Dental restoration preparation

Abstract

A method for evaluating a dental preparation for a dental restoration is disclosed. A relevant point cloud is assembled from a dental preparation point cloud. A trained inference model, which may include a Dynamic Graph Convolutional Neural Network and a transformer-based encoder-decoder, predicts an inference cloud from the relevant point cloud. A dental preparation score is computed from the inference cloud. The score may include components such as scan quality, margin line, undercut, thickness, occlusal fit, taper, corner-rounding, gap analysis, contact area, gingival proximity, alignment, and abutment spacing. Inferences may be refined using spline fitting, outlier removal, and clearance analysis without necessarily generating a restoration mesh. The preparation may be classified as acceptable or not, and actions to improve the preparation may be suggested. A non-transitory computer-readable medium, a system, and a device configured to perform the method and display color-coded risk indicators are also disclosed.

Description

DENTAL RESTORATION PREPARATIONTECHNICAL FIELD

[0001] The present invention relates to dentistry techniques and, more particularly, to dental restoration preparation.BACKGROUND

[0002] Dental restoration treatments, such as dental crowns, implants, and dental bridges, are among the most common procedures in dentistry. Crowns are typically produced by dental technicians in dental laboratories based on specifications provided by the dentist. Dentists prepare teeth needing crown treatment by reshaping them and removing sections thereof. A die represents the prepared tooth that requires a crown restoration. In modern practice, technicians scan the die and use computer software to design a crown that fits the scanned die.

[0003] Similarly, dental implants involve the surgical placement of metal posts into the jawbone. These posts serve as a foundation for mounting artificial teeth. Dental bridges, used to bridge gaps left by missing teeth, consist of one or more artificial teeth (pontics) anchored to adjacent natural teeth or implants. The design and fabrication of dental implants and dental bridges also benefit from the use of computer-aided design to ensure proper fit and function.

[0004] Dental preparation is sometimes assisted by software to detect appropriate actions to prepare teeth for dental restoration. For example, margin lines may be detected to assist in avoiding undercuts when preparing a tooth for a crown restoration. However, due to variability of teeth among patients and differences in scan quality and anatomy, detected margin lines and related features may not always be sufficiently precise. Consequently, dental practitioners often need to manually verify and adjust outputs provided by digital tools before proceeding with fabrication, which may introduce variability and affect consistency and repeatability of the workflow.

[0005] Inconsistencies or suboptimal dental preparation can result in issues discovered later, upon execution of the dental restoration. When such situations arise, dentists must undertake a process to rectify the fit. The process typically begins with an assessment to pinpoint specific preparation flaws, such as undercuts or incorrect tapering, which may impede seating of the dental restoration. Adjustments are then made to the tooth, reshaping it to eliminate imperfections and create an acceptablebase for the restoration. In some cases, a new impression or scan may be necessary to fabricate an adjusted dental restoration that conforms to the corrected preparation.

[0006] Additionally, when an initial dental preparation is found to be incorrect or suboptimal, it is often necessary to schedule a follow-up appointment with the patient to allow the practitioner to make the necessary adjustments to the tooth preparation. The need for a new appointment can delay the overall treatment timeline and may contribute to patient inconvenience.

[0007] These inconveniences and corrective measures underscore the need for a proactive approach in evaluating preparation design at an early stage.SUMMARY

[0008] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

[0009] In a first aspect, the invention described herein relates to a method for evaluating a dental preparation for a dental restoration. A relevant point cloud is assembled from a dental preparation point cloud of the dental preparation. An inference cloud comprising inference points is predicted using a trained inference model and the relevant point cloud. A dental preparation score is computed from the inference cloud, thereby evaluating the dental preparation.

[0010] In embodiments, the relevant point cloud comprises at least one surface point on a surface of a tooth and at least one internal point internal to the tooth.

[0011] In embodiments, the relevant point cloud comprises at least one surface point on a surface of a tooth and at least one external point external to the tooth.

[0012] In embodiments, the relevant point cloud may be a decimated point cloud, and the dental preparation point cloud may be decimated into the decimated point cloud, thereby reducing the complexity of the relevant point cloud. The decimated point cloud may comprise between 250 and 15,000 points.

