Scholarlink

<p>One of the emerging technology of biomedical imaging is Optical Coherence Topography (OCT), which provides highly scattering tissues based super resolution imaging and non-invasive real-time imaging system. For the retina and anterior eye structure, the diagnostic imaging is performed with commonly utilized ophthalmology. For few retinal diseases, the OCT images are assessed to perform clinical studies. Due to morphological changes, different kinds of eye diseases with its early detection are provided by means of OCT images. Two publicly available dataset images to provide the OCT images are taken for the proposed work. Pre-process the input images based on Fuzzy Transform (FT) to neglect the noises and background surfaces non-ophthalmological elimination thereby utilizes feature extraction procedure using Modified Transfer Learning (MTL) followed by the significant features are selected using Prairie Dog Optimization (PDO). An adaptive sand cat swarm optimization (ASCSO) algorithm along with Equivalent depth-wise separable convolutional neural network (EDSCNN) is to classify the normal class along with the abnormalities such as Diabetic macular edema (DME), Choroidal neovascularization (CNV) and Drusen (D) from the assessed OCT images. The proposed work performance beats the performances of other existing works. The classification accuracy of proposed system is higher for all the classes and is represented as 97, 96, 98, and 95% while comparing the performance of proposed work with existing works for the classes Normal, DME, CNV, and DURSEN respectively. At 180th iterations, accuracy, specificity and sensitivity of the proposed work is 97, 92 and 94% than existing works.</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

Classification of Retinal Abnormalities using Adaptive Sand Cat Swarm Optimization Algorithm with Equivalent Depth-Wise Separable Convolutional Neural Network based on OCT Images

  • 期刊论文

  • 发表日期：

    2025年3月31日

查看全文
  • R. Pugal Priya,
  • L. Raja Saviour

摘要

One of the emerging technology of biomedical imaging is Optical Coherence Topography (OCT), which provides highly scattering tissues based super resolution imaging and non-invasive real-time imaging system. For the retina and anterior eye structure, the diagnostic imaging is performed with commonly utilized ophthalmology. For few retinal diseases, the OCT images are assessed to perform clinical studies. Due to morphological changes, different kinds of eye diseases with its early detection are provided by means of OCT images. Two publicly available dataset images to provide the OCT images are taken for the proposed work. Pre-process the input images based on Fuzzy Transform (FT) to neglect the noises and background surfaces non-ophthalmological elimination thereby utilizes feature extraction procedure using Modified Transfer Learning (MTL) followed by the significant features are selected using Prairie Dog Optimization (PDO). An adaptive sand cat swarm optimization (ASCSO) algorithm along with Equivalent depth-wise separable convolutional neural network (EDSCNN) is to classify the normal class along with the abnormalities such as Diabetic macular edema (DME), Choroidal neovascularization (CNV) and Drusen (D) from the assessed OCT images. The proposed work performance beats the performances of other existing works. The classification accuracy of proposed system is higher for all the classes and is represented as 97, 96, 98, and 95% while comparing the performance of proposed work with existing works for the classes Normal, DME, CNV, and DURSEN respectively. At 180th iterations, accuracy, specificity and sensitivity of the proposed work is 97, 92 and 94% than existing works.