Validation of a fully automatic assessment of volume changes in the mandibular condyles following bimaxillary surgery.

Abstract

This study was performed to propose and validate a fully automatic assessment of volume changes in the mandibular condyles following orthognathic surgery. Two sets of cone beam computed tomography scans were included: one with segmentations of complete mandibles and the other with pre- and postoperative segmentations of the mandibular rami. Two convolutional neural networks predicted a segmentation of the mandible and its ramus segments. Each preoperative ramus segment was registered to the postoperative mandible, and the pre- and postoperative condylar volumes were determined. For validation, the agreement between the fully automatic assessment and a validated semi-automated method was calculated using mean absolute differences (MAD) and intraclass correlation coefficients (ICC). Forty condyles in 20 patients (16 female, four male; mean age 27.6 years) with maxillomandibular retrognathia, who underwent bimaxillary surgery, were assessed. A small difference in condylar volume change measurements was observed between the two methods (MAD 2.7%); the ICC, at 0.993, was excellent. The fully automatic method was considerably faster than the semi-automated method (3 min vs 30 min) and demonstrated high precision and excellent reliability for quantifying condylar volume changes. A fast and reliable assessment of condylar changes can identify volume changes sooner, leading to improved personalized patient care.