Deep learning for automated segmentation of the temporomandibular joint.

OBJECTIVE: Quantitative analysis of the volume and shape of the temporomandibular joint (TMJ) using cone-beam computed tomography (CBCT) requires accurate segmentation of the mandibular condyles and the glenoid fossae. This study aimed to develop and validate an automated segmentation tool based on a deep learning algorithm for accurate 3D reconstruction of the TMJ. MATERIALS AND METHODS: A three-step deep-learning approach based on a 3D U-net was developed to segment the condyles and glenoid fossae on CBCT datasets. Three 3D U-Nets were utilized for region of interest (ROI) determination, bone segmentation, and TMJ classification. The AI-based algorithm was trained and validated on 154 manually segmented CBCT images. Two independent observers and the AI algorithm segmented the TMJs of a test set of 8 CBCTs. The time required for the segmentation and accuracy metrics (intersection of union, DICE, etc.) was calculated to quantify the degree of similarity between the manual segmentations (ground truth) and the performances of the AI models. RESULTS: The AI segmentation achieved an intersection over union (IoU) of 0.955 and 0.935 for the condyles and glenoid fossa, respectively. The IoU of the two independent observers for manual condyle segmentation were 0.895 and 0.928, respectively (p<0.05). The mean time required for the AI segmentation was 3.6 s (SD 0.9), whereas the two observers needed 378.9 s (SD 204.9) and 571.6 s (SD 257.4), respectively (p<0.001). CONCLUSION: The AI-based automated segmentation tool segmented the mandibular condyles and glenoid fossae with high accuracy, speed, and consistency. Potential limited robustness and generalizability are risks that cannot be ruled out, as the algorithms were trained on scans from orthognathic surgery patients derived from just one type of CBCT scanner. CLINICAL SIGNIFICANCE: The incorporation of the AI-based segmentation tool into diagnostic software could facilitate 3D qualitative and quantitative analysis of TMJs in a clinical setting, particularly for the diagnosis of TMJ disorders and longitudinal follow-up.

Three-dimensional analysis of condylar remodeling and skeletal relapse following LeFort-I osteotomy: A one-year follow-up bicenter study.

The aim of the study was to quantify the postoperative condylar remodeling after Le Fort I surgery. Patients treated with a Le Fort I osteotomy were investigated. CBCT scans were acquired preoperatively, one week and one year postoperatively. A preoperative 3D cephalometric analysis was performed on the preoperative CBCT. Surgical movements were quantified using a voxel-registration based method (OrthoGnaticAnalyser). After rendering of the condyles from the CBCT, a volumetric analysis was performed. The correlation between the surgical movement of the maxilla and the postoperative condylar volume changes was determined with analysis of variance. RESULTS: A total of 45 subjects were included in this study. 47 of 90 condyles (52%) showed a mean volume reduction of 93 mm3 (4.9 volume-%) postoperatively. The maxilla was impacted in 12 patients (2.44 ± 2.49 mm) and extruded in 33 patients (1.78 ± 1.29 mm). The maxillary impaction group showed a volume reduction of 50 ± 122 mm3 and the extrusion group showed a mean volume gain of 21 ± 139 mm3 (p = 0.028). CONCLUSION: Clinicians should be aware of potential condylar remodeling following solitary Le Fort I osteotomies, particularly in female patients with maxillary impaction.

Landmark-Based Versus Voxel-Based 3-Dimensional Quantitative Analysis of Bimaxillary Osteotomies: A Comparative Study.

PURPOSE: We compared the accuracy of landmark-based and voxel-based 3-dimensional (3D) analysis to quantify the osseous movements of the maxilla and mandible after bimaxillary osteotomy. MATERIALS AND METHODS: Cone beam computed tomography (CBCT) scans of 15 patients who had undergone bimaxillary osteotomy were randomly selected from the database. Before surgery, CBCT scanning was performed and an individualized 3D virtual surgical plan made for all patients. During surgery, the mandibular and maxillary segments were positioned as planned using 3D-milled interocclusal splints. At 1 week after surgery, a postoperative CBCT scan was acquired. All pre- and postoperative CBCT data were rendered in 3 dimensions. The 3D virtual head models were superimposed on the cranial base. The 3D surgical movements of the maxilla and mandible were quantified using conventional landmark-based 3D cephalometric analyses and voxel-based 3D analyses (OrthoGnathicAnalyser). This process was performed by the same observer 3 times. The intraclass correlations and Bland-Altman plots were computed to quantify the measurement errors and reproducibility of both methods. RESULTS: High intraclass correlation coefficients were found for both methods. The voxel-based analyses yielded a higher correlation concerning the maxilla and distal mandible (r = 0.98) compared with the landmark-based cephalometric analyses (r = 0.90). CONCLUSIONS: The use of voxel-based 3D analyses in the quantification of osseous movements was more reliable and reproducible than the use of conventional landmark-based 3D analyses.

Three-dimensional analysis of condylar remodeling and skeletal relapse following bimaxillary surgery: A 2-year follow-up study.

PURPOSE: To quantify the postoperative condylar remodelling and its role in skeletal relapse after bimaxillary surgery. MATERIALS AND METHODS: 50 patients with mandibular hypoplasia who underwent bimaxillary surgery were analyzed. CBCT scans were acquired preoperatively, one week postoperatively and two years postoperatively. 3D cephalometric analysis was carried out for each CBCT scan, after which the condylar volume analysis was performed. RESULTS: The maxilla was advanced by a mean of 2.1 mm with a corresponding mean relapse of 0.3 mm. The maxilla was impacted in 23 and extruded in 27 patients. The mean mandibular advancement was 7.8 mm. Two years after surgery a mean mandibular skeletal relapse of 1.3 mm was observed. 78% of condyles exhibited a postoperative reduction in volume of 179 mm3 (mean), equivalent to 12.5 volume%. Postoperative condylar volume loss was correlated with mandibular skeletal relapse (r = 0.42, p < 0.01), but not with maxilla relapse. Linear regression analysis identified age, gender, amount of surgical mandibular advancement and postoperative condylar volume loss as predictive factors for mandibular relapse. CONCLUSION: A significant correlation between postoperative condylar volume loss and skeletal relapse was found. Young female patients who underwent large bimaxillary advancement and postoperative reduction in condylar volume were particularly at risk for skeletal relapse.