PMCanalSeg: A dataset for automatic segmentation of the pterygopalatine and mandibular canals from 3D CBCT images

In orthognathic surgery, nerve canal segmentation in maxillofacial cone beam computed tomograp (CBCT) is an important link to provide key auxiliary information, which is of great significance for preventing surgical complications such as nerve damage. However, due to the high cost of data collection, patient privacy protection and ethical restrictions, high-quality, publicly available maxillofacial CBCT datasets are extremely scarce. At the same time, the nerve canal has a low contrast with the background area in the image and has a complex topological structure. These characteristics increase the difficulty of the segmentation task, making it difficult for existing deep learning segmentation models to obtain sufficient segmentation accuracy and unable to meet the needs of clinical applications. To solve the above problems and promote the development of the field of maxillofacial nerve canal segmentation, this paper constructs and publishes a large-scale dataset specifically for maxillofacial CBCT nerve canal segmentation. This dataset covers various patient cases and provides a detailed reflection of the anatomical structures of the pterygopalatine canal in the maxilla and the mandibular canal in the mandible involved in orthognathic surgery.