Reduction of truncation artifacts in rapid 3D articular cartilage imaging

Abstract

Purpose: To reduce Gibbs ringing artifact in three-dimensional (3D) articular knee cartilage imaging with linear prediction (LP). Materials and Methods:A reconstruction method using LP in 3D was applied to truncated data sets of six healthy knees. The technique first linearizes the data before applying the prediction algorithm. Three radiologists blindly reviewed and ranked images of the full, truncated, and predicted data sets. Statistical analysis of the radiologists' reviews was performed for image quality, clinical acceptability of the images, and equivalence with the gold standard.Results: LP applied to 3D knee cartilage imaging allows for 40% decreased scan time while providing image quality with statistical equivalence to a full data set. Conclusion:3D spoiled gradient echo imaging (SPGR) knee cartilage imaging requires significant scan time. This 40% reduction in scan time will allow such scans to be more feasible without sacrificing clinical acceptability.

Common Mistakes and Pitfalls in Magnetic Resonance Imaging of the Knee

“…Interpretation errors due to technical artifacts such as magic angle phenomenon on imaging with low echo time (TE) [ 9 ], truncation artefacts on gradient echo imaging [ 10 ], blurring artifact on fast spin echo T2-WI [ 11 ], phase-artifacts and susceptibility artifacts [ 12 ] will not discussed in this short overview.…”

Removal of truncation artefacts in magnetic resonance images by recovering missing spectral data

“…For reconstruction-based truncation artefact removal, a typical example is the class of classical reconstruction technique based on phase compensation [13]. Other examples are the methods based on extrapolating truncated k-space data using various techniques such as the autoregressive moving average (ARMA) model [14] or linear predication [15][16][17][18]. Directly extrapolating kspace data (or one-dimensionally transformed k-space) has to be undertaken with caution since the k-space data is rather decorrelated in comparison with the spatial data in the original image domain.…”