Scholarlink

<p>The human retina serves as a critical diagnostic tool for detecting various health issues, including diabetic retinopathy (DR), cardiovascular disease, hypertension, and age-related conditions. Early detection of DR is crucial for preventing vision loss, yet traditional methods often suffer from low accuracy. Recent advancements in deep learning (DL) have enhanced diagnostic capabilities, albeit at the expense of increased computational requirements. This study proposes a deep convolutional neural network (DCNN) framework for DR detection, consisting of four key stages: pre-processing, segmentation, feature extraction, and classification. In the pre-processing stage, Gamma correction is applied to enhance image contrast and reduce noise, improving the quality of the input data. The segmentation stage utilizes the Coye filter to isolate retinal blood vessels, essential for accurate DR detection. The feature extraction phase incorporates several methods, including correlation, discrete cosine transform (DCT), local binary pattern (LBP), and gray level co-occurrence matrix (GLCM) features. These features are then fed into various DCNN models, such as AlexNet, VGG16, VGG19, GoogleNet, and InceptionV3, for classification. To optimize model performance, the Aquila optimization algorithm (AOA) is employed to fine-tune hyperparameters. The proposed model is validated through a comprehensive performance analysis, demonstrating high accuracy in detecting DR.</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

An enhanced model for high accuracy diabetic retinopathy detection using deep convolutional neural network (DCNN)

  • 期刊论文

  • 发表日期：

    2026年3月24日

查看全文
  • M. R. Padmapriya,
  • Vijayakumar Adaickalam

摘要

The human retina serves as a critical diagnostic tool for detecting various health issues, including diabetic retinopathy (DR), cardiovascular disease, hypertension, and age-related conditions. Early detection of DR is crucial for preventing vision loss, yet traditional methods often suffer from low accuracy. Recent advancements in deep learning (DL) have enhanced diagnostic capabilities, albeit at the expense of increased computational requirements. This study proposes a deep convolutional neural network (DCNN) framework for DR detection, consisting of four key stages: pre-processing, segmentation, feature extraction, and classification. In the pre-processing stage, Gamma correction is applied to enhance image contrast and reduce noise, improving the quality of the input data. The segmentation stage utilizes the Coye filter to isolate retinal blood vessels, essential for accurate DR detection. The feature extraction phase incorporates several methods, including correlation, discrete cosine transform (DCT), local binary pattern (LBP), and gray level co-occurrence matrix (GLCM) features. These features are then fed into various DCNN models, such as AlexNet, VGG16, VGG19, GoogleNet, and InceptionV3, for classification. To optimize model performance, the Aquila optimization algorithm (AOA) is employed to fine-tune hyperparameters. The proposed model is validated through a comprehensive performance analysis, demonstrating high accuracy in detecting DR.