Content uploaded by P. Manju Bala
Author content
All content in this area was uploaded by P. Manju Bala on Jan 28, 2022
Content may be subject to copyright.
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 1
Usage of Deep learning in Bio informatics and
biomedical images: A study of Applications
S.Usharani1, P. Manju Bala2, G. Leema Roselin3, A. Bala Chandar4
Department of Computer Science and Engineering,
IFET College of Engineering, Villupuram, Tamilnadu, India.
Abstract
Deep learning (DL) has grown rapidly in bioinformatics, demonstrating an
exhilaratingly potential ability to explore the intricate relationships buried
in large scale biomedical and pharmaceutical information and images. A
variety of extensive assessments of such programmes were published,
especially in high studies including future directions to demonstrations.
These studies have offered an interesting counterpoint to and guidance for
the use of deep learning (DL) in bioinformatics, encompassing a wide
range of Machine learning(ML) issues, Deep learning architectures, and
biomedical challenges. However, researchers cover either an apophatic
explanation to deep learning and specific details and prototypes of its
typical applications in bioinformatics and biomedical images in this
chapter. Furthermore, researchers discuss modern DL, recent trends, and
potential opportunities in the pragmatic DL area, and discuss potential and
substantial bioinformatics and biomedical applications.
Keywords: Deep learning, Machine learning, Bio-informatics.
Introduction
Machine learning is frequently often charity to execute a simple or sequences of steps with no
need for coders to expressly analyze them, demonstrating artificial intelligence's potential. Rather
of accepting the developer's commands, the technology uses quantitative methods and analytics
to do jobs[1-3]. As a result, it is a viable approach for learning and reacting to the event itself. It
offers a variety of social networking, online client service development, and prediction-related
services, among other things [4-6]. By prediction-related opportunities, researchers imply that the
bioinformatics field has many applications for disease prediction, and that is why deep learning
produces excellent results for bioinformatics challenges.
Deep Learning
The primary concept of machine learning techniques is to comprehend the real-world challenges
that humans face. The labelling component in bioinformatics databases is denoted by
technological interpretation like semi supervised, supervised, and unsupervised learning[7]. The
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 2
goal is to obtain information from the past behavior of statistics and then act on it. This
necessitates the use of possible techniques and algorithms, which is a time-consuming procedure.
Neural networks play a role to help with this intricacy.The challenge of studying enormous
datasets can be overcome by utilizing deep neural networks (DNNs). Only huge datasets allow us
to thoroughly examine the DNN's[8] basic possibilities. The more data utilized for training, the
more accurate the testing will be. Deep learning applications include voice recognition, video
processing, recommendation systems, illness prediction, medication development, speech
recognition, web content filtering, and more. Deep learning applications are rapidly expanding as
the spectrum of learning algorithms expands[9-11].
Bioinformatics Deep Learning Applications
Bioinformatics, often known as computational biology, is the study of using computers to
understand biological data[12]. A great quantity of biomedical information is being created as a
result of the major expansion of protein sequence, genomes, 3D modelling of biomolecules and
biological processes, and so on. To make conclusions out of this massive amount of biomedical
will need to have a good understanding of molecular biology and engineering. As the amount of
data generated by genome, proteome, and other database systems has expanded, evaluating that
data has become increasingly important[13]. Those datasets are analyzed using data mining
approaches. The results of huge data analysis should matter in terms of the organization
perceived by the data. Cancerous cells classification, genetic classification, and microarray text
categorization are some of the applications covered by image classification. Membrane protein
prediction, mathematical modelling of protein–protein interrelations[14], gene finding, protein
structure domain identification, function pattern detection, template matching inference,
diagnostics, disease progression, diseases prevention optimization, nutrients and genomic
functional gene reconstruction, information extraction, and protein cell surface position
prognostication are some of the techniques used. As a result, the interaction between deep
learning and bioinformatics is growing and developing[15]. Microarray data, as instance, is
utilized to forecast a patient's outcome. Patients' survival time and probability of tumour growth
or relapse can be assessed using genetic microarray data. It is highly appealing to have an
effective way that takes into account all correlated information.
