e-ISSN: 2582-5208

International Research Journal of Modernization in Engineering Technology and Science

( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:06/June-2022 Impact Factor- 6.752 www.irjmets.com
PREDICTION OF HEART DISEASE USING RETINAL IMAGES
M. Rupadevi*1, P. Sandhya Rani*2, K. Mounika*3, G. Lochana*4,
Bhavana*5, Ms. D. Ramya*6
*1,2,3,4,5B. Tech IV, Computer Science and Engineering, Vignan’s Institute of Engineering for
Women, Visakhapatnam, India.
*6Assistant Professor, Computer Science and Engineering, Vignan’s Institute of Engineering for
Women, Visakhapatnam, India.
ABSTRACT
Heart disease increases the mortality rate in recent years across the world. So, it is necessary to develop a
model to predict heart disease occurrence as early as possible with a higher rate of accuracy. Till now the
detections are gone through blood tests, ECGs, and invasive stress tests. In this project, heart disease is
predicted by a non-invasive method with the retinal image data. A Chase image dataset is considered, as the
health of our eyes is connected to the health of our heart. Here, Heart problems can be detected from the
changes in the microvasculature, which is imaged from the retina. The prediction of disease is by considering
features like the size of blood vessels, non-uniform background illumination, etc. We use Image processing for
identifying patterns in images and the Support Vector Machine (SVM) and Random Forest Classifier (RFC)
algorithm for classification. The main objective of the proposed system is to predict the occurrence of heart
disease from retinal fundus images with a higher rate of accuracy.
Keywords: Image processing, Heart disease, Retina, Microvasculature, SVM, RFC.
I. INTRODUCTION
Heart disease is one in each of the key causes of to increase in the fatality rate at intervals in the developed
world. Therefore, the prediction of heart conditions is unbelievably necessary and up to decrease the fatality
rate. There area unit many techniques on the market to sight the prevalence of heart conditions. Yet, they are
expensive to sight the unwellness and to boot take longer. The relation between heart and eye is high. The
membrane is one in each of the required choices that facilitate direct microcirculation. The membrane provides
a window for the detection of changes in the microvasculature in the event of a heart condition. throughout this
project machine learning, classification algorithms unit of measurement was involved in a method of structure
photos. As the health of our eyes is connected to the health of our heart. The upset area unit is usually detected
from the changes in blood vessels, that are imaged from the membrane. The prediction of unwellness is
completed by extracting the choices like the size of blood vessels, non-uniform background illumination, etc.
from the retinal image.
In the recent years, the image method that deals with extracting useful data from a digital image play a unique
II. LITERATURE SURVEY
A study undertaken and enhanced by Melillo et al. [1]; have used a machine learning algorithm named CART
which stands for Classification and Regression in which sensitivity and specificity are achieved at 93.3 percent
and 63.5 percent respectively. Then to improve performance, an electrocardiogram (ECG) approach which was
suggested by Rahhal et al. [2] in which deep neural networks are used for choosing the best features and then
using them. Then, to detect heart failure, a clinical decision support system by Guidi et al. [3] for preventing it at
an early stage. They tried to compare different machine learning models and deep learning models, especially
neural networks such as Support Vector Machine, Random Forest, and CART algorithms. An accuracy of 87.6
percent was achieved by Random Forest and CART, which outperformed all used in the classification. SVM
techniques are used for binary classification i.e., to detect the patients who already have diabetes and then
predict heart disease by Parthiban and Srivatsa [4] achieved an accuracy rate of 94.60 and the characteristics
collected were common such as blood glucose level, the patient’s age, and their blood pressure.
Heart disease is very deadly and should not be taken lightly. According to Harvard Health Publishing [12], heart
disease is more common in men than women.

www.irjmets.co @International Research Journal of Modernization in Engineering, Technology and

