Related Experiment Video
Updated: Jun 26, 2026
Reliability of Artificial Intelligence-Based Cone Beam Computed Tomography Integration with Digital Dental Images
Published on: February 23, 2024
Measurements of radial angulation using computer-aided design software provide equal or superior reliability to
Regan M Stoneburner1, Lucas B Lassinger2, Galina Hayes1
1Department of Clinical Sciences, Cornell University Hospital for Animals, Ithaca, NY.
Objective:
To compare CT measurements of radial geometry obtained using computer-aided design (CAD) software to those obtained using multiplanar reconstruction (MPR) and to compare their intraobserver reliability. We hypothesized that CAD and MPR measurements would be different and that the CAD method would have higher reliability.
Methods:
CT images from normal thoracic limbs of dogs scanned from January 1, 2018, through March 1, 2023, were retrospectively collected and analyzed. The CAD and MPR-based measurements in frontal (proximal/distal), sagittal, and torsional planes were made by 1 observer, in triplicate, and compared. Biases and reliability were calculated.
Results:
10 right-sided thoracic limbs were included. Median proximal frontal angulation was 3.6° (IQR, -0.7° to 6.2°; valgus) using CAD and -6.8° (IQR, -8.9° to -5.3°; varus) using MPR. The reliability of CAD measurements of sagittal and distal frontal angulation was excellent, whereas the reliability of MPR measurements was moderate for sagittal angulation and poor for distal frontal angulation. There was no difference in reliability between CAD and MPR for proximal frontal or torsional measurements. Agreement between CAD and MPR was poor in all planes.
Conclusions:
Measurements of proximal frontal angulation differed between CAD and MPR groups, most likely due to differences in the anatomic features used. The reliability of CAD measurements was equal to or superior to the reliability of MPR measurements.
Clinical Relevance:
Understanding the relationship between CAD and MPR measurements is important as CAD-derived measurements are likely not equivalent to MPR-derived measurements.