Deep Learning and Bioinformatics
Computer vision, image processing, clinical images, Gene editing, RNA detection, gene structure
prediction, systems biology, infectious diseases, farming, weather prediction, criminology,
immunology, nutritional research, and other applications of deep learning and bioinformatics
operate hand in hand[16-18]. After integrating convolutional, strong belief, and recurring neural
network models to bioinformatics challenges, they are in the early stages of development.
Bioinformatics is used in a variety of ways, as seen in Figure 1. Microbiology, genomes, and
other bioinformatics applications are unsurprising. However, the development of bioinformatics
employing information systems, database, and machine drug design should be emphasized.
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 3
Figure 1.Application of Bioinformatics
Bioinformatics Applications
In bioinformatics applications, the efficiency of all deep learning methods is apparent. As a
result, deep learning appears to be most effective of the techniques used in this sector[19]. It
does, however, need the proper position of the framework for the issue, and also the parameters.
Analyze Sequences
In the discipline of computational biology, analyzing a sequence is a fairly fundamental activity.
In clinical research and genomic mapping, it is utilized to detect related biological sequences and
regions. It is possible to properly align a sequence by examining it. Sequences that are often
searched are saved in the database and may be accessed from the computers on a regular
basis[20, 22].
Annotation of the Genome
In 1995, Dr. Owen White created the very first genome annotation software model. The labelling
of genes and associated biological properties in a DNA sequence is really what genomics is all
about[21].
Gene Expression Analysis
The activity of many of these genes may be assessed using methods like microarrays, DNA
sequencing, genome serial analysis, concurrent signature sequencing, and so on. All of the
strategies discussed are absolutely noise isolating and influence the surroundings intuitively. In
genomic investigations, progress is being made in developing methods that can discern between
transmitter and receiver[23,24].
Bioinformatics
Molecular
Databases/
Datasets
Genomes/
X-omics
Information
Technology
Computationa
l Resources
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 4
Protein Expression Analysis
There are several methods for measuring expression of genes, but protein levels is the greatest
since it provides a platform. Integumentary system, proteomics, and extremely high usage
capability can provide a picture of the protein for study[25].
Mutation Analysis in Cancer
The genes of cancer patients' bodies are changed at random in a perplexing manner. To locate the
mutation spot, which would be unknown, extremely elevated genomic approaches are necessary.
As the population expands, the number of genomic sequences grows dramatically, necessitating
high-level algorithms to appropriately recognize the sequence[26].
Prediction of Protein Structure
The DNA sequence is the most important sequence for predicting the structure of a protein. It
may be deduced from the human gene, which aids in the identification of the gene's distinct
structure. The understanding of this one-of-a-kind structure is crucial[27] in understanding how
proteins work. Membrane protein predictions is being used to develop new medicines and
enzyme.
Biological System Modeling
In the field of cognitive biological sciences, modelling biological processes is more important.
For cellular systems, computerized simulations are performed, and genetic regulation networks
are utilized to discover complicated shapes in cellular networks. The interconnections among
cells can be highly delineated, yet computer modelling can easily overlook them. The goal of
artificial intelligence is to comprehend real-world problems by designing a system that works
similarly[28].
Image Analysis with a High Throughput
Biomedical images have had the ability to provide a wealth of information. A researcher will
make a proactive judgment on the advantages and downsides by studying these medical images.
The computer-modeled system can create applications the scientist's work of observation.
Clinical image processing, Genetic clone overlapped detection, and other applications are only a
few examples.
Microarrays
Microarrays are extremely handy for gathering additional enormous volumes of data. Deep
learning can aid in the study of microarrays, allowing for the recognition of patterns and
relationships between genes. Microarrays [29]detect the expression of genes involved in a
genome, allowing tumors to be diagnosed. The most commonly used methods for analyzing data
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 5
include radial basis functions, neural networks, gradient boosting, random forest, and regression
trees.
Biology System
The behaviors of a biological system may be deduced by monitoring the parts of the system.
