stagea supervised model has been
reformulated formulae. This architecture calculates from selected, apply to tune on parameters and finally
layer 1 to layer 152 having minimum complexity. This analyze the prediction data.
pre-trained model consumes less time and obtains more
features. This will train the input dataset and maintain III. RESULT ANALYSIS
all its weights and bias that reflects the corresponding In this paper, the malaria disease dataset was collected
features. Furthermore, these are transferred and tested from the website, which has images of Parasitized and
by other datasets. Special features of ResNet are to Uninfected blood samples. Parasitized and Uninfected
avoid network and rapid use of batch normalization. datasets each consists of 13,779 cell images of blood
This further avoids fully connected layers at the samples. Considered malaria disease dataset was
extreme. In Convolution Neural Networks for the image implemented over a Python environment using
reorganization, uses ConvNets. Procedure for Anaconda.
Lydia et al., International Journal on Emerging Technologies 10(4): 66-70(2019) 68
Fig. 1 demonstrates the plot using lr.find() function
between learning rate and loss to achieve a good
learning rate. Fig. 2 demonstrates the plot for the
learning rate over fit_one_cycle at stage_1. Fig. 3
demonstrates the plot on losses at stage-2
Table 1 shows overall accuracies for fit_one_cycle for
six epochs. Table 2 shows overall accuracies for
fit_one_cycle for four epochs with maximum learning
rate.

Fig. 4. Plot on losses at stage_2.

IV. CONCLUSION
Image classification on blood samples for the detection
of malaria disease. This paper proposed Convolution
neural networks for the detection of malaria over
13,779 Parasitized and Uninfected blood cell images.
The learning rate at different stages is calculated and
identified that 97% of the learning rate is accurate for
classification. This involves the identification of parasites
in blood sample images. The proposed Deep
Convolution Neural Networks predicts the patient
Fig. 2. Plot for Learning rate versus Loss using Learning conditions for further diagnosis. Therefore, computer
Rate Finder. predictive models took prior place in medical diagnosis
and treatment. Implementation of Deep Convolution
Table 1: Over all accuracy for fit_one_cycle for six Neural Networks results ina good performance on
epochs. image datawhile estimating its optimal features,
orientation and location of the cells (object).
epoch Train-Loss Valid-loss accuracy
1 0.171754 0.129417 0.956632 Conflict of Interest. Nil.
2 0.154249 0.116933 0.958447
3 0.146585 0.153310 0.950100
REFERENCES
4 0.130996 0.105276 0.961350 [1]. Vijayalakshmi, A. (2019). Deep learning approach to
5 0.101642 0.091105 0.966975 detect malaria from microscopic images. Multimedia
6 0.090302 0.083954 0.969153 Tools and Applications, 1-21.
[2]. Poostchi, M., Ersoy, L., McMenamin, K., Gordon, E.,
Palaniappan, N., Pierce, S., Maude, R. J., Bansai, A.,
Srinivasan, P., Miller, L., Palaniappan, K., Thoma, G., &
Jaeger, S. (2018). Malaria parasite detection and cell
counting for human and mouse using this blood smear
microscopy”, Journal of medical imaging, 5(4): 044506-
13.
[3]. Devi, S. S., Sheikh, S. A., Talukdar, A., & Laskar, R.
H. (2016). Malaria infected erythrocyte classification
based on the histogram features using microscopic
images of thin blood smear. Indian Journal of Science
and Technology, 9(45), 1-10.
[4]. Bibin, D., Nair, M. S., & Punitha, P. (2017). Malaria
parasite detection from peripheral blood smear images
using deep belief networks. IEEE Access, 5, 9099-9108.
[5]. Parveen, R., Jalbani, A. H., Shaikh, M., Memon, K.
H., Siraj, S., Nabi, M., & Lakho, S. (2017). Prediction of
Malaria using Artificial Neural Network. International
Fig. 3. Learning rate fit_one_cycle at stage_1.
