Loading...
Improving the Generalizability of Medical Imaging Segmentation and Foundational Models
Kunhimon, Shahina Karuvattuparambil
Kunhimon, Shahina Karuvattuparambil
Files
Supervisor
Department
Machine Learning
Embargo End Date
Type
Dissertation
Date
2024
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
This thesis focuses on the domain of medical image segmentation and foundational models, addressing the critical challenge of enhancing model generalizability across multiple imaging modalities, such as CT, MRI, and X-rays. Medical images from different modalities exhibit distinct spectral and dimensional properties, contributing to significant variability in the data. This variability, combined with challenges such as data scarcity and domain shifts caused by differences in imaging protocols, equipment, and patient demographics, complicates the application of models trained on one dataset when applied to new or unseen clinical environments. To address these issues, the thesis focuses on self-supervised learning techniques, which reduce the dependence on large labeled datasets, a major constraint in medical imaging.
Citation
S. K. Kunhimon, "Improving the Generalizability of Medical Imaging Segmentation and Foundational Models", PhD Dissertation, Machine Learning, MBZUAI, Abu Dhabi, UAE, 2024