PROJECT SUMMARY:
This proposal is devoted to the development, clinical implementation and evaluation of a novel, methodology
for generating brain sodium images of unparalleled resolution and signal-to-noise-ratio. This methodology is a
result or recent developments in Bayesian image reconstruction and manifold mapping techniques that
together make it possible to robustly utilize anatomical information from high-resolution brain scans to guide
the reduction of partial voluming effects from lower resolution, lower signal-to-noise ratio (SNR) brain scans.
Though such “constrained reconstruction” schemes have been long-advocated in the MRI imaging community,
their computational demands, limited performance and heavy operator input had previously limited their
evaluation and practical characterization thus rendering them impractical for use within the confines of a
clinical environment. Recently, we have shown that the combined use of segmentation free Bayesian
approaches together with manifold mapping techniques can be used to provide a fast framework for
Anatomically Guided Reconstruction where the information form high-resolution brain scans is used, routinely,
to improve spatial resolution and SNR of concurrently acquired brain PET scans leading to improved sensitivity
for the detection of low-contrast lesions in the brain. Our initial experience with extensions of this approach for
sodium MRI of the brain has produced images of spatial resolution and SNR that far exceed what had been
previously achieved using high and ultra-high magnetic field strengths. Establishing the limits of this approach
and characterizing its performance on clinically relevant cases is the thrust of this proposal.