Journal of Computer Science and Network Security,
Table 2: Accuracy for fit_one_cycle on four epochs 17(12), 79-86.
with maximum learning rate. [6]. Chakrabortya, K., Chattopadhyayb, A., Chakrabarti,
A., Acharyad, T., & Dasguptae, A. K. (2015). A
epoch Train-Loss Valid-loss Accuracy combined algorithm for malaria detection from thick
1 0.100960 0.081042 0.970967 smear blood slides. Journal of Health Medical
2 0.082964 0.972600 Informatics, 6(1), 645-652.
0.103197
[7]. Loddo, A., Di Ruberto, C., & Kocher, M. (2018).
3 0.086624 0.080855 0.972600
Recent advances of malaria parasites detection
4 0.087902 0.079658 0.973870
systems based on mathematical morphology. Sensors,
18(2), 513.
Lydia et al., International Journal on Emerging Technologies 10(4): 66-70(2019) 69
[8]. Pattanaik, P. A., & Swarnkar, T. (2019). Vision- IEEE EMBS International Conference on Biomedical &
based malaria parasite image analysis: a systematic Health Informatics (BHI) (pp. 101-104). IEEE.
review. International Journal of bioinformatics research [21]. Rajaraman, S., Antani, S. K., Poostchi, M., Silamut,
and applications, 15(1), 1-32. K., Hossain, M. A., Maude, R. J., Jaeger, S., & Thoma,
[9]. Bashir, A., Mustafa, Z. A., Abdelhameid, I., & G. R. (2018). Pre-trained convolutional neural networks
Ibrahem, R. (2017). Detection of malaria parasites using as feature extractors toward improved malaria parasite
digital image processing. In 2017 International detection in thin blood smear images. Peer J., 6, e4568.
Conference on Communication, Control, Computing and [22]. Rajaraman, S., Jaeger, S., & Antani, S. K. (2019).
Electronics Engineering (ICCCCEE) (pp. 1-5). IEEE. Performance evaluation of deep neural ensembles
[10]. Mehrjou, A., Abbasian, T., & Izadi, M. (2013). toward malaria parasite detection in thin-blood smear
Automatic malaria diagnosis system. In 2013 First images. Peer J., 7, e6977.
RSI/ISM International Conference on Robotics and [23]. Var, E., & Tek, F. B. (2018). Malaria parasite
Mechatronics (ICRoM) (pp. 205-211). IEEE. detection with deep transfer learning. In 2018 3rd
[11]. Mohammed, H. A., & Abdelrahman, I. A. M. (2017). International Conference on Computer Science and
Detection and classification of malaria in thin blood thin Engineering (UBMK) (pp. 298-302). IEEE.
images. International conference on communication, [24]. Quinn, J. A., Nakasi, R., Mugagga, P. K. B.,
control, computing, and electronics engineering, pp.1- Byanyima, P., Lubega, W., & Andama, A., (2016). Deep
5.doi:10.1109/iccccee.2017.7866700. Convolutional neural networks for microscopy-based
[12]. Pattanaik, P. A., Swarnkar, T., & Sheet, D. (2017). point of care diagnostics. Proceedings of the
Object detection technique for malaria parasite in thin International conference on machine learning for
blood smear images. In 2017 IEEE International healthcare, 56, 1-12.
Conference on Bioinformatics and Biomedicine (BIBM) [25]. Raut, B. P., & Marathe, R. V., (2019). Image
(pp. 2120-2123). IEEE. Retrieval and classification using Image text extraction.
[13]. Peñas, K. E. D., Rivera, P. T., & Naval, P. C. International Journal of research in electronics and
(2017). Malaria parasite detection and species computer engineering, (IJRECE), 7(1), 413-417.
identification on thin blood smears using a convolutional [26]. Singh, R., & Singh, B. (2019). Improved leaf
neural network. In 2017 IEEE/ACM International disease detection and classification convolution-based
Conference on Connected Health: Applications, particle swarm optimization. International Journal of
Systems and Engineering Technologies (CHASE) (pp. research in electronics and computer engineering
1-6). IEEE. (IJRECE), 7(2), 3124-3132.