[0013] In embodiments, the dental preparation score may include a scan quality score. A dental preparation mesh of the dental preparation may be computed from adental preparation scan. The dental preparation mesh may be converted into the dental preparation point cloud. The scan quality score may be computed from the dental preparation scan. The scan quality score may include, a completeness score, and a registration score.

[0014] In embodiments, the relevant point cloud may be sampled using Furthest Point Sampling, thereby improving accuracy of the inference cloud.

[0015] In embodiments, the trained inference model may include a trained Dynamic Graph Convolutional Neural Network and a trained transformer-based encoderdecoder. A feature cloud may be computed by extracting a plurality of features from the relevant point cloud using the trained Dynamic Graph Convolutional Neural Network and the inference cloud may be predicted by processing the feature cloud using the trained transformer-based encoder-decoder.

[0016] In embodiments, the dental preparation score may be computed using an inference spline. The inference points may be ordered into an ordered point cloud. The inference spline may be generated from the ordered point cloud. The inference points may be ordered using a nearest neighbor algorithm. The inference spline may be a b- spline.

[0017] In embodiments, generating the inference spline may include computing a predicted inference spline by fitting the predicted inference spline to the ordered point cloud and adjusting one or more control points of the predicted inference spline, thereby generating the inference spline.

[0018] In embodiments, inference outliers may be discarded from the inference cloud, thereby improving an accuracy of the inference cloud. A local point density may be evaluated for a candidate point from the inference points by calculating an average distance therefrom to one or more nearest inference points. An alignment of the candidate point may be assessed with a main trend of one or more nearby inference points. An outlier score may be computed from the local point density and the alignment. The candidate point may be discarded when the outlier score is above a maximum outlier score threshold.

[0019] In embodiments, the dental preparation score may include a confidence score. The confidence score may be computed from a percentage of the inference outliers.

[0020] In embodiments, a visual representation of the dental preparation may be displayed, with areas communicating the confidence score. Actions to improve the confidence score may be suggested.

[0021] In embodiments, a dental restoration mesh prepared for the dental preparation may be obtained. The dental restoration mesh may be aligned with the dental preparation point cloud. Computing the dental preparation score may be performed by measuring an alignment between the inference cloud and the dental restoration mesh. The dental restoration mesh may be a generated dental restoration mesh generated from the preparation point cloud.

[0022] In embodiments, the dental preparation score may include a gap analysis score. The gap analysis score may be computed from an average deviation and a maximum deviation between a dental preparation surface and the dental restoration mesh. A restoration distance may be calculated between the dental restoration mesh and the inference cloud. Optionally, the restoration distance may include at least one of a Chamfer Distance (CD), a Hausdorff Distance, or an Earth Mover’s Distance (EMD).

[0023] In embodiments, the dental preparation score may include a contact area score. The contact area score may be computed from a percentage of the inference cloud in contact with the dental restoration mesh.

[0024] In embodiments, the dental preparation score comprises an occlusal fit score. A dental occlusion of the dental restoration mesh aligned onto the preparation point cloud may be simulated. The occlusal fit score may be computed from the dental occlusion. The occlusal fit score may include at least one of an occlusal clearance score, a contact points score, and an alignment score.

[0025] In embodiments, actions to improve the dental preparation score may be suggested. Optionally, the actions may include at least one of a tooth reduction, a tooth contouring, a margin definition, and a core buildup.

[0026] In embodiments, the dental restoration may include at least one of a crown, an implant, and a dental bridge.

[0027] In embodiments, the dental preparation score may include at least one of a margin line score, a gingival proximity score, an undercut score, an occlusal reduction score, an axial reduction score, a smoothness score, a thickness score, aninterproximal clearance score, an alignment score, a path of insertion score, a contact area score, and an abutment spacing score.

[0028] In embodiments, the trained inference model may be trained for detecting margin line points located on a margin line of the dental preparation and the dental preparation score may include a margin line score.

[0029] In embodiments, the trained inference model may be trained to predict undercut points lying within an undercut from the relevant point cloud and the margin line points and the dental preparation score may include an undercut score. Optionally, an undercut area from the undercut points may be generated. An undercut profile comprising an undercut height measured from the undercut area may be computed. The undercut score may be computed from the undercut profile.

[0030] In embodiments, the trained inference model may be trained to detect points located on a gingival margin of the dental preparation, and the dental preparation score may include a gingival proximity score.

