Brazilian Dataset for Retinal Lesion Analysis: A Deep Learning Diagnostic Pipeline
Palavras-chave:
Diabetic retinopathy, classification, segmentation, deep learningResumo
Diabetic retinopathy (DR) is a leading cause of vision loss in working-age adults, and early diagnosis is essential. However, limited infrastructure and a shortage of specialists restrict access to proper screening. To address this, we created the BDR-iD dataset using anonymized fundus images collected from an eye clinic in Pelotas, Brazil. From a total of 13,131 images collected between 2012 and 2024, 150 were selected and annotated based on DR presence and associated lesions. Deep learning models were evaluated, achieving an overall accuracy of 0.6667 in DR classification, but with poor performance in lesion segmentation and detection. The BDR-iD dataset aims to support the development of automated tools for DR diagnosis.
Downloads
Referências
International Council of Ophthalmology. Updated 2017 ICO Guidelines for Diabetic Eye Care [Internet]. San Francisco: ICO; 2017 [citado 2025 jun 6]. Disponível em: http://www.icoph.org/downloads/ICOGuidelinesforDiabeticEyeCare.pdf
Nayak J, Bhat PS, Acharya UR, Lim CM, Kagathi M. Automated identification of diabetic retinopathy stages using digital fundus images. J Med Syst. 2008;32(2):107–15.
Ting DSW, Cheung GCM, Wong TY. Diabetic retinopathy: global prevalence, major risk factors, screening practices and public health challenges: a review. Clin Exp Ophthalmol. 2016;44(4):260–77.
Chakrabarti R, Harper CA, Keeffe JE. Diabetic retinopathy management guidelines. Expert Rev Ophthalmol. 2012;7(5):417–39. Disponível em: https://doi.org/10.1586/eop.12.52
Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition [Internet]. 2015 [citado 2025 jun 6]. Disponível em: https://arxiv.org/abs/1409.1556
He K, Zhang X, Ren S, Sun J. Deep residual learning for image recognition [Internet]. 2015 [citado 2025 jun 6]. Disponível em: https://arxiv.org/abs/1512.03385
Szegedy C, Liu W, Jia Y, Sermanet P, Reed S, Anguelov D, et al. Going deeper with convolutions [Internet]. 2014 [citado 2025 jun 6]. Disponível em: https://arxiv.org/abs/1409.4842
Huang G, Liu Z, van der Maaten L, Weinberger KQ. Densely connected convolutional networks [Internet]. 2018 [citado 2025 jun 6]. Disponível em: https://arxiv.org/abs/1608.06993
Tan M, Le QV. EfficientNet: Rethinking model scaling for convolutional neural networks [Internet]. 2020 [citado 2025 jun 6]. Disponível em: https://arxiv.org/abs/1905.11946
Radosavovic I, Kosaraju RP, Girshick R, He K, Dollár P. Designing network design spaces [Internet]. 2020 [citado 2025 jun 6]. Disponível em: https://arxiv.org/abs/2003.13678
Iandola FN, Han S, Moskewicz MW, Ashraf K, Dally WJ, Keutzer K. SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size [Internet]. 2016 [citado 2025 jun 6]. Disponível em: https://arxiv.org/abs/1602.07360
Liu Z, Lin Y, Cao Y, Hu H, Wei Y, Zhang Z, et al. Swin Transformer: Hierarchical vision transformer using shifted windows [Internet]. 2021 [citado 2025 jun 6]. Disponível em: https://arxiv.org/abs/2103.14030
Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation. Lect Notes Comput Sci. 2015;9351:234–41.
Oktay O, Schlemper J, Le Folgoc L, Lee M, Heinrich M, Misawa K, et al. Attention U-Net: Learning where to look for the pancreas. arXiv Preprint. 2018; arXiv:1804.03999.
Alom MZ, Hasan M, Yakopcic C, Taha TM, Asari VK. Recurrent residual convolutional neural network based on U-Net (R2U-Net) for medical image segmentation. arXiv Preprint. 2018; arXiv:1802.06955.
Downloads
Publicado
Como Citar
Edição
Secção
Licença
Direitos de Autor (c) 2026 Carlos Santos, Marcelo Dias, Alejandro Pereira, Marilton Aguiar, Daniel Welfer, Laura Bernardes, Artur Heckler
Este trabalho encontra-se publicado com a Licença Internacional Creative Commons Atribuição-NãoComercial-CompartilhaIgual 4.0.
A submissão de um artigo ao Journal of Health Informatics é entendida como exclusiva e que não está sendo considerada para publicação em outra revista (Declaração de Exclusividade). A permissão dos autores para a publicação de seu artigo no J. Health Inform. implica na exclusiva autorização concedida aos editores para incluí-lo na revista. Ao submeter um artigo, ao autor será solicitada a permissão de um Termo de Transferência de Direitos de Publicação. Uma mensagem eletrônica será enviada ao autor correspondente confirmando o recibo do manuscrito e o aceite da Declaração de Direito de Publicação.
