Proceedings of the 2nd International conference on Electronics, Communication and Aerospace Technology (ICECA 2018)
IEEE Conference Record # 42487; IEEE Xplore ISBN:978-1-5386-0965-1
Machine Learning in Healthcare: A Review
K. Shailaja B. Seetharamulu M. A. Jabbar
M. Tech Scholar Professor Professor
Centre for Data Science Department of CSE Centre for Data Science
Vardhaman College of Engineering Vardhaman College of Engineering Vardhaman College of Engineering
Hyderabad, Telangana Hyderabad, Telangana Hyderabad, Telangana
[email protected] [email protected] [email protected]Abstract-Machine Learning is modern and highly sophisticated B. Semi-supervised Learning
technological applications became a huge trend in the industry.
Machine Learning is Omni present and is widely used in various It is the method of identifying the best classifier from the
applications. It is playing a vital role in many fields like finance, each unlabeled and labeled information. By using unlabeled
Medical science and in security. Machine learning is used to
information it transfers high performance of classification.
discover patterns from medical data sources and provide
The success of this method totally depends on few
excellent capabilities to predict diseases. In this paper, we review
various machine learning algorithms used for developing underlying assumptions.
efficient decision support for healthcare applications. This paper
helps in reducing the research gap for building efficient decision Machine Learning
support system for medical applications.
Keywords- Machine Learning; Medical Science; Disease
Prediction; Healthcare; Decision Support System; Heart Disease. Supervised Unsupervised Reinforcement Semi-
Supervised
I. INTRODUCTION
Machine learning could be a broad multi-disciplinary Fig 1: Machine Learning Classification Techniques
filed that has its roots in statistics, algebra, data processing,
C. Unsupervised Machine Learning
and knowledge analytics etc., that makes it tough to return
up with a novel definition [1]. ML is a particular method of In Unsupervised learning, there is no training given to
artificial intelligence it collects information from training the labels. The technique used for unsupervised learning is
data. In this learning, we are not informing the machines clustering, fuzzy clustering, hierarchical clustering, K
where to look, and it is at the foundation of the tree and has means clustering, association rule mining of the algorithm
a lot of branches and sub-branches. Machine learning is split that comes under this learning [4]. Here a cluster is formed
into the following categories as represented by figure 1 [2]. on the availability of trained data with unknown labels.
These algorithms are used for a developing framework with
A. Supervised Learning.
the help of data samples.
B. Semi-supervised Learning
D. Reinforcement Learning
C. Unsupervised Learning
In this learning, access is given by computer program to
D. Reinforcement Learning the dynamic environment for performing a specific goal.
Feedback in terms of rewards and punishments is provided
A. Supervised Machine Learning to the program because it navigates its drawback [5].
This learning is taken under training and algorithm II. NEED OF DECISION SUPPORT SYSTEM FOR
develops an exercise that matches inputs to related outputs. HEALTHCARE
One common establishment of the supervised learning task
is the classification issue [3]. Most of the American citizens die every year because of
errors present in the health care system, and thousands of
people deteriorate from nonfatal burns owing to the constant
cause. Health Information Technology (IT) Framework [6]
recommended few strategies, like collaboration, knowing
978-1-5386-0965-1/18/$31.00 ©2018 IEEE 910
Proceedings of the 2nd International conference on Electronics, Communication and Aerospace Technology (ICECA 2018)
IEEE Conference Record # 42487; IEEE Xplore ISBN:978-1-5386-0965-1
consumer selection of clinicians and organizations, and IT analyze in a database and they will carry a set of statistics
adoption. and classes as well as however not restricted to patient
weights, and even generic symptoms like stomach pain,
A. Decision support headache, etc. [10]. The bulk of medical knowledge is
unstructured information within the variety of numerous
An efficient Machine Learning based health protection
completely different notes, images, audio and video
system capitalizes on the computer’s large computing
recording, reports, and discharge summaries. It’s terribly
capability and also the doctor’s reasoning ability. Both
exhausting to quantify and analyze a conversation between
machine and doctor are searching for patterns, but the
the supplier and the patient; the conversation is incredibly
doctor cannot evaluate the heartbeat of each patient or be
personalized and might take many alternative directions
aware of every disease’s nuances. The machine will do all
[11].
