Patent Application
20250064562
The present disclosure generally relates to a system and method for designing a dental prosthesis. In particular, the present disclosure relates to a method for designing a dental prosthesis based on a digital 3D representation, wherein at least a part of the surface of the dental prosthesis is determined independent of any knowledge or estimate of a hidden part of the preparation surface.
Prosthodontics is the dental specialty concerned with the design, manufacture and fitting of artificial replacements (prostheses) for teeth. In preparing for a dental prosthesis, a dental practitioner may grind down existing dentition to create a tooth preparation for mounting a dental prosthesis. The surface of this tooth preparation is known as the preparation surface. The dentist may further shape the tooth preparation to ensure that the prosthesis can be designed with sufficient thickness. The lower outer boundary of the preparation surface is typically denoted the preparation line. When the prosthesis is mounted on the preparation, the lower edge of the prosthesis is also demarcated by this line. If the seal between the prosthesis and the preparation is not good, there is a risk that gaps will collect dirt and bacteria, which can lead to further tooth decay and that the prosthesis will not last long. For that reason, it is important to achieve a good fit of the prosthesis at the preparation line.
In digital dentistry, a digital 3D model of the tooth preparation and the surrounding dentition may be obtained. The digital model may be acquired using an intraoral scanning device. The prosthesis may then be digitally designed and manufactured using computer aided design and manufacturing (CAD/CAM). The prosthesis is typically manufactured in a mill. The need for a good seal between prosthesis and preparation, as discussed above, typically implies a need for a high accuracy 3D scan in the area surrounding the preparation line.
In order to achieve an aesthetically pleasing result, the preparation line is often located sub-gingivally, i.e., slightly below the surface of the gingiva. By doing so, the boundary of the prosthesis is hidden by the gingiva when mounted on the preparation. One problem with this approach is that the preparation line, or parts of the preparation line, may become hidden/obscured by e.g. the gingiva, which consequently means that in some cases the entire preparation line is not visible in the corresponding digital 3D model. To circumvent this problem, the surrounding gingiva is sometimes retracted, such that the preparation surface itself can be scanned. The dentist performs this retraction by pushing retraction cord(s) into the sulcus. The retraction cord is then removed immediately before scanning such that the gingiva does not have time to move back to a position where it covers the preparation line. It is often an issue that the gingiva collapse onto the preparation before the scan is performed, preventing the acquisition of a scan wherein the preparation line is visible. Another problem with this approach is that the procedure is unpleasant to the patient.
Another procedure to mitigate the problem of collapsed gingiva or foreign objects is to take an impression without any physical remedy, i.e., with parts of the preparation surface hidden. The hidden parts of the preparation surface may in some cases be estimated afterwards. Traditionally, dental technicians have used their experience to physically remove material from a gypsum cast of the impression. However, the quality of the outcome of the physical procedure depends on the level of experience of the dental technician.
More recently, computer algorithms for the estimation of hidden dental surfaces have been described. The computer algorithms suggested, e.g. in US 2019/321145 A1, rely on extrapolation of surfaces believed to represent the dentition. The distinction is however difficult to make, particularly when no other relevant data such as color photographs are available. Furthermore, such extrapolation methods depend on the choice of a rather simple mathematical model of only two-dimensional cross sections of surfaces that may not be appropriate generally.
US 2021/321872 A1 describes a method of determining a 3D model of a dental prosthesis. The described method relies on determining the margin line in order to determine a 3D model of the dental prosthesis. The document further describes a correction that may be performed to remove the representation of foreign material and show the underlying tooth surface and/or margin line.
US 2021/0059796 A1 describes a method to mark and/or correct margin lines, the method utilizing a trained machine learning model that has been trained to identify margin lines of preparation teeth, wherein the trained machine learning model outputs a probability map comprising, for each pixel in the image, a probability that the pixel depicts a margin line.
