Diabetic Retinopathy: Severity Level Classification based on Object Detection (Microaneurysms, Hemorrhages, and Hard Exudates) using Mathematical Morphology and Neural Networks

Fifi Diah Rosalina, Dian C. Rini Novitasari, Ahmad Hanif Asyhar, Abdulloh Hamid, Muhammad Firmansjah

2018

Abstract

According to WHO, the number of people with diabetes had reached approximately 422 million people in 2014. Most cases of diabetes in the world occur as type 2 diabetes which can result in serious complications of vital organs such as the eyes, kidneys, and heart, causing death. Complications of diabetes often causes people with Diabetic Retinopathy (DR) to be unaware of the disease for several years, which can lead to permanent blindness. Early detection of DR is indicated by the presence of microaneurysms, bleeding, and hard exudates. Five DR severity classifications which include normal, mild-NPDR, moderate-NPDR, severe-NPDR and proliferative DR are performed using the backpropagation method. Detection methods for microaneurysms and hemorrhages are based on diamond disc and morphology opening methods, while detection of hard exudate features is based on morphology reconstruction methods and minimal area of images. The area and perimeter of each feature is applied as a backpropagation input with 80% of training data and 20% of test data from the total of 53 image data. Four inputs with 150 hidden layers are arranged as a network structure capable of producing MSE values of 0.000190 and an accuracy rate of 90.90%.

Paper Citation

in Harvard Style

Rosalina F., Novitasari D., Asyhar A., Hamid A. and Firmansjah M. (2018). Diabetic Retinopathy: Severity Level Classification based on Object Detection (Microaneurysms, Hemorrhages, and Hard Exudates) using Mathematical Morphology and Neural Networks. In Proceedings of the Built Environment, Science and Technology International Conference - Volume 1: BEST ICON, ISBN 978-989-758-414-5, pages 272-280. DOI: 10.5220/0008906300002481

in Bibtex Style

@conference{best icon18,
author={Fifi Diah Rosalina and Dian C. Rini Novitasari and Ahmad Hanif Asyhar and Abdulloh Hamid and Muhammad Firmansjah},
title={Diabetic Retinopathy: Severity Level Classification based on Object Detection (Microaneurysms, Hemorrhages, and Hard Exudates) using Mathematical Morphology and Neural Networks},
booktitle={Proceedings of the Built Environment, Science and Technology International Conference - Volume 1: BEST ICON,},
year={2018},
pages={272-280},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0008906300002481},
isbn={978-989-758-414-5},
}

in EndNote Style

TY - CONF

JO - Proceedings of the Built Environment, Science and Technology International Conference - Volume 1: BEST ICON,
TI - Diabetic Retinopathy: Severity Level Classification based on Object Detection (Microaneurysms, Hemorrhages, and Hard Exudates) using Mathematical Morphology and Neural Networks
SN - 978-989-758-414-5
AU - Rosalina F.
AU - Novitasari D.
AU - Asyhar A.
AU - Hamid A.
AU - Firmansjah M.
PY - 2018
SP - 272
EP - 280
DO - 10.5220/0008906300002481