“Exploring Myelin Water Imaging: from Application to Atlas to Algorithm”

Event Date:
2020-06-16T14:00:00
2020-06-16T17:00:00
Event Location:
via Zoom
Speaker:
HANWEN (KEVIN) LIU
Related Upcoming Events:
Grad Theses
Intended Audience:
Public
Local Contact:

Physics and Astronomy

Abstract:

 

Event Information:

Departmental Doctoral Oral Examination

Abstract:
Myelin water imaging (MWI) is a quantitative magnetic resonance (MR) method that specifically measures the myelin content in the central nervous system. MWI operates on the principle that the MR signal of water trapped between myelin bilayers can be extracted from the total MR signal based on a characteristic short T2 relaxation time. The ratio of myelin water signal relative to the total signal is termed myelin water fraction (MWF), used as a quantitative biomarker for myelin. This thesis explores three aspects of MWI: application, atlas, and algorithm.

Firstly, the MWI was applied to study cervical spondylotic myelopathy (CSM), which is a common spinal cord neurodegenerative disease. The function of the spinal cord conduction was assessed by an electrophysiologic technique called somatosensory evoked potentials (SSEP). Significant MWF reduction was observed in those CSM patients with functional deficits (e.g. delayed SSEP latency). A linear correlation between the MWF and the SSEP latency was discovered in CSM.

Secondly, the MWI atlases, which represent the MWI normative references of the normal myelin distribution in the brain and spinal cord, were created by coregistering and averaging the MWI images acquired from many healthy volunteers. These resulting atlases were utilized to demonstrate areas of demyelination in individuals with pathological conditions such as multiple sclerosis. The MWI atlases have been uploaded on the Internet and made publicly available. 

Thirdly, the current MWI data analysis, based on the non-negative least squares (NNLS) method, was accelerated by implementing the neural network (NN) algorithm. A NN model was trained by the ground truth labels produced by the commonly used NNLS method. The trained NN model achieved to yield a whole-brain MWF map in 33 seconds, which is 150 faster than the NNLS method.

Finally, a novel T2 data analysis method, namely the spectrum analysis for multiple exponentials via experimental condition oriented simulation (SAME-ECOS), was proposed. SAME-ECOS is a simulation-derived solver that tailored for different MR experimental conditions. When dealing with the MWI data, it is found that SAME-ECOS largely surpassed the NNLS method in terms of calculation accuracy and speed.

Add to Calendar 2020-06-16T14:00:00 2020-06-16T17:00:00 “Exploring Myelin Water Imaging: from Application to Atlas to Algorithm” Event Information: Departmental Doctoral Oral Examination Abstract: Myelin water imaging (MWI) is a quantitative magnetic resonance (MR) method that specifically measures the myelin content in the central nervous system. MWI operates on the principle that the MR signal of water trapped between myelin bilayers can be extracted from the total MR signal based on a characteristic short T2 relaxation time. The ratio of myelin water signal relative to the total signal is termed myelin water fraction (MWF), used as a quantitative biomarker for myelin. This thesis explores three aspects of MWI: application, atlas, and algorithm. Firstly, the MWI was applied to study cervical spondylotic myelopathy (CSM), which is a common spinal cord neurodegenerative disease. The function of the spinal cord conduction was assessed by an electrophysiologic technique called somatosensory evoked potentials (SSEP). Significant MWF reduction was observed in those CSM patients with functional deficits (e.g. delayed SSEP latency). A linear correlation between the MWF and the SSEP latency was discovered in CSM. Secondly, the MWI atlases, which represent the MWI normative references of the normal myelin distribution in the brain and spinal cord, were created by coregistering and averaging the MWI images acquired from many healthy volunteers. These resulting atlases were utilized to demonstrate areas of demyelination in individuals with pathological conditions such as multiple sclerosis. The MWI atlases have been uploaded on the Internet and made publicly available.  Thirdly, the current MWI data analysis, based on the non-negative least squares (NNLS) method, was accelerated by implementing the neural network (NN) algorithm. A NN model was trained by the ground truth labels produced by the commonly used NNLS method. The trained NN model achieved to yield a whole-brain MWF map in 33 seconds, which is 150 faster than the NNLS method. Finally, a novel T2 data analysis method, namely the spectrum analysis for multiple exponentials via experimental condition oriented simulation (SAME-ECOS), was proposed. SAME-ECOS is a simulation-derived solver that tailored for different MR experimental conditions. When dealing with the MWI data, it is found that SAME-ECOS largely surpassed the NNLS method in terms of calculation accuracy and speed. Event Location: via Zoom