US 2007/0015111 A1 describes a method wherein a 3D virtual dental model of an intra oral cavity, in which at least a part of a finish line of a preparation is obscured, is manipulated to create, recreate or reconstruct finish line data corresponding to the obscured part. The document further describes that an auxiliary 3D virtual model corresponding to the obscured portion of the finish line is created.
All of the methods described in the state of the art involve one or more steps of recreating or determining the hidden part(s) of the preparation surface, such as the obscured part(s) of the margin line. However, in some cases, such approaches may be inaccurate, difficult, and cumbersome to carry out.
Therefore, there is a need for improved methods for designing dental prostheses; in particular, in situations where at least a part of the preparation surface is hidden or obscured from view.
The present disclosure addresses the above-mentioned challenges by providing a system and method for designing a dental prosthesis, wherein at least a part of the surface of the dental prosthesis is determined independent of any estimate of hidden part(s) of the preparation surface. In other words, the presently disclosed method is able to automatically generate the design of a dental prosthesis, such as a dental crown, ideally without full knowledge of the margin line and/or preparation surface.
As previously mentioned, in many cases a foreign object covers, or partially covers, a part of the preparation surface. The preparation surface may be understood as the resulting surface of a tooth/teeth, which has been prepared for receiving a dental prosthesis. One way to obtain a preparation surface is to grind a tooth down, whereby a tooth preparation is obtained. The surface of the tooth preparation is also referred to as the preparation surface. The lower outer boundary of the preparation surface, hidden or not, is referred to herein as the preparation line. In the technical field of prosthodontics, this line is sometimes also referred to as the finish line or the margin line. Examples of foreign objects that could cover parts of the preparation surface, such as parts of the preparation line, include soft tissue such as the gingiva of the patient, debris from the grinded tooth/teeth, or saliva. The obscured/covered part of the preparation surface is referred to herein as the hidden part of the preparation surface.
One advantage of the presently disclosed method is that it does not require full knowledge of the preparation surface, nor does it require an estimation of the hidden parts of said surface. In other words, the method does not require any step to reconstruct any hidden surface of the tooth and/or preparation surface. Furthermore, the method ideally does not require any preliminary step of determining whether a foreign object, such as gingiva, hides any part of the preparation surface. Such a step could however be advantageous in some embodiments.
The presently disclosed system and method is useful for both intraoral scans, scans of dental impressions, and scans of stone models cast from dental impressions. All of the above can be affected by gingiva or other objects covering some part of the preparation surface. This is an advantage over the known art, such as US 2019/321145 A1, which is limited to handheld scanning devices, which in effect implies intraoral scanning devices.
The present disclosure provides an improved method for computer-aided design (CAD) of dental prostheses, based on a 3D scan of the relevant dentition including the preparation surface. The design of a dental prosthesis refers to the process of creating the shape and form of the prosthesis, generally described by its surface in 3D space. Examples of dental protheses comprise: crowns, bridges, inlays, onlays, and others. In computer-aided design, shape, form, and surface can be stored as data files that then can be used in computer-aided manufacturing (CAM).
The bottom of a dental prosthesis is understood to be the part of the prosthesis that is the furthest in the apical direction, when it is mounted to the preparation surface. In other words, the bottom of a dental prosthesis is the surface which is adapted to be fitted/mounted to the preparation surface, typically via a cement gap or cement layer. The wording ‘to be fitted’ may be understood as the ability to place and mount the dental prosthesis on the tooth preparation (comprising the preparation surface) in a dental procedure. As an example, the dental prosthesis may be a dental crown which is digitally designed and manufactured such that it can be fitted on the preparation (e.g. a stump) inside the mouth of a subject. The design of the dental prosthesis may result in a variety of different marginal finish lines such as knife edge, chamfer, deep chamfer, radial shoulder, radial shoulder with a bevel, or classic shoulder. In general, the shape of the tooth preparation may influence the digital design of the dental prosthesis and the resulting marginal finish line.
