An analytical phantom-based scheme for assessing the accuracy of high-resolution 3D MRI motion-artifact simulations [1]
Simulated motion-corrupted MRI is widely used to develop motion detection and correction methods. Yet different simulation algorithms produce different, visually plausible artifacts from the same inputs. How can we tell which is more accurate?
APHABAMAS uses an analytical phantom [2] to compute a ground-truth simulation free from sampling-induced error. The algorithm under assessment receives the phantom's motion-free image; both paths use identical motion and acquisition parameters. Comparing their outputs measures simulation accuracy.
The experiments follow two phases:
- Demonstrate the need for assessment. Show that algorithms produce different artifacts on brain and phantom images, even when all outputs look plausible.
- Evaluate simulation accuracy. Compare phantom simulations with analytical ground truth to quantify image and signal errors.
All experiment code is in experiments/. Figure generators and saved figures are in manuscript_figures/.
Across the evaluated motion conditions, Type-2 NUFFT gives the most consistent agreement with ground truth. Image-based simulations generally show more blurring, while Type-1 NUFFT accuracy deteriorates with increasing motion severity.
If you use APHABAMAS for your work, please cite [1] (and potentially also [2]).
[1] Wu T and Zhang H. APHABAMAS: An analytical phantom-based scheme for assessing the accuracy of high-resolution 3D MRI motion-artifact simulations. arXiv:2607.09945.
[2] Koay CG, Sarlls JE, Özarslan E. Three-dimensional analytical magnetic resonance imaging phantom in the Fourier domain. Magnetic Resonance in Medicine. 2007;58(2):430–436. doi:10.1002/mrm.21292.