of these tasks then present its outcomes to the doctor for
confirmation. B. Applications of ML in Healthcare
B. Decision Support System in Healthcare The algorithms of Machine learning are useful in
identifying complicated patterns within prosperous and huge
Decision Support System helps the finance branch of
data. This facility is especially well-suited to clinical
the clinic keep track charge, accounts receivable, outlay and
applications, particularly those people who rely on advanced
accounts payable. This method additionally helps to keep
genomics and proteomics measurements. It is often used
the patient insurance policy, different refund choices. It’s an
utilized in numerous illness diagnosing and detection. In
organization that gives completely different modules for
medical applications, machine learning algorithms will
DSS in healthcare [7].
manufacture higher decisions regarding treatment plans for
Decision Support System helps in disease investigation patients by suggestions of implementing useful health-care
by arranging a perception of health issues to the doctors or system [12].
else by exposing background knowledge about individual
Healthcare management is utilizing this method to
patients. It additionally gives help with identifying a
forecast wait times for patients in exigency department
patient’s situation and a guide suggest the patient when to
waiting for places. These models use factors like patient
use the proper drug at a specific time, a web-based structure
information, discomfort levels, exigency department charts,
that has been involved with computerized
and even the layout of the hospital room itself to conclude
patient/therapeutic record.
wait times.
III. HEALTHCARE USING MACHINE LEARNING
Using the prognostic model, clinics will think hospital
A. Introduction room admissions. So machine learning application could
profit patients by decreasing price, rising accuracy, or
The hazardous improvement of health-related information diffusing experience that is in brief offer.
given new opportunities for developing recuperate of a
patient, Machine learning shows a vital performance in C. Different Techniques Used by ML
health-care and these are mostly enforced to healthcare,
1. Support Vector Machine
which includes computer-aided diagnosis, image
registration, image annotation, image-guided medical aid, Support Vector Machine (SVM) which is designed in
and image database retrieval, multimodal image fusion, 1990’s. To achieve machine learning (ML) tasks support
medical image segmentation, where deficiency might be vector machine (SVM) is used, and it is a simple and
incurable. prominent process. During this technique, a collection of
training samples is given each sample is divided into
Machine learning has probably limited social impacts in
different categories. Support vector machine (SVM) mainly
the health-care field [8]. Machine learning provides the
used for classification and regression problems [13].
solution for decreasing the increasing price of health-care
and serving to create an improved patient-clinician 2. Naive Bayes classification
communication. ML solutions will be used for an
inordinateness of health-relevant uses; some include serving Statistical classifiers are the example for Bayesian
to clinicians identify a lot of customized prescriptions and classifiers. Naive Bayes identify the class membership
therapy for patients and additionally serving to patients probabilities based on given class label [14]. It performs one
identify once and if they must record follow up scanning of data and hence classification is easy.
appointments.
3. Decision Tree
Currently, in health-care, a huge quantity of information
has become accessible. It contains EMRs that consist of Decision Tree (DT) is mostly used technique for
information which may be either unstructured or structured classification containing internal node and one leaf node
[9]. Structured health information is the data that’s simple to with a class label. The top nodes of the decision tree (DT)
911
Proceedings of the 2nd International conference on Electronics, Communication and Aerospace Technology (ICECA 2018)
IEEE Conference Record # 42487; IEEE Xplore ISBN:978-1-5386-0965-1
are called as root nodes. The decision tree is very popular C. Machine Learning in Disclosure of Breast cancer
because the construction is very simple, which won’t
require any parameters [15]. It is one of the top cancer that occurs in a woman and it is
the second main leading reason for woman in the United
4. K-nearest neighbor States and in Asia countries. Some machine learning
algorithms are used to predict breast cancer. The data
K-nearest neighbor is frequently used approach for considered from WISCONSIN dataset UCI machine
classification of samples. By using this technique we can learning repository.
calculate distance measure from N number of training
samples [16]. Williams et al. used a j48, Naive Bayes to identify breast
cancer risks in the United States. The experiment is
5. Fuzzy Logic performed through WEKA tool. They conclude j48 is the
best algorithm for the prediction of breast cancer it gives
Fuzzy Logic which is evolved from Fuzzy set theory.
