Author: Li Tong, Chuyan Wang, Zhengkui Wang, Yingli Yang, Jie Zhang 👨🔬
Affiliation: Shanghai United imaging Healthcare Advanced Technology Research Institute, Shanghai United Imaging Healthcare Co., LTD, Department of Radiology, Ruijin Hospital, Institute for Medical Imaging Technology, Ruijin Hospital 🌍
Purpose:
Pelvic radiotherapy (RT)-induced bone marrow (BM) damage affects patient prognosis by causing hematologic toxicity. However, consensus on BM-sparing (BMS) RT is still lacking, owing to the absence of standardized BM quantification methods and the inherent limitations of the imaging gold standard, magnetic resonance spectroscopy (MRS). Recently, MR proton density fat fraction (PDFF) imaging is becoming the mainstream for BMS RT given its sensitivity to RT-induced BM changes. This study aims to evaluate different PDFF techniques for spatialized BM quantification and potential applications in RT.
Methods:
Three PDFF sequences (mDIXON Quant, FACT, and WFI) from two vendors were optimized and compared to establish our clinical PDFF protocol. Ten fat phantoms, with fat fractions ranging from 0% to 100%, were constructed to benchmark MRS quantification accuracy against thermogravimetric analysis (TGA), which was then used as the in-vivo reference for PDFF sequences. Data from 30 volunteers were collected to determine the optimal parameter settings for each PDFF sequence. Additionally, reproducibility was studied with intra-class correlation coefficients (ICCs), and image quality was evaluated with signal-to-noise ratio, high-contrast spatial resolution, and low-contrast object detectability using the ACR phantom.
Results:
For quantification accuracy with phantom measurements, comparing to TGA, the mean absolute error (MAE) for MRS is 3.48%, while the highest MAE across three PDFF sequences with optimized parameters is 5.88%. For in-vivo measurements, linear regression between MRS and each optimized PDFF sequences resulted in a minimum R² of 0.93 and maximum MAE of 7.57%. For reproducibility analysis on the three optimized PDFF sequences, a minimum ICC of 0.97 was achieved for every sequence, both in-vivo and in-vitro. All three sequences meet clinical standard by image quality with no specific technique standing out.
Conclusion:
With parameter optimization, PDFF techniques demonstrate great accuracy, reproducibility, and image quality for BM imaging, showing considerable potential for BMS RT.