Science
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:06/June-2022 Impact Factor- 6.752 www.irjmets.com
For decreasing the dimensionality of the dataset, there are various feature engineering and feature selection
techniques that can be used to remove that data that does not have that much importance in the dataset [6].
Harvard Medical School [12]2020Hungarian-Cleveland datasets were used for predicting heart disease using
various machine learning classifiers and PCA was used for dimensionality reduction and feature selection. A
common problem in machine learning is the high dimensionality of the data. The datasets that can be used
contain huge amounts of data and sometimes we can't even view the data in 3D, also known as the curse of
dimensionality [5].
The previous studies are mainly based on a 13-feature dataset. Classification is common in any study to predict
whether or not a patient has heart disease, and the most common pattern seen is the dataset commonly used in
Cleveland [6]. The results obtained achieved great accuracies like a Random forest with 89.2 percent accuracy
[7]; decision tree with 89.1 percent accuracy [8]; ANN with 92.7 percent accuracy [9], 89 percent [9], and 89.7
percent accuracy [10]; and SVM accuracy with 88 percent [10]. A hybrid model is created which achieved an
accuracy of 94.2 percent by GA þ NN [11].
III. PROPOSED WORK
The goal of this system is to a develop Heart Disease Prediction system that can recognize heart disease using
retinal images, especially in children. In this project, a CHASE image dataset (Child Heart and Health Study in
England dataset) is considered for detecting heart disease via the retina. The changes in the eye can be an
indication of many symptoms. The classifiers like Support Vector Machine Classifier (SVM) and Random Forest
Classifier (RFC) are used to identify and the results are compared.

Fig 3.1: Content Diagram of Proposed System

a. Input Database
CHASE-DB1 is also a publicly available dataset produced by Child health and study England (CHASE). This has
taken retinal images from 14 children and prepared a dataset. It maintains 28 retinal images of children for
both testing and training purposes.
The retinal image suffers from imperfections such as poor contrast and noise, that need to be reduced or
eliminated before processing. The Retinal background is pre-processed with a Gaussian filter. The image
dataset is input and then pre-processed by subtracting the original image with the Gaussian value to remove
the Gaussian noise.

www.irjmets.co @International Research Journal of Modernization in Engineering, Technology and

Science
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:06/June-2022 Impact Factor- 6.752 www.irjmets.com
b. Histogram Equalization
Histogram Equalization is a computer image processing technique used to improve contrast in images. It
accomplishes this by effectively spreading out the most frequent intensity values, i.e., stretching out the
intensity range of the image. The contrast adjustment is done using the image histograms. To improve the
quality of the image this step is very much essential and also helpful in extracting the features.
c. Image segmentation/Extraction of blood vessels
Retinal blood vessel extraction is a process to get a binary vessel map where retinal blood vessels are labeled as
logic 1 (white) and retinal background pixels are labeled as logic 0 (black) or vice versa. In this step, the
thickness of the blood vessel is identified. The area of the blood vessel is calculated to identify the stage of the
disease. If the blood vessels are found to be thick then the person is more prone to heart disease. After
calculating the area, the next step is the extraction of features.
d. Feature Extraction
One of the most important step is the extraction of features. The extracted features will be given to the classifier
for purpose of classification. The features like contrast, entropy, auto-correlation, homogeneity, energy, mean,
variance, standard deviation, dissimilarity, and clustering. Based on these features the heart disease will be
identified by the classifier.
e. Classification
Finally, two types of classifiers are used to identify heart disease. One is a Support Vector Machine classifier in
which training and testing of the dataset are performed. A non-linear SVM class model is used to determine the
disease. The other classifier is the Random Forest Classifier(RVC). The helps in identifying more accurately by
forming a tree structure and evaluating the input data by taking a random decision based on the features of the
image.
f. Identification of Disease
After the completion of classification, a performance matrix will be displayed for both classifiers, which are the
Support Vector Machine Classifier (SVM) and Random Forest Classifier (RFC). Later, a comparison graph of
both classifiers will be displayed. Finally, the output will be displayed as either diseased or not diseased.
IV. RESULTS

Figure 4.1: After Image Preprocessing

www.irjmets.co @International Research Journal of Modernization in Engineering, Technology and

Science
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:06/June-2022 Impact Factor- 6.752 www.irjmets.com

Figure 4.2: After Histogram Equalization

Figure 4.3: After Image Segmentation

Figure 4.4: SVM Classifier Performance Analysis

www.irjmets.co @International Research Journal of Modernization in Engineering, Technology and

Science
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:06/June-2022 Impact Factor- 6.752 www.irjmets.com