94.2% accuracy, and Naive Bayes gives 82.6% [23].
These values are lies in between 0 and 1. It is a very popular
method which is used in engineering applications [17]. To predict breast cancer Senturk et al. used several
classification models like Support Vector Machine (SVM),
6. CART
Naive Bayes (NB), K-nearest neighbor, and Decision tree
(DT). K-NN gives 95.15% accuracy and SVM gives
Classification and Regression Tree Methodology is
96.40% accuracy [24].
called as the CART. In classification and regression trees
the target variable represented as categorical and Majali et al. used decision tree and Frequent Pattern in
continuous. These variables are used to predict values in the data mining to predict the breast cancer. They conclude
tree [18].
decision tree gives 94% accuracy [25].
D. Diagnosis of Thyroid Disorder
IV. ML TECHNIQUES USED FOR PREDICTION
OF VARIOUS DISEASES To predict thyroid diseases machine learning techniques
are used. Classification algorithms that are support vector
A. Detection of Heart Disease machines and Decision tree are used and dataset considered
from UCI repository.
To increase the accuracy of diagnosing in the Heart
disease Machine Learning Techniques are frequently used. Papageorgiou EI, Papandrianos NI proposed advanced
Dataset considered from the UCI Machine Learning approaches for thyroid diagnosing diseases using fuzzy map
Repository. utilizing data mining algorithms [26]. The below table
summarizes different ML techniques used for diagnosis of
Parthiban and Srivatsa proposed a machine learning
various diseases.
algorithm for detection and analysis of heart disease by
utilizing Naive Bayes algorithm, Support vector machine
[19]. By utilizing Naïve Bayes algorithm provides 74%
accuracy and SVM offer 94.60%.
Otoom has performed Support Vector Machine, Bayes
Net to predict coronary heart disease [20]. The accuracy
provided by SVM is 88.3%, Bayes Net provides 84%.
B. Analysis of diabetic Diseases
To increase the accuracy of scrutiny of diabetic diseases
various machine learning techniques are used. Dataset
considered from the UCI Machine Learning Repository.
Iyer proposed a machine learning algorithm to predict
diabetic disorder by using Naive Bayes and Decision trees.
Naive Bayes gives 79.56% accuracy and decision tree
provides 76.95% accuracy [21].
Dash and Sen performed Machine learning algorithms for
diagnosing diabetes disease. Logiboost, CART algorithms
are used and Logiboost provides the correctness of 77.479%
[22].
978-1-5386-0965-1/18/$31.00 ©2018 IEEE 912
Proceedings of the 2nd International conference on Electronics, Communication and Aerospace Technology (ICECA 2018)
IEEE Conference Record # 42487; IEEE Xplore ISBN:978-1-5386-0965-1
TABLE 1: ML Techniques for Diagnosis of Various Diseases V. CONCLUSION
S. Reference MLTechni Disease Dataset Accura Machine learning (ML) techniques are crucial in different
n que used cy business fields. Healthcare field facing more problems and
o it is becoming more expensive. Several ML techniques are
1 Chaurasia and J48 Heart 84.35% used to rectify them. This paper presents various ML
Pal [27] SVM Disease UCI 85.03% techniques for prediction of various diseases like heart
disease, breast cancer, diabetic disease and thyroid disease.
From the earlier study, it is recognized that naive Bayes
2 Vembandasamy Naïve Heart Diabetic 86.41%
Researc provides 86% of accuracy for the diagnosis of heart disease.