[0031] In embodiments, the trained inference model may be a thickness model trained to predict thin points exhibiting insufficient thickness from the relevant point cloud, and the dental preparation score may include a thickness score. A thin surface from the thin points may be generated. A thickness profile may be computed from the thin surface. The thickness score may be computed from the thickness profile. The thickness score may be computed according to a property of a material used for the dental restoration.

[0032] In embodiments, when the dental restoration is a dental bridge, the trained inference model may be an abutment alignment model trained to detect alignment points describing an alignment of abutment teeth. The dental preparation score may include an alignment score. An alignment spline may be fitted through the alignment points. A curvature of the alignment spline may be evaluated against an alignment threshold. The alignment score may be computed from the curvature of the alignment spline.

[0033] In embodiments, when the dental restoration is a dental bridge, the trained inference model may be an abutment spacing model trained to detect abutment spacing points located between abutment teeth. The dental preparation score may include an abutment spacing score. An abutment spacing spline may be fitted through the abutment spacing points. A distance along the abutment spacing spline may becalculated to determine an available abutment spacing between abutment teeth. The abutment spacing score may be computed from the available abutment spacing.

[0034] In embodiments, when the dental restoration is a dental bridge, the trained inference model may be a gingival clearance model trained to detect gingival clearance points located in a space between an edge of the dental restoration and a gingival margin. The dental preparation score may include a gingival proximity score. A gingival clearance spline may be fitted through the gingival clearance points. A gingival clearance distance may be calculated along the gingival clearance spline to determine an available clearance. The gingival proximity score may be computed from the gingival clearance distance.

[0035] In embodiments, when the dental restoration is a dental bridge, the dental preparation score may include a load distribution score. Occlusal forces on the dental restoration mesh aligned onto the preparation point cloud may be simulated. A stress distribution assessment from the occlusal forces across abutment teeth and the dental bridge may be computed. The load distribution score from the stress distribution assessment may be computed. Optionally, the stress distribution assessment may be computed in consideration of a property of a material of the dental bridge.

[0036] In embodiments, when the dental restoration is a dental bridge, the dental preparation score may include an alignment score. An alignment spline may be computed from the dental restoration mesh aligned onto the preparation point cloud. The alignment score may be computed from the alignment spline.

[0037] In embodiments, when the dental restoration is a dental bridge, the dental preparation score may include an abutment space score. An available abutment space between abutment teeth from the dental restoration mesh aligned onto the preparation point cloud may be determined. The abutment space score may be computed from the available abutment space.

[0038] In embodiments, when the dental restoration is a dental bridge, dental preparation score may include a gingival proximity score. Gingival margin clearance may be simulated from the dental restoration mesh aligned onto the preparation point cloud. The gingival proximity score may be computed from the gingival margin clearance.

[0039] In embodiments, the dental preparation score may further be computed based on at least one clearance distance between the dental preparation andsurrounding anatomy, and the at least one clearance distance may be evaluated against at least one clearance threshold while generating an actual dental restoration mesh is unnecessary. A first-stage feasibility outcome may be computed from the at least one clearance distance. A second-stage outcome may subsequently be computed after generating an actual dental restoration mesh aligned with the dental preparation point cloud.

[0040] In embodiments, the dental preparation score may comprise a taper score. A path of insertion for a dental restoration may be determined. A taper angle of at least one axial wall of the dental preparation relative to the path of insertion may be determined. The taper angle may be evaluated against at least one taper threshold or taper range. The dental preparation score may further comprise a corner-rounding score. At least one transition region between an occlusal surface and an axial surface of the dental preparation may be identified. A sharpness metric at the at least one transition region may be evaluated against a minimum rounding threshold. The minimum rounding threshold may be based on a characteristic dimension of a milling tool used to fabricate the dental restoration.

[0041] In embodiments, the method may further comprise classifying the dental preparation as being within an acceptable range or outside the acceptable range based on at least one component of the dental preparation score. The method may be performed in at least one of a training mode and a practitioner mode, wherein at least one evaluation feature is enabled by default in the training mode and is user-activatable in the practitioner mode.

[0042] In a second aspect, the invention described herein relates to a non-transitory computer-readable medium storing a set of instructions for evaluating a dental preparation for a dental restoration. When executed by one or more processors of a device, the set of instructions causes the device to communicate a dental preparation score. Color-coded risk indicators may be displayed for at least one of an undercut, a thickness, and a margin line.