Genes, RNA, and enzymes are only a few of the components that must be monitored. For these,
probability models are created, which are then employed in simulated annealing, which are based
on Markov models. Enzymatic function predictions, elevated microarray signal processing,
assessment of genome-wide correlation[30]trainings are needed to improvedcomprehend
indicators of multiple sclerosis, protein structure prediction, and identifying of NCR-sensitive
genes in yeast are just a few examples of uses.
Text Exploration
Text Exploration is particularly useful in genomic data, papers, and other places. To recover the
prospective information from the dataset, to detect the molecular mechanism of a proteins, to
assess Genome arrays[31], and to study huge protein and molecular interaction, further
approaches are required. Text Exploration may also be used to discover and visualize unique
DNA regions if enough reference material is available.
Deep learning models
Deep Learning techniques (Fig. 3) allow for the identification and extraction of image
information to improve performance of the model for the job at hand. DL is an area of machine
learning that uses convolutional neural networks to handle raw information effectively[32].
Furthermore, deep learning methods make it possible to build end-to-end forecasting analytics by
automating all of the procedures needed in the creation of a traditional machine learning model,
such as image retrieval and training (see Fig. 2). Deep learning models are projection techniques
made up of a stack of heterogeneous deep architectures with a finite amount of regressive
components. The inlet and outlet stages of a system are designated as the initial and last levels,
accordingly, with any layers placed between them being referred to as hidden nodes. Neural
networks multi-layered approach allows them to behave as complex functional replicates,
learning distinct interpretations of the input data at several abstraction levels. A DL system can
accurately approach billions of training examples to predict during the training phase, based on
the amount of levels and components per layer[33]. As a result, Deep learning models are
vulnerable to computational complexity, particularly when combined with tiny training samples,
and should only be used on datasets with thousands of photos. Deep Learning has been widely
used in medicine imaging and interventional radiology due to its capacity to represent very
complicated relationships within huge datasets, with particular solicitations in medical imaging
sector comprising both big and minor image datasets, as well by differing consequences.
Convolutional neural network are the most commonly utilized artificial neural design for medical
image analysis. The existence of convolution operation between both the layers in the neural
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 6
distinguishes these systems, which convolve an input data with a predefined activation
functions[34]. Because the values of convolution operation gained during development can
recover image features appropriate to the researched task, multiple convolution layers can be
used in CNNs as per the desired application. In contrast to completely interconnected neural
networks, CNNs apply the similar kernel strictures to the given dataset, minimizing the number
of training examples and speeding up the learning process[35]. Two-, one-, or tri dimensional
convolutional kernels can be used depending on the dimensions of the inputs and outputs. Some
other important part of the CNN design is the sharing layers, which diminish feature vector
resolution and introduce translational invariance. Furthermore, by combining fully convolutional
layers, geographic hierarchies among image patterns can be learned.
Figure 2.Traditional Machine learning model
The feature selection method is made up of a stack of regular (linear interpolation) and quadratic
(stimulation) computational layers that gradually increase the amount of complexity,
dimensionality, and exclusionary capability between levels[36]. Following this, these features are
integrated using either a sequence of convolution layers or other traditional machine learning
methods to execute the learning strategy (Fig. 3). Classification algorithms can include more than
just convolutional, sharing, and activating layers. Because of the system architecture of
Convolutional Neural Networks, numerous topologies integrating Convolutional Neural
Networks with some other artificial neural networks have indeed been developed. Image
recognition tasks have been performed using end-to-end Classification methods that immediately
map pictures to a classification model for both prediction and treatment. Many Convolutional
Neural Networks architectures, like as ImageNet, that were developed on huge natural images
based data take been used for medical image processing by adjustment pre - trained levels to deal
with data nonlinearity difficulties. The U-Net design, which was first published in 2015, is one of
the most used Convolutional Neural Networks designs for clinical image segmentation. The basic
U-Net architecture is made up of symmetric encoder and the decoder routes that are linked
together by back propagation [37]. It was initially designed to handle two-dimensional images,
but it has since been tweaked to produce vector graphics classification from three-dimensional
images. After then, numerous variations of this net were created by adding remainder,
attentiveness, or DenseNetchunks to train bottomless networks, pick key elements, and handle
Segmented
biomedical
images
Feature
Extraction Normalization Feature
Selection
Predictive
modeling
Prediction
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 7
slope disappearing concerns, accordingly. The designs shown above are just a sampling of the
wide range of options: Although it is beyond the scope of this thesis, a thorough classification of
Classification algorithms may be discovered in a recent analysis. To retrieve geographical
properties from neuroimaging data series, deep neural networks have been integrated with
Convolutional Neural Networks. By exchanging node values over time, these systems can
analyze new data while remaining aware of past inputs and outputs. RNNs, on the other hand, are
important to implement and vulnerable to overloading because model difficulty is openlyrelative
to the size of input data. Gated recurrent units and LSTM were developed to overcome shortest
path difficulties and to enable for the memorizing of long-term information (LSTM)[38].
