Analysis of human retinal morphology using random forest classifier

Diabetic retinopathy is a possible complication of diabetes that causes severe eye problems and, in some cases, blindness in adulthood as a result of an increase in glucose levels in the physical body fluids. The diabetic retinopathy, if it is detected early will avoid the harm to retina and eye sig...

Full description

Saved in:
Bibliographic Details
Main Authors: Vimala, G. S. Annie Grace, Sungeetha, D., Preetha, R., Kishanlal, M. Samayaraj Murali
Format: Conference Proceeding
Language:English
Subjects:
Blindness
Body fluids
Classifiers
Datasets
Diabetes
Diabetic retinopathy
Digital imaging
Hemorrhage
Pictures
Retinal images
Screening
Support vector machines
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Diabetic retinopathy is a possible complication of diabetes that causes severe eye problems and, in some cases, blindness in adulthood as a result of an increase in glucose levels in the physical body fluids. The diabetic retinopathy, if it is detected early will avoid the harm to retina and eye sight damage or a minimum of reducing its development. Most important objective of the current research was to produce a CAD system so as to spot diabetic retinopathy by removing red lesion both Micro aneurysm and Hemorrhage. The extracted retinal images will be validated with several image databases. Digital retinal fundus pictures got from sixty Indian subjects where forty images from Messidor information base and another forty pictures from DIARETDB1 data set. They were used for various examinations, and a hand-drawn ’ground truth’ result was gathered for each image. The proposed calculation has been evaluated quantitatively using Support Vector Machine and Random Forest classifiers. The results are connected against the ’ground truth’. The highlights like mean region of the segmented area, Eccentricity and Solidity were extricated for all test dataset. The CAD framework for Diabetic Retinopathy screening will be approved with public fundus image dataset, containing the ground certainties. The images are gathered from numerous expert doctors and contrasted with privately accessible data set. The proposed Random Forest classifier gives the Sensitivity of 95%, Specificity of 90% and Accuracy of 94%. In this way, the proposed CAD framework may be helpful for Diabetic Retinopathy screening.
ISSN:0094-243X
1551-7616
DOI:10.1063/5.0072536