Assessment of bone segmentation quality of cone-beam CT versus multislice spiral CT: a pilot study

Abstract

The objective of this study was to quantitatively assess the quality of jawbone models generated from cone beam computed tomography (CBCT) by comparison with similar models obtained from multislice spiral computed tomography (MSCT). Three case studies were performed involving images of anthropomorphic head phantoms and real patients acquired with 3 CBCT (NewTom 9000 DVT, Accuitomo 3D, and i-CAT) and 2 MSCT scanners (Somatom VolumeZoom and Lightspeed). Bone was segmented from the CBCT and MSCT images using global thresholding. CBCT vs MSCT segmentation differences were assessed by comparing bone thickness measurements at anatomically corresponding sites, identified automatically by CBCT to MSCT image registration. There was a statistically significant difference between the MSCT and CBCT segmented bone thicknesses, varying from 0.05 +/- 0.47 mm (i-CAT) up to 1.2 +/- 1.00 mm (3D Accuitomo, posterior maxilla). An automated, reproducible, and observer-independent method has been developed to assess the quality of CBCT bone models using MSCT as a clinically established method of reference. Our validation method is generally applicable in cases where no geometric ground-truth is available.