Sivan Unit (SU) for standardization of voxel intensity in CBCT: A multi-vendor, multi-centric validation study.
Authors
Affiliations (1)
Affiliations (1)
- Professor and HOD, Department of Oral Medicine and Radiology, Teerthanker Mahaveer Dental College and Research Centre, Teerthanker Mahaveer University, Moradabad Uttar Pradesh, 244001, India. Electronic address: [email protected].
Abstract
To develop and validate the Sivan Unit (SU), a biologically referenced voxel intensity harmonization framework for cone-beam computed tomography (CBCT) designed to improve cross-vendor grayscale consistency in quantitative maxillofacial imaging. This retrospective multi-vendor, multi-centric study analyzed 160 CBCT datasets from ten commercial scanner systems. Raw voxel intensities (VI<sub>raw</sub>) from seven maxillofacial tissues were first processed using global Z-score transformation to reduce gross scaling differences. These were subsequently mapped using two internal biological anchors: dental pulp (0 SU) and cortical bone (1000 SU). Inter-scanner consistency was evaluated using one-way ANOVA, coefficient of variation (CV), intraclass correlation coefficients (ICC), and Bland-Altman analysis. Pre-harmonization intensities demonstrated significant inter-scanner variability across all tissues (p < 0.001), with a mean CV of 14.7%. Following SU transformation, voxel intensity distributions showed no significant inter-vendor differences (all p > 0.05), with the mean CV reduced to 2.1%. Reliability was excellent across all categories (ICC: 0.89-0.97). Bland-Altman analysis confirmed a marked reduction in inter-scanner bias and a narrowing of 95% limits of agreement. The intrinsic biological hierarchy (enamel > cortical bone > dentin > trabecular bone > pulp > soft tissue > air) was preserved post harmonization (r = 0.992, p < 0.001). The Sivan Unit (SU) provides a biologically anchored harmonization framework that effectively eliminates vendor-dependent variability without external phantoms or absolute physical calibration. By ensuring cross-platform reproducibility, the SU framework provides a standardized quantitative foundation that may support future radiomics and AI-based applications in maxillofacial radiology.