MRI measurement of bone marrow cellularity for radiation dosimetry.

The current gold standard for measuring marrow cellularity is the bone marrow (BM) biopsy of the iliac crest. This measure is not predictive of total marrow cellularity, because the biopsy volume is typically small and fat fraction varies across the skeleton. MRI and localized MR spectroscopy have been demonstrated as noninvasive means for measuring BM cellularity in patients. The accuracy of these methods has been well established in phantom studies and in the determination of in vivo hepatic fat fractions but not for in vivo measurement of BM cellularity. Spoiled gradient-echo in vivo images of the femur, humerus, upper spine, and lower spine were acquired for 2 dogs using a clinical 3-T MRI scanner. Single-peak iterative decomposition of water and fat with echo asymmetry and least squares (SP-IDEAL) was used to derive BM fat fractions. Stimulated-echo acquisition mode spectra were acquired in order to perform multipeak IDEAL with precalibration (MP-IDEAL). In vivo accuracy was validated by comparison with histology measurements. Histologic fat fractions were derived from adipocyte segmentation. Bland-Altman plots demonstrated excellent agreement between SP-IDEAL and histology, with a mean difference of -0.52% cellularity and most differences within ±2% cellularity, but agreement between MP-IDEAL and histology was not as good (mean difference, -7% cellularity, and differences between 5% and -20%). Adipocyte segmentation of histology slides provides a measure of volumetric fat fraction (i.e., adipocyte volume fraction [AVF]) and not chemical fat fraction, because fat fraction measured from histology is invariant to the relative abundances of lipid chemical species. In contrast, MP-IDEAL provides a measure of chemical fat fraction, thus explaining the poor agreement of this method with histology. SP-IDEAL measures the relative abundance of methylene lipids, and this measure is shown to be equivalent to AVF. AVF provides the appropriate parameter to account for patient-specific cellularity in BM mass predictive equations and is consistent with current micro-CT-based models of skeletal dosimetry.