Convolutional neural networks for diabetic retinopathy detection
The early detection of diabetic retinopathy remains a critical challenge in medical diagnostics, with deep learning techniques in artificial intelligence offering promising solutions for identifying pathological patterns in retinal images. This study evaluates and compares the performance of three...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Universidad Politécnica Salesiana
2025-01-01
|
| Series: | Ingenius: Revista de Ciencia y Tecnología |
| Subjects: | |
| Online Access: | https://revistas.ups.edu.ec/index.php/ingenius/article/view/8846 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1825201947791065088 |
|---|---|
| author | Darwin Patiño-Pérez Luis Armijos-Valarezo Luis Chóez-Acosta Freddy Burgos-Robalino |
| author_facet | Darwin Patiño-Pérez Luis Armijos-Valarezo Luis Chóez-Acosta Freddy Burgos-Robalino |
| author_sort | Darwin Patiño-Pérez |
| collection | DOAJ |
| description |
The early detection of diabetic retinopathy remains a critical challenge in medical diagnostics, with deep learning techniques in artificial intelligence offering promising solutions for identifying pathological patterns in retinal images. This study evaluates and compares the performance of three convolutional neural network (CNN) architectures ResNet-18, ResNet-50, and a custom, non-pretrained CNN using a dataset of retinal images classified into five categories. The findings reveal significant differences in the models' ability to learn and generalize. The non-pretrained CNN consistently outperformed the pretrained ResNet-18 and ResNet-50 models, achieving an accuracy of 91% and demonstrating notable classification stability. In contrast, ResNet-18 suffered severe performance degradation, with accuracy dropping from 70% to 26%, while ResNet-50 required extensive tuning to improve its outcomes. The non-pretrained CNN excelled in handling class imbalances and capturing complex diagnostic patterns, emphasizing the potential of tailored architectures for medical imaging tasks. These results underscore the importance of designing domain-specific architectures, demonstrating that model complexity does not necessarily guarantee better performance. Particularly in scenarios with limited datasets, well-designed custom models can surpass pre-trained architectures in diagnostic imaging applications.
|
| format | Article |
| id | doaj-art-ce28b1ed6b85465c803cb6bd0a34b987 |
| institution | Kabale University |
| issn | 1390-650X 1390-860X |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Universidad Politécnica Salesiana |
| record_format | Article |
| series | Ingenius: Revista de Ciencia y Tecnología |
| spelling | doaj-art-ce28b1ed6b85465c803cb6bd0a34b9872025-02-07T16:30:08ZengUniversidad Politécnica SalesianaIngenius: Revista de Ciencia y Tecnología1390-650X1390-860X2025-01-013310.17163/ings.n33.2025.08Convolutional neural networks for diabetic retinopathy detectionDarwin Patiño-Pérez0https://orcid.org/0000-0001-9850-0393Luis Armijos-Valarezo1https://orcid.org/0000-0002-9264-0448Luis Chóez-Acosta2https://orcid.org/0000-0003-3370-7793Freddy Burgos-Robalino3https://orcid.org/0000-0002-2518-9212Universidad de GuayaquilUniversidad de GuayaquilUniversidad de GuayaquilUniversidad de Guayaquil The early detection of diabetic retinopathy remains a critical challenge in medical diagnostics, with deep learning techniques in artificial intelligence offering promising solutions for identifying pathological patterns in retinal images. This study evaluates and compares the performance of three convolutional neural network (CNN) architectures ResNet-18, ResNet-50, and a custom, non-pretrained CNN using a dataset of retinal images classified into five categories. The findings reveal significant differences in the models' ability to learn and generalize. The non-pretrained CNN consistently outperformed the pretrained ResNet-18 and ResNet-50 models, achieving an accuracy of 91% and demonstrating notable classification stability. In contrast, ResNet-18 suffered severe performance degradation, with accuracy dropping from 70% to 26%, while ResNet-50 required extensive tuning to improve its outcomes. The non-pretrained CNN excelled in handling class imbalances and capturing complex diagnostic patterns, emphasizing the potential of tailored architectures for medical imaging tasks. These results underscore the importance of designing domain-specific architectures, demonstrating that model complexity does not necessarily guarantee better performance. Particularly in scenarios with limited datasets, well-designed custom models can surpass pre-trained architectures in diagnostic imaging applications. https://revistas.ups.edu.ec/index.php/ingenius/article/view/8846diabetic retinopathyblindnessearly detectionartificial intelligenceconvolutional neural networksocular image analysis |
| spellingShingle | Darwin Patiño-Pérez Luis Armijos-Valarezo Luis Chóez-Acosta Freddy Burgos-Robalino Convolutional neural networks for diabetic retinopathy detection Ingenius: Revista de Ciencia y Tecnología diabetic retinopathy blindness early detection artificial intelligence convolutional neural networks ocular image analysis |
| title | Convolutional neural networks for diabetic retinopathy detection |
| title_full | Convolutional neural networks for diabetic retinopathy detection |
| title_fullStr | Convolutional neural networks for diabetic retinopathy detection |
| title_full_unstemmed | Convolutional neural networks for diabetic retinopathy detection |
| title_short | Convolutional neural networks for diabetic retinopathy detection |
| title_sort | convolutional neural networks for diabetic retinopathy detection |
| topic | diabetic retinopathy blindness early detection artificial intelligence convolutional neural networks ocular image analysis |
| url | https://revistas.ups.edu.ec/index.php/ingenius/article/view/8846 |
| work_keys_str_mv | AT darwinpatinoperez convolutionalneuralnetworksfordiabeticretinopathydetection AT luisarmijosvalarezo convolutionalneuralnetworksfordiabeticretinopathydetection AT luischoezacosta convolutionalneuralnetworksfordiabeticretinopathydetection AT freddyburgosrobalino convolutionalneuralnetworksfordiabeticretinopathydetection |