The shape of an ideal preparation near the bottom of the dental prosthesis typically depends on the material of the prosthesis. For example, for precious metal alloys, preparations should have feather edge, whereas for ceramic materials and especially for glass ceramic materials, preparations should have a shoulder edge. In dental CAD software, the main part of the outer surface of a dental prosthesis, such as an upper part of the outer surface, is typically designed interactively by a user, using digital shape manipulation tools. The outer surface should achieve good contacts to neighboring and antagonist dentition. Interaction between software and user is typically via a computer screen, where the software renders a 3D scene of dentition and prosthesis, wherein the user can manipulate the shape of the prosthesis, receiving immediate visual feedback.
The design of prosthesis bottoms in dental CAD software known in the art typically also requires the virtual marking of a preparation line, which is the (lower) outer boundary of the preparation surface. Inwards from the preparation line, cement gap is typically narrower than elsewhere on the intaglio. Especially for ceramic materials, outwards from the preparation line, the bottom surface of the prosthesis often has a collar consisting of a nearly flat band up to an angled transition to the main outer surface of the prosthesis, often called the die extension.
Dental CAD software known in the art often applies mechanistic design near the preparation surface. Mechanistic design typically requires some parameters as input but does not require interactive user input. Mechanistic design comprises offsetting the preparation surface to arrive at the inner surface of the prosthesis to realize a cement gap, with parameters describing, e. g., the desired thickness of the cement gap, any width and location relative to the boundary of the preparation surface of zones with smaller thickness, etc. Additional offset may be required due to the finite radius of a milling burr or drill, which may be larger than inner corner radius of the preparation surface. The radius would be another parameter entering mechanistic design. Mechanistic design can also comprise designing a collar outwards from the boundary of the preparation surface, where parameters are collar width, transition angles, and possibly others.
A further advantage of the presently disclosed method is that it does not require any manual marking nor any automatic determination or estimate of the entire preparation surface/line. At least one part of the surface of the dental prosthesis, such as the prosthesis bottom, is predicted from the scanned surface of the dentition alone. The scan should preferably include the area of the intended prosthesis, surrounding dentition and surrounding gingiva. Preferably, the scan should include all the visible parts of the preparation surface.
The method disclosed herein can be implemented in dental CAD software. As mentioned above, the present disclosure provides a method of designing a dental prosthesis, wherein at least one part of the prosthesis, such as the bottom part, is determined automatically without estimating any hidden parts of the preparation surface. Other parts of the prosthesis may be designed by other methods such as interactive design by a user in the CAD software, or such parts may be generated by other shape design algorithms, or by mechanistic algorithms.
Accordingly, the present disclosure relates to a computer-implemented method for designing a dental prosthesis, such as a dental crown, the method comprising the steps of:
The phrase ‘without any knowledge’ or ‘independent of any knowledge’ may be understood as the ability of e.g. the statistical model or trained machine-learning model to automatically generate at least a part of the dental prosthesis without knowing where e.g. the margin line is. Thus, the phrase primarily applies to the situation when the trained model is applied, and not during the training phase.
In some embodiments, the computer-implemented method comprises the steps of:
In some embodiments, the computer-implemented method comprises the steps of:
To calibrate a method according to the present disclosure, it may be advantageous to make use of at least one data set of 3D surfaces where hidden parts of preparation surfaces are known or estimated and annotated accordingly. As an example, said data set may constitute a training data set for a machine-learning model, such as a neural network.
The present disclosure further relates to a data processing system comprising one or more processors configured for carrying out the steps of any of the methods disclosed herein. The present disclosure further relates to a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the steps of any of the methods disclosed herein. The computer program may be stored on a computer-readable medium.
The present disclosure further relates to a system for digitally designing a dental prosthesis, such as a dental crown, the system comprising:
The present disclosure further relates to a method for manufacturing a dental prosthesis, the method comprising the steps of obtaining a digital design of the dental prosthesis, wherein the design is obtained by the computer-implemented method disclosed herein; and manufacturing the dental prosthesis based on the digital design, wherein the dental prosthesis is manufactured using a computer-aided manufacturing (CAM) device, such as a milling machine or a 3D printer.