Classifiers are also important in unstructured DL architectures, as they learn how to recreate the
data input in an unsupervised fashion. The use of fewer hidden units in the encoding route,
normalization, and nonlinearity restrictions in these networks enables the network can study a
lower-dimensional description of the information, avoiding the net from learning the
identification transform. Due to their capacity to predict distribution of data and produce accurate
datasets, recursive neural networks have lately become important in medical image processing.
Figure 3.Deep learning model
GANs are made up of two key strategies: one produces fresh genuine data by learning dispersal
of data from training images, while the other distinguishes among manipulated data. The
interplay of various key strategies improves the overall performance of the GAN and produces
genuine image data. Although their unique strategy, these systems are difficult to train due to
fading away gradient difficulties and are susceptible to producing fresh data that looks identical.
After deciding on the best network infrastructure, parameters tweaking is a time-consuming
process. The amount of multilayer units, the convolution layer size, and the training algorithm
can all have a significant impact on system performance, making it difficult to design the right
architecture[39-41]. The network parameters are determined during learning to tackle a certain
goal. To do this, a back - propagation error method adjusts the network's variables to reduce a
Biomedical
images
Deep
Feature
Extraction Prediction
Predictive
modeling
a) Image feature extraction model
Prediction
Biomedical
images End-to-End Learning
b) End-to-End learning model
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 8
lower bound that represents the network's cost function. The modification is dependent on the
loss function's gradient changing when routing protocols change. Many optimization techniques
have been designed to improve this procedure. To develop quality minimum identification in
complicated optimization methods, most of them use effective learning rates in addition to loss
function. Furthermore, both input picture normalization and the use of convolutional layers to
standardize the dynamically derived deep features have been demonstrated to aid training
junction and minimize response variable shift. The network's depth should rise in proportion to
the task's difficulty. Furthermore, very neural networks are susceptible to the problem of gradient
descent, which basically stops the weights from index reflects throughout training, resulting in
either a long training period or an inability to converge[42,44]. This problem can be mitigated in
part by using the Rectified Linear Unit activation function, suitable initialization approaches, and
skip connection[43]. Because an excess increase in network size can lead to overfitting,
normalization techniques like as R1 and R2 normalization, batch normalization, washout, early
halting, and feature extraction approaches can be employed to critical evaluations generalization.
As discussed in a detailed analysis of the most commonly used normalization strategies and their
impacts on DL system performance, these strategies can be coupled to reap the benefits of the
complementing impacts of diverse approaches. The "no free lunch" thesis argues that each
method needs a certain hyper - parameters configuration to optimize its effectiveness on a certain
job when it comes to design decisions. As a result, modulating is a crucial, if time wasting, step
that necessitates the uninterrupted assessment of model error rate on calibration and testing
datasets in order to determine an appropriate compromise between over-fitting and under-fitting.
Several ways can be utilized to find the ideal number of hidden layers configuration.
