Naomi R. Mraz1, Michael T. Nelson1, Jessica Kuehn-Hajder1, Greg John Metzger2, Patrick John Bolan2
1Radiology, University of Minnesota, Minneapolis, MN, United States; 2Radiology/CMRR, University of Minnesota, Minneapolis, MN, United States
In order for single-voxel MRS measurements to be a useful clinical tool for characterizing breast lesions, voxel placement must be accurate and consistent. This project tested the consistency of voxel placement between four different operators using a retrospective study to simulate voxel placement on a MR scanner. The relative geometric intersection was used as a measure of concordance between readers. Overall concordance was lower than expected (ranging 0%-82.5%, mean 31.2%). Concordance was higher for masses than for non-mass-like lesions. These results suggest that further training, guidance, or software support should be used to improve voxel placement consistency.