Image registration and statistical analysis for quantitative in vivo spin-lock magnetic resonance imaging of the intervertebral disc response to compression

Joey Ann Kimdon

EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2007-95
August 2, 2007

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2007/EECS-2007-95.pdf

Low back pain decreases the quality of life and affects almost 80% of the population. Quantitative spin-lock magnetic resonance imaging of the T1rho relaxation parameter is a promising marker of early degeneration in intervertebral discs. However, T1rho measurements may be affected by disc compression, and understanding this effect is necessary for studying abnormal dynamics of disc response to compression or correlations between degeneration and T1rho values.

We developed an algorithm to register images of intervertebral discs in different compression states and to calculate the statistical significance of local changes in T1rho. Our procedure includes automatic registration during image acquisition to view the same location across exams, segmentation of intervertebral discs with minimal user intervention, automatic registration of discs using a non-rigid transformation guided by rigid transformations of vertebrae, and a method of investigating the statistical significance of changes in local neighborhoods using resampling hypothesis testing with variable confidence levels.

Validation tests on phantoms and volunteer data indicate sub-pixel registration accuracy and precision. Registration results agree with a manual gold standard. Hypothesis testing is sensitive to registration accuracy, indicating the need for registration even with patient movements of less than a millimeter. Tests confirm expected trade-offs between type I and type II statistical significance errors depending on neighborhood size, significance level, and confidence level.

We demonstrate the algorithm on varying compression states caused by lying down after 30 minutes of standing with a 20 lb backpack or by supporting 55 pounds while lying using a leg-press-like device. T1rho varies locally depending on the distribution of compression within the discs. The coefficient of variation was 7.6% while relaxing and 13.9% when applying and releasing the compression device, indicating an effect of the device. Variation was 48% less in degenerated versus healthy discs (p<0.04). T1rho in the lower discs increased by 20% (p<0.05) when the compression device pressure was released. Local comparisons showed that T1rho tended to decrease under compression and increase after compression was released. Changes were not significant during an hour of supine resting, suggesting that variable pre-scan compression from normal daily activities does not considerably complicate T1rho measurements.

Advisor: Thomas F. Budinger


BibTeX citation:

@phdthesis{Kimdon:EECS-2007-95,
    Author = {Kimdon, Joey Ann},
    Title = {Image registration and statistical analysis for quantitative in vivo spin-lock magnetic resonance imaging of the intervertebral disc response to compression},
    School = {EECS Department, University of California, Berkeley},
    Year = {2007},
    Month = {Aug},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2007/EECS-2007-95.html},
    Number = {UCB/EECS-2007-95},
    Abstract = {<p>
Low back pain decreases the quality of life and affects almost 80&#37; of the population. Quantitative spin-lock magnetic resonance imaging of the T1rho relaxation parameter is a promising marker of early degeneration in intervertebral discs. However, T1rho measurements may be affected by disc compression, and understanding this effect is necessary for studying abnormal dynamics of disc response to compression or correlations between degeneration and T1rho values.
</p><p>
We developed an algorithm to register images of intervertebral discs in different compression states and to calculate the statistical significance of local changes in T1rho. Our procedure includes automatic registration during image acquisition to view the same location across exams, segmentation of intervertebral discs with minimal user intervention, automatic registration of discs using a non-rigid transformation guided by rigid transformations of vertebrae, and a method of investigating the statistical significance of changes in local neighborhoods using resampling hypothesis testing with variable confidence levels.
</p><p>
Validation tests on phantoms and volunteer data indicate sub-pixel registration accuracy and precision. Registration results agree with a manual gold standard. Hypothesis testing is sensitive to registration accuracy, indicating the need for registration even with patient movements of less than a millimeter. Tests confirm expected trade-offs between type I and type II statistical significance errors depending on neighborhood size, significance level, and confidence level.
</p><p>
We demonstrate the algorithm on varying compression states caused by lying down after 30 minutes of standing with a 20 lb backpack or by supporting 55 pounds while lying using a leg-press-like device. T1rho varies locally depending on the distribution of compression within the discs. The coefficient of variation was 7.6&#37; while relaxing and 13.9&#37; when applying and releasing the compression device, indicating an effect of the device. Variation was 48&#37; less in degenerated versus healthy discs (p&#60;0.04). T1rho in the lower discs increased by 20&#37; (p&#60;0.05) when the compression device pressure was released. Local comparisons showed that T1rho tended to decrease under compression and increase after compression was released. Changes were not significant during an hour of supine resting, suggesting that variable pre-scan compression from normal daily activities does not considerably complicate T1rho measurements.}
}

EndNote citation:

%0 Thesis
%A Kimdon, Joey Ann
%T Image registration and statistical analysis for quantitative in vivo spin-lock magnetic resonance imaging of the intervertebral disc response to compression
%I EECS Department, University of California, Berkeley
%D 2007
%8 August 2
%@ UCB/EECS-2007-95
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2007/EECS-2007-95.html
%F Kimdon:EECS-2007-95