The present disclosure further relates to a computer-implemented method of training a machine-learning model for designing a dental prosthesis, comprising: receiving an input training dataset comprising pairs of 3D scans of a preparation surface and corresponding prostheses designed by experienced dental technicians, where, in at least some of the pairs, at least a part of the preparation surface is obscured e.g. by gingiva or foreign material.
Accordingly, the present disclosure provides a system and method for designing a dental prosthesis, wherein the dental prosthesis is generated or designed to fit a preparation surface without any knowledge or estimate of potentially hidden surfaces of the preparation surface.
In preferred embodiments of the disclosed method, statistical design is employed to predict a part of the surface of the dental prosthesis. Statistical design may be understood as using a statistical model, such as a trained machine-learning model, to predict at least part of the shape of the dental prosthesis.
Preferably, a first step of the disclosed method is to obtain a digital 3D representation. This may be done by scanning a surface of interest, e.g., the preparation surface, using a 3D scanning device. The scan may be an intraoral scan, a scan of a dental impression, or a scan of a stone/gypsum model, which is cast from a dental impression. The scan may be acquired using an intraoral 3D scanning device; in particular if the preparation surface is to be scanned intraorally. As previously mentioned, oftentimes one or more foreign objects cover some parts of the preparation surface. Examples of foreign objects are soft tissue such as the gingiva, debris from the grinded tooth/teeth, or saliva. Therefore, the obtained 3D scan/3D representation may comprise regions wherein the preparation surface is partly hidden because the foreign object(s) obscured the view when the surface was scanned.
The 3D scanning device may acquire scan data, such as images of the surface, and it may further process the scan data/images to obtain depth information of the surface. The scan data may be transmitted to a computer or processing unit, which is configured for generating the digital 3D representation based on the scan data/images. The generation of a digital 3D representation from scan data is generally well-known within the field of digital dentistry. It is further described in WO 2019/158442 A1 and WO 2014/000745 A1 by the same applicant, which are incorporated herein by reference in their entirety.
In preferred embodiments, the digital 3D representation comprises 3D data of at least a part of the preparation surface adapted to receive the dental prosthesis, wherein the preparation surface comprises a visible part and a hidden part which is obscured by a foreign object such as soft tissue.
A next step of the disclosed method is to determine at least one part of a surface of the dental prosthesis using a statistical design model. The at least one part is determined based on the digital 3D representation such that the dental prosthesis can be fitted to the preparation surface.
The statistical design model can be expressed mathematically as
where Sp is the predicated shape of the dental prosthesis or part thereof, Ss is the scanned surface, F is a function, and X is a set of conditions, which may in some embodiments be empty. The scanned surface may be represented by the digital 3D representation mentioned above. The surface may be represented mathematically as a point cloud, or a mesh of points and facets, or by basis functions of points' coordinates, or in other ways. The scanned surface/representation can be obtained by an intraoral scanning device, or from a scan of an impression, or from a scan of a stone model cast from the impression.
Statistical design comprises defining and applying a function F of equation (1). Preferably, statistical design further comprises calibration of the function F. The function F may be a machine-learning based function, such as a machine learning algorithm or a neural network. A definition typically comprises a choice of mathematical form and parameterization. Examples of mathematical forms are regression models and neural networks. Examples of parameterization are the choice of degree of a polynomial, or the choice of neural network structure, activations functions, and others. The calibration can include, e.g., least-squares fitting or neural network training, and others. Applying a function can be evaluation of the function in some software.
