PCA-based groupwise image registration for quantitative MRI

Area of Science:

• Medical Imaging
• Biophysics
• Computational Biology

Background:

• Quantitative magnetic resonance imaging (qMRI) estimates tissue properties like T1/T2 relaxation times and diffusion coefficients.
• Accurate qMRI requires precise image registration to correct motion and distortion artifacts.
• Existing registration methods struggle with the appearance variations inherent in qMRI data.

Purpose of the Study:

• To develop and evaluate a novel groupwise image registration method for qMRI.
• To overcome challenges in registering qMRI images with significant appearance differences.
• To eliminate registration bias by avoiding the need for a reference image.

Main Methods:

• Proposed a groupwise image registration framework utilizing a principal component analysis (PCA)-based cost function.
• Exploited low-dimensional intensity variation models in qMRI without assuming specific acquisition physics.
• Evaluated the method on diverse qMRI applications (T1/T2 mapping, ADC, DTI, DCE) and 4D CT data.

Main Results:

• The proposed PCA-based groupwise method outperformed or matched state-of-the-art registration techniques across all tested qMRI applications.
• Demonstrated superior accuracy in qMRI parameter estimation, segmented structure overlap, landmark correspondence, and deformation smoothness.
• Showcased that conventional pairwise registration results are sensitive to reference image selection, unlike the proposed groupwise approach.

Conclusions:

• The PCA-based groupwise registration method is a robust and preferred technique for compensating misalignments in qMRI.
• This approach enhances the reliability and precision of quantitative tissue property measurements.
• Eliminates reference image bias, leading to more consistent and accurate qMRI analysis.