To test a proof-of-concept that the accuracy and reliability of alveolar bone height measurements from orthodontic grade (large field-of-view [FOV], large voxel-size) cone-beam computed tomography (CBCT) images may be improved by using pixel gray values.

Twenty fresh cadaver pig heads underwent CBCT scans (17 × 23 cm FOV, 0.4-mm voxel size). Buccal alveolar bone heights of maxillary first molars were measured using the conventional vision-based (VB) and the proposed gray value–assisted (GVA) methods. The GVA methods entailed localization of landmarks through observation of gray value pattern changes across tissue boundaries followed by mathematical calculation of distances between landmark pixels. Interrater reliability and accuracy of CBCT measurements made by all methods were statistically analyzed by comparing with physical measurements (gold standards).

The interrater reliability of CBCT measurements made by GVA methods was comparable to physical measurements but higher than those made by the VB method. The GVA (bend-down pattern) method yielded average measurements similar to physical measurements, while those obtained by the VB and the GVA (straight pattern) methods were significantly larger (repeated measures analysis of variance, P < .001). The GVA (bend-down pattern) method also produced significantly more measurements within one voxel size of physical measurements than did the VB and GVA (straight pattern) methods (Chi-square tests, P < .017).

These data confirm a concept that local gray value change patterns may be used to improve the accuracy and reliability of alveolar bone height measurement from large FOV and large voxel-size CBCT images.