ConKeD: multiview contrastive descriptor learning for keypoint-based retinal image registration
Medical & Biological Engineering & Computing · 2024-12 · 62 (12) · 3721-3736
DOI: 10.1007/s11517-024-03160-6 ↗Kurz gefasst
Fachartikel in der Zeitschrift Medical & Biological Engineering & Computing, online erschienen am 06.07.2024. Verfasst von David Rivas-Villar und 3 weiteren Autor:innen. Beteiligte Einrichtung: Grupo VARPA, Instituto de Investigacion Biomédica de A Coruña (INIBIC), Universidade da Coruña, A Coruña, 15006, A Coruña, Spain. david.rivas.villar@…. Hauptthemen laut MEDLINE: Retina, Deep Learning. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Ministerio de Ciencia e Innovación, Ministerio de Ciencia e Innovaciónn, Consellería de Cultura, Educación e Universidade, Xunta de Galicia.
Abstract (Originalsprache)
Retinal image registration is of utmost importance due to its wide applications in medical practice. In this context, we propose ConKeD, a novel deep learning approach to learn descriptors for retinal image registration. In contrast to current registration methods, our approach employs a novel multi-positive multi-negative contrastive learning strategy that enables the utilization of additional information from the available training samples. This makes it possible to learn high-quality descriptors from limited training data. To train and evaluate ConKeD, we combine these descriptors with domain-specific keypoints, particularly blood vessel bifurcations and crossovers, that are detected using a deep neural network. Our experimental results demonstrate the benefits of the novel multi-positive multi-negative strategy, as it outperforms the widely used triplet loss technique (single-positive and single-negative) as well as the single-positive multi-negative alternative. Additionally, the combination of ConKeD with the domain-specific keypoints produces comparable results to the state-of-the-art methods for retinal image registration, while offering important advantages such as avoiding pre-processing, utilizing fewer training samples, and requiring fewer detected keypoints, among others. Therefore, ConKeD shows a promising potential towards facilitating the development and application of deep learning-based methods for retinal image registration. Graphical abstract
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Grupo VARPA, Instituto de Investigacion Biomédica de A Coruña (INIBIC), Universidade da Coruña, A Coruña, 15006, A Coruña, Spain. david.rivas.villar@udc.es.
- Departamento de Ciencias de la Computación y Tecnologías de la Información, Universidade da Coruña, A Coruña, 15071, A Coruña, Spain. david.rivas.villar@udc.es.
- Grupo VARPA, Instituto de Investigacion Biomédica de A Coruña (INIBIC), Universidade da Coruña, A Coruña, 15006, A Coruña, Spain.
- Departamento de Ciencias de la Computación y Tecnologías de la Información, Universidade da Coruña, A Coruña, 15071, A Coruña, Spain.
Förderung
- Ministerio de Ciencia e Innovación
- Ministerio de Ciencia e Innovaciónn
- Consellería de Cultura, Educación e Universidade, Xunta de Galicia
- Instituto de Salud Carlos III
Zitieren
APA
Rivas-Villar, D., Hervella, Á. S., Rouco, J., & Novo, J. (2024). ConKeD: multiview contrastive descriptor learning for keypoint-based retinal image registration. Medical & Biological Engineering & Computing, 62(12), 3721-3736. https://doi.org/10.1007/s11517-024-03160-6
Vancouver
Rivas-Villar D, Hervella ÁS, Rouco J, Novo J. ConKeD: multiview contrastive descriptor learning for keypoint-based retinal image registration. Medical & Biological Engineering & Computing. 2024;62(12):3721-3736. doi:10.1007/s11517-024-03160-6
BibTeX
@article{RivasVillar2024,
author = {Rivas-Villar, David and Hervella, Álvaro S and Rouco, José and Novo, Jorge},
title = {ConKeD: multiview contrastive descriptor learning for keypoint-based retinal image registration},
journal = {Medical & Biological Engineering & Computing},
year = {2024},
volume = {62},
number = {12},
pages = {3721--3736},
doi = {10.1007/s11517-024-03160-6}
}
EUVIMED