Conventional methods include comprehensive, randomized, and multiphase dimensionality
reduction searches, while more newly suggested methods comprise automated hyperpara meter
optimization techniques, which decrease the burden of parameters adjustment on the model
construction process. Reinforcement learning and meta-heuristic methods, in this case, are
potential replacements to trial-and-error methodologies. Nonetheless, the test set must be used to
evaluate DL performance of the model because it is the only impartial and exterior data set that
can verify modeling generalization
Deep learning in Bio-medical images
The unavailability of a significant amount of well-labeled therapeutic picture data makes learning
and assessing deep neural networks in therapeutic data more difficult than texture - based
investigation with machine learning. Image enhancement and image classification approaches
can be utilized to overcome this problem[44]. GANs will be utilized to make synthetic
continuing education cases in this light. Deep transfer learning techniques, on the other hand,
allow bypassing the need to train DL models from start by relaxing the presumption that testing
and training data comes from the same normal distribution. Deep transfer learning methods
consists of four types such as instances-based, adversarial-based, mapping-based and network-
based. A semi-supervised[45] or weakly-supervised approach can also be used to solve the
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 9
absence of correctly labeled data. Completely supervised learning uses labeled examples to learn,
verify, and test a deep learning method, whereas weakly-supervised learning uses partially- or
weakly-labeled data. Partial labeled dataset (unfinished supervision), fine grained annotated
datasets (incomplete supervision), and datasets with labeling other than ground-truth labels are
examples of such techniques (inaccurate supervision). Furthermore, recent advances in DL
investigation have demonstrated the value of self-supervised or unsupervised pre-training
techniques: Labeling are dynamically recovered from data in self-supervised algorithms, whereas
imaging characteristics are recovered without labeling in unsupervised algorithms. Another key
difficulty for both DL and ML in the increasing paradigm of customized and medical
applications is the combination of social data modalities elements into a single product. Though
imaging and medical data must be combined with additional genic data in a single Deep learning
model, this problem becomes especially acute. The research includes all available strategies,
starting with ML and moving on to multisensory and DL integration strategies. Adversarial
attacks should be included with adversarial transfer learning to transfer learning and transfer
learning. DL-based computer-assisted diagnostic systems, as well as radiomics-based models,
may be affected by the production of adversary samples, which involves making tiny changes to
diagnostic imaging examples that are near to the training samples acquired by a classification.
Minorvariations to the pixel data could, in fact, affect the standards of some radiomic properties
that affect downstream studies[40,45]. This issue must be addressed in any trustworthy
computer-assisted diagnostics system that must be used in clinical settings.
Conclusion
Deep learning is a sophisticated and valuable approach that has aided the advancement of a
variety of industries, including bioinformatics. Deep learning and bioinformatics applications
include a range of learning approaches in this multidisciplinary subject. The area of
bioinformatics is well adapted to the development of large volumes of data, which is ideal for
deep learning, but still it lacked molecular order. Because deep learning and bioinformatics are
fast emerging fields in modern environment, it's critical to solve research challenges in these
fields. In this chapter, researchers discussed various current and systematic DL approaches, many
of which have already been deployed to bioinformatics and others yet to be used. This viewpoint
may provide new insight on the use of recent DL approaches in bioinformatics in the near future.
References
1. Castiglioni, Isabella, Leonardo Rundo, Marina Codari, Giovanni Di Leo, Christian Salvatore,
Matteo Interlenghi, Francesca Gallivanone, Andrea Cozzi, Natascha Claudia D'Amico, and
Francesco Sardanelli. "AI applications to medical images: From machine learning to deep
learning." PhysicaMedica 83 (2021): 9-24.
2. S.Sasubilli, A. Kumar, V.Dutt, ''Machine Learning Implementation on Medical Domain to
Identify Disease Insights using TMS", 2020, Sixth International Conference on Advances in
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 10
Computing & Communication Engineering Las Vegas USA ICACCE 2020 (22-24 June)
ISBN: 978-1-7281-6362-8
3. S. A. Kumar, A. Kumar, V. Dutt and R. Agrawal, "Multi Model Implementation on General
Medicine Prediction with Quantum Neural Networks," 2021 Third International Conference
on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV),
Tirunelveli, India, 2021, pp. 1391-1395, doi: 10.1109/ICICV50876.2021.9388575.
