Purpose: To demonstrate the feasibility of using the keyhole technique to minimize error in a least squares regression estimation of T(1rho) from magnetic resonance (MR) image data.
Materials and methods: The keyhole method of partial k-space acquisition was simulated using data from a virtual phantom and MR images of ex vivo bovine and in vivo human cartilage. T(1rho) maps were reconstructed from partial k-space (keyhole) image data using linear regression, and error was measured with relation to T(1rho) maps created from the full k-space images. An error model was created based on statistical theory and fitted to the error measurements.
Results: T(1rho) maps created from keyhole images of a human knee produced levels of error on the order of 1% while reducing standard image acquisition time approximately by half. The resultant errors were strongly correlated with expectations derived from statistical theory.
Conclusion: The error model can be used to analytically optimize the keyhole method in order to minimize the overall error in the estimation of the relaxation parameter of interest. The keyhole method can be generalized to significantly expedite all forms of relaxation mapping.
Copyright 2003 Wiley-Liss, Inc.