Development of a Model for Diabetic Retinopathy Image Classification

EasyChair Preprint 12986

Abstract

Diabetic retinopathy is a disease capable of resulting to loss of vision in humans if not treated early enough. The study aimed at predicting the occurrence of diabetic retinopathy through the application of stacking ensemble classification technique on features extracted from output of different retinal images. The proposed system was implemented on embedded system using Python-based integrated development environment and executed in an Apple M1 computer with 3.20 GHz CPU and 8 GB RAM under Mac OS Mojave. The prediction of diabetic retinopathy was done by comparing the performances of decision tree singled classifiers, support vector machine, stochastic gradient descent, and XG Boost with their ensemble through Stacking classification technique. The novel ensemble model performs better than singled models with accuracy, sensitivity and specificity of 85%, 81 % and 86 % respectively.

Keyphrases: Diabetic Retinopathy, Ensemble model, XG Boost, stochastic gradient descent

@booklet{EasyChair:12986,
  author    = {Olusogo Adetunji and Oladapo Ibitoye and Ayobami Olusesi and Oluwafunso Osaloni},
  title     = {Development of a Model for Diabetic Retinopathy Image Classification},
  howpublished = {EasyChair Preprint 12986},
  year      = {EasyChair, 2024}}