[14]. Demirev, P. A., Feldman, A. B., Kongkasuriyachai, [27]. Rao, N. Y., Kumari, V. T., Satyavathi, C., Mounika,
D., Scholl, P., Sullivan, D., & Kumar, N. (2002). L., G., & Sandeep, P. (2019). A deep learning enhanced
Detection of malaria parasites in blood by laser technique for classification of blood cell images.
desorption mass spectrometry. Analytical chemistry, International Journal of research in electronics and
74(14), 3262-3266. computer engineering (IJRECE), 7(1), 1857-1862.
[15]. Oguntimilehin, A., Adetunmbi, A. O., & Abiola, O. [28]. Fegade, G. K., Namdeo, V., & Gupta, R. (2017).
B. (2015). A review of predictive models on diagnosis Predictive Modelling for E-commerce Data Classification
and treatment of malaria fever. International Journal of Tasks: An Azure Machine Learning Approach.
Computer Science and Mobile Computing, 4(5), 1087- International Journal of Electrical, Electronics and
1093. Computer Engineering (IJEECE), 6(1), 45-50.
[16]. Poostchi, M., Silamut, K., Maude, R. J., Jaeger, S., [29]. Lydia, L. E., Prasad, B., Mahesh, B. C., Shankar,
& Thoma, G. (2018). Image analysis and machine K., & Kumar, V. K. (2019). An unsupervised deep
learning for detection malaria. Translational research, learning methods for fabricating text mining analysis
2018. doi:10.1016/j.trsl.2017.12.004. based on topic modeling and document clustering
[17]. Dharpal, S. N., & Malviya, A. V. (2018). Diagnosis techniques. International Journal on Emerging
of Malaria parasite based on Image processing. Technologies, 10(2), 103-109.
International research journal of engineering technology, [30]. Lydia, L. E., Sharmili, N., Shankar, K., &
5(5), pp.2962-2965. Masseleno, A. (2019). Analysing the performance of
[18]. Sajana, T., & Narasingarao, M. R. (2017). Machine classification algorithms on diseases datasets.
learning techniques for malaria disease diagnosis - a International Journal on Emerging Technologies, 10(3):
review. Journal of advanced research in dynamical and 224-230.
control systems, 9, 349-369. [31]. Lydia, L. E., Pandi Selvam, B., Saranya, R.,
[19]. Ersoy, I., Bunyak, F., Higgins, J. M., & Kirutikaa, U. S., IIayaraja, M., Shankar, K., & Maseleno,
Palaniappan, K. (2012). Coupled edge profile active A. (2019). Data Integration and Data Privacy through.
contours for red blood cell flow analysis. In 2012 9th Pay-As-You-Go, Approach. International Journal on
IEEE International Symposium on Biomedical Imaging Emerging Technologies, 10(2), 167-173.
(ISBI) (pp. 748-751). [32]. Lydia, L. E., Pandi Selvam, B., Kirutikaa, U. S.,
[20]. Dong, Y., Jiang, Z., Shen, H., Pan, W. D., Williams, Saranya, R., IIayaraja, M., & Shankar, K. (2019).
L. A., Reddy, V. V., ... & Bryan, A. W. (2017). Transforming healthcare Big Data Implementing through
Evaluations of deep convolutional neural networks for Aprior-Map Reduce. International Journal on Emerging
automatic identification of malaria infected cells. In 2017 Technologies, 10(2), 71-77.

How to cite this article: Lydia, E. L., Moses, G. J., Sharmili, N., Shankar, K. and Maseleno, A. (2019). Image
Classification using Deep Neural Networks for Malaria Disease Detection. International Journal on Emerging
Technologies, 10(4): 66–70.

Lydia et al., International Journal on Emerging Technologies 10(4): 66-70(2019) 70