Scholarlink

<p>Early detection and timely treatment of retinal diseases were made easier by the processing of optical coherence tomography (OCT) images. Classifying retinal diseases is a challenging task and deep learning architectures have shown remarkable results in classifying retinal diseases using OCT images. This research put forward three deep learning models; OctDeepNet, OctDeepNet1, OctDeepNet2 and investigates the impact of various architectural components such as the number of layers, kernel sizes, and pooling strategies on classification accuracy. Evaluation metrics are employed to quantify the performance of each architecture. Results demonstrate the effectiveness of the deep learning architecture OCTDeepNet2 with 50 layers achieved high accuracy when compared to the OCTDeepNet and OCTDeepNet1 architectures with 30 and 17 layers respectively The outputs were analyzed for various batch sizes of 8, 16, and 32 and for different epochs 25, 50 and 100. OCTDeepNet 2 showed better accuracy of 98% with a batch size of 32 for 100 epochs with precision, recall as 0.98, 1.00 and F score as 0.99. The study's findings provide significant details for selecting the best deep-learning architecture for classifying retinal disorders from OCT scans. The purpose of the suggested framework is to support continuous efforts in the field of ophthalmology to improve diagnostic accuracy and clinical process.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep Learning Architectures for OCT Images Retinal Disease Classification

  • 期刊论文

  • 发表日期：

    2025年2月12日

查看全文
  • Ranjitha Rajan,
  • S. N. Kumar

摘要

Early detection and timely treatment of retinal diseases were made easier by the processing of optical coherence tomography (OCT) images. Classifying retinal diseases is a challenging task and deep learning architectures have shown remarkable results in classifying retinal diseases using OCT images. This research put forward three deep learning models; OctDeepNet, OctDeepNet1, OctDeepNet2 and investigates the impact of various architectural components such as the number of layers, kernel sizes, and pooling strategies on classification accuracy. Evaluation metrics are employed to quantify the performance of each architecture. Results demonstrate the effectiveness of the deep learning architecture OCTDeepNet2 with 50 layers achieved high accuracy when compared to the OCTDeepNet and OCTDeepNet1 architectures with 30 and 17 layers respectively The outputs were analyzed for various batch sizes of 8, 16, and 32 and for different epochs 25, 50 and 100. OCTDeepNet 2 showed better accuracy of 98% with a batch size of 32 for 100 epochs with precision, recall as 0.98, 1.00 and F score as 0.99. The study's findings provide significant details for selecting the best deep-learning architecture for classifying retinal disorders from OCT scans. The purpose of the suggested framework is to support continuous efforts in the field of ophthalmology to improve diagnostic accuracy and clinical process.