SU-F-J-165: Spectral Characterization of Tissues in High Spectral and Spatial Resolution MR Images in Vitro: Implications for Classification-Based Synthetic CT Algorithm

Authors


Abstract

Purpose:

To investigate features of spectral parameters of tissue for use in a synthetic CT algorithm.

Methods:

A phantom of fresh beef leg embedded in 1% agarose gel was imaged at 1.5T on a Siemens Aero using HiSS MRI with 1.0mm resolution and 11.1Hz spectral resolution. Post-processing was performed off-line and included labeling of fat and water peaks, and fitting the spectra. All spectra were normalized by the noise within each voxel; thus all parameters are reported in units of signal-to-noise (SNR). Images of the peak height (PH) and peak integral (PI) of both the water and fat resonances were analyzed. Several regions-of-interest (ROIs) were identified: bone marrow, cortical bone, adipose tissue, muscle, agar gel, and air. Distributions and scatterplots were generated. Water PH in cortical bone was compared to air using a t-test. Composition of the various ROIs was also reported.

Results:

Bone marrow was 97% fat (mean PH=4.5SNR). Adipose tissue was mostly mixed fat/water voxels (66%); mean fat and water PH were 4.1 and 7.8 SNR, respectively. Cortical bone was 70% water (PH=2.3SNR). Muscle was 85% water (PH=4.2SNR, with fat PH=1.6SNR). For the air ROI, mean fat and water PH were 1.1SNR and 1.1SNR, respectively. The scatterplot of PH (water vs. fat) showed fair separation of bone marrow and adipose tissue. Water vs. fat PI scatterplot showed good separation of muscle and cortical bone. Cortical bone water PH was significantly different than air (p<0.05). A scatterplot of the water PH vs. pixel value of a summation of the first three echoes showed a different distribution with slightly better separation of air and bone.

Conclusion:

This study shows promising results for utilizing HiSS imaging in a classification-based syn-CT algorithm. Reducing early echo times would improve SNR of bone and will be investigated along with further optimization.

Ancillary