[0043] In embodiments, an assessment comprising removal of additional material to improve the dental preparation score may be provided. Display areas where material should be added may be displayed. Areas where material should be removed may be displayed.

[0044] In embodiments, the device may suggest a rescan of a highlighted dental preparation.

[0045] In embodiments, areas of contact and a contact score may be displayed.

[0046] In embodiments, the display areas where material should be removed may include surrounding teeth.

[0047] In a third aspect, the invention described herein relates to a system for evaluating a dental preparation for a dental restoration. The system may include a data acquisition module, a processing module and an evaluation module. The data acquisition module may assemble a relevant point cloud from a dental preparation point cloud of the dental preparation. The processing module may predict an inference cloud comprising inference points using a trained inference model and the relevant point cloud. The evaluation module may compute a dental preparation score from the inference cloud, thereby evaluating the dental preparation. The system may be configured to perform the method of the first aspect.

[0048] In a fourth aspect, the invention described herein relates to device for evaluating a dental preparation for a dental restoration. The device may include one or more processors and a display. The one or more processors may assemble a relevant point cloud from a dental preparation point cloud of the dental preparation, predict an inference cloud comprising inference points using a trained inference model and the relevant point cloud, compute a dental preparation score from the inference cloud, thereby evaluating the dental preparation. The display may be configured to show color- coded risk indicators for at least one of an undercut, a thickness, and a margin line. The device may be configured to perform the method of the first aspect.BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Further features and exemplary advantages of the present invention will become apparent from the following detailed description, taken in conjunction with the appended drawings, in which:

[0050] Figure 1 is an activity diagram of an exemplary method for evaluating a dental preparation for a dental restoration in accordance with the teachings of the present invention;

[0051] Figure 2 is an activity diagram of an exemplary method for assembling a relevant point cloud from a dental preparation in accordance with the teachings of the present invention;

[0052] Figure 3 is an activity diagram of an exemplary method for predicting an inference cloud from the relevant point cloud in accordance with the teachings of the present invention;

[0053] Figure 4 is an activity diagram of an exemplary method for discarding inference outliers in accordance with the teachings of the present invention;

[0054] Figure 5 is an activity diagram of an exemplary method for computing a dental preparation score from the inference cloud in accordance with the teachings of the present invention;

[0055] Figure 6 shows a logical modular representation of an exemplary system 2000 for evaluating a dental preparation for a dental restoration in accordance with the teachings of the present invention; and

[0056] Figures 7 A, 7B, 7C, 7D, 7E, 7F and 7G, hereinafter referred to as Figure 7, show exemplary three-dimensional representations of dental anatomy in accordance with the teachings of the present invention.DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

[0057] Certain preparation deficiencies may be identifiable prior to completing a restoration design. For example, inadequate clearance between a prepared tooth and adjacent teeth and / or opposing dentition may prevent fabrication of a restoration with sufficient thickness or may increase the likelihood of undesirable contacts. However, some workflows only identify such feasibility issues after additional steps have been performed, including generating a restoration design or transmitting the case for laboratory review.

[0058] Further, preparation quality may be affected not only by geometric fit considerations but also by fabrication and durability constraints. For example, sharp corners or abrupt transitions between preparation surfaces may be difficult or impossible to reproduce with common subtractive manufacturing tools and may also create stress concentrations that increase the risk of fracture of the restoration and / or the underlying tooth structure. Accordingly, a preparation evaluation that accounts formanufacturability constraints, including corner rounding, may improve clinical and fabrication outcomes.

[0059] Additionally, preparation evaluation is relevant in educational and training contexts. Certain approaches assess a student’s preparation by comparison to a reference preparation made by an instructor. Such approaches may be time-consuming and may yield variable outcomes due to differences between instructors and the practical need to produce and scan multiple reference preparations. There remains a need for objective and repeatable evaluation tools that provide actionable feedback, optionally by reference to a target preparation representation and / or by verifying compliance with a set of exclusion criteria.

[0060] Systems and methods that evaluate dental preparations using objective criteria are desirable. In some embodiments, the methods and systems comprise (i) an early feasibility assessment that can be performed without generating, obtaining, or aligning a restoration mesh, and (ii) a more detailed assessment that may incorporate an obtained or generated restoration mesh, thereby reducing rework, improving efficiency, and enhancing consistency.

