Logotipo do repositório
 

Publicação:
A Fully Unsupervised Deep Learning Framework for Non-Rigid Fundus Image Registration

dc.contributor.authorBenvenuto, Giovana A. [UNESP]
dc.contributor.authorColnago, Marilaine
dc.contributor.authorDias, Maurício A. [UNESP]
dc.contributor.authorNegri, Rogério G. [UNESP]
dc.contributor.authorSilva, Erivaldo A. [UNESP]
dc.contributor.authorCasaca, Wallace [UNESP]
dc.contributor.institutionUniversidade Estadual Paulista (UNESP)
dc.contributor.institutionUniversidade de São Paulo (USP)
dc.date.accessioned2023-03-01T21:13:09Z
dc.date.available2023-03-01T21:13:09Z
dc.date.issued2022-08-01
dc.description.abstractIn ophthalmology, the registration problem consists of finding a geometric transformation that aligns a pair of images, supporting eye-care specialists who need to record and compare images of the same patient. Considering the registration methods for handling eye fundus images, the literature offers only a limited number of proposals based on deep learning (DL), whose implementations use the supervised learning paradigm to train a model. Additionally, ensuring high-quality registrations while still being flexible enough to tackle a broad range of fundus images is another drawback faced by most existing methods in the literature. Therefore, in this paper, we address the above-mentioned issues by introducing a new DL-based framework for eye fundus registration. Our methodology combines a U-shaped fully convolutional neural network with a spatial transformation learning scheme, where a reference-free similarity metric allows the registration without assuming any pre-annotated or artificially created data. Once trained, the model is able to accurately align pairs of images captured under several conditions, which include the presence of anatomical differences and low-quality photographs. Compared to other registration methods, our approach achieves better registration outcomes by just passing as input the desired pair of fundus images.en
dc.description.affiliationFaculty of Science and Technology (FCT) São Paulo State University (UNESP)
dc.description.affiliationInstitute of Mathematics and Computer Science (ICMC) São Paulo University (USP)
dc.description.affiliationScience and Technology Institute (ICT) São Paulo State University (UNESP)
dc.description.affiliationInstitute of Biosciences Letters and Exact Sciences (IBILCE) São Paulo State University (UNESP)
dc.description.affiliationUnespFaculty of Science and Technology (FCT) São Paulo State University (UNESP)
dc.description.affiliationUnespScience and Technology Institute (ICT) São Paulo State University (UNESP)
dc.description.affiliationUnespInstitute of Biosciences Letters and Exact Sciences (IBILCE) São Paulo State University (UNESP)
dc.identifierhttp://dx.doi.org/10.3390/bioengineering9080369
dc.identifier.citationBioengineering, v. 9, n. 8, 2022.
dc.identifier.doi10.3390/bioengineering9080369
dc.identifier.issn2306-5354
dc.identifier.scopus2-s2.0-85137360100
dc.identifier.urihttp://hdl.handle.net/11449/241615
dc.language.isoeng
dc.relation.ispartofBioengineering
dc.sourceScopus
dc.subjectcomputer vision applications
dc.subjectdeep learning
dc.subjectfundus image
dc.subjectimage registration
dc.titleA Fully Unsupervised Deep Learning Framework for Non-Rigid Fundus Image Registrationen
dc.typeArtigo
dspace.entity.typePublication
unesp.author.orcid0000-0002-0531-1284[1]
unesp.author.orcid0000-0003-1599-491X[2]
unesp.author.orcid0000-0002-1361-6184[3]
unesp.author.orcid0000-0002-4808-2362[4]
unesp.author.orcid0000-0002-7069-0479[5]
unesp.author.orcid0000-0002-1073-9939[6]
unesp.departmentEstatística - FCTpt

Arquivos