In Vivo Bone Mineral Density Assessment With Spectral Localizer Radiographs From Photon-Counting Detector CT: Prospective Comparison With DXA

Abstract

Purpose: The aim of this study was to determine in a prospective patient study the accuracy of areal bone mineral density (aBMD) measurements with spectral localizer radiographs obtained with a clinical photon-counting detector computed tomography (PCD-CT) scanner in comparison with dual-energy x-ray absorptiometry (DXA).

Methods: In this institutional review board-approved, prospective study, 41 patients (15 females, 26 males; mean age 61.3 years, age range 35-78 years) underwent PCD-CT of the abdomen with a spectral localizer radiograph (tube voltage 140 kVp, tube current 30 mA) and DXA within a median of 45 days. aBMD values were derived for lumbar vertebrae L1-L4 from both methods and were compared with linear regression, Pearson correlation, intraclass correlation coefficients (ICCs), and Bland-Altman plots. T-scores were calculated on a patient level and were compared between methods.

Results: DXA and spectral localizer radiographs showed strong correlation in aBMD measurements ( R = 0.97, P < 0.001) and patient level T-scores ( R = 0.99, P < 0.001). There was a strong agreement between aBMD from both methods (ICC, 0.96; 95% CI, 0.94-0.97). Bland-Altman analysis revealed a very small mean difference in aBMD between methods (mean absolute error 0.019 g/cm 2 ) with narrow limits of agreement (-0.083 g/cm 2 to 0.121 g/cm 2 ). Similarly, there were small differences in regard to the T-score (mean absolute error 0.156) with narrow limits of agreement (-0.422 to 0.734) between methods. ICCs indicated an excellent agreement between T-scores from DXA and spectral localizer radiographs (ICC, 0.98; 95% confidence interval, 0.95-0.99).

Conclusions: Our prospective patient study indicates that spectral localizer radiographs obtained with a clinical PCD-CT system enable accurate quantification of the lumbar bone areal mineral density. This opens up the opportunity for opportunistic screening of osteoporosis in patients who undergo CT for other indications.

Keywords: bone mineral density; computed tomography; localizer radiographs; photon-counting detector; spectral imaging.

FIGURE 1

Illustration of the segmentation process of aBMD measurements on spectral localizer radiographs from PCD-CT. Lumbar DXA scan (A) of a 59-year-old male patient (BMI 26.8 kg/m²) with corresponding spectral localizer radiography imaged in conventional mode (B), as hydroxyapatite material decomposition map (C), and as water material decomposition map (D). Segmentations in images C and D correspond to vertebral body L2 (cyan) and adjacent soft tissue (yellow).

FIGURE 2

Study flowchart.

FIGURE 3

Bland-Altman plot for aBMD measurements between spectral localizer radiographs and DXA for each vertebra. Note the small mean error between modalities and the trend toward higher aBMD values at higher bone mineral densities for spectral localizer radiographs as compared with DXA. LoA, limits of agreement representing 95% confidence interval.

FIGURE 4

Scatterplot for DXA and spectral localizer radiography-derived aBMD values for each vertebra. Blue line: linear regression visualizing linear relationship between DXA and localizer derived aBMD (g/cm²). Formula shows function for linear regression. R value shows Pearson R.

FIGURE 5

Bland-Altman plot for T-scores from spectral localizer radiographs and DXA. LoA, limits of agreement representing 95% confidence interval.

FIGURE 6

Scatterplot for DXA and spectral localizer radiography-derived T-scores. Blue line: linear regression visualizing linear relationship between DXA and localizer derived T-scores. Formula shows function for linear regression. R value shows Pearson R.

FIGURE 7

Categorization of T-scores with DXA and with spectral localizer radiographs. Note that in 2 patients the categorization changed with spectral localizer radiographs as compared with DXA. Alluvial plot. Shaded areas between bars correspond to the change in category from DXA to spectral localizer.