Scholarlink

In this research, we introduce about the EEC-IGE framework, which encompasses data preprocessing, data augmentation, transfer learning, and fine-tuning of a pre-existing convolutional neural network (CNN) model tailored for the classification of eye diseases. The implications of eye ailments extend beyond physical harm, frequently inducing discomfort, limited functionality, and diminished quality of life among afflicted persons. Recent years have witnessed notable progress in the realm of machine learning methodologies, particularly in their potential to address the complexities associated with diagnosing and categorizing eye diseases. By harnessing the capabilities of deep learning frameworks, investigators strive to devise precise and effective mechanisms for the automated identification and classification of various eye pathologies based on medical imaging data. The present investigation highlights the efficacy of the EEC-IGE framework in leveraging pre-trained CNN models (EfficientNetB3) for the classification of Color fundus photography (CFP) images depicting eye diseases, achieving classification accuracies of 99%, 99.5%, 97.25%, and 97.75% across four distinct experimental scenarios. These outcomes stem from the utilization of transfer learning and fine-tuning methodologies on an expanded dataset comprising 4217 CFP images classified into discrete categories: cataract, diabetic retinopathy, glaucoma, and normal. Moreover, the incorporation of Integrated Gradients explanation techniques plays a pivotal role in enhancing the interpretability and performance optimization of the image classification model. A comprehensive evaluation of each model’s performance is undertaken through a detailed analysis encompassing metrics such as accuracy, recall, F1-score, and precision.

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

EEC-IGE: Diagnosing Eye Diseases with DL-CNN and Integrated Gradients

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Huong Hoang Luong,
  • Quy Thanh Lu,
  • Triet Minh Nguyen

摘要

In this research, we introduce about the EEC-IGE framework, which encompasses data preprocessing, data augmentation, transfer learning, and fine-tuning of a pre-existing convolutional neural network (CNN) model tailored for the classification of eye diseases. The implications of eye ailments extend beyond physical harm, frequently inducing discomfort, limited functionality, and diminished quality of life among afflicted persons. Recent years have witnessed notable progress in the realm of machine learning methodologies, particularly in their potential to address the complexities associated with diagnosing and categorizing eye diseases. By harnessing the capabilities of deep learning frameworks, investigators strive to devise precise and effective mechanisms for the automated identification and classification of various eye pathologies based on medical imaging data. The present investigation highlights the efficacy of the EEC-IGE framework in leveraging pre-trained CNN models (EfficientNetB3) for the classification of Color fundus photography (CFP) images depicting eye diseases, achieving classification accuracies of 99%, 99.5%, 97.25%, and 97.75% across four distinct experimental scenarios. These outcomes stem from the utilization of transfer learning and fine-tuning methodologies on an expanded dataset comprising 4217 CFP images classified into discrete categories: cataract, diabetic retinopathy, glaucoma, and normal. Moreover, the incorporation of Integrated Gradients explanation techniques plays a pivotal role in enhancing the interpretability and performance optimization of the image classification model. A comprehensive evaluation of each model’s performance is undertaken through a detailed analysis encompassing metrics such as accuracy, recall, F1-score, and precision.