[0061] In some embodiments, the early feasibility assessment is configured to be less computationally intensive than the more detailed assessment. Broadly, the systems and methods described herein late to the evaluation of dental preparations for the creation of dental restorations, such as crowns, bridges, and implants. The evaluation may involve the assembly of a point cloud and may use a trained inference model o extract features and predict the inference cloud.

[0062] Reference is now made to the drawings and concurrently to Figure 1 and Figure 7. Figure 1 depicts an exemplary method 1000 for evaluating a dental preparation for a dental restoration in accordance with the teachings of the present invention. Figure 7 provides exemplary three-dimensional representations from various stages of the method 1000. For example, Figure 7A illustrates a three-dimensional representation of a dental arch 7010 in which a prepared tooth 7020 is visually distinguished from surrounding anatomy. Figure 7B illustrates a surface mesh representation 7030 of a dental arch 7010 including a prepared tooth 7020. Figure 7C illustrates a representation of a prepared tooth 7020. Figures 7D, 7E and 7F illustrate further exemplary representations usable for alignment, segmentation, and margin-related evaluations, and Figure 7G illustrates a representation including a restoration 7040 positioned on a prepared tooth 7020.

[0063] A first aspect of the techniques described herein relates to a method 1000 for evaluating a dental preparation for a dental restoration. The method may include assembling 1100 a relevant point cloud 100 from a dental preparation point cloud 101 of the dental preparation; predicting 1200 an inference cloud 102 comprising inference points 103 using a trained inference model 104 and the relevant point cloud 100; and computing 1300 a dental preparation score 105 from the inference cloud 102, thereby evaluating the dental preparation.

[0064] The dental restoration may include, for example, a crown, an implant, and / or a dental bridge. Broadly, the dental restoration discussed herein may refer to a prosthetic or corrective dental structure that may be designed to restore function, integrity, and aesthetics to a tooth or group of teeth

[0065] A crown may be described as a type of dental restoration that may fully encase the visible portion of a tooth. Crowns may be used to restore a tooth suffering from significant structural damage or decay, potentially providing protection and support while restoring shape, size, and appearance. Crowns may be manufactured from various materials, such as porcelain, metal, or ceramic, and may be designed to fit precisely over the prepared tooth, thereby ensuring a secure and functional restoration. A dental bridge may serve as a restoration to replace one or more missing teeth. Dental bridges may comprise artificial teeth (pontics) anchored to adjacent natural teeth or implants (abutments), thereby bridging the gap left by missing teeth. Dental bridges may restore the functionality of the dental arch, potentially allowing for normal chewing and speaking, and may help maintain the alignment of adjacent teeth by preventing them from shifting into the gap.

[0066] While particularly applicable to crowns and dental bridges, the method described herein may also be adapted for use with a wide range of other dental restorations. The range of dental restorations may include, but is not limited to, implants, inlays, onlays, and veneers. Implants may serve as artificial roots for supporting crowns or bridges and may require precise alignment with the jawbone and surrounding teeth. Inlays and onlays may be types of restorations used to repair damaged or decayed areas within the cusps of teeth or on their surfaces, respectively.Veneers may be thin shells placed over the front surface of teeth to improve aesthetics and function.

[0067] The method may be used to assess the readiness and suitability of a tooth or dental structure to receive a restorative dental appliance. Evaluating, in this context, may encompass the analysis and examination of various aspects of the dental preparation to ensure that it meets predefined criteria necessary for successful restoration. Evaluation may include checking for proper shape, alignment, margin lines, and the absence of undercuts or other irregularities that may impede the proper fitting of a dental restoration.

[0068] In embodiments, evaluating a dental preparation comprises an intrinsic validity assessment performed directly on the dental preparation point cloud and / or the relevant point cloud. The intrinsic validity assessment may verify whether one or more exclusion criteria are satisfied, including, for example, whether the preparation exhibits one or more undercuts, whether a margin line is sufficiently identifiable, whether clearance to surrounding anatomy is sufficient to permit fabrication of a restoration having at least a minimum thickness, and whether one or more transition regions exhibit excessive sharpness.

