Motion‐compensated image reconstruction for improved kidney function assessment using dynamic contrast‐enhanced MRI
Cemre Ariyurek,Aziz Koçanaoğulları,Onur Afacan,Sila Kurugol
DOI: https://doi.org/10.1002/nbm.5116
2024-02-17
NMR in Biomedicine
Abstract:The proposed motion‐correction (MoCo) reconstruction technique models the motion navigator signal by a Gaussian process, and solves for motion correction images and motion fields iteratively, employing a computationally efficient nonrigid registration. The proposed technique improves depiction of the kidneys and the temporal signal‐to‐noise ratio of DCE‐MR images and enables more reliable renal function estimation in the presence of motion. Accurately measuring renal function is crucial for pediatric patients with kidney conditions. Traditional methods have limitations, but dynamic contrast‐enhanced magnetic resonance imaging (DCE‐MRI) provides a safe and efficient approach for detailed anatomical evaluation and renal function assessment. However, motion artifacts during DCE‐MRI can degrade image quality and introduce misalignments, leading to unreliable results. This study introduces a motion‐compensated reconstruction technique for DCE‐MRI data acquired using golden‐angle radial sampling. Our proposed method achieves three key objectives: (1) identifying and removing corrupted data (outliers) using a Gaussian process model fitting with a k ‐space center navigator, (2) efficiently clustering the data into motion phases and performing interphase registration, and (3) utilizing a novel formulation of motion‐compensated radial reconstruction. We applied the proposed motion correction (MoCo) method to DCE‐MRI data affected by varying degrees of motion, including both respiratory and bulk motion. We compared the outcomes with those obtained from the conventional radial reconstruction. Our evaluation encompassed assessing the quality of images, concentration curves, and tracer kinetic model fitting, and estimating renal function. The proposed MoCo reconstruction improved the temporal signal‐to‐noise ratio for all subjects, with a 21.8% increase on average, while total variation values of the aorta, right, and left kidney concentration were improved for each subject, with 32.5%, 41.3%, and 42.9% increases on average, respectively. Furthermore, evaluation of tracer kinetic model fitting indicated that the median standard deviation of the estimated filtration rate (σFT ), mean normalized root‐mean‐squared error (nRMSE), and chi‐square goodness‐of‐fit of tracer kinetic model fit were decreased from 0.10 to 0.04, 0.27 to 0.24, and, 0.43 to 0.27, respectively. The proposed MoCo technique enabled more reliable renal function assessment and improved image quality for detailed anatomical evaluation in the case of bulk and respiratory motion during the acquisition of DCE‐MRI.
radiology, nuclear medicine & medical imaging,biophysics,spectroscopy