Skip to main content
eScholarship
Open Access Publications from the University of California

Spinal Cord Imaging in Multiple Sclerosis Patients: Applications of Machine Learning and Computer Vision Methods

  • Author(s): Datta, Esha
  • Advisor(s): Henry, Roland G
  • et al.
Abstract

During the last few decades, researchers have struggled to find reliable biomarkers for multiple sclerosis (MS) that could aid in diagnosis, measurement of disease progression, evaluation of treatments in clinical trials, and prediction of treatment effect. Traditional metrics, such as brain and lesion volumes, are poor contenders since they do not reliably reflect clinical metrics. Until recently, spinal cord metrics were also poor contenders, due to the quality limitations of spinal cord imaging. However, with recent technological advances, we are now able to acquire better quality spinal cord images and capture these metrics more accurately. This thesis investigates the potential of using spinal cord images clinically in MS through four different studies. The first study investigates different spinal cord metrics and shows how spinal cord PSIR gradient independently predicts EDSS in RRMS patients. The second study demonstrates two different methods for how spinal cord gray matter can be automatically segmented so that metrics can be easily obtained in a clinical setting. The third study is an investigation of how spinal cord metrics change longitudinally. The fourth study is a voxel-wise analysis of spinal cord metrics that shows local patterns of intensity, gradient, and deformation.

Main Content
Current View