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 scr...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Wiley
2024-02-01
|
Series: | IET Image Processing |
Subjects: | |
Online Access: | https://doi.org/10.1049/ipr2.12987 |
_version_ | 1797311386153385984 |
---|---|
author | Peter Macsik Jarmila Pavlovicova Slavomir Kajan Jozef Goga Veronika Kurilova |
author_facet | Peter Macsik Jarmila Pavlovicova Slavomir Kajan Jozef Goga Veronika Kurilova |
author_sort | Peter Macsik |
collection | DOAJ |
description | 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. |
first_indexed | 2024-03-08T01:57:54Z |
format | Article |
id | doaj.art-723bf6c9b0934e4fb53de857d5620ac8 |
institution | Directory Open Access Journal |
issn | 1751-9659 1751-9667 |
language | English |
last_indexed | 2024-03-08T01:57:54Z |
publishDate | 2024-02-01 |
publisher | Wiley |
record_format | Article |
series | IET Image Processing |
spelling | doaj.art-723bf6c9b0934e4fb53de857d5620ac82024-02-14T07:53:25ZengWileyIET Image Processing1751-96591751-96672024-02-0118380782810.1049/ipr2.12987Image preprocessing‐based ensemble deep learning classification of diabetic retinopathyPeter Macsik0Jarmila Pavlovicova1Slavomir Kajan2Jozef Goga3Veronika Kurilova4Faculty of Electrical Engineering and Information Technology Slovak University of Technology Bratislava SlovakiaFaculty of Electrical Engineering and Information Technology Slovak University of Technology Bratislava SlovakiaFaculty of Electrical Engineering and Information Technology Slovak University of Technology Bratislava SlovakiaFaculty of Electrical Engineering and Information Technology Slovak University of Technology Bratislava SlovakiaFaculty of Electrical Engineering and Information Technology Slovak University of Technology Bratislava SlovakiaAbstract 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.https://doi.org/10.1049/ipr2.12987biomedical optical imagingcomputer visionconvolutional neural netsimage classificationmedical image processing |
spellingShingle | Peter Macsik Jarmila Pavlovicova Slavomir Kajan Jozef Goga Veronika Kurilova Image preprocessing‐based ensemble deep learning classification of diabetic retinopathy IET Image Processing biomedical optical imaging computer vision convolutional neural nets image classification medical image processing |
title | Image preprocessing‐based ensemble deep learning classification of diabetic retinopathy |
title_full | Image preprocessing‐based ensemble deep learning classification of diabetic retinopathy |
title_fullStr | Image preprocessing‐based ensemble deep learning classification of diabetic retinopathy |
title_full_unstemmed | Image preprocessing‐based ensemble deep learning classification of diabetic retinopathy |
title_short | Image preprocessing‐based ensemble deep learning classification of diabetic retinopathy |
title_sort | image preprocessing based ensemble deep learning classification of diabetic retinopathy |
topic | biomedical optical imaging computer vision convolutional neural nets image classification medical image processing |
url | https://doi.org/10.1049/ipr2.12987 |
work_keys_str_mv | AT petermacsik imagepreprocessingbasedensembledeeplearningclassificationofdiabeticretinopathy AT jarmilapavlovicova imagepreprocessingbasedensembledeeplearningclassificationofdiabeticretinopathy AT slavomirkajan imagepreprocessingbasedensembledeeplearningclassificationofdiabeticretinopathy AT jozefgoga imagepreprocessingbasedensembledeeplearningclassificationofdiabeticretinopathy AT veronikakurilova imagepreprocessingbasedensembledeeplearningclassificationofdiabeticretinopathy |