To present an image registration method for renal blood oxygen level-dependent (BOLD) measurements that enables semiautomatic assessment of parenchymal and medullary R2* changes under a functional challenge.
In a series of breath-hold acquisitions, three-dimensional data were acquired initially for prospective image registration of subsequent BOLD measurements. An algorithm for kidney alignment for BOLD renal imaging (KALIBRI) was implemented to detect the positions of the left and right kidney so that the kidneys were acquired in the subsequent BOLD measurement at consistent anatomical locations. Residual in-plane distortions were corrected retrospectively so that semiautomatic dynamic R2* measurements of the renal cortex and medulla become feasible. KALIBRI was tested in six healthy volunteers during a series of BOLD experiments, which included a 600- to 1000-mL water challenge.
Prospective image registration and BOLD imaging of each kidney was achieved within a total measurement time of about 17 s, enabling its execution within a single breath-hold. KALIBRI improved the registration by up to 35% as found with mutual information measures. In four volunteers, a medullary R2* decrease of up to 40% was observed after water ingestion.
KALIBRI improves the quality of two-dimensional time-resolved renal BOLD MRI by aligning local renal anatomy, which allows for consistent R2* measurements over many breath-holds.
Definition of cROIl and cROIr. The rectangular regions are manually defined in dynamic coronal and transverse 2D FLASH images to ensure that the kidneys are included in the bounding boxes for all possible breathing positions. The ROIs are then transferred to the prospective RIGR to reduce the calculation time.
Examples of manual segmentation of the kidney for quantification analysis, shown in prospective motion-corrected BOLD images (at first echo time) acquired before water loading and at 12, 28, and 44 min after water loading.
Images of volunteer 3 acquired at different time points during the BOLD experiment. (a) Reference images BOLDref,r/l. (b–d) Baseline images before water challenge (b) and 20 min (c) and 55 min (d) after water loading of both kidneys. For visualization of the registration improvements, moving images of Ibefore,l/r without any registration (first column) and the corresponding prospective aligned BOLDcurr,r/l (third column) and retrospective (fifth column) registered images BOLDnrig,r/l are shown as well as the subtractions to the references (normalized).
R2* weighted images of both kidneys acquired at an echo time of 13.06 ms during baseline (left) and R2* maps of volunteer 6 at (a) baseline and at (b) 18 min, (c) 22 min, (d) 29 min, and (e) 42 min after water loading. For comparison, R2* maps without retrospective registration are shown in the first row (calyx system excluded by thresholding). Individual ROIs were defined in the renal cortex and medullar regions in the baseline data and were then copied to all subsequent time frames. As shown, the positions of the ROIs vary noticeably before registration, whereas they remain in constant anatomical positions after correction (second and third row), so that a regional R2* analysis can be performed.