4. Madiajagan, M., and S. Sridhar Raj. "Parallel machine learning and deep learning approaches
for bioinformatics." In Deep Learning and Parallel Computing Environment for
Bioengineering Systems, pp. 245-255. Academic Press, 2019.
5. S. M. Sasubilli, A. Kumar and V. Dutt, "Improving Health Care by Help of Internet of
Things and Bigdata Analytics and Cloud Computing," 2020 International Conference on
Advances in Computing and Communication Engineering (ICACCE), Las Vegas, NV, USA,
2020, pp. 1-4, doi: 10.1109/ICACCE49060.2020.9155042.
6. Azizah, Kurniawati, and MirnaAdriani. "Hierarchical Transfer Learning for Text-to-Speech
in Indonesian, Javanese, and Sundanese Languages." In 2020 International Conference on
Advanced Computer Science and Information Systems (ICACSIS), pp. 421-428. IEEE, 2020.
7. Li, Yu, Chao Huang, Lizhong Ding, Zhongxiao Li, Yijie Pan, and XinGao. "Deep learning in
bioinformatics: Introduction, application, and perspective in the big data era." Methods 166
(2019): 4-21.
8. Li, Haoyang, ShuyeTian, Yu Li, Qiming Fang, Renbo Tan, Yijie Pan, Chao Huang, Ying Xu,
and XinGao. "Modern deep learning in bioinformatics." Journal of molecular cell biology
(2020).
9. Abhishek Kumar, TvmSairam, Vishal Dutt, “Machine Learning Implementation For Smart
Health Records: A Digital Carry Card”, Global Journal on Innovation, Opportunities and
Challenges in AAI and Machine Learning Vol. 3, Issue 1 – 2019.
10. Larranaga, Pedro, Borja Calvo, Roberto Santana, Concha Bielza, Josu Galdiano, InakiInza,
José A. Lozano et al. "Machine learning in bioinformatics." Briefings in bioinformatics 7, no.
1 (2006): 86-112.
11. Usharani, S., D. Jayakumar, Dr U. Palani, D. Raghuraman, R. Parthiban, D. Saravanan, and
Dr D. Stalin David. "Industrialized Service Innovation Platform Based On 5g Network And
Machine Learning." European Journal of Molecular & Clinical Medicine 7, no. 11 (2020):
5684-5703.
12. SwarnAvinash Kumar, Harsh Kumar, SrinivasaRaoSwarna, Vishal Dutt, “Early Diagnosis
and Prediction of Recurrent Cancer Occurrence in a Patient Using Machine Learning”,
European Journal of Molecular & Clinical Medicine, 2020, Volume 7, Issue 7, Pages 6785-
6794.
13. R. Raturi and A.Kumar " An Analytical Approach for Health Data Analysis and finding the
Correlations of attributes using Decision Tree and W-Logistic Modal Process", 2019,
IJIRCCE Vol 7, Issue 6, ISSN(Online): 2320-9801 ISSN (Print) : 23209798.
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 11
14. Bala, P. Manju, S. Usharani, R. Rajmohan, and G. Leemaroselin. "Applications of Machine
Learning and Deep Learning for maintaining Electronic Health Records." Global Journal of
Internet Interventions and IT Fusion [ISSN: 2582-1385 (online)] 3, no. 1 (2020).
15. Usharani, S., P. ManjuBala, R. Rajmohan, and M. Pavithra. "Application of ML/DL for
improved Radiotherapy." Global Journal of Computing & Software Testing [ISSN: 2581-
7981 (online)] 2, no. 1 (2020).
16. Narmadha, S., S. Gokulan, M. Pavithra, R. Rajmohan, and T. Ananthkumar. "Determination
of various Deep Learning Parameters to Predict Heart Disease for Diabetes Patients." In 2020
International Conference on System, Computation, Automation and Networking (ICSCAN),
pp. 1-6. IEEE, 2020.
17. SwarnAvinash Kumar, Harsh Kumar, Vishal Dutt, Himanshu Swarnkar, “Role of Machine
Learning in Pattern Evaluation of COVID-19 Pandemic: A Study for Attribute Explorations
and Correlations Discovery among Variables”, (2020): Global Journal on Application of
Data Science and Internet of Things, Vol 4 No 2, [ISSN: 2581-4370].
