Quantification of bone mineral, collagen, and water using a robust DECT-based algorithm: addressing attenuation similarity and CT imaging noise

Bone consists of hydroxyapatite, collagen, and water, each essential to mechanical integrity. Dual-energy computed tomography (DECT) offers a non-invasive way to quantify these components, but attenuation similarity between collagen and water and CT noise undermine stable decomposition. This study develops and evaluates a robust, constraint-based DECT algorithm to improve stability and accuracy under these conditions. The method was first verified using digital CT phantoms with prescribed compositions and then validated using 28 cylindrical bovine specimens scanned at 45/90 keV. Stepwise drying (110 °C, 6h) and ashing (600 °C, 9h) provided reference fractions of hydroxyapatite, collagen, and water. Simulations demonstrated high voxel-wise accuracy with negligible deviation from reference values. Experimentally, DECT-derived and ashing-measured fractions correlated moderately for hydroxyapatite (r = 0.61, p < 0.001) and collagen (r = 0.46, p = 0.02), but poorly for water (r = 0.02, p = 0.91), reflecting attenuation similarity and dehydration-related bonded-water loss. After excluding seven specimens with severe beam-hardening and streak artefacts, hydroxyapatite accuracy improved markedly (r = 0.89, p < 0.001). The algorithm enhances the reliability of DECT-based bone-composition assessment under realistic noise, providing robust hydroxyapatite quantification. Collagen–water separation remains limited, and future work will integrate advanced denoising and multi-energy CT.