A three-material model for dual-layer detector spectral computed tomography measurements of marrow adipose tissue and bone mineral density
Single-energy QCT (SEQCT) scans to measure volumetric BMD (vBMD) are susceptible to errors caused by variability in the amount of marrow adipose tissue (MAT). We developed a three-material model that uses dual-layer spectral CT (DLCT) technology to measure bone matrix (BM), yellow marrow (YM), and red marrow (RM) and compared the results with measurements of proton density fat fraction (PDFF) by MRI and vBMD by SEQCT. Hounsfield units (HU) were measured in the L1-3 vertebrae on 50 and 150 keV mono-energy images in a training set of 100 Chinese adults. The densities of YM and RM in the three-material model were adjusted so that the mean and SD of the YM volume as a fraction of total marrow volume matched historical bone histology data. A validation set of 125 adults was scanned, and the findings were compared with measurements of L1-3 MRI PDFF and SEQCT vBMD. We evaluated the sensitivity, specificity, and area under the ROC curve (AUROC) for DLCT vBMD measurements to predict osteoporosis and investigated the relationship between SEQCT vBMD, DLCT vBMD, and YM volume fraction. The mean (range) of the YM volume as a fraction of total marrow volume averaged 0.471 (0.190-0.674) and 0.480 (0.258-0.760) in men and women. The corresponding results for MRI PDFF were 0.487 (0.224-0.675) and 0.477 (0.238-0.745). The coefficient of determination was <i>r</i> <sup>2</sup> = 0.696 (<i>p</i> < .0001; SEE = 0.059). A L1-3 DLCT vBMD of 100 mg/cm<sup>3</sup> gave a sensitivity of 100.0% and a specificity of 94.3% for predicting osteoporosis (AUROC = 0.986). A multiple linear regression model to predict L1-3 SEQCT vBMD from DLCT vBMD and the YM fraction gave a coefficient of determination of <i>r</i> <sup>2</sup> = 0.989 (<i>p</i> < .0001; SEE = 5.2 mg/cm<sup>3</sup>). In conclusion, we developed a three-material model for analyzing DLCT scans that correlates with MRI measurements of MAT PDFF and offers a potentially improved method of using CT to measure vBMD.
Authors
- Jian Geng (ORCID: https://orcid.org/0000-0002-3248-1094)
- Kangkang Ma
- Glen M. Blake (ORCID: https://orcid.org/0000-0002-1530-5267)
- Fengyun Zhou (ORCID: https://orcid.org/0000-0001-9234-1635)
- Wenshuang Zhang (ORCID: https://orcid.org/0000-0002-9291-2538)
- Ling Wang (ORCID: https://orcid.org/0000-0003-3220-6065)
- Qingyu Zhang (ORCID: https://orcid.org/0000-0001-9739-3899)
- Yi Yuan (ORCID: https://orcid.org/0009-0007-7472-0376)
- Zitong Cheng (ORCID: https://orcid.org/0009-0003-1509-5858)
- Bo Hu
- Yandong Liu (ORCID: https://orcid.org/0000-0002-4752-9484)
- Zhe Guo (ORCID: https://orcid.org/0000-0002-5776-4699)
- Dong Yan (ORCID: https://orcid.org/0000-0003-4517-3530)
- Xiaoguang Cheng
Institutions
- St Thomas' Hospital (GB)
- King's College London (GB)
- Peking University (CN)
- Beijing Jishuitan Hospital (CN)
- Beijing Institute of Neurosurgery (CN)
Publication Details
- Journal
- JBMR Plus
- Published
- 2025-04-16
- DOI
- https://doi.org/10.1093/jbmrpl/ziaf066
- Citations
- 1
- Primary Topic
- Infrared Thermography in Medicine
- Type
- article
- Field-Weighted Citation Impact
- 1.36