18. Aiswarya, S., S. Usharani, K. Dhanalakshmi, and M. Roberts Masillamani. "A Cti Based Big
Data Sanitization Using Darknet Preproessing Model." International Journal of Pure and
Applied Mathematics 119, no. 14 (2018): 1095-1100.
19. Gokulan, S., S. Narmadha, M. Pavithra, R. Rajmohan, and T. Ananthkumar. "Determination
of Various Deep Learning Parameter for Sleep Disorder." In 2020 International Conference
on System, Computation, Automation and Networking (ICSCAN), pp. 1-6. IEEE, 2020.
20. S. Chandrasekaran and A. Kumar Implementing Medical Data Processing with Ann with
Hybrid Approach of Implementation Journal of Advanced Research in Dynamical and
Control Systems – JARDCS issue 10, vol.10, page 45-52, ISSN-1943-023X. 2018/09/15.
21. Usharani, S., P. ManjuBala, and R. Balamurugan. "Voice Based Form Filling System For
Visually Challenged People." In 2020 International Conference on System, Computation,
Automation and Networking (ICSCAN), pp. 1-5. IEEE, 2020.
22. SwarnAvinash Kumar, Harsh Kumar, Vishal Dutt, Himanshu Swarnkar, “Contribution Of
Machine Learning Techniques To Detect Disease In-Patients: A Comprehensive Analysis of
Classification Techniques”, Global Journal on Innovation, Opportunities and Challenges in
AAI and Machine Learning, Vol. 3, Issue 1 -2019, ISSN: 2581-5156.
23. Pavithra, M., Rajmohan, R., Kumar, T. A., &Ramya, R. (2021). Prediction and Classification
of Breast Cancer Using Discriminative Learning Models and Techniques. Machine Vision
Inspection Systems, Volume 2: Machine Learning‐Based Approaches, 241-262.
24. G. Sasubilli and A. Kumar, "Machine Learning and Big Data Implementation on Health Care
data," 2020 4th International Conference on Intelligent Computing and Control Systems
(ICICCS), 2020, pp. 859-864, doi: 10.1109/ICICCS48265.2020.9120906.
25. Vishal Dutt, Sriramakrishnan Chandrasekaran, Vicente García-Díaz, (2020). “Quantum
neural networks for disease treatment identification.”, European Journal of Molecular &
Clinical Medicine, 7(11), 57-67
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 12
26. Usharani, S., and S. G. Sandhya. "Detection of ransomware in static analysis by using
Gradient Tree Boosting Algorithm." In 2020 International Conference on System,
Computation, Automation and Networking (ICSCAN), pp. 1-5. IEEE, 2020.
27. Usharani, S., P. ManjuBala, and M. Martina Jose Mary. "Dynamic Analysis on Crypto-
ransomware by using Machine Learning: Gand Crab Ransomware." In Journal of Physics:
Conference Series, vol. 1717, no. 1, p. 012024. IOP Publishing, 2021.
28. Swarn Avinash Kumar, Harsh Kumar, Vishal Dutt, Pooja Dixit, “The Role of Machine
Learning in COVID-19 in Medical Domain: A Survey”, Journal on Recent Innovation in
Cloud Computing, Virtualization & Web Applications, Vol 4 No 1 (2020), [ISSN: 2581-
544X]
29. Usharani, S., and R. Ramya. "Security Based Novel Context Aware Mobile Computing
Scheme Via Crowdsourcing." (2017): 2395-1990.
30. Deiva Ragavi, M., and S. Usharani. "Social data analysis for predicting next event." In
Information Communication and Embedded Systems (ICICES), 2014 International
Conference. 2014.
31. Vikas Kumar Singh, Dr. Sanjay Pawar, Lohit Shekam, Vishal Dutt (2020),” Impact Of Covid
19 On Fmcg Sector.” Journal of Critical Reviews, 7 (12), 4477-4484. doi:10.31838/
jcr.07.12.640.
