Rapid determination of vertebral fat fraction over a large range of vertebral bodies

Abstract

Vertebral body fat fraction (FF) has been found to vary between lumbar vertebrae using magnetic resonance spectroscopy (MRS). We aim to more quickly assess a larger number of adjacent vertebrae using a single T2-weighted iterative decomposition of water and fat with echo asymmetry and least-squares estimation (IDEAL) sequence. Five men had dual-energy X-ray absorptometry (DEXA) and 1.5-T MR scans performed. MRS was performed at L3, and a sagittal IDEAL sequence was also performed, resulting in separate fat-only and water-only readings from T10 to S2. For the IDEAL measurements, two independent observers followed a set reading protocol, with five observations each per vertebra. Intra- and interobserver variability were assessed as deviations from the mean within and between observers, respectively. For FF measurements there was limited intra-observer variation, with observers being on average within 3.4% of the pooled mean value. Similarly, there was good interobserver agreement, with an average variation of 2.1%. All men showed a reduction in FF of 1.6-7% between L5 and S1. Otherwise, there was a trend of increasing FF moving inferiorly from T10 to S2. This averaged 2.7% per vertebra (range 1.1-3.8%) and may not have been dependent on MRS-measured FF at the L3 level. There was poor correlation between MRS-measured FF at L2-4 and bone mineral density measured using DEXA (R(2) = 0.06). IDEAL measurements are generally reproducible between observers following a set protocol. There appears to be a gradient in FF moving from T10 to S2, with S1 showing a consistent decrease. This variation may better describe overall marrow function than a single-vertebra reading.