Abstract
BACKGROUND AND PURPOSE: Pelvic radiotherapy-induced bone marrow (BM) damage adversely affects patient prognosis. Progress in BM-sparing radiotherapy is limited by the lack of standardized BM quantification and the inherent constraints of magnetic resonance spectroscopy (MRS), the current gold standard for BM magnetic resonance imaging (MRI). Proton density fat fraction (PDFF), derived from DIXON-based MRI, has emerged as an imaging biomarker for detecting BM changes. This study evaluated the potential of DIXON-based MRI in pelvic BM for radiotherapy. MATERIALS AND METHODS: Three existing DIXON-based techniques were optimized and compared to establish clinical protocols. In vitro measurements were performed using fat phantoms calibrated against thermogravimetric analysis, while in vivo measurements were based on data from 30 volunteers with MRS serving as the reference standard. Quantitative accuracy was assessed using mean absolute error (MAE), repeatability via intra-class correlation coefficients (ICCs), and image quality using an ACR phantom. RESULTS: Comprehensive evaluation identified optimal parameters for each DIXON-based sequence. For in vitro measurements, the MAE for MRS was 3.5 % and the highest MAE across three optimized DIXON-based sequences was 5.9 %. For in vivo measurements, linear regressions between MRS and each of the optimized DIXON-based sequence resulted in R(2) ≥ 0.93 and MAE ≤ 7.6 %. All three optimized DIXON-based sequences demonstrated high repeatability (ICCs ≥ 0.97) and clearly visualized BM with varying fat fractions, with no consistently outperforming in image quality. CONCLUSION: For BM assessment, this study demonstrated DIXON-based PDFF quantification achieved high accuracy, repeatability, and image quality, supporting its potential for radiotherapy.