The accuracy of cone-beam computed tomography in assessing maxillary molar furcation involvement.