顎関節の顆状形態変化に対する人工知能支援自動定量的評価システム
Yuan Li1, Mu-Qing Liu2, Zi-Neng Xu3
1Department of Oral & Maxillofacial Radiology, Peking University School & Hospital of Stomatology, Beijing 100081, China; Department of Prosthodontics, Peking University School & Hospital of Stomatology, Beijing, 100081, China; National Center for Stomatology & National Clinical Research Center for Oral Diseases, Beijing 100081, China; National Engineering Research Center of Oral Biomaterials and Digital Medical Devices, Beijing 100081, China; Beijing Key Laboratory of Digital Stomatology, Beijing 100081, China.
Objectives:
To develop an artificial intelligence (AI)-assisted system for the automated quantitative evaluation of temporomandibular joint (TMJ) condylar morphological changes.
Methods:
This study utilized 212 cone-beam computed tomography (CBCT) scans of the TMJ from 114 patients, with radiologist-validated annotations. The proposed system comprised three modules: (1) an automated segmentation model based on the U-Net architecture, trained on 120 TMJs (60 normal and 60 osteoarthritis condyles) and tested on 32 TMJs (16 normal, 8 type I osteoarthritis condyles with surface erosion/destruction, and 8 type II condyles with osteophyte formation, sclerosis, deformity, or cyst-like lesions); (2) an automated registration model using mutual information similarity, validated on 20 paired pre- and post-treatment TMJ scans of 10 patients with normal condyles; and (3) an automated quantitative measurement model for assessing condylar morphological changes based on anatomical landmarks. Segmentation accuracy was evaluated using the Dice similarity coefficient. Furthermore, pre- and post-treatment bilateral CBCT data from 10 patients with unilateral osteoarthritis were used to assess system accuracy. Consistency between automated and manual measurements was evaluated using the intra-class correlation coefficient. Additionally, time efficiency was assessed.
Results:
The Dice similarity coefficient values for 16 normal, 8 type I, and 8 type II osteoarthritis condyles were 0.988, 0.986, and 0.981, respectively. The intraclass correlation coefficients of condyle and condylar head volumetric changes between the automated and manual methods were 0.983 and 0.995, respectively. The automated system completed measurements in 85 seconds versus 100 minutes for manual measurement.
Conclusions:
The AI-assisted quantitative evaluation system demonstrated initial technical feasibility on a single-center CBCT dataset, providing an intuitive, user-friendly, and efficient assessment of the radiographic manifestations of condylar morphological changes. However, further validation using diverse cases is required.