Figure 4.5: RFC Classifier Performance Analysis

Figure 4.6: Collation Graph

Figure 4.7: Non Diseased Input Image

www.irjmets.co @International Research Journal of Modernization in Engineering, Technology and

Science
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:06/June-2022 Impact Factor- 6.752 www.irjmets.com

Figure 4.8: Output of Not Diseased

Figure 4.9: Diseased Input Image

Figure 4.10: Output of Diseased Image

V. CONCLUSION
Prediction of heart disease occurrence for a patient by a non-invasive method is very essential with the
increased population and to reduce the mortality rate in the developing world. Prediction of risk for the
occurrence of cardiovascular disease from the given input of retinal dataset with high accuracy is developed in
this study. The classifiers like SVM and RFC are used for the identification of heart disease. The results are
compared using both the classifier, in which RFC is shown to be good in identifying the heart disease. In this
work, a Chase image dataset is considered for detecting heart disease via the retina. The changes in the eye can
be an indication of many symptoms.
Further, we discuss, the process of preprocessing of data, the process of feature extraction, and the process of
SVM and Random Forest classifiers, Followed by experimental results evaluated.
VI. REFERENCES
[1] P. Melillo, M. Bracale, and L. Pecchia, “For risk assessment in patients a classification tree with
congestive heart failure via long-term heart rate variability,” IEEE Journal of Biomedical and Health
Informatics, vol. 17, no. 3, pp. 727–733, 2013.
[2] M. M. A. Rahhal, Y. Bazi, H. Alhichri, N. Alajlan, F. Melgani, and R. R. Yager, “For active classification of
electrocardiogram signals he uses a Deep learning approach” Information Sciences, vol. 345, pp. 340–
354, 2016.

www.irjmets.co @International Research Journal of Modernization in Engineering, Technology and

Science
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:06/June-2022 Impact Factor- 6.752 www.irjmets.com
[3] G. Guidi, M. C. Pettenati, P. Melillo, and E. Iadanza, “A machine learning system to improve heart failure
patient assistance,” IEEE Journal of Biomedical and Health Informatics, vol. 18, no. 6, pp. 1750–1756,
2014.
[4] G. Parthiban and S. K. Srivatsa, “Applying machine learning methods in diagnosing heart disease for
diabetic patients,” International Journal of Applied Information Systems, vol. 3, no. 7, pp. 25–30, 2012.
[5] A. Mueen, and E. Keogh “Curse of dimensionality,” in Encyclopedia of Machine Learning and Data
Mining, C. Sammut and G. I. Webb, Eds., Springer, Cham, Switzerland, 2017.
[6] H. Liu and H. Motoda, Feature Extraction, Construction and Selection, Springer, Cham, Switzerland,
1998.
[7] S. S. Khan and S. M. K. Quadri, “Prediction of angiographic disease strategies using rule-based data
mining techniques,” Biological Forum—An International Journal, vol. 8, no. 2, pp. 103–107, 2016.
[8] S. Kumar, “Predict and diagnose of heart disease using machine learning algorithms,” International
Journal of Engineering and Computer Science, vol. 6, no. 6, pp. 2319–7242, 2017.
[9] R. Das, I. Turkoglu, and A. Sengur, “Effective diagnosis of heart disease through neural networks
ensembles,” Expert Systems with Applications, vol. 36, no. 4, pp. 7675–7680, 2009.
[10] K. Srinivas, A. Govardhan, and G. Raghavendra Rao, “Analysis of coronary artery disease and prediction
of a myocardial infarction in coal mining regions using data mining techniques,” in Proceedings of 2010
5th International Conference on Computer Science & Education, pp. 1344–1349, IEEE, Hefei, China,
August 2010.
[11] N. G. B. Amma, “Cardiovascular disease prediction system using genetic algorithm and neural network,”
in Proceedings of the 2012 International Conference on Computing, Communication, and Applications,
IEEE, Dindigul, India, February 2018.
[12] Harvard Medical School, “Throughout life, heart attacks are twice as common in men than women,”
2020, https://www.health.harvard.edu/heart-health/throughout-life-heart-attacks-are-twice-as-
common-in-men-than-women.

www.irjmets.co @International Research Journal of Modernization in Engineering, Technology and

Science