Scholarlink

<p>The World Health Organization warns that visual impairment is one of the most widespread health problems globally. Estimates indicate that around 253 million people have some degree of visual impairment, of whom around 36 million have completely lost their sight. This picture is consistent with the findings of the Expert Group on Vision Loss and the Global Burden of Disease study, which put the number of cases of moderate to severe visual impairment at 206 million. It should not be forgotten that these conditions have a direct impact on the daily autonomy and overall health of those who suffer from them. In this context, the work focused on the identification and classification of different eye diseases using a convolutional neural network (CNN) based on the VGG16 architecture. It should be noted that this type of model has been widely used in medical image processing, which justifies its choice for ophthalmological analysis. For training and evaluation, a set of 6,392 images was used, accompanied by 19 variables associated with the patients’ clinical information, which allowed for a combined approach between visual data and descriptive records. The results obtained show that the model achieved 90% accuracy, with an average recall of 89% and an F1 score of 89%. These metrics allow us to consider its usefulness as a support for diagnosis and therapeutic decision-making in the field of ophthalmology, even under different test conditions. It should be noted that, despite these values, the work leaves open the need to incorporate broader databases, as well as new variables and modeling strategies, to refine predictions and capture nuances that purely quantitative measurements may leave out of clinical analysis.</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

Application of convolutional neural networks in the diagnosis of ocular diseases: evaluation of the VGG16 model

  • 期刊论文

  • 发表日期：

    2026年3月21日

查看全文
  • Orlando Iparraguirre-Villanueva,
  • Cleoge Paulino-Moreno,
  • Rosalynn Ornella Flores-Castañeda,
  • María Suxe-Ramírez

摘要

The World Health Organization warns that visual impairment is one of the most widespread health problems globally. Estimates indicate that around 253 million people have some degree of visual impairment, of whom around 36 million have completely lost their sight. This picture is consistent with the findings of the Expert Group on Vision Loss and the Global Burden of Disease study, which put the number of cases of moderate to severe visual impairment at 206 million. It should not be forgotten that these conditions have a direct impact on the daily autonomy and overall health of those who suffer from them. In this context, the work focused on the identification and classification of different eye diseases using a convolutional neural network (CNN) based on the VGG16 architecture. It should be noted that this type of model has been widely used in medical image processing, which justifies its choice for ophthalmological analysis. For training and evaluation, a set of 6,392 images was used, accompanied by 19 variables associated with the patients’ clinical information, which allowed for a combined approach between visual data and descriptive records. The results obtained show that the model achieved 90% accuracy, with an average recall of 89% and an F1 score of 89%. These metrics allow us to consider its usefulness as a support for diagnosis and therapeutic decision-making in the field of ophthalmology, even under different test conditions. It should be noted that, despite these values, the work leaves open the need to incorporate broader databases, as well as new variables and modeling strategies, to refine predictions and capture nuances that purely quantitative measurements may leave out of clinical analysis.