Deep Learning Architectures for Diagnosis of Diabetic Retinopathy
Applied Sciences · 31.03.2023 · 13 (7) · 4445
DOI: 10.3390/app13074445 ↗Kurz gefasst
Fachartikel in der Zeitschrift Applied Sciences, online erschienen am 31.03.2023. Verfasst von Alberto Solano und 5 weiteren Autor:innen. Die Arbeit wurde bisher 15-mal zitiert (Crossref) und nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: MICIIN/FEDER/UE, Spanish MIU, GVA.
Abstract (Originalsprache)
For many years, convolutional neural networks dominated the field of computer vision, not least in the medical field, where problems such as image segmentation were addressed by such networks as the U-Net. The arrival of self-attention-based networks to the field of computer vision through ViTs seems to have changed the trend of using standard convolutions. Throughout this work, we apply different architectures such as U-Net, ViTs and ConvMixer, to compare their performance on a medical semantic segmentation problem. All the models have been trained from scratch on the DRIVE dataset and evaluated on their private counterparts to assess which of the models performed better in the segmentation problem. Our major contribution is showing that the best-performing model (ConvMixer) is the one that shares the approach from the ViT (processing images as patches) while maintaining the foundational blocks (convolutions) from the U-Net. This mixture does not only produce better results (DICE=0.83) than both ViTs (0.80/0.077 for UNETR/SWIN-Unet) and the U-Net (0.82) on their own but reduces considerably the number of parameters (2.97M against 104M/27M and 31M, respectively), showing that there is no need to systematically use large models for solving image problems where smaller architectures with the optimal pieces can get better results.
Förderung
- MICIIN/FEDER/UE
- Spanish MIU
- GVA
Zitieren
APA
Solano, A., Dietrich, K. N., Martínez-Sober, M., Barranquero-Cardeñosa, R., Vila-Tomás, J., & Hernández-Cámara, P. (2023). Deep Learning Architectures for Diagnosis of Diabetic Retinopathy. Applied Sciences, 13(7), 4445. https://doi.org/10.3390/app13074445
Vancouver
Solano A, Dietrich KN, Martínez-Sober M, Barranquero-Cardeñosa R, Vila-Tomás J, Hernández-Cámara P. Deep Learning Architectures for Diagnosis of Diabetic Retinopathy. Applied Sciences. 2023;13(7):4445. doi:10.3390/app13074445
BibTeX
@article{Solano2023,
author = {Solano, Alberto and Dietrich, Kevin N. and Martínez-Sober, Marcelino and Barranquero-Cardeñosa, Regino and Vila-Tomás, Jorge and Hernández-Cámara, Pablo},
title = {Deep Learning Architectures for Diagnosis of Diabetic Retinopathy},
journal = {Applied Sciences},
year = {2023},
volume = {13},
number = {7},
pages = {4445},
doi = {10.3390/app13074445}
}
EUVIMED