Deep learning for atrophy quantification in brain magnetic resonance imaging

The quantification of cerebral atrophy is fundamental in neuroinformatics since it permits diagnosing brain diseases, assessing their progression, and determining the effectiveness of novel treatments to counteract them. However, this is still an open and challenging problem since the performance 2/...

Descripción completa

Detalles Bibliográficos
Autor: Bernal Moyano, Jose
Tipo de recurso: tesis doctoral
Estado:Versión publicada
Fecha de publicación:2020
País:España
Institución:CBUC, CESCA
Repositorio:TDR. Tesis Doctorales en Red
OAI Identifier:oai:www.tdx.cat:10803/671699
Acceso en línea:http://hdl.handle.net/10803/671699
Access Level:acceso abierto
Palabra clave:Aprenentatge profund
Aprendizaje profundo
Deep learning
Atròfia cerebral
Atrofia cerebral
Cerebral atrophy
Quantificació
Cuantificación
Quantification
Xarxes neuronals convolucionals
Redes neuronales convolucionales
Convolutional neural networks
Segmentació de teixits
Segmentación de tejidos
Tissue segmentation
Imatges per ressonància magnètica
Imágenes por resonancia magnética
Magnetic resonance imaging
Ressonància magnètica cerebral
Resonancia magnética cerebral
Brain MRI
004
615
616.8
Descripción
Sumario:The quantification of cerebral atrophy is fundamental in neuroinformatics since it permits diagnosing brain diseases, assessing their progression, and determining the effectiveness of novel treatments to counteract them. However, this is still an open and challenging problem since the performance 2/2 of traditional methods depends on imaging protocols and quality, data harmonisation errors, and brain abnormalities. In this doctoral thesis, we question whether deep learning methods can be used for better estimating cerebral atrophy from magnetic resonance images. Our work shows that deep learning can lead to a state-of-the-art performance in cross-sectional assessments and compete and surpass traditional longitudinal atrophy quantification methods. We believe that the proposed cross-sectional and longitudinal methods can be beneficial for the research and clinical community