EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Next‐Gen AI for Diabetic Retinopathy: Unified Multi‐Trait Fundus Analysis Using Vision Transformers and Federated Learning

Raghavendra M. Devadas, Preethi, Sowmya T, Sapna R

The Journal of Engineering · 2026-01 · 2026 (1)

DOI: 10.1049/tje2.70213 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Journal of Engineering, online erschienen am 28.07.2026. Verfasst von Raghavendra M. Devadas und 3 weiteren Autor:innen. Die Arbeit nennt 62 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Diabetic Retinopathy (DR), the most common reason for vision loss globally, presents extremely difficult diagnostic challenges such as lesion detection, disease grading and comorbidity classification. The recent innovations of deep learning (i.e., multi‐task learning [MTL], vision transformers [ViTs] and federated learning [FL]) introduced a new wave of work in the last decade. This survey extensively integrates 31 new DR studies between 2015 and 2025 across five fundamental areas: MTL, ViTs, FL, lesion segmentation, and DR grading. Comparative assessment of methodologies, datasets and performance measures (e.g., AUC, Dice and IoU) indicates future trends like hybrid models, self‐supervised ViTs and privacy‐enhancing FL workflows. Despite competitive accuracies (>90%) being reached on public databases such as EyePACS, Messidor and IDRiD by most of the models, integration of tasks, generalisability and clinical validation are issues to be solved. We present future directions as integrated multi‐trait FL architectures, lesion‐aware ViT encoders and uncertainty‐guided optimisation strategies to close the gap between technical developments and real‐world implementation in the clinic.

Zitieren

APA

Devadas, R. M., Preethi, T, S., & R, S. (2026). Next‐Gen AI for Diabetic Retinopathy: Unified Multi‐Trait Fundus Analysis Using Vision Transformers and Federated Learning. The Journal of Engineering, 2026(1). https://doi.org/10.1049/tje2.70213

Vancouver

Devadas RM, Preethi, T S, R S. Next‐Gen AI for Diabetic Retinopathy: Unified Multi‐Trait Fundus Analysis Using Vision Transformers and Federated Learning. The Journal of Engineering. 2026;2026(1). doi:10.1049/tje2.70213

BibTeX
@article{Devadas2026,
  author = {Devadas, Raghavendra M. and Preethi and T, Sowmya and R, Sapna},
  title = {Next‐Gen AI for Diabetic Retinopathy: Unified Multi‐Trait Fundus Analysis Using Vision Transformers and Federated Learning},
  journal = {The Journal of Engineering},
  year = {2026},
  volume = {2026},
  number = {1},
  doi = {10.1049/tje2.70213}
}
Kostenlos starten →