Downloads: 6 | Views: 276 | Weekly Hits: ⮙1 | Monthly Hits: ⮙1
Research Paper | Computer Science and Information Technology | India | Volume 12 Issue 11, November 2023 | Rating: 5.3 / 10
Grasshopper Optimization Technique with Deep Learning Driven Retinal Fundus Image Grading and Retrieval
S. Syed Mahamood Shazuli, A. Saravanan
Abstract: In the domain of ophthalmology and medical image analysis, the accurate estimation of retinal fundus images is important for analyzing and treating different eye diseases, comprising Diabetic Retinopathy (DR). This study introduced an innovative architecture that integrates the capability of deep learning with content-based image retrieval (CBIR) to transform the evaluation of retinal fundus images. Consequently, this study establishes a Grasshopper Optimization Algorithm with Deep Learning Driven Retinal Fundus Image Grading and Retrieval (GOADL-RFIGR) method. The introduced GOADL-RFIGR technique is a comprehensive system for retinal fundus image retrieval, incorporating advanced technologies for hyper parameter optimization, pre-processing, similarity measurement, and feature extraction. We apply Bilateral Filtering (BF) as a pre-processing stage for improving the quality of retinal fundus images. We introduce a Lightweight Convolutional Neural Network (CNN) developed for effectively removing distinctive features from retinal fundus images. For more improvement of the retrieval performance, we employ the ability of the Least Square Support Vector Machine (LS-SVM) for classification. Moreover, we leverage the Grasshopper Optimization Algorithm for optimizing hyperparameters over the CBIR technique. This nature-inspired optimization method supports in fine-tuning diverse modules of the system and lastly, maximizing retrieval efficiency and accuracy. The simulated validation of the GOADL-RFIGR model on benchmark dataset represents a better performance over other systems.
Keywords: Deep learning, Retinal fundus images, Image classification, Parameter tuning, Image retrieval
Edition: Volume 12 Issue 11, November 2023,
Pages: 23 - 29