[0069] In embodiments, evaluating a dental preparation comprises comparison of the dental preparation to a target preparation representation. As used herein, a “target preparation representation” may refer to a computed reference associated with a preparation objective and is not limited to a single ideal geometry. In embodiments, the target preparation representation comprises an acceptable preparation envelope defining a range of acceptable geometries, thereby enabling classification of the preparation as being within an acceptable range (safe zone) or outside an acceptable range (out-of-bounds), and optionally as being within the acceptable range while still improvable.

[0070] The process of dental preparation for this method may involve reshaping and conditioning a tooth or dental structure to create a foundation suitable for a dental restoration. This process may include removing decayed or damaged areas and shaping the remaining structure to provide a preparation surface having appropriate taper, clearance, and smoothness for adherence and support of the restoration. Dental restoration, in this context, may involve repairing or replacing parts of a tooth to restore function, integrity, and morphology. These restorations may include crowns, bridges,implants, and other prosthetic devices designed to replace missing or damaged tooth structures. The goal of the dental preparations discussed herein may be to restore the appearance and function of a tooth, thereby supporting normal chewing, speaking, and overall oral health.

[0071] Broadly, a point cloud, such as the dental preparation point cloud 101 or the relevant point cloud 100, may be understood as a collection of data points defined in a three-dimensional coordinate system, which typically, but not necessarily, represents the external surface of an object, such as a dental structure. Each point in the point cloud may have associated X, Y, and Z coordinates, providing spatial information about the surface it represents. In addition to spatial coordinates, points in a point cloud may include additional attributes such as density, material, and reliability. Initially, these point clouds may be generated through 3D scanning technologies, such as laser scanners or structured light scanners. In some embodiments, scan data may initially be represented as a surface mesh and converted to a point cloud representation for processing. By way of example, Figure 7B illustrates a surface mesh representation 7030 of a dental arch 7010 that may be converted into the dental preparation point cloud 101 and / or used to assemble the relevant point cloud 100.

[0072] The dental preparation point cloud 101 discussed herein may refer to a point cloud representing the three-dimensional surface of a prepared tooth or dental structure. The dental preparation point cloud 101 may include detailed information about the contours, angles, and features of the tooth that has been reshaped to receive a dental restoration. The dental preparation point cloud 101 may be a subset of points, focusing on the dental surfaces that have been modified for restorative purposes. By way of example, Figure 7 A illustrates a dental arch representation 7010 in which a prepared tooth is visually distinguished from surrounding anatomy, which may correspond to identifying a subset of scan data 7035 to be represented as the dental preparation point cloud 101 and / or used to assemble the relevant point cloud 100.

[0073] The process of assembling 1100 the relevant point cloud 100 from the dental preparation point cloud 101 discussed herein may involve selecting and organizing specific data points from the dental preparation point cloud 101 that are pertinent to the evaluation process. The assembly may include filtering, decimating, or otherwise manipulating the data to ensure that the resulting relevant point cloud 100 accurately and efficiently represents the important features needed for subsequent analysis. Assembling the relevant point cloud 100 from the dental preparation point cloud 101may involve a reduction in data complexity and a concentration of computational efforts on the most significant aspects of the dental preparation. As an illustrative example, the relevant point cloud 100 may be assembled from a three-dimensional representation such as the mesh of Figure 7B and / or from a representation in which the prepared tooth 7020 is identified in context, such as in Figure 7A. In embodiments, assembling the relevant point cloud includes selecting points corresponding to a prepared tooth 7020 and optionally selecting additional points corresponding to surrounding anatomy 7060 to support subsequent evaluations.

[0074] The relevant point cloud 100 may include surface points on the surface of a tooth, internal points 107 internal to the tooth, and / or external points 207 that may capture spatial relationships and features surrounding the tooth. Surface points may refer to data points representing the external geometry of the tooth, capturing contours, edges, and overall shape to provide information about visible structure. Internal points 107 may refer to data points representing internal aspects of the tooth, potentially derived from imaging techniques that penetrate beyond the surface, such as CBCT scans or similar volumetric imaging technologies. Internal points 107 may offer insight into the internal architecture of the tooth, potentially revealing features such as pulp chambers, root canals, or underlying structures not apparent from the surface alone. Internal points 107 may be used in assessing the depth and available material of the tooth, thereby informing whether additional material may be safely removed during preparation without compromising structural integrity and tooth viability. External points 207 may capture positioning of adjacent teeth, gingival contour, and interproximal spaces. External points 207 may provide context for understanding how the prepared tooth fits within the overall dental arch and interacts with neighboring anatomical structures. External points 207 may be useful in assessing spacing for restorations, ensuring sufficient clearance and alignment with adjacent teeth to prevent interference during placement. Labeled external points 207 of the tooth surroundings may aid in defining margin locations more accurately by referencing the gingival architecture and landmarks. Using external points 207 may support the design of dental restorations that accommodate both functional performance and aesthetic integration within the oral cavity.

