Image preprocessing‐based ensemble deep learning classification of diabetic retinopathy

Abstract

Diabetic retinopathy (DR) can cause irreversible eye damage, even blindness. The prognosis improves with early diagnosis. According to the International Classification of Diabetic Retinopathy Severity Scale (ICDRSS), DR has five stages. Modern, cost‐effective techniques for automatic DR screening and staging of fundus images are based on deep learning (DL). To obtain higher classification accuracy, the combination of several diverse individual DL models into one ensemble could be used. A new approach to model diversity in an ensemble is proposed by manipulating the training input data involving original and four variants of preprocessed image datasets. There are publicly available datasets with labels for all five stages, but some contain poor‐quality images. In contrast, this algorithm was trained on images from a six‐class DDR dataset, including the class of poor‐quality ungradable images, to enhance the classification performance. The solution was evaluated on the APTOS dataset, containing only ICDRSS classes. Classification results of the ensemble model were presented on two different ensemble convolutional neural network (CNN) models, based on Xception and EfficientNetB4 architectures using two fusion approaches. Our proposed ensemble models outperformed all other single deep learning architectures regarding overall accuracy and Cohen's Kappa, with the best results using the EfficientNetB4 architecture.

Comparison of Preprocessing Methods Impact on Detection of Soldering Splashes Using Different YOLOv8 Versions

“…Mixing the independent color channels of RGB image can increase the features of specific object for its localization. Mácsik et al [18] created their own 3-channel color image highlighting the object of interest features in images. In this study, the maximal value channel, green channel and grayscale channel were used.…”

A review of optic disc and optic cup segmentation based on fundus images

“…However, manual segmentation of OD and OC is very time-consuming and necessitates subjective judgment, making it difficult for inexperienced physicians to achieve accurate segmentation. Given the rapid development of smart healthcare, computer technology has been widely applied in the field of medical image analysis and processing, such as electrocardiogram analysis [3,4], computerized tomography image lesion segmentation [4,5], and fundus image processing [6][7][8][9][10][11][12]. In the domain of fundus image processing, research in areas like diabetic retinopathy grading [6], retinal vessel segmentation [7,9] and glaucoma classification [8,[10][11][12] has significantly propelled its advancement.…”

“…Given the rapid development of smart healthcare, computer technology has been widely applied in the field of medical image analysis and processing, such as electrocardiogram analysis [3,4], computerized tomography image lesion segmentation [4,5], and fundus image processing [6][7][8][9][10][11][12]. In the domain of fundus image processing, research in areas like diabetic retinopathy grading [6], retinal vessel segmentation [7,9] and glaucoma classification [8,[10][11][12] has significantly propelled its advancement. These studies have provided valuable references and insights for the research related to optic disc and cup segmentation, offering crucial experiential support for further expanding the frontiers of medical image analysis and clinical applications.…”

“…FIGURE6 GAN model and its modified models. In (c), S and T represent the source domain and the target domain, respectively.…”