Accuracies of four simulation approaches in reproducing motion artifacts and morphometric parameter biases.

Objective: Despite widespread uses in MRI research, the relative accuracies of different motion artifact simulation approaches in reproducing artifacts and artifact-induced changes (AIC) of morphometric parameters in structural MRI remain largely unknown. We aim to evaluate the performances of four simulation approaches in reproducing artifacts and AIC of brain morphometric parameters.

Methods: Within-session repeated T₁-weighted scans were acquired on ten volunteers with their heads remaining still or undergoing intentional motion monitored by fat navigators. Four simulation approaches were adopted, which differed in terms of whether channel-combined magnitude image or complex multi-channel k-space data were utilized, and whether motion effects were introduced by modifying k-space data value (MDV) or modifying k-space coordinates and data phase (MCP). By means of simulation, the dependence of morphometric parameter changes on motion pattern and severity was studied.

Results: Multi-channel k-space database simulation achieved higher artifact similarity and AIC consistency with measured motion scan images than magnitude image-based simulation. MDV- and MCP-based simulations achieved comparable results. From k-space database simulation employing MDV, the motion-induced biases in morphometric parameters were found to vary linearly with motion severity with motion pattern-dependent slopes.

Conclusions: Simulations based on multi-channel complex k-space data outperformed those based on channel-combined magnitude images in reproducing artifacts and AICs. Head motion caused imaging artifacts and systematic biases in morphometric parameters which can be equally reproduced by simulations using two different motion effect introduction strategies.