[28] Bayes Disease
h SVM gives 96.40% of accuracy for the breast cancer
Institute
diagnosis, and CART provides 79% of accuracy for the
in
detection of diabetic disease. In future, we are trying to
Chennai
improve the accuracy of breast cancer prediction by using
3 Kumari and SVM Diabetic 78%
different machine learning algorithms.
Chitra [29] Disease UCI
4 Sarwar and Naive Diabetes Differen 95% References
Sharma [30] Bayes type-2 t Sectors [1] T. Mitchell, “ Machine Learning”, McGraw Hill. P. 2, 1997.
of
Society [2] Apoorva Mishra and Anupam Shukla, “From Machine Learning to
Deep Learning Trends and Challenges”, CSI Communications, December
in India 2018.
5 Shrivastava et al Breast UCI
[31] J48 Cancer machine 98.14% [3] Allix, N. M. “Epistemology And Knowledge Management Concepts
And Practices”. Journal of Knowledge Management Practice.
Learnin
g [4] Anil Mathur, G. P. “Socialization influences on preparation for later
Reposit life”. Journal of Marketing Practice: Applied Marketing Science, vol.5,
163 – 176, 2007.
ory
6 Venkatesan et Decision Breast Swami 97% [5] LP. Kaelbling and Aw. Moore, “Reinforcement Learning A Survey”,
Vivekan Journal of Artificial Intelligence Research, vol-4, pp.237-285, 1996.
Tree Cancer
al [32] anda
Diagnos [6] T.G. Thompson and D.J. Brailer, “The decade of health information
tic
technology: Delivering consumer-centric and information-rich health care:
Centre
Hospital Framework for strategic action”, 2007.
7 Sivakami [33] DT+SVM Breast UCI 91% [7] K. Rajalakshmi, S. Chandra Mohan, S. Dhinesh Babu, “Decision
Cancer Machine Support System in Healthcare Industry”, International Journal of Computer
Applications, vol.9, 2013.
Learnin
g [8] Rohan Bhardwaj, Ankitha R. Nambiar, “A Study of Machine Learning
in Healthcare”, IEEE 41st Annual Computer Software and Applications
Reposit
Conference, 2017.
ory
8 Shajahaan et al CART Breast Universi 92.42% [9] M. Hauskrecht , Visweswaran S., Cooper G., Clermont G.: Clinical
Naïve Alerting of Unusual Care that Is Based on Machine Learning , 2015.
[34] Cancer ty of 97.42%
Bayes [10] M. Kohn, “Real World Data and Clinical Decision Support”,
Wiscons
ELSEVIER, 2009.
in
hospital [11] D. Maddux, “The Human Condition in Structured and Unstructured
s Data”.
9 M.R.Nazari SVM Thyroid UCI 98.62%
[12] Mrs.J. Sukanya, “Applications of Big Data Analytics and Machine
Kousarrizi et al Disease Learning Techniques in Health Care Sectors”, International Journal of
[35]. Engineering and Computer Science, vol.6, pp.21963-21967, 2017.
[13] K .P Murphy, Machine Learning: A Probabilistic Perspective, The
MIT Press, 2012.
[14] Animesh et al., “Study and analysis of Breast cancer Cell Detection
using Naïve Bayes, SVM and Ensemble Algorithms”, International Journal
of Computer Applications, vol.2, 2016
[15] P. Sharma and A. P. R. Bhartiya, “Implementation of Decision Tree
Algorithm to Analysis the Performance”, vol. 10, pp. 861–864, 2012.
[16] C.M.Bishop(1995),” Neural networks for pattern recognition”,
England: Oxford University .
913
Proceedings of the 2nd International conference on Electronics, Communication and Aerospace Technology (ICECA 2018)
IEEE Conference Record # 42487; IEEE Xplore ISBN:978-1-5386-0965-1
[17] H. Zimmermann, Fuzzy set theory and its applications.
Kluwer Academic Publishers. 2001.
