Automatic segmentation of brain structures in magnetic resonance images using deep learning techniques
Year: 2020
Type: dissertation
Abstract: This PhD thesis focuses on the development of deep learning based methods for \naccurate segmentation of the sub-cortical brain structures from MRI. First, \nwe have proposed a 2.5D CNN architecture that combines convolutional and \n2/2 \nspatial features. Second, we proposed a supervised domain adaptation \ntechnique to improve the robustness and consistency of deep learning model. \nThird, an unsupervised domain adaptation method was proposed to eliminate the \nrequirement of manual intervention to train a deep learning model that is \nrobust to differences in the MRI images from multi-centre and multi-scanner \ndatasets. The experimental results for all the proposals demonstrated the \neffectiveness of our approaches in accurately segmenting the sub-cortical \nbrain structures and has shown state-of-the-art performance on well-known \npublicly available datasets
Author: Kaisar Kushibar
Institution: Universitat de Girona
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funders: Nvidia, Fundació la Marató de TV3
Awards: G2217399429, G3712997785, G5997417279
OpenAlex ID: W3158057394
