K -space acquisition method for dynamic contrast -enhanced MRI: Application to breast tumors.

Abstract

Dynamic contrast-enhanced (DCE)-MRI is increasingly being used for detection and diagnosis of tumors. The objective of DCE-MRI is to elicit diagnostically significant architectural and pharmacokinetic features of lesions. Hence, DCE-MRI of tumors is ideally performed at high spatial resolution while sampling a time-varying process at high temporal resolution. A variety of variable rate sampling strategies and associated reconstructive schemes have been developed to resolve the conflicting demands of simultaneous high resolution sampling of temporal and spatial detail. In this work a novel method, termed spatio-temporal bandwidth-based (STBB) acquisition, is developed to address the trade-off inherent in DCE-MRI. This technique is constrained only by the overall scan duration, within which the temporal changes are expected to reach steady-state, the imaging sequence repetition time (TR), and a targeted spatio-temporal object. Neither the spatial nor temporal resolution is preselected. The STBB formalism, which is applicable to any dynamic contrast-enhanced imaging condition, is demonstrated using <italic>a priori</italic> modeling of breast tumors. The k-space traversal scheme is obtained by maximizing energy coverage of the Fourier space that encompasses the instantaneous spectral energy of the temporally enhancing object during the DCE-MRI experiment. A method to use the energy maximization concept in designing an acquisition scheme that is adequate for a class of <italic>space-time</italic> objects is also demonstrated in this work. This concept was tested in computer simulations under a range of object spatial features and enhancement conditions. In addition, two issues that closely impact the accuracy in quantification of pharmacokinetic parameters measured using DCE-MRI are addressed: motion artifact and B1-field inhomogeneity. A linear three-dimensional motion-correction algorithm to compensate for patient motion over the course of the dynamic acquisition is developed. The errors in parameter estimation due to B1-field inhomogeneity are investigated and a correction method is proposed.