Because of their flexibility, neural networks are particularly suitable choices for the function F. The network could for instance have an encoder/decoder structure with skip connections. A U-Net architecture would be one such structure. Thus, in preferred embodiments, one or more steps of the disclosed method are carried out by a trained machine-learning model, such as a neural network, which has been trained to determine at least a part of the shape of a dental prosthesis based on a digital surface scan/digital representation of a preparation surface. In some embodiments, the at least one part of the shape corresponds to a bottom part of the dental prosthesis, said bottom part being configured to fit to the preparation surface. In some embodiments, the neural network is trained to output a full 3D representation of the dental prosthesis based on a digital representation of the preparation surface. In some cases, the digital surface scans are converted to a 2D data structure before being input to the neural network.
Inputs for the calibration should preferably be pairs [Sp, Ss], i.e., designed prostheses for given surface scans. Preferably, the designs of the prostheses have been executed by experienced dental professionals. These dental professionals will recognize some gingiva or artefacts in the surfaces Ss and design a prosthesis compensating for such artefact, geometrically penetrating that part of the observed surface. Thus, the neural network may have been trained using a training data set including a plurality of 3D surface representations and corresponding dental prostheses designed by one or more experienced dental technicians. The dental prostheses designed by the technicians may constitute the ground truth. The 3D surface representations may include a preparation surface, where, in at least some of the representations, the preparation surface is partly covered by a foreign object, such as gingiva. In some embodiments, the training data set comprises at least 10,000 pairs, each pair comprising a 3D surface representation and a corresponding designed dental prosthesis. The training of the neural network may be an iterative process, which may be repeated until the output from the neural network reaches a given accuracy or until the neural network is no longer able to improve its output. Accordingly, after a number of iterations, the output of the neural network may resemble the ground truth within a given precision and/or accuracy.
One way of calibration uses triplets [Sp, Ss, Ssx], where at least some of the scans Ss include gingiva and other scans Ssx are modifications of the Ss, with any gingiva or foreign objects removed in software based on assessments performed by one or more dental professionals. Thus, in some embodiments, the neural network has been trained using a training data set including a plurality of triplets of 3D surface representations.
An advantageous way of calibration uses triplets [Sp, Ss, Ssc], where at least some of the scans Ss intentionally include gingiva and other scans Ssc are re-scans of the Ss, but with any gingiva or foreign objects removed mechanically. In other words, in some cases the patients' dentition is scanned twice: both with and without gingiva or foreign objects covering some part of the preparation surface. The prosthesis is then designed based on Ssc, while the calibration remains based on Ss. The function F thus learns to compensate for gingiva or foreign objects. The quality of predictions made with such a calibrated function should be higher than that of a function calibrated based on visual recognition of gingiva or foreign objects alone. Constraints X may be introduced to impose some property on the predicted surface, e.g., smoothness or continuity.
In some embodiments, calibration exploits knowledge of a cement gap in pairs [Sp, Ss] or triplets [Sp, Ss, Ssc] to provide for a function F that is normalized by a cement gap. For example, the scan surface Ss in every dental design order to be used in the calibration could first be expanded by the cement gap known for that order. The calibration would then be normalized to a cement gap of zero. When using the calibrated model for a prediction, said prediction would also apply to a cement gap of zero. If the order for the prediction requests another cement gap, the first prediction could then be contracted by that gap.
In some embodiments of the presently disclosed method, the 3D representations Ss are transformed by some transform T, and the predictions and the conditions are likewise transformed values,
where Tp and Tx are the transforms of the predictions and the conditions, respectively; F′ is a function not necessarily derived from F but related to F by the domain transforms Tp and Tx.
The transformation may form part of a pre-processing step of the method. The desired predicted surface Sp is then found via an inverse transformation Tp−1 as
In some embodiments, the existence of Tp−1 is provided by suitable choice of at least one of F′ and X. In some embodiments, Tp−1 is provided by approximation. If some F′(Ts(Ss)) is found not to be invertible, an approximate inverse Tp−1 can be found by choosing function {tilde over (F)}′ for which {tilde over (F)}′(a)≈F′(a) for arguments a in a range relevant for the intended application. In some embodiments where an approximate Tp−1 gives a sufficiently good estimate, it is practical to use said approximation even if F′ (Ts(Ss)) is not strictly invertible.
