Quantitative assessment of vertebral bone marrow adipose tissue using photon-counting CT compared with MRI

To evaluate the feasibility and accuracy of photon-counting CT (PCCT)-derived parameters for quantifying vertebral bone marrow adipose tissue (BMAT), using MRI proton density fat fraction (PDFF) as the reference standard.

Adult participants with low back pain were prospectively recruited to undergo both PCCT and MRI. PCCT-derived parameters, including virtual non-calcium (VNCa) CT values, VNCa CT fat fraction (CTFF), and standardized 70 keV CT values, were obtained. MRI PDFF maps were generated using a 6-echo q-Dixon sequence. Correlation and agreement between PCCT-derived parameters and MRI PDFF were assessed at both vertebral and subject levels.

A total of 73 participants were included, and 339 vertebral bodies were analyzed. VNCa CTFF showed strong positive correlation with MRI PDFF at both vertebral (r = 0.863) and subject (r = 0.895) levels. In contrast, VNCa CT values (r = -0.858 to -0.888) and standardized 70 keV CT values (r = -0.791 to -0.833) showed significant negative correlations with MRI PDFF. Bland-Altman plots showed small mean differences but relatively wide limits of agreement between calibrated VNCa CTFF and MRI PDFF.

PCCT-derived VNCa CTFF showed strong correlation with MRI PDFF, and linear calibration reduced the mean difference between the two measurements, although the limits of agreement remained relatively wide. Standardized 70 keV CT values were also correlated with MRI PDFF, but should be interpreted cautiously as attenuation-based surrogates influenced by both trabecular mineral content and marrow fat.