H. Michael Gach1,
Julie C. Price2, Denise K. Davis2, Sungkyu Jung3,
Jonathan P. Carney2, James A. Ruszkiewicz2, Brian J.
Lopresti2, Charles M. Laymon2, Chester A. Mathis2
1Radiology
& Bioengineering, University of Pittsburgh, Pittsburgh, PA, United
States; 2Radiology, University of Pittsburgh, Pittsburgh, PA,
United States; 3Statistics, University of Pittsburgh, Pittsburgh,
PA, United States
Positron emission tomography (PET) requires photon attenuation correction to accurately reconstruct the PET images. PET-MR systems use MRI to image and differentiate tissue and generate attenuation or mu-maps. Tissues are typically delineated into four categories (air, bone, fat, and water) using Dixon fat/water or ultra-short TE (UTE) pulse sequences and then assigned the applicable mu-value. Unfortunately, these sequences have limitations in differentiating the tissue types and cannot image air (due to its negligible mass and proton densities) and compact bone (due to its T2 < 20 s). A regression analysis based on brain tissue physical parameters and MRI signals indicates that tissue mu values can be accurately calculated for all brain tissues based on a combination of the data, particularly the tissues proton density and magnetic susceptibility.