Reproducibility of functional MR imaging results using two different MR systems

Abstract

Background and purpose: In the application of functional MR imaging for presurgical planning, high reproducibility is required. We investigated whether the reproducibility of functional MR imaging results in healthy volunteers depended on the MR system used.

Methods: Visual functional MR imaging reproducibility experiments were performed with 12 subjects, by using two comparable 1.5-T MR systems from different manufacturers. Each session consisted of two runs, and each subject underwent three sessions, two on one system and one on the other. Reproducibility measures D (distance in millimeters) and R(size) and R(overlap) (ratios) were calculated under three conditions: same session, which compared runs from one session; intersession, which compared runs from different sessions but from the same system; and intermachine, which compared runs from the two different systems. The data were averaged per condition and per system, and were compared.

Results: The average same-session values of the reproducibility measures did not differ significantly between the two systems. The average intersession values did not differ significantly as to the volume of activation (R(size)), but did differ significantly as to the location of this volume (D and R(overlap)). The average intermachine reproducibility did not differ significantly from the average intersession reproducibility of the MR system with the worst reproducibility.

Conclusion: The location of activated voxels from visual functional MR imaging experiments varied more between sessions on one MR system than on other MR system. The amount of the activated voxels is independent of the MR system used. We suggest that sites performing functional MR imaging for presurgical planning measure the intersession reproducibility to determine an accurate surgical safety margin.

Fig 1.

Example of functional MR imaging activation, in one subject after visual stimulation (checkerboard, 8 Hz), superimposed on an anatomic image (MP-RAGE). Pixels with a z-score above 4.0 are coded as to the bar on the right. Top row shows runs 1 and 2 from session 1 with system B, middle row shows both runs from session 2 with system A, and bottom row shows the runs from session 3 with system B.

Fig 2.

Averages for the reproducibility measures D, R_size, and R_overlap for same-session and intersession reproducibility for system A and system B, and intermachine reproducibility. D is the distance between the centers of the activated areas, R_size is the ratio of the volumes of activated areas, and R_overlap is the ratio of the common activated areas. R_size and R_overlap must be read from the left y axis, and D must be read from the right y axis where the order is inverted.

Fig 3.

Averages for the reproducibility measure D, separated for the frequency- and phase-encoding directions and the section direction, for same-session and intersession reproducibility for system A and system B, and intermachine reproducibility. D is the distance between the centers of the activated areas.