Robust deep learning for eye fundus images: Bridging real and synthetic data for enhancing generalization

Registro completo de metadados
MetadadosDescriçãoIdioma
Autor(es): dc.contributorUniversidade Estadual Paulista (UNESP)-
Autor(es): dc.contributorRoyal Melbourne Institute of Technology-
Autor(es): dc.contributorUniversity of Melbourne-
Autor(es): dc.creatorOliveira, Guilherme C.-
Autor(es): dc.creatorRosa, Gustavo H.-
Autor(es): dc.creatorPedronette, Daniel C.G.-
Autor(es): dc.creatorPapa, João P.-
Autor(es): dc.creatorKumar, Himeesh-
Autor(es): dc.creatorPassos, Leandro A.-
Autor(es): dc.creatorKumar, Dinesh-
Data de aceite: dc.date.accessioned2025-08-21T15:25:25Z-
Data de disponibilização: dc.date.available2025-08-21T15:25:25Z-
Data de envio: dc.date.issued2025-04-29-
Data de envio: dc.date.issued2024-08-01-
Fonte completa do material: dc.identifierhttp://dx.doi.org/10.1016/j.bspc.2024.106263-
Fonte completa do material: dc.identifierhttps://hdl.handle.net/11449/304275-
Fonte: dc.identifier.urihttp://educapes.capes.gov.br/handle/11449/304275-
Descrição: dc.descriptionDeep learning applications for assessing medical images are limited because the datasets are often small and imbalanced. The use of synthetic data has been proposed in the literature, but neither a robust comparison of the different methods nor generalizability has been reported. Our approach integrates a retinal image quality assessment model and StyleGAN2 architecture to enhance Age-related Macular Degeneration (AMD) detection capabilities and improve generalizability. This work compares ten different Generative Adversarial Network (GAN) architectures to generate synthetic eye-fundus images with and without AMD. We combined subsets of three public databases (iChallenge-AMD, ODIR-2019, and RIADD) to form a single training and test set. We employed the STARE dataset for external validation, ensuring a comprehensive assessment of the proposed approach. The results show that StyleGAN2 reached the lowest Fréchet Inception Distance (166.17), and clinicians could not accurately differentiate between real and synthetic images. ResNet-18 architecture obtained the best performance with 85% accuracy and outperformed the two human experts (80% and 75%) in detecting AMD fundus images. The accuracy rates were 82.8% for the test set and 81.3% for the STARE dataset, demonstrating the model's generalizability. The proposed methodology for synthetic medical image generation has been validated for robustness and accuracy, with free access to its code for further research and development in this field.-
Descrição: dc.descriptionDepartment of Biotechnology, Ministry of Science and Technology, India-
Descrição: dc.descriptionFundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)-
Descrição: dc.descriptionConselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)-
Descrição: dc.descriptionStiftelsen Promobilia-
Descrição: dc.descriptionEngineering and Physical Sciences Research Council-
Descrição: dc.descriptionSchool of Sciences São Paulo State University-
Descrição: dc.descriptionSchool of Engineering Royal Melbourne Institute of Technology-
Descrição: dc.descriptionCentre of Eye Research University of Melbourne-
Descrição: dc.descriptionSchool of Sciences São Paulo State University-
Descrição: dc.descriptionFAPESP: #2013/07375-0-
Descrição: dc.descriptionFAPESP: #2014/12236-1-
Descrição: dc.descriptionFAPESP: #2018/15597-6-
Descrição: dc.descriptionFAPESP: #2019/00585-5-
Descrição: dc.descriptionFAPESP: #2019/02205-5-
Descrição: dc.descriptionFAPESP: #2019/07665-4-
Descrição: dc.descriptionFAPESP: #2023/10823-6-
Descrição: dc.descriptionCNPq: #307066/2017-7-
Descrição: dc.descriptionCNPq: #309439/2020-5-
Descrição: dc.descriptionCNPq: #427968/2018-6-
Descrição: dc.descriptionCNPq: #88887.606573/2021-00-
Descrição: dc.descriptionStiftelsen Promobilia: 2019-
Descrição: dc.descriptionEngineering and Physical Sciences Research Council: EP/T021063/1-
Descrição: dc.descriptionStiftelsen Promobilia: P-134-
Idioma: dc.languageen-
Relação: dc.relationBiomedical Signal Processing and Control-
???dc.source???: dc.sourceScopus-
Palavras-chave: dc.subjectAge-related macular degeneration-
Palavras-chave: dc.subjectData augmentation-
Palavras-chave: dc.subjectDeep learning-
Palavras-chave: dc.subjectGenerative Adversarial Networks-
Palavras-chave: dc.subjectMedical images-
Palavras-chave: dc.subjectStyleGAN2-
Título: dc.titleRobust deep learning for eye fundus images: Bridging real and synthetic data for enhancing generalization-
Tipo de arquivo: dc.typelivro digital-
Aparece nas coleções:Repositório Institucional - Unesp

Não existem arquivos associados a este item.