Authors:

Dave Tailor

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

For select magnetic resonance (MR) acquisitions in the cardiothoracic region, the separation of water and fat signals by means of chemical shift correction is paramount for quantitative characterization. Conventional methods utilize iterative optimization-based separation, and while effective in most anatomical regions, cardiac water fat separation suffers from suboptimal initialization. Here, we examine a multistep graph cut initialization method known as Simultaneous Phase Unwrapping and Removal of Chemical Shift (SPURS) [1] applied to data obtained across a cardiac cycle. Inherent is the physiological consistency of fat between timepoints in motion-invariant regions, that can be exploited by constructing a single low-rank initialization seed point derived across multiple cardiac phase volumes. A series of low-rank plus sparsity (L+S) decomposition algorithms was optimized to separate beating motion and temporal outliers. Singular Value Decomposition (SVD) was employed across the temporal dimension of the cardiac MR data to compute SPRUS using the primary non-sparse channels, thereby lessening the weight of sparse channels that contain outliers. The proposed processing was compared against conventional SPURS processing per volume frame. Region-of-interest measurement analysis was performed by AHA-classified ROIs (n=960) using both SPURS outputs. Computation benchmarking was additionally performed. Three quantitative measurements were derived after water fat separation occurred to evaluate the multivariable optimization at every time point. This similarity was achieved with 2. 13+0. 31 acceleration on an Apple M4 Pro System. Pearson's correlation yielded R=0.93 and 0.97 (n=960) in derived quantitative ROI measures, where numerical outliers suggested algorithmic correction of failed measures using our method. This approach with two-fold computation reduction, yielded comparable outputs while further addressing initialization errors associated with SPURS.

Source:

Illinois Institute of Technology

Topics:

No topics listed

Co-authors:

Dave Tailor