In some embodiments, the transform Ts transforms a scanned 3D surface to a 2D representation. The 2D representation could comprise a scalar such as a distance of the surface Ss to some reference plane, or a curvature, or other. The 2D representation can also comprise a vector such as a surface normal of the surface Ss, or other.
In some embodiments, the transform Ts creates multiple independent representations, e.g., several 2D representations as per above, where each representation is restricted to some subset of the full surface. In an example of such an embodiment, the transform Ts generates a set of 2D depth images taken from various viewpoints and angles relative to the 3D surface Ss.
It can be advantageous to transform to 2D space because the lower dimensionality can reduce computational requirements. It can also be advantageous to transform to 2D space because more standard algorithms and software exists for those, in particular neural networks. An example of a method for transforming a 3D mesh to 2D is mesh flattening as described in US 2020/0273248 A1, which is incorporated herein by reference in its entirety. Thus, the trained neural network may be configured to process 2-dimensional data wherein the 2-dimensional data may be obtained by converting 3-dimensional point clouds, e.g. by mesh flattening algorithms, to 2-dimensional data.
A 2D representation can be sampled in an equidistant grid to form a pseudo-image Is with pixel values. A pseudo-image has the same data structure as a traditional image but is not necessarily intelligible to a human. In cases where the 2D representation comprises vectors, each pixel will represent vectors. This can also be seen as the pseudo-image having multiple channels. Pseudo-images in the context of neural networks have been further described in WO 2020/161245 A1, which is incorporated herein by reference in its entirety. Many neural network structures have been published for 2D images and pseudo-images, e.g., Ronneberger, Olaf; Fischer, Philipp; Brox, Thomas (2015): U-Net: Convolutional Networks for Biomedical Image Segmentation, in: Medical Image Computing and Computer-Assisted Intervention—MICCAI 2015, Springer International Publishing, p. 234-241, ISBN 978-3-319-24574-4. Alternatively, the trained neural network may be capable of directly handling 3-dimensional (3D) point clouds. An example of such a network architecture is a PointNet neural network (“PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation”, Charles R. Qi, Hao Su, Kaichun Mo, Leonidas J. Guibas). Alternatively, PointNet++ or PointNet-like network architectures may be used for directly handling 3D point clouds.
In some embodiments, Tp and Ts are defined on the same 2D grid. For example, if Ts yields pseudo-images Is, and the predictions F′(Ts(Ss)) are other pseudo-images of the same dimension as the Is, then the transform Tp could be defined to also output pseudo-images of that same dimension.
In some embodiments, the transforms Tp and Ts are the same, which implies that Tp−1=Ts−1. In some embodiments, the transform Tp includes a representation of a conversion. For example, if the predictions Tp(Sp)=F′(Ts(Ss)) are differences of distances between an observed and a desired surface, then Tp−1 can include an operation of offsetting the observed surface by the predicted values F′(Ts(Ss)). In some embodiments, Tp−1 includes an operation of computing one or more distance fields. In some embodiments, Tp−1 includes interpolation.
In some embodiments, distance fields are converted back to 3D meshes, e.g., by a marching cubes algorithm as described, e.g., in William E. Lorensen and Harvey E. Cline. 1987. Marching cubes: A high resolution 3D surface construction algorithm. SIGGRAPH Comput. Graph. 21, 4 (July 1987), 163-169.
In some embodiments, the conditions can be transformed in the same way as the scanned surface, i.e., Tx=Ts. In some situations, it may not be possible to transform a constraint, for example it may not be possible to transform a constraint in 3D space to 2D space. If so, the original constraint X of equation (2) can be applied after the inverse transform Tp−1.
In some embodiments, statistical design is combined with mechanistic design. The latter may be applied preferentially to intaglio and outer surface of the prosthesis away from the bottom.
Statistical design in some embodiments is used to predict only a bottom region of the prosthesis. In other embodiments, statistical and mechanistical design are both applied to the regions near the transition from or to the prosthesis bottom.
