Severity Assessment Of Retina Disease Using Pel-Dcnn Approach

Authors

  • Sheeja Mary F Research Scholar, Dept. of Computer Science and Engineering Annamalai University Annamalainagar – 608002
  • Dr. V. Asanambigai Assistant Professor, Dept. of Computer Science and Engineering, Annamalai University, Annamalainagar – 608 002
  • Dr. A. Lenin Fred Principal, Mar Ephraem College of Engineering and Technology, Marthandam

Keywords:

Interval Contrast Limited Adaptive Histogram Equalization (ICLAHE), Position Encoded Layer based Deep Convolutional Neural Network (PEL-DCNN), retinal disease, Sorensen Dice Similarity with Mask based FCM (SDSM-FCM), feature segmentation, severity classification and retinal fundus images

Abstract

The automatic identification of retinal disorders has proven to be more difficult and has received a lot of interest recently. Since visual impairments take many different forms, an efficient approach for screening the retina is needed. The majority of the studies were carried out by the researchers to categorize the retinal illnesses, and the ophthalmologists often use colored fundus images to diagnose the abnormalities. However, low contrast difficulties and uneven lighting that impair the overall classification accuracy cause the fundus imagequality to degrade. Most studies show lower convergence rates, potential over fitting problems, and higher classification error rates. This study employs retinal fundus imagesfrom an online dataset that have been preprocessed using Interval Contrast Limited Adaptive Histogram Equalization (ICLAHE) to classify retinal illnesses at an early stage. For feature segmentation, the pre-processed imageswere input into a Sorensen Dice Similarity with Mask based FCM (SDSM-FCM). This study approach employs
Dominant Principal Component Analysis (DPCA) after dimensionality reduction. Final step, Position Encoded Layer based Deep Convolutional Neural Network (PEL-DCNN) classifies the severity of the segmented image. Using the position encoded method, the generated feature vector was combined and the SoftMax classifier was used to successfully classify eleven types of retinal disorders.This improved the network's accuracy and speed. Implementing the suggested strategy yields results of 0.99% accuracy, 0.965% precision, 0.96% recall, and 0.97% F measure. During experimentation when the new technique is compared to the current approaches, the proposed method performs better.

Downloads

Published

2023-09-28

How to Cite

Sheeja Mary F, Dr. V. Asanambigai, & Dr. A. Lenin Fred. (2023). Severity Assessment Of Retina Disease Using Pel-Dcnn Approach. Chinese Journal of Computational Mechanics, (5), 62–74. Retrieved from http://jslxxb.cn/index.php/jslxxb/article/view/4336

Issue

Section

Articles