Normalization of Trabecular Bone Attenuation across Kvp Values: A Physics-Based Correction for Opportunistic Screening of Osteoporosis
Abstract
Purpose
To develop and validate a physics‑informed model that converts trabecular bone attenuation (HU) measured at non‑standard tube potentials into 120‑kVp–equivalent CT numbers.
Methods
X‑ray spectra at 100, 120, and 140 kVp were simulated using the TASMICS model with 3.6 mm Al + 0.1 mm Cu inherent filtration and 0.5 mm Cu added filtration (120‑kVp mean energy: 71 keV). Effective linear attenuation coefficients for trabecular bone (solid bone and marrow components) were computed by weighting NIST‑tabulated mass attenuation coefficients by these spectra. Model validation used an age‑thresholded cohort (>60 years) of non‑contrast abdominal CT scans (N=16,747) acquired across multiple scanner vendors, multiple generations of hardware, and encompassing all abdominal imaging protocols. Trabecular attenuation at L1 was extracted using an automated deep‑learning pipeline. Model performance was evaluated by comparing 100‑kVp and 140‑kVp trabecular attenuation distributions against 120‑kVp references using Welch’s t‑tests (α=0.05).
Results
The correction model successfully shifted non‑standard tube‑potential distributions toward the 120‑kVp baseline (120‑kVp mean: 140.9 HU). For the 140‑kVp cohort, the mean increased from 131.3 to 138.3 HU, with pre‑correction statistical differences (p<0.001) eliminated after correction (p=0.058). For the 100‑kVp cohort, the mean decreased from 148.1 to 137.3 HU, and significant pre‑correction differences (p=0.032) were similarly eliminated post‑correction (p=0.270).
Conclusion
This physics‑informed correction model effectively normalizes trabecular bone attenuation acquired at higher or lower tube potentials to 120‑kVp–equivalent values. The approach supports standardized bone density monitoring and enables robust opportunistic screening for bone disease using CT studies acquired under non‑standard imaging protocols.