Highly accelerated magnetic resonance acoustic radiation force imaging for in vivo transcranial ultrasound focus localization: A comparison of three reconstruction methods.

Magnetic resonance acoustic radiation force imaging (MR-ARFI) is a promising tool for transcranial neurosurgery planning and monitoring. However, the ultrasound dose during ARFI is quite high due to the high intensity required and the repetitive ultrasound sonication. To reduce the ultrasound deposition and prevent unwanted neurological effects, undersampling in k-space data acquisition is adopted in the current study. Three reconstruction methods, keyhole, k-space hybrid and temporal differences (TED) compressed sensing, the latter two of which were initially proposed for MR thermometry, were applied to the in vivo transcranial focus localization based on MR-ARFI data in a retrospective way. The accuracies of the three methods were compared with the results from the fully sampled data as reference. The results showed that the keyhole method tended to smooth the displacement map and underestimate the peak displacement. The K-space hybrid method was better at recovering the displacement map and was robust to the undersampling pattern, while the TED method was more time efficient under a higher image resolution. For an image of a lower resolution, the K-space hybrid and TED methods were comparable in terms of accuracy when a high undersampling rate was applied. The results reported here facilitate the choice of appropriate undersampled reconstruction methods in transcranial focal localization based on MR-ARFI.