Abstract
PurposeDual-energy X-ray absorptiometry (DXA) is the clinical gold standard for areal bone mineral density (aBMD) assessment for the detection of osteoporosis. The advancement of photon-counting CT (PCCT) provides an opportunity to use spectral localizer radiographs (LRs) for opportunistic aBMD assessment. This study explored that feasibility.ApproachThe study used a combination of virtual imaging trials (VITs) and physical phantom experiments. For the VIT experiment, a 50th percentile BMI patient model was conditioned to represent three bone health statuses (normal, osteopenic, and osteoporotic) and was imaged with PCCT LR at six different tube currents (30, 60, 80, 100, 120, and 140 mAs). The simulation was repeated for two additional body sizes at BMI of 20 and 31 kg/m2. All datasets were evaluated for aBMD quantification derived from PCCT LRs. To confirm accuracy, a lumbar spine phantom was scanned on a clinical PCCT system and on a DXA device, and aBMD was measured for the L1 to L4 vertebrae.ResultsThe VIT study demonstrated stable aBMD accuracy across varying tube current settings, with fluctuations in error limited to ∼0.046 g/cm2 across varying dose levels, within the range of reported lumbar spine DXA variability. Average mean absolute error (MAE) across all dose levels and vertebrae was 0.031 g/cm2±0.012, 0.047 g/cm2±0.016, and 0.069 g/cm2±0.020 for normal, osteopenic, and osteoporotic bone, corresponding to percent errors of 2.77%±0.40%, 4.85%±0.31%, and 8.36%±0.57%. Patient size influenced aBMD quantification accuracy, with mean MAE across bone-density conditions increasing from 0.053 g/cm2 at BMI 20 to 0.113 g/cm2 at BMI 31 corresponding to mean percent errors of ∼5.75% and 11.83%. The physical phantom experiments showed close aBMD measurements between the PCCT LRs and DXA scans, with percent differences across vertebrae ranging from ±4.95%.ConclusionsPCCT LR-based aBMD estimation demonstrated quantitative agreement with DXA, while VIT results showed that reliability depended on bone density, patient size, and acquisition settings. Misclassification risk was driven primarily by patient size, with lower bone density further reducing accuracy. These findings highlight the role of VITs in identifying performance limiting conditions for PCCT LR-derived aBMD assessment.