Comparing CBCT With Multi-Slice CT in Diagnosing Condylar Osseous Defects

Objectives: This study investigated whether the inaccuracy involved in diagnosing small (<2mm) condylar osseous defects from CBCT can be improved by using CT or segmentation-assisted diagnostic methods.
Methods: Nine fresh pig heads (18 condyles, 36 medial/lateral condylar regions) were used. After exposed by dissection and demarcated by gutta percha markers, small osseous defects (diameter and depth <1.5mm) were created with 0-4 defects per condylar region predetermined randomly. Then with dissected soft tissues restored, pig heads underwent orthodontic-grade CBCT scans (i-CAT, 0.4mm voxel-size) and medical-grade CT scans (GE LightSpeed, 0.625mm voxel-size). Two calibrated raters blinded of defect information, diagnosed defect number in each condylar region from CBCT and CT images using Dolphin-3D software, initially without image segmentation, then 1-week later with proprietary image segmentation designed by Dolphin-3D. Condylar PVS impressions were collected and evaluated by the same raters, which produced physical gold-standard diagnoses. CBCT or CT diagnoses were compared with physical diagnoses to derive true/false positive or negative diagnoses and calculate classification functions.
Results: Classification functions of both raters demonstrated consistent patterns (Table 1): diagnoses using CT images were better than CBCT images, although the differences were below 10% and statistically insignificant (McNemar’s tests, p>0.05). Diagnoses of CBCT images with the segmentation tool tended to lower diagnostic sensitivity, NPV and accuracy, slightly increase specificity and PPV, but all to a statistically insignificant level (p>0.05).
Conclusions: Even when scanned at a lower voxel-size than CT, orthodontic-grade CBCT images of mandibular condyles are more difficult than CT images for diagnosing small condylar defects, and this limitation cannot be overcome by using image segmentation.