Scholarlink

<p>Diabetic Retinopathy (DR) is an advanced eye disease of diabetes and a main reason of blindness in the world. The diagnosis at the early stage is vital in reducing the risks of blindness. This research suggests a new Deep Learning (DL)-based diagnostic model DRNet-X, which automatically detects Diabetic Retinopathy Image Detection (DRID) in Macula Fundus Images (MFI). The proposed approach incorporates a hybrid preprocessing algorithm that merges CLAHE (Contrast Limited Adaptive Histogram Equalization) along with Gaussian filter to increase the visibility of vessels and reduce noise. Proposed U-Net-based segmentation module separates pathological areas of the retina, with a particular emphasis on microaneurysms, exudates, and hemorrhages. Feature extraction is carried out using DR-ResNet50X, a modified and fine-tuned version of ResNet50 tailored to DR-specific image features. A hybrid feature selection mechanism HippoMaster Optimization (HMO) is applied to minimize dimensionality and save the most discriminative features by combining Masterpiece Optimization Algorithm (MOA), and Hippopotamus Optimization (HO). In the proposed DRNet-X, a stacked CNN-LSTM hybrid model is used to exploit the spatial and sequential patterns in the images and a Vision Transformer (ViT) model is applied to classify them finally. The dataset is divided into 80% for training and 20% for testing to ensure robust evaluation of the model’s performance. The experimental findings show that the accuracy, sensitivity, and specificity are much higher than those of the conventional CNN methods. The suggested DRNet-X framework provides a precise, and thorough solution to the early-stage DR recognition along with grading.</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

DRNET-X: a deep learning-based framework for diabetic retinopathy image detection using hybrid preprocessing and CNN-based classification

  • 期刊论文

  • 发表日期：

    2025年12月4日

查看全文
  • J. Josephine Sahaya Vergin,
  • R. M. Vidhyavathi

摘要

Diabetic Retinopathy (DR) is an advanced eye disease of diabetes and a main reason of blindness in the world. The diagnosis at the early stage is vital in reducing the risks of blindness. This research suggests a new Deep Learning (DL)-based diagnostic model DRNet-X, which automatically detects Diabetic Retinopathy Image Detection (DRID) in Macula Fundus Images (MFI). The proposed approach incorporates a hybrid preprocessing algorithm that merges CLAHE (Contrast Limited Adaptive Histogram Equalization) along with Gaussian filter to increase the visibility of vessels and reduce noise. Proposed U-Net-based segmentation module separates pathological areas of the retina, with a particular emphasis on microaneurysms, exudates, and hemorrhages. Feature extraction is carried out using DR-ResNet50X, a modified and fine-tuned version of ResNet50 tailored to DR-specific image features. A hybrid feature selection mechanism HippoMaster Optimization (HMO) is applied to minimize dimensionality and save the most discriminative features by combining Masterpiece Optimization Algorithm (MOA), and Hippopotamus Optimization (HO). In the proposed DRNet-X, a stacked CNN-LSTM hybrid model is used to exploit the spatial and sequential patterns in the images and a Vision Transformer (ViT) model is applied to classify them finally. The dataset is divided into 80% for training and 20% for testing to ensure robust evaluation of the model’s performance. The experimental findings show that the accuracy, sensitivity, and specificity are much higher than those of the conventional CNN methods. The suggested DRNet-X framework provides a precise, and thorough solution to the early-stage DR recognition along with grading.