Three-dimensional tooth image display apparatus and method, and method for training three-dimensional tooth image display apparatus
A three-dimensional tooth image display apparatus, according to an embodiment of the present invention, comprises: a tooth division unit which divides a boundary of each object in a dental cross-section image generated by dividing an input three-dimensional dental image in an axial direction; a tooth detection unit which recognizes a tooth area among objects in the dental cross-sectional image and detects a tooth number of each tooth belonging to the tooth area; and a tooth image obtainment unit which generates a three-dimensional image of each tooth on the basis of the boundary of each object in each dental cross-sectional image and the tooth number of each tooth, which are obtained through the tooth division unit and the tooth detection unit, wherein the three-dimensional image of each tooth may have a different display method according to the tooth number.
1 . A three-dimensional tooth image display apparatus comprising:
a tooth division unit which divides a boundary of each object in a dental cross-section image generated by dividing an input three-dimensional dental image in an axial direction;
a tooth detection unit which recognizes a tooth area among objects in the dental cross-sectional image and detects a tooth number of each tooth belonging to the tooth areal; and
a tooth image obtainment unit which generates a three-dimensional image of each tooth on the basis of the boundary of each object in each dental cross-sectional image and the tooth number of each tooth, which are obtained through the tooth division unit and the tooth detection unit, wherein the three-dimensional image of each tooth has a different display method according to the tooth number, wherein the display method comprises displaying at least some teeth in different colors according to a tooth number of each tooth.
2 . The three-dimensional tooth image display apparatus of claim 1 , wherein the tooth division unit comprises a tooth division neural network, and the tooth division neural network is trained using a plurality of training dental cross-sectional images as training input data and as well as using, as label data, a masking image in which a boundary region of each object in the training dental cross-sectional images is masked.
3 . The three-dimensional tooth image display apparatus of claim 1 , wherein the tooth detection unit comprises a tooth detection neural network, and the tooth detection neural network is trained using a plurality of training dental cross-sectional images as training input data and as well as using, as label data, a tooth area where a tooth exists in the training dental cross-sectional images and a tooth number for each tooth within the tooth area.
4 . The three-dimensional tooth image display apparatus of claim 1 , wherein the tooth image obtainment unit comprises a tooth image obtainment neural network, and the tooth image obtainment neural network is trained to generate individual three-dimensional images for each tooth with the same tooth number within the three-dimensional image on the basis of training dental cross-sectional images generated by dividing a three-dimensional image along one axis and boundary information of each tooth and a tooth number of each tooth within each training dental cross-sectional image.
5 . The three-dimensional tooth image display apparatus of claim 4 , wherein the tooth image obtainment neural network is trained to generate the individual three-dimensional images by stacking tooth-specific divided images extracted based on the boundary information within each of the divided training dental cross-sectional images and tooth-specific divided images with the same tooth number for the extracted tooth-specific divided images in the axial direction.
6 . The three-dimensional tooth image display apparatus of claim 1 , wherein the display method comprises displaying teeth in a same color when a relationship of tooth numbers for each tooth is a mirroring or antagonist relationship, or displaying teeth in different colors when a relationship of tooth numbers for each tooth is not a mirroring or antagonist relationship.
7 . The three-dimensional tooth image display apparatus of claim 1 , wherein the three-dimensional dental image comprises a cone-beam computed tomography (CBCT), and the divided dental cross-sectional image is an image perpendicular to the axial direction.
8 . A three-dimensional tooth image display method comprising:
dividing a boundary of each object in a dental cross-section image generated by dividing an input three-dimensional dental image in an axial direction;
recognizing a tooth area among objects in the dental cross-sectional image and detecting a tooth number of each tooth belonging to the tooth area; and
generating a three-dimensional image of each tooth on the basis of the obtained boundary of each object in each dental cross-sectional image and the obtained tooth number of each tooth, wherein the three-dimensional image of each tooth has a different display method according to the tooth number, wherein the display method comprises displaying at least some teeth in different colors according to a tooth number of each tooth.
9 . A method for training a three-dimensional tooth image display apparatus, comprising: training a tooth division neural network to divide a boundary of each object in a dental cross-section image generated by dividing an input three-dimensional dental image in an axial direction;
training a tooth detection neural network to recognize a tooth area among objects in the dental cross-sectional image and detect a tooth number of each tooth belonging to the tooth area; and training a tooth image obtainment neural network to generate a three-dimensional image of each tooth on the basis of the boundary of each object in each dental cross-sectional image and the tooth number of each tooth, which are obtained through the tooth division neural network and the tooth detection neural network,
wherein the three-dimensional image of each tooth has a different display method according to the tooth number, wherein the display method comprises displaying at least some teeth in different colors according to a tooth number of each tooth.
10 . The method of claim 9 , wherein the training dental cross-sectional image is an image generated by dividing a training three-dimensional image along one axis.
11 . The method of claim 9 , wherein the training of the tooth image obtainment neural network comprises training the tooth image obtainment neural network to generate individual three-dimensional images for each tooth with the same tooth number within the three-dimensional image on the basis of boundary information of each tooth and a tooth number of each tooth within each training dental cross-sectional image.