When combining statistical with mechanistic design, the two approaches may yield differing surfaces at or near the transition between the regions to which each was applied. Typically, this is the transition from bottom to remainder surface. The two different surfaces can then be merged. In some embodiments, constraints X can be used to enforce continuity. In other embodiments, the two predicted surfaces can be stitched together. In other embodiments where both approaches are applied to the same region, some form of average surface can be found at or near the transition. Other operations such as smoothing, e.g., Laplacian smoothing, can be applied in merging, too.
In some embodiments, both statistical (e.g. machine-learning based) and mechanistic design are applied in the bottom region. This is relevant, e.g., when the material intended for the prosthesis must follow a particular shape in some sub-region of the bottom region as per manufacturer prescription. In such constellations, the mechanistic design should override the statistic one. While doing so, some characteristics of the statistical design may be preserved, e.g., by a least-squares or volume-preserving fit of a mechanistically designed sub-region to the statistical design of the overall bottom region. In some embodiments, a mechanistic design, or a part of the mechanistic design, is used as the input, or a part of the input, to the statistical design.
To calculate Sp from the network output Δ=F′(Ts(Ss)), Tp−1, the inverse of the transform Tp needs to be defined. First the input surface Ss can be defined as a signed distance field Ds
where C(p) tells whether a point p in 3D space is inside the surface of Ss
The network parameters for the neural network of
Because input 801 and output 802 are defined on the same 2D grid, i.e., because pixels with same 3D coordinates correspond to same points in 3D space, the inverse transform Ts−1 can be used for de-flattening, resulting in a 3D distance field Dp, representing the offset from the input surface.
The distance field representation D of the predicted surface is then the sum of the de-flattened offset distance field Dp and the distance field computed from the input surface Ds
The predicted surface can then be reconstructed using a meshing algorithm , such as the Marching Cube algorithm,
where the points p are chosen in a 3D grid chosen large enough to encapsulate the surface Sp.
In step 1002, at least one part of the surface of the dental prosthesis is determined using statistical design. In preferred embodiments, the at least one part is determined automatically using a neural network, which has been trained to determine the part even in situations where parts of the preparation surface were hidden in the 3D representation. In step 1003, the dental prosthesis is designed based on the at least one part determined in step 1002. In preferred embodiments, the at least one part is a bottom part of the dental prosthesis, wherein the bottom part is understood to be the part configured to interface the preparation surface, possibly via a cement gap. The dental prosthesis may be designed by statistical design, mechanistic design, or a combination hereof. In a final step 1004, the dental prosthesis may be manufactured according to the design. The manufacturing may be performed by a computer-aided manufacturing (CAM) device, such as a milling machine or a 3D printer.
Although some embodiments have been described and shown in detail, the disclosure is not restricted to such details, but may also be embodied in other ways within the scope of the subject matter defined in the following claims. In particular, it is to be understood that other embodiments may be utilized, and structural and functional modifications may be made without departing from the scope of the present disclosure. Furthermore, the skilled person would find it apparent that unless an embodiment is specifically presented only as an alternative, different disclosed embodiments may be combined to achieve a specific implementation and such specific implementation is within the scope of the disclosure.
A claim may refer to any of the preceding claims, and “any” is understood to mean “any one or more” of the preceding claims.
It should be emphasized that the term “comprises/comprising/including” when used in this specification is taken to specify the presence of stated features, integers, operations, steps or components but does not preclude the presence or addition of one or more other features, integers, steps, components or groups thereof.
In claims enumerating several means, several of these means can be embodied by one and the same item of hardware. The mere fact that certain measures are recited in mutually different dependent claims or described in different embodiments does not indicate that a combination of these measures cannot be used to advantage.
| Number | Date | Country | Kind |
|---|---|---|---|
| 21213700.4 | Dec 2021 | EP | regional |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2022/085189 | 12/9/2022 | WO |
