Image-stitching artefacts and distortion in CCD-based cephalograms and their association with sensor type and head movement: ex vivo study

To assess presence and severity of image-stitching artefacts and distortion in lateral cephalograms acquired by CCD-based sensors and their association with movement.

A human skull was mounted on a robot simulating five head movement types (anteroposterior translation/lifting/nodding/lateral rotation/tremor), at three distances (0.75/1.5/3 mm), based on two patterns (skull returning/not returning to the initial position, except for tremor). Three cephalometric units, two ProMax-2D (Planmeca Oy, Finland), one with Dimax-3 (D-3) and one with Dimax-4 (D-4) sensor, and one Orthophos-SL (ORT, Dentsply-Sirona, Germany), acquired cephalograms during the predetermined movements, in duplicate (54 with movement and 28 controls with no movement per unit). One observer assessed the presence of an image-stitching line (none/thin/thin with vertical stripes or thick), misalignment between the anatomical structure display (none/<1/1-3/>3 mm), and distortion in each image quadrant (present/absent), in duplicate. Severe image-stitching artefacts were defined for images scored with a thin line with vertical stripes or thick line and/or misalignment between anatomical structure display ≥1 mm. Severe distortion was defined for images scored with distortion in both anterior quadrants of the skull. κ-statistics provided intraobserver agreement.

Intraobserver reproducibility was >0.8 (all assessed parameters). Severe image-stitching artefacts were scored in 70.4 and 18.5% of D-3 and D-4 movement images, respectively. Severe distortion was scored in 64.8% of D-3, 5.6% of D-4 and 37% of ORT movement images. Neither severe image-stitching artefacts nor severe distortion were observed in control images.

Sensor type, movement type, distance and pattern affected presence and severity of image-stitching artefacts and distortion in CCD-based cephalograms.