MO-FG-CAMPUS-IeP3-02: Gradient Nonlinearity Calibration and Correction for Full-Volume Imaging of a Compact Asymmetric MRI Gradient System




Due to engineering limitations, the spatial encoding gradient fields generated by an MRI scanner are inherently nonlinear and images must undergo geometric correction prior to display. The gradient nonlinearity (GNL) of a conventional symmetric whole-body gradient is typically characterized using a spherical harmonic polynomial (SHP) expansion of magnetic gradient fields with up to fifth-order terms. Only odd-order terms are required due to design symmetry (i.e., 3rd/5th-order). Recently, a high-performance asymmetric gradient system was developed which exhibits more complex GNL. Here, the GNL across the 26cm diameter-spherical-volume (DSV) of this system is characterized using a phantom-based iterative calibration procedure.


The Alzheimer's Disease Neuroimaging Initiative (ADNI) phantom was scanned with a 3D IR-FSPGR sequence (sagittal acquisition, Nx=Ny=256, Nz=196, Δx=Δy=1.05mm, Δz=1.3mm, BW=±125kHz) on the asymmetric gradient system. The phantom (20cm diameter with 160 spherical fiducials of 1.0/1.5cm diameter) was imaged at isocenter and at spatial offsets of ±3cm along the superior/inferior, right/left, anterior/posterior axes to cover the 26cm DSV. The spatial positions of fiducials were measured from MR images and compared with the design positions to extract distortion field, which is then used to fit the SHP model using an iterative calibration procedure. The effect of higher-order terms was tested by incrementally increasing the model order from 5 to 10. The calibrated coefficients were then used to geometrically correct the images, and for each case the residual root-mean-square-error (RMSE) was determined. For comparison, RMSE was also determined for images corrected using coefficients obtained from electromagnetic simulation.


The calibrated coefficients (10th-order) reduce RMSE from 0.96mm (using a 10th-order correction from electromagnetic simulation) to 0.36mm.


The model parameters for GNL correction over the 26cm DSV of an asymmetric gradient system were successfully determined. The use of calibrated coefficients up to 10thorder coefficients including even-order terms yielded improved geometric accuracy (RMSE) comparable to conventional symmetric gradients.

Funding support: NIH R01EB010065