Impact of motion artefacts and motion-artefact correction on diagnostic accuracy of apical periodontitis in CBCT images: an ex vivo study in human cadavers.

Abstract

To assess the impact of motion artefacts and motion-artefact correction on diagnostic accuracy of apical periodontitis (AP) in CBCT images. Based on clinical and radiographic inspection of 40 formalin-fixated human jaw specimens, 77 roots in 45 teeth (molars and premolars), with various disease and treatment state, were selected. The specimens were mounted on a robot simulating 3-mm movement types (nodding, lateral rotation and tremor). CBCT images with and without (controls) movements were acquired in four CBCT units: without motion-artefact correction in Cranex 3Dx, Orthophos SL 3D, and Promax 3D Mid, and with motion-artefact correction in Promax 3D Mid and X1. Three observers blindly assessed (i) whether the images were interpretable and (ii) if AP was present (5-step probability index). Histopathology provided the reference standard for presence of AP. Weighted Kappa statistics described inter-observer agreement. Estimates of diagnostic accuracy were assessed by means of receiver operator characteristic (ROC) curve analysis. Area under the curve (AUC) provided a measure of accuracy, and paired-sample AUC difference tests compared differences amongst the CBCT units and movement types. Observer agreement was substantial for control images, moderate for motion-artefact corrected images and fair for images without motion-artefact correction. When movement was present, motion-artefact correction reduced the percentage of images scored as noninterpretable or with uncertain disease state (score 3 in the 5-step probability index). Control images were not perfectly accurate (both false-positive and false-negative results were present; AUC 0.750-0.799). Images acquired with movement and without motion-artefact correction (AUC 0.541-0.709) were associated with significantly lower accuracy than control images (P<0.05). With motion-artefact correction, accuracy was comparable to that observed in control images (AUC 0.732-0.790). Diagnostic accuracy of apical periodontitis in CBCT images was dependent on the presence of motion artefacts (i.e. lower accuracy associated with the presence of movement). Motion-artefact correction systems positively influenced image interpretability and diagnostic accuracy.