Errors in quantitative dynamic three‐dimensional keyhole MR imaging of the breast

Abstract

Three-dimensional (3D) keyhole magnetic resonance (MR) imaging has been proposed as a means of providing dynamic monitoring of contrast agent uptake by breast lesions, with complete breast coverage and high spatial and temporal resolution. The 3D keyhole technique dynamically samples the central regions of k-space in both phase-encoding directions and provides high-frequency data from a precontrast acquisition. Errors due to data truncation with two-dimensional and 3D region-of-interest measurements are estimated from a numerical simulation of various implementations of the 3D keyhole technique. Errors were found to increase with increasing temporal resolution and reduced object size. Errors of 75% are possible for objects with a diameter approaching 1 pixel when a 3D keyhole implementation that samples 50% of phase-encoding data in each direction is used. Preliminary clinical images with this approach illustrate artifacts consistent with inadequate k-space sampling.