Manifold Learning for Brain Tumor MRI Image Retrieval and Classification

Registro completo de metadados
MetadadosDescriçãoIdioma
Autor(es): dc.contributorUniversidade Estadual Paulista (UNESP)-
Autor(es): dc.creatorde Antonio, André Lara Temple-
Autor(es): dc.creatorPedronette, Daniel Carlos Guimarães-
Data de aceite: dc.date.accessioned2025-08-21T20:44:51Z-
Data de disponibilização: dc.date.available2025-08-21T20:44:51Z-
Data de envio: dc.date.issued2025-04-29-
Data de envio: dc.date.issued2022-12-31-
Fonte completa do material: dc.identifierhttp://dx.doi.org/10.1109/BIBE60311.2023.00014-
Fonte completa do material: dc.identifierhttps://hdl.handle.net/11449/309985-
Fonte: dc.identifier.urihttp://educapes.capes.gov.br/handle/11449/309985-
Descrição: dc.descriptionThe evolution of image acquisition and storage technologies has been fundamental in numerous medical fields, supporting doctors to deliver more precise diagnoses and, consequently, recommend more effective treatments for their patients. Recently, deep learning techniques have played a key role in more accurate medical image analysis, mainly due to the capacity to effectively represent the image visual content. However, in spite of tremendous advances, deep-learning techniques commonly require huge quantities of data for training, that are not available in many scenarios, especially in the medical domain. Conversely, manifold learning techniques have been successfully applied in unsupervised and semi-supervised scenarios for more effective encoding of similarity relationships between multimedia data in the absence or restriction of labeled data. In this work, we propose to exploit jointly the representation power of deep-learning strategies with the ability of unsupervised manifold learning in delivering more effective similarity measurement. Convolutional Neural Networks (CNNs) and Transformer-based models trained through transfer learning are combined by unsupervised manifold learning methods, which define a more effective similarity among images. The output can be used for unsupervised retrieval and semi-supervised classification based on a k-NN strategy. An experimental evaluation was conducted on different datasets of MRI brain tumor images, considering different features. Effective results were obtained on both retrieval and classification tasks, with significant gains obtained by manifold learning approaches. In scenarios with limited training data, our approach achieves results that are competitive or superior to state-of-the-art deep learning approaches.-
Descrição: dc.descriptionFundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)-
Descrição: dc.descriptionDepartment of Statistics Applied Mathematics and Computing São Paulo State University (UNESP)-
Descrição: dc.descriptionDepartment of Statistics Applied Mathematics and Computing São Paulo State University (UNESP)-
Descrição: dc.descriptionFAPESP: #2018/15597-6-
Formato: dc.format36-42-
Idioma: dc.languageen-
Relação: dc.relationProceedings - 2023 IEEE 23rd International Conference on Bioinformatics and Bioengineering, BIBE 2023-
???dc.source???: dc.sourceScopus-
Palavras-chave: dc.subjectbrain tumor-
Palavras-chave: dc.subjectfeature extraction-
Palavras-chave: dc.subjectfusion-
Palavras-chave: dc.subjectknn classification-
Palavras-chave: dc.subjectmedical images-
Palavras-chave: dc.subjectMRI-
Palavras-chave: dc.subjectranking-
Palavras-chave: dc.subjectunsupervised learning-
Título: dc.titleManifold Learning for Brain Tumor MRI Image Retrieval and Classification-
Tipo de arquivo: dc.typeaula digital-
Aparece nas coleções:Repositório Institucional - Unesp

Não existem arquivos associados a este item.