Neural-conduit automatic-detection algorithm based on deep learning

Abstract

The invention discloses a neural-conduit automatic-detection algorithm based on deep learning. The algorithm adopts the following steps: step 1, reading CBCT image data, reconstructing a three-dimensional model through the CBCT image data, and automatically fitting a dental arch curve; step 2, using a manner of manual interaction to delineate a corresponding curve along the dental arch curve in the three-dimensional model, and reconstructing mandibular canals in three-dimensional space; and step 3, using the dental arch curve as a reference line to slice the three-dimensional model along the dental arch curve at an interval of a specified distance in a three-dimensional space coordinate system to obtain slicing images, recording a space position of each slicing image in the three-dimensional space coordinate system, and forming an image sequence of a three-dimensional image by all the slicing images. By using an advanced deep-learning detection model, the algorithm can adapt to patientdata of different ages and different sclerotin conditions, and thus has very high stability.

Description

technical field [0001] The invention relates to the field, in particular to an automatic detection algorithm for nerve conduits based on deep learning. Background technique [0002] Defect analysis of this technology in related products [0003] Implant software products involving mandibular canal detection technology: [0004] [0005] [0006] Defect summary: [0007] At present, there is no fully automatic mandibular canal detection technology in the market, and most of the cavity implant software uses manual interaction to outline the mandibular canal. When using it, the user needs to draw the complete curve of the mandibular nerve conduit on the panoramic image, and then generate a three-dimensional nerve conduit. To ensure accuracy, the user usually needs to constantly adjust the panorama to fully visualize the mandibular canal, which is very tedious and requires a lot of interactive work. Very few products provide semi-automatic mandibular canal detection fu...

AI Technical Summary

Problems solved by technology

[0013] Disadvantages: The method requires statistical reconstruction of the entire mandible and mandibular canal, followed by shortest path refinement of the initial mandibular canal

[0017] Deficiencies: This method uses morphological methods to detect the entrance and exit of the nerve conduit, and then uses a local matching based method to track the nerve conduit from the entrance and exit respectively

[0021] Defect: This method needs to manually determine the starting point and end point of the mandibular canal, and then determine the mandibular canal by the shortest path method

However, because the mandibular canal will show great differences with the age, gender, and bone condition of the patient, the traditional method cannot deal with such a complicated situation, so the existing methods have limitations in automation and stability. big deficiency

Images