Non‐Rigid Motion‐Compensated MR Multitasking for Free‐Breathing Low‐Dose Dynamic Contrast‐Enhanced MRI in the Abdomen

ABSTRACT Purpose To improve free‐breathing low‐dose quantitative dynamic contrast‐enhanced (DCE) MRI for imaging pancreatic ductal adenocarcinoma. Methods We introduced non‐rigid (NR) motion compensation (MoCo) into MR Multitasking using a novel “in‐subspace” formulation that directly applies motion fields to spatial feature maps (eigenimages), reducing computational cost relative to warping in the spatiotemporal domain. We evaluated NR‐MoCo DCE Multitasking in n  = 11 healthy volunteers scanned using 20% (0.02 mmol/kg) of the standard Gd dose; 10 of these volunteers underwent a repeat 20%‐dose scan after exiting and re‐entering the scanner. Data were reconstructed without MoCo (No‐MoCo), translation MoCo (Tr‐MoCo), and NR‐MoCo Multitasking. We assessed motion by measuring respiratory motion energy ratio and liver dome position across time within the end‐expiration respiratory bin. We estimated pharmacokinetic parameters from all reconstruction frameworks, evaluating the test–retest repeatability of parameters in the head, body, and tail of the pancreas. Results Non‐rigid MoCo of eigenimages was compatible with MR Multitasking, using > 3000× fewer Fourier transforms and motion deformations. Compared with Tr‐MoCo and No‐MoCo, NR‐MoCo showed the best performance in reducing artifactual intra‐bin motion along DCE dimension, indicated by the smallest liver‐dome position variation ( p  = 0.026). NR‐MoCo lowered the effective tensor rank and improved repeatability of quantitative kinetic vascular properties across healthy volunteers as compared to No‐MoCo. The Bland–Altman analysis indicated that NR‐MoCo narrowed K trans limits of agreement by 2.1× relative to No‐MoCo and Tr‐MoCo. Conclusion Non‐rigid motion compensation reduces respiratory motion and improves K trans repeatability in low‐dose free‐breathing, whole‐abdomen quantitative dynamic contrast enhanced (DCE) MR Multitasking.