The effect of quality of temporomandibular joint MR images on interrater agreement.

OBJECTIVES

Effects of calibration on interrater agreement in evaluating magnetic resonance (MR) images of the temporomandibular joint (TMJ) have already been examined. The objectives of the present study were to assess to what extent the quality of MR images of the TMJ influences interrater agreement and to evaluate interrater agreement with respect to image quality assessment.

METHODS

Two non-calibrated medical radiologists and two general dentists evaluated sagittal images of 100 TMJs for both a rating of the image quality and the performance of five diagnostic tasks. The agreement between these raters with respect to the diagnoses was evaluated. Additionally, two additional raters, calibrated during a 5 h training including the evaluation of 70 MR images, also evaluated the diagnostic aspects and the image quality, on the basis of objective criteria. The agreement between the subjective diagnoses of the non-calibrated raters and the objective diagnoses of the calibrated raters was evaluated. Afterwards, the subjective and the objective quality assessments were compared using kappa statistics.

RESULTS

When good quality images were evaluated, higher kappa values were obtained for all diagnostic categories by the non-calibrated raters (mean Deltak for making diagnoses >0.1). This finding was confirmed by the value obtained for the agreement between the non-calibrated and the calibrated raters. The non-calibrated raters were in good agreement (k=0.67, standard error +/-0.09) with the calibrated raters for assessment of image quality.

CONCLUSION

The present study shows that it is possible even without calibration to obtain a better interrater agreement when higher quality MR images of the TMJ are evaluated.