Accuracy of cone-beam computed tomography imaging of the temporomandibular joint: Comparisons with panoramic radiology and linear tomography

Introduction: Cone-beam computed tomography (CBCT) is increasingly being used as an imaging modality, particularly in the assessment of the temporomandibular joint (TMJ). A blinded observational cross-sectional in-vitro study was conducted to compare the diagnostic accuracy of observers viewing images made with CBCT, panoramic radiography, and linear tomography. The task was to detect cortical erosions affecting the mandibular condylar head. Methods: The sample consisted of 37 TMJ articulations from 30 skulls with either normal condylar morphology (n = 19) or erosion of the lateral pole (n = 18). The articulations were imaged by using corrected angle linear tomography (TOMO), normal (Pan-N) and TMJ-specific (Pan-TM) panoramic radiography, and CBCT. Digital images were obtained with photostimulable phosphor plates for all modalities except CBCT. The CBCT detector used an amorphous silicon flat-panel array combined with cesium iodide. Images and 10 rereads were presented to 10 observers on a flat-panel display at a pixel-to-monitor ratio of 1:1. CBCT multi-planar images were presented both statically (CBCT-S) and interactively (CBCT-I). The observers were permitted to scroll through axial (0.4 mm) and para-sagittal (1 mm) sections and then independently rate their confidence about the presence or absence of cortical erosion. Intraobserver reliability was determined by weighted kappa and diagnostic accuracy by the fitted area under the ROC curve. Means were compared by using ANOVA (P <=.05). Results: Intraobserver reliability was moderate (0.57 +/- 0.22; range, 0.34-0.78). Pan-N (0.72 +/- 0.15), CBCT-I (0.65 +/- 0.21), and CBCT-S (0.65 +/- 0.17) reliability was significantly greater than TOMO (0.44 +/- 0.25). The diagnostic accuracy of CBCT-I (0.95 +/- 0.05) and CBCT-S (0.77 +/- 0.17) was significantly greater than all other modalities (Pan-N [0.64 +/- 0.11], Pan-TM [0.55 +/- 0.111, TOMO [0.58 +/- 0.15]). CBCT-I was also more accurate than CBCT-S, and Pan-N was more accurate than Pan-TM and TOMO. Conclusions: CBCT images provide superior reliability and greater accuracy than TOMO and TMJ panoramic projections in the detection of condylar cortical erosion.