32. Usharani, S., Dhanalakshmi, K., Dhanalakshmi, N. “De-duplication techniques in centralised
billing systems”, International Journal of Scientific and Technology Research, 2020, 9(1),
pp. 2602-2606.
33. Swarn Avinash Kumar, Harsh Kumar, Vishal Dutt, Pooja Dixit, “Deep Analysis of COVID-
19 Pandemic using Machine Learning Techniques”, (2020): Global Journal on Innovation,
Opportunities and Challenges in AAI and Machine Learning, Vol 4 No 2, [ISSN: 2581-
5156].
34. S. A. Kumar, H. Kumar, V. Dutt and H. Soni, "Self-Health Analysis with Two Step
Histogram based Procedure using Machine Learning," 2021 Third International Conference
on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV),
Tirunelveli, India, 2021, pp. 794-799, doi: 10.1109/ICICV50876.2021.9388427.
35. Bala, P. Manju, S. Usharani, and V. Abarna. "Detect the Replication Attack on Wireless
sensor network by Using Intrusion Detection System." In Journal of Physics: Conference
Series, vol. 1717, no. 1, p. 012023. IOP Publishing, 2021.
36. SwarnAvinash Kumar, Harsh Kumar, Vishal Dutt, Himanshu Swarnkar, “COVID-19
Pandemic analysis using SVM Classifier: Machine Learning in Health Domain”, Global
Journal on Application of Data Science and Internet of Things, 2020, Vol 4 No. 1.
37. Bala, P. Manju, S. Usharani, and M. Aswin. "IDS based fake content detection on Social
Network Using Bloom Filtering." In 2020 International Conference on System, Computation,
Automation and Networking (ICSCAN), pp. 1-6. IEEE, 2020.
38. Bala, P. Manju, J. Kayalvizhi, S. Usharani, and D. Jayakumar. "A decentralized file
shareing& data transmission in peer to peer communication using edonkey protocol."
International Journal of Pure and Applied Mathematics 119, no. 14 (2018): 1027-1032.
Global Journal on Innovation, Opportunities and Challenges in AAI and Machine Learning
Vol. 5, Issue 1 – 2021
ISSN: 2581-5156
© Eureka Journals 2021. All Rights Reserved. Page 13
39. S. R. Swarna, S. Boyapati, V. Dutt and K. Bajaj, "Deep Learning in Dynamic Modeling of
Medical Imaging: A Review Study," 2020 3rd International Conference on Intelligent
Sustainable Systems (ICISS), Thoothukudi, India, 2020, pp. 745-749, doi: 10.1109/
ICISS49785. 2020.9315990.
40. Rajendirane Rajmohan, Kuttathati SrinivasanVishvaksenan & Rajakani Kalidoss (2021)
Compact four port slot based MIMO antenna with isolation enhancement, International
Journal of Electronics, DOI: 10.1080/00207217.2021.1908625
41. S. Boyapati, S. R. Swarna, V. Dutt and N. Vyas, "Big Data Approach for Medical Data
Classification: A Review Study," 2020 3rd International Conference on Intelligent
Sustainable Systems (ICISS), Thoothukudi, India, 2020, pp. 762-766, doi: 10.1109/ICISS
49785. 2020.9315870.
42. Jayasri, K., Rajmohan, R., & Dinagaran, D. (2015, May). Analyzing the query performances
of description logic based service matching using Hadoop. In 2015 International Conference
on Smart Technologies and Management for Computing, Communication, Controls, Energy
and Materials (ICSTM) (pp. 1-7). IEEE.
43. Vishal Dutt, Rohit Raturi, Vicente García-Díaz, Sreenivas Sasubilli, “Two-Way Bernoulli
distribution for Predicting Dementia with Machine Learning and Deep Learning
Methodologies”, Solid State Technology, 63(6), pp.: 9528-9546.
44. Rajmohan, R., & Padmapriya, N. (2012, July). A Domain ontology based service matching
for CHORD based Super Peer network. In 2012 International Conference on Data Science &
Engineering (ICDSE) (pp. 214-219). IEEE.