[18] Jalpaiguri et al.,” Heart Disease Diagnosis and Prediction Using
Machine Learning and Data Mining Techniques: A Review“, Advances in
Computational Sciences and Technology, vol.10, pp. 2137-2159, 2017.
[19] Parthiban, G. and Srivatsa, S.K., “Applying Machine Learning
Methods in Diagnosing Heart Disease for Diabetic Patients”, International
Journal of Applied Information Systems, vol.3, pp.25-30, 2012.
[20] Otoom A.F., Abdallah E.E., Kilani Y., Kefaye A., “Effective
Diagnosis and Monitoring of Heart Disease”, International Journal of
Software Engineering and Its Applications, vol.9, pp.143-156, 2015.
[21] Iyer A., Jeyalatha S. and Sumbaly R, “Diagnosis of Diabetes Using
Classification Mining Techniques”, International Journal of Data Mining &
Knowledge Management Process (IJDKP), vol.5, pp. 1-14,2015.
[22] Sen S.K. and Dash S, “Application of Meta Learning Algorithms for
the Prediction of Diabetes Disease”, International Journal of Advance
Research in Computer Science and Management Studies, vol.2, pp.396-
401, 2014.
[23] Williams et al., “Breast cancer risk prediction using data mining
classification techniques”, Transactions on Networks and
Communicatons,vol.2, pp.1-11,2015.
[24] Senturk et al., "Breast cancer Diagnosis via Data Mining: Performance
Analysis of Seven Different Algorithms ", Computer Science &
Engineering”, vol.1, pp.1-10, 2014.
[25] Majali et al., ”Data Mining Techniques for Diagnosis and Prognosis of
Cancer”, International Journal of Advanced Research in Computer and
Communication Engineering, vol.3, pp. 613-616, 2015.
[26] Papageorgiou EI, Papandrianos NI, Apostolopoulos DJ, Vassilakos
PJ., “ Fuzzy cognitive map based decision support system for thyroid
diagnosis management”, International Conference on Fuzzy Systems pp.
1204-1211,2008.
[27] Chaurasia, V. and S. Pal,” Data Mining Approach to Detect Heart
Disease”, International Journal of Advanced Computer Science and
Information Technology, vol.2,pp.56-66,2018.
[28] Vembandasamy et al., “Heart Diseases Detection Using Naive Bayes
Algorithm”, vol.2, pp. 441-444, 2015.
[29] Kumari, V.A. and R. Chitra, “Classification of Diabetes Disease Using
Support Vector Machine”, International Journal of Engineering Research
and Applications, vol.3, pp. 1797-1801, 2013.
[30] Sarwar, A. and Sharma, V. , ”Intelligent Naïve Bayes Approach to
Diagnose Diabetes Type-2. Special Issue”, International Journal of
Computer Applications and Challenges in Networking, Intelligence and
Computing Technologies, vol. 3, pp.14-16, 2012.
[31] Shrivastava et al., “An Overview on Data Mining Approach on Breast
Cancer data”, International Journal of Advanced Computer Research,
vol.3, pp. 256-262, 2013.
[32] E. Venkatesan et al., “Performance analysis of decision tree
algorithms for breast cancer classification”, Indian Journal of Science and
Technology, vol.8, pp.1-8, 2015.
[33] K. Sivakami, “Mining Big Data: Breast Cancer Prediction using DT-
SVM Hybrid Model”, International Journal of Scientific Engineering and
Applied Science vol.1, pp.418-429, 2015.
[34] S. Shajahaan et al., “Application of Data Mining techniques to model
breast cancer data” International Journal of Emerging Technology and
Advanced Engineering, vol.3, pp.362-369, 2013.
[35] M. R. Nazari Kousarrizi et al., “An Experimental Comparative Study
on Thyroid Disease Diagnosis Based on Feature Subset Selection and
classification”, International Journal of Electrical& Computer Sciences,
vol.1,pp.13-19,2012.
978-1-5386-0965-1/18/$31.00 ©2018 IEEE 914
