Bone Microstructure Segmentation and Prediction of Condyle Based on Fusion of Texture and Geometric Features

The mandibular condyle is an important part of the temporomandibular joint. Temporomandibular arthritis (TMJOA) patients often have abnormal condylar changes, suggesting the progression of TMJOA lesions. Due to the lack of in-depth research on the microstructures of condylar bone in existing studies and the lack of prediction of condylar morphology before pathological changes, a method for segmenting the microstructures of condylar bone in cone-beam computed tomography (CBCT) images is proposed. First, the geometric features of the condyle are combined with a data augmentation module to increase the number of training samples; then, the edge extraction module effectively captures the boundary information of the condyle, and the texture feature extraction module obtains texture features from the image; finally, the boundary information is fused with the feature map obtained from the decoder to obtain edge-enhanced region feature maps of different scales, enhancing the model’s perception of texture and edges. The experimental results of condylar bone microstructure segmentation on the condylar CBCT dataset show that the proposed method has achieved good results in segmenting the microstructures of the condylar bone, with a Dice coefficient of 0.901 for the cortical part and 0.958 for the cancellous part. The experimental results of morphology prediction of the cortical part before pathological changes show that the prediction result has a Dice coefficient of 0.948, verifying that the proposed method has better performance than the comparative methods. The proposed method provides strong support and guidance for the imaging diagnosis and treatment of TMJOA lesions.