[0075] In some embodiments, the external points 207 are used to perform an early feasibility assessment without generating a restoration mesh. For example, distances between a prepared tooth surface and one or more adjacent tooth surfaces may becomputed to determine interproximal clearance, and distances between the prepared tooth surface and an opposing dentition surface may be computed to determine occlusal clearance. The computed clearances may be evaluated against one or more thresholds associated with feasibility of fabricating a restoration having at least a minimum thickness and / or associated with avoidance of undesirable contacts. Figure 7C illustrates an example representation 7035 suitable for such clearance-based feasibility evaluations prior to generating or aligning a dental restoration mesh 7030, including evaluation of clearance relative to adjacent teeth 7080 and / or opposing dentition 7090.

[0076] In embodiments, the relevant point cloud 100 may be a decimated point cloud 108, wherewith assembling 1100 the relevant point cloud 100 may involve decimating 1110 the dental preparation point cloud 101 into the decimated point cloud 108. A decimated point cloud 108 may refer to a point cloud that has been processed to reduce the number of data points while preserving the essential geometric features of the original structure. The reduction in data points may be achieved by selectively removing points that contribute the least to the overall shape and form, thereby simplifying the dataset without significantly compromising its accuracy or representation. Decimating 1110 may involve the process of systematically reducing the number of points in the dental preparation point cloud 101 , accomplished through algorithms designed to maintain the integrity and detail of relevant features while eliminating redundant or non-essential data points. The purpose of decimating may be to create a more manageable and efficient dataset that retains the essential characteristics of the original point cloud needed for subsequent analysis and processing.

[0077] Reducing complexity by decimating the point cloud may lead to a more streamlined and efficient workflow, potentially allowing for faster processing and analysis while maintaining the necessary detail for accurate evaluation of the dental preparation. A reduction in complexity may facilitate the use of the point cloud in realtime or near real-time applications and enhance the overall efficiency of the evaluation process. The selection of points according to relevancy and the filtering of outliers may also contribute to an increase in the reliability of method 1000.

[0078] In embodiments, the decimated point cloud 108 may include between 250 and 15,000 points. It was found through experimentation that a range of 250 to 15,000 points may balance detail and manageability within the point cloud. A decimated pointcloud 108 within the range of 250 to 15,000 points may retain sufficient detail to accurately represent the geometric features of the dental preparation, while being reduced enough to ensure efficient processing and analysis. Using between 250 and 15,000 points may be sufficient in most situations for preserving relevant features necessary for evaluating the dental preparation, such as contours, edges, and key structural aspects, without overwhelming computational resources. Persons skilled in the art will readily recognize that using fewer than 250 points, or more than 15,000 points may also be valid according to specific situations and embodiments. For example, a point cloud covering only one tooth may require fewer points, while a point cloud covering several may require more. As computing power increases and becomes more affordable, embodiments may be capable of managing higher complexity without additional costs, potentially pushing the range of recommended number of points being processed upward

[0079] In embodiments, assembling 1100 the relevant point cloud 100 may include sampling 1120 the relevant point cloud 100 using Furthest Point Sampling (FPS) to improve the accuracy of the inference cloud 102. Furthest Point Sampling (FPS) is a technique used to select a subset of points from a larger point cloud. The primary goal of FPS is to ensure that the selected points are distributed evenly across the entire surface of the object represented by the point cloud. The method may operate by iteratively selecting the point that is furthest away from the set of already selected points, starting from an initial random point. FPS may help in capturing the overall geometry of the object by maintaining a uniform distribution, which is particularly useful for reducing the complexity of the point cloud while preserving its essential features. Sampling 1120 in this context may refer to the process of selecting a representative subset of points from the relevant point cloud 100. The process of sampling may be important for managing data size and computational load, as it reduces the number of points that need to be processed while retaining the geometric integrity of the original point cloud. Sampling 1120 using techniques like FPS may ensure that the most informative points are chosen, improving accuracy. The accuracy, in this context, may refer to the degree to which the inference cloud 102 represents the true geometry and features of the dental preparation.

