Scholarlink

<p>This study aims to identify retinal disease and abnormalities at an early stage, thereby preventing serious visual loss or blindness using a Machine Learning (ML)-based method by analysing their fundus images. Initially, the Preprocessing is performed by converting Red, Green, and Blue (RGB) images into grayscale, followed by the application of Gaussian, median, and Contrast Limited Adaptive Histogram Equalization (CLAHE) filters to enhance the quality of the fundus images and emphasise vital features. Structural, non-structural, and textural features are extracted in sequence, and classified using three ML models: Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), and Bagged Trees (BT). Among these, SVM showed more consistent performance after addressing class imbalance using Synthetic Minority Oversampling Technique (SMOTE), achieving 96.80% accuracy in identifying diseased eyes when trained and tested on the proposed feature set that integrates structural, non-structural, and textural features. This demonstrates a potential path for reliable early diagnosis in ophthalmological applications. The one-way Analysis of Variance (ANOVA) method is used to determine the p-value and F-statistic of the extracted features, thereby aiding in the selection of features for the proposed feature set. Additionally, the ResNet50 model is evaluated for comparative performance for a robust clinical assessment of the proposed ML framework. SHapley Additive exPlanations (SHAP) is utilised to identify the most influential features.Hence, this ML approach aims to support clinical decision-making and reduce the risk of vision loss through timely diagnosis of retinal diseases and abnormalities. The source code used in this study is available at: <a href="https://github.com/raghavbadri06-gif/An-Explainable-Machine-Learning-Framework-for-Retinal-Disease-and-Abnormality-Classification">https://github.com/raghavbadri06-gif/An-Explainable-Machine-Learning-Framework-for-Retinal-Disease-and-Abnormality-Classification</a>.</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 Explainable Machine Learning Framework for Retinal Disease and Abnormality Classification from Fundus Images

  • 期刊论文

  • 发表日期：

    2026年7月3日

查看全文
  • R. Rohini,
  • B. Raghavan,
  • B. Madhavan,
  • S. Venkatesh,
  • Rengarajan Amirtharajan

摘要

This study aims to identify retinal disease and abnormalities at an early stage, thereby preventing serious visual loss or blindness using a Machine Learning (ML)-based method by analysing their fundus images. Initially, the Preprocessing is performed by converting Red, Green, and Blue (RGB) images into grayscale, followed by the application of Gaussian, median, and Contrast Limited Adaptive Histogram Equalization (CLAHE) filters to enhance the quality of the fundus images and emphasise vital features. Structural, non-structural, and textural features are extracted in sequence, and classified using three ML models: Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), and Bagged Trees (BT). Among these, SVM showed more consistent performance after addressing class imbalance using Synthetic Minority Oversampling Technique (SMOTE), achieving 96.80% accuracy in identifying diseased eyes when trained and tested on the proposed feature set that integrates structural, non-structural, and textural features. This demonstrates a potential path for reliable early diagnosis in ophthalmological applications. The one-way Analysis of Variance (ANOVA) method is used to determine the p-value and F-statistic of the extracted features, thereby aiding in the selection of features for the proposed feature set. Additionally, the ResNet50 model is evaluated for comparative performance for a robust clinical assessment of the proposed ML framework. SHapley Additive exPlanations (SHAP) is utilised to identify the most influential features.Hence, this ML approach aims to support clinical decision-making and reduce the risk of vision loss through timely diagnosis of retinal diseases and abnormalities. The source code used in this study is available at: https://github.com/raghavbadri06-gif/An-Explainable-Machine-Learning-Framework-for-Retinal-Disease-and-Abnormality-Classification.