Deep Learning aplicado ao diagnóstico assistido de retinopatia diabética
Revista ft · 27.06.2026 · 30 (159) · 01-12
DOI: 10.69849/th866h42 ↗Kurz gefasst
Fachartikel in der Zeitschrift Revista ft, online erschienen am 27.06.2026. Verfasst von Matheus Pedroza da Cunha und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The assessment of diabetic retinopathy (DR) severity by human specialists is inherently limited by substantial interobserver variability, particularly in decentralized screening settings, where diagnostic agreement may range from fair to moderate among professionals with different levels of training. In Brazil, this challenge is further exacerbated by structural barriers, including long waiting times, limited access to specialists and medical resources, and the disproportionate concentration of healthcare professionals in major urban centers. These factors systematically delay the diagnosis of a condition that remains asymptomatic in its early stages, precisely when intervention is most effective in preventing progression to irreversible blindness. In this context, automated systems based on deep learning have emerged as a promising alternative, demonstrating high sensitivity, area under the receiver operating characteristic curve (AUC) values of up to 0.879, and the potential to safely extend follow-up intervals between retinal examinations. This study proposes the development of a computer-aided diagnostic tool for diabetic retinopathy based on a deep learning model for the classification of retinal fundus photographs (fundus images). The development of the proposed model is grounded on three main pillars: (i) diabetic retinopathy severity classification systems and their limitations in the context of emerging imaging platforms and artificial intelligence; (ii) medical image processing techniques applied to diabetic retinopathy detection; and (iii) the current state of the art in deep learning–based diabetic retinopathy diagnosis, with particular emphasis on the remaining challenges related to model interpretability and clinical reliability.
Zitieren
APA
Cunha, M. P. D., Cicero, T. S., Martins, M., Tognetti, D. M., & Noriega, C. L. (2026). Deep Learning aplicado ao diagnóstico assistido de retinopatia diabética. Revista ft, 30(159), 01-12. https://doi.org/10.69849/th866h42
Vancouver
Cunha MPD, Cicero TS, Martins M, Tognetti DM, Noriega CL. Deep Learning aplicado ao diagnóstico assistido de retinopatia diabética. Revista ft. 2026;30(159):01-12. doi:10.69849/th866h42
BibTeX
@article{Cunha2026,
author = {Cunha, Matheus Pedroza da and Cicero, Thiago Souza and Martins, Matheus and Tognetti, Danilo Miranda and Noriega, Carlos Lopez},
title = {Deep Learning aplicado ao diagnóstico assistido de retinopatia diabética},
journal = {Revista ft},
year = {2026},
volume = {30},
number = {159},
pages = {01--12},
doi = {10.69849/th866h42}
}
EUVIMED