[0080] From the relevant point cloud 100, an inference cloud 102 of inference points 103 may be predicted 1200 using a trained inference model 104. In this context, a trained inference model 104 may refer to a machine learning model developed andoptimized for specific predictive tasks related to dental preparations. Such a model may be trained using a dataset comprising examples of dental structures and their corresponding features, enabling the model to learn patterns and make accurate predictions on new data. The trained inference model 104 may include architectures such as a Dynamic Graph Convolutional Neural Network (DGCNN) and a transformerbased encoder-decoder, which are designed to extract meaningful features from the complexities of the relevant point cloud 100 for evaluation purposes. In embodiments, the trained inference model 104 may be trained to identify points from the relevant point cloud 100 located on features of interest, such as a margin line. The trained inference model 104 may, in some embodiments, be a collection of multiple models trained or configured for different tasks. For example, a margin line model may be tuned or trained specifically to recognize margin lines, while an undercut model may be tuned or trained specifically to recognize undercut areas. Together, these specialized models are referred to herein as the trained inference model 104. Each specialized model may use or combine different techniques; for example, an undercut model may be mostly based on reinforcement machine learning, while an undercut model may rely on applying thresholds to geometrical properties. Combinations of machine learning and geometrical processing, involving one or several steps such as pre-processing or postprocessing, may be considered for each of the specialized models.

[0081] Predicting 1200 may involve the process of using the trained inference model 104 to analyze the relevant point cloud 100 and generate an output known as the inference cloud 102. During prediction 1200, the model may apply its learned knowledge to assess the input data, identifying key features and generating predictions based on the patterns it has been trained to recognize.

[0082] The inference cloud 102 may be the collection of inference points 103 generated by the prediction 1200 of the trained inference model 104. The inference points 103 may represent the predictions of the trained inference model 104 about specific features or characteristics of the dental preparation, such as margin lines, undercut areas, or other aspects required for dental restoration. Each inference point may provide additional information derived from the analysis of the trained inference model 104, contributing to the overall understanding of the condition and suitability of the dental preparation.

[0083] The inference cloud 102 may represent a synthesized view of the dental preparation, highlighting areas of interest and potential concerns identified by themodel. The inference cloud 102 may serve as a basis for computing the dental preparation score 105, providing a comprehensive overview of the readiness of the preparation for restoration and facilitating the identification of necessary adjustments or improvements.

[0084] In one embodiment, the trained inference model 104 may include a trained Dynamic Graph Convolutional Neural Network (DGCNN) and a trained transformerbased encoder-decoder. Predicting 1200 the inference cloud 102 may involve computing 1210 a feature cloud 114 by extracting a plurality of features 115 from the relevant point cloud 100 using the trained DGCNN, and predicting 1220 the inference cloud 102 by processing the feature cloud 114 using the trained transformer-based encoder-decoder.

[0085] Dynamic Graph Convolutional Neural Network (DGCNN) refers to a type of neural network architecture specifically designed to handle data represented in the form of graphs, which may be particularly suitable for processing point clouds. Unlike traditional convolutional neural networks that operate on structured grid data, such as images, a DGCNN may dynamically compute the connectivity of points in a point cloud, treating them as nodes in a graph. The edges of the graph formed by these nodes may be determined based on spatial proximity or feature similarity, allowing the network to capture local geometric relationships. During training, a DGCNN may learn to extract features by aggregating and transforming information from neighboring points, effectively capturing both local and global structures in the data.

[0086] The extracted features may refer to the distinctive attributes or patterns within the point cloud data that characterize the geometric and spatial properties of the dental preparation. Features may include aspects such as curvature, edges, surface normals, and local shape descriptors, which may be relevant for understanding the structure and form of the dental preparation. Extracting the plurality of features 115 involves the process of identifying and isolating these relevant attributes from the point cloud. The DGCNN may perform the extraction by dynamically constructing graphs based on the spatial relationships between points and then applying convolutional operations to the constructed graphs. The DGCNN may then aggregate information from neighboring points, thereby capