Operator Performance of Voxel Placement for Single-Voxel MRS of Breast Lesions

Introduction: Single-voxel H MR spectroscopy of human breast cancer has become increasingly common, with clinical applications in diagnosis and treatment monitoring, and multi-site clinical trials currently in progress. In order for these measurements to be useful in characterizing lesions, the MRS voxel placement must be accurate and consistent. Poor voxel placement creates partial volume errors and can introduce artifacts, reduce signal-to-noise, and lower measurement reproducibility. To date, the potential problem of voxel placement inconsistency has not been directly addressed. This is difficult to test in the true setting as voxel placement is performed mid-study with the subject in the scanner. For this work, we performed a retrospective observer-performance study of voxel placement to measure the consistency of voxel placement by expert operators. Note that this study design only tests the agreement of voxel placement and does not determine which placements are more correct or lead to betterquality spectra. We also compared performance between different lesion types, with the expectation that the concordance between operators would be higher for masses than non-mass-like enhancement