Detecção dos Estágios Iniciais da Retinopatia Diabética Utilizando Redes Neurais Profundas

• João Victor G. Martins IF Sudeste MG
• Alessandra M. Coelho IF Sudeste MG
• Matheus F. O. Baffa USP

DOI: https://doi.org/10.5753/ercas.2021.17435

Abstract

Diabetic Retinopathy is one of the main complications related to Diabetes Mellitus. It occurs when excess glucose in the blood damages the blood vessels within the retina. The diagnosis of the disease uses fundus eye images collected by detecting abnormal signs, either in the vascular structures or in the retina. The detection in early stages promotes a better prognosis and can prevent blindness in the patient. This paper addresses a Computer Vision methodology in order to detect the early stages of Diabetic Retinopathy. The proposed method, using Deep Neural Networks and the extraction of radiomicbased features, obtained an average accuracy of 91.06% in the experiments.

References

Gama, Í., Coelho, A., and Baffa, M. (2020). Fundus eye images classication for diabetic retinopathy detection using very deep convolutional neural network. In Anais do XVI Workshop de Visão Computacional, pages 24–29. SBC.

Kumar, V., Gu, Y., Basu, S., Berglund, A., Eschrich, S. A., Schabath, M. B., Forster, K., Aerts, H. J., Dekker, A., Fenstermacher, D., et al. (2012). Radiomics: the process and the challenges. Magnetic resonance imaging, 30(9):1234–1248.

Ministério da Saúde (2021). Diabetes (diabetes mellitus). Disponível em: [link]. Acesso em: 30 maio 2021.

Nentwich, M. M. and Ulbig, M. W. (2015). Diabetic retinopathy-ocular complications of diabetes mellitus. World journal of diabetes, 6(3):489.

Rath, S. R. (2021). Diabetic retinopathy 224x224 gaussian ltered. Disponível em: <https://www.kaggle.com/sovitrath/diabetic-retinopathy-224x224-gaussian-ltered>. Acesso em: 30 maio 2021.