A geometric anthropomorphic phantom for quantitative susceptibility mapping: accuracy and repeatability

The research details the creation and testing of an MRI-compatible anthropomorphic phantom designed to mimic the electromagnetic properties and morphology of human tissue. This phantom, composed of cylindrical and spheroid compartments with realistic susceptibilities, was used to study Quantitative Susceptibility Mapping (QSM) error in vivo.

The study focused on assessing the accuracy and repeatability of MEDI-susceptibility measurements within regions of interest (ROIs). It was found that ellipsoids performed better than straws in achieving reliable susceptibility measurements, with error values ranging from 0.007 to 0.083 parts per million at 3T and 0.050 to 0.118 ppm at 7T.

Repeatability, a measure of consistency in measurements, was also a key focus. The study found that the repeatability coefficient was significantly higher for ellipsoid ROIs (0.168 ppm at 3T and 0.141 ppm at 7T) compared to straw ROIs (0.652 ppm at 3T and 0.459 ppm at 7T).

The study further compared the accuracy of QSM reconstructions at two different magnetic field strengths, 3T and 7T. It was observed that the accuracy was better at 3T (bias = -0.002 ppm, RMSE = 0.082 ppm) than at 7T (bias = -0.056 ppm, RMSE = 0.092 ppm). At 3T, there was excellent agreement between susceptibility maps generated using COSMOS and MEDI algorithms, with a linear regression of y=1.00x-0.01 and an r-value of 0.99.

However, at 7T, there was some underestimation of MEDI susceptibility maps relative to COSMOS, with a linear regression of y=0.93x-0.04 and an r-value of 0.99. The results suggest that while QSM reconstructions are reliable at 3T scanners, they can be more challenging at 7T scanners when dealing with high magnetic susceptibilities.