Deep semi-supervised learning for automatic segmentation of inferior alveolar nerve using a convolutional neural network

The study aimed to develop an automated method for imaging and tracking the position of the inferior alveolar nerve (IAN) using artificial intelligence (AI) in cone-beam computed tomography datasets. A customized 3D nnU-Net was used for image segmentation, and active learning was carried out in iterations for 83 datasets. The accuracy of the model for IAN segmentation was evaluated using the remaining 50 datasets, and the dice similarity coefficient (DSC) value and the segmentation time for each learning step were compared. The deep active learning framework was found to be a fast, accurate, and robust clinical tool for demarcating IAN location. The overall segmentation was performed using AVIEW Modeler software (version 1.0.3, Coreline Software, Seoul, Korea), and the ground truth of the IAN was provided by three specialists.