Meeting Banner
Abstract #0360

A Bayesian Algorithm Using Spatial Priors for Multi-Exponential T2 Relaxometry from Multi-Echo Spin Echo MRI

MAGNA25Dushyant Kumar1, Thanh Nguyen2, Susan Gauthier3, Ashish Raj2

1Neuroradiology, University of Hamburg, Hamburg, Germany; 2Radiology, Weill Cornell Medical College, Newyork, NY, United States; 3neurology, Weill Cornell Medical College, Newyork, NY, United States

Problem: Because of the ill-posedness of the inverse problem with more unknowns than number of echoes, the accurate quantification of myelin water fraction (MWF) from T2-relaxometry requires high SNR (~500-1000). Methods: The voxelwise conventional regularization is performed followed by implementation of the spatial smoothness over local neighborhood using proposed spatial approach. Results: The inferred MWF-map has significantly reduced spatial variations resulting in better tissue-differentiation and is superior to conventionally regularized version based on various criteria. Conclusions: Spatial constraints allow the handling of lower SNR data which may allow better MWF reproducibility for longitudinal or multi-site studies and warrants further evaluation.

Keywords

able achieve agreement allow allowed anatomical approaches assuming better blocks brain capsule central chosen collected college column comparable compartment consisted consistent consisting constant constraints contributions conventionally corpus decode detection developed diagonal diseases distribution distributions echoes error evaluate evaluation expectations exploits exponential extracted fidelity fitting flair form forward function graham hamburg healthy implies imposes imposing improve improved improvement inferred internal inversion least lesions logarithmically maps masks matrix measured medical minimization minimized noise norm numerical optimum parametric particular particularly patient penalizes performance phantom pools preliminary press priors problem proposed quantities reconstructed reconstruction recovered reduced regarding regularization regularized relevant reported residual resolve respective review robustness sclerosis score sensitive separate setting seven significantly simulated site slice smooth smoothness software spaced spatial spatially spin square stability structures studies superior surrounding symmetric system table temporal term theory tissue tissues typical underlying uniformity variance various vary varying vector vectors version visually volunteers warrants water weak whereby whose wise