Scholarlink

<p>Diabetic retinopathy (DR) is one of the leading causes of vision loss, often associated with diabetes. It affects a significant number of individuals worldwide. Early detection is very vital in the timely prevention of severe visual disability caused by diabetic retinopathy. This paper introduces an advanced automated technique using the YOLO-v10 (You Only Look Once, version 10) deep-learning model for diagnosing diabetic retinopathy, optimized through a hybrid approach of Satin Bowerbird Optimization (SBO) and Aquila Optimizer (AO). The YOLO-v10 model is a real-time object detection model implemented on retinal images that uses Contrast Limited Adaptive Histogram Equalization (CLAHE) to improve local contrast. The assessment of the model performance utilized key indicators: Precision (P), Recall (R), mean Average Precision at an Intersection over Union (IoU) threshold of 0.5 (mAP50), and (mAP50-95). The optimized model achieved a precision of 0.938, a recall of 0.941, a mAP50 of 0.976, and a mAP(50–95) of 0.976. These results thus show the model is highly accurate, robust, and suitable for real-time DR detection. The proposed methodology shows the clear potential for its deployment and use in the clinic and telemedicine in providing a more cost-efficient tool for the early diagnosis of diabetic retinopathy.</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

Real-Time Diabetic Retinopathy Detection Using YOLO-v10 with Nature-Inspired Optimization

  • 期刊论文

  • 发表日期：

    2025年5月7日

查看全文
  • Ajita Arvind Mahapadi,
  • Vishal Shirsath,
  • Ajitkumar Pundge

摘要

Diabetic retinopathy (DR) is one of the leading causes of vision loss, often associated with diabetes. It affects a significant number of individuals worldwide. Early detection is very vital in the timely prevention of severe visual disability caused by diabetic retinopathy. This paper introduces an advanced automated technique using the YOLO-v10 (You Only Look Once, version 10) deep-learning model for diagnosing diabetic retinopathy, optimized through a hybrid approach of Satin Bowerbird Optimization (SBO) and Aquila Optimizer (AO). The YOLO-v10 model is a real-time object detection model implemented on retinal images that uses Contrast Limited Adaptive Histogram Equalization (CLAHE) to improve local contrast. The assessment of the model performance utilized key indicators: Precision (P), Recall (R), mean Average Precision at an Intersection over Union (IoU) threshold of 0.5 (mAP50), and (mAP50-95). The optimized model achieved a precision of 0.938, a recall of 0.941, a mAP50 of 0.976, and a mAP(50–95) of 0.976. These results thus show the model is highly accurate, robust, and suitable for real-time DR detection. The proposed methodology shows the clear potential for its deployment and use in the clinic and telemedicine in providing a more cost-efficient tool for the early diagnosis of diabetic retinopathy.