Purpose
To estimate the susceptibility and the geometry of metallic implants from Multi-Spectral Imaging (MSI) information, to separate the metal implant region from the surrounding signal loss region.
Methods
The susceptibility map of signal-void regions is estimated from MSI B0 field maps using total variation (TV) regularized inversion. Voxels with susceptibility estimates above a predetermined threshold are identified as metal. The accuracy of the estimated susceptibility and implant geometry was evaluated in simulations, and phantom and in vivo experiments.
Results
The proposed method provided more accurate susceptibility estimation compared with a previous method without TV regularization, in both simulations and phantom experiments. In the phantom experiment where the actual implant was 40% of the signal-void region, the mean estimated susceptibility was close to the susceptibility in literature, and the precision and recall of the estimated geometry was 85% and 93%. In vivo studies in subjects with hip implants also demonstrated that the proposed method can distinguish implants from surrounding low-signal tissues, such as cortical bone.
Conclusion
The proposed method can improve the delineation of metallic implant geometry by distinguishing metal voxels from artificial signal voids and low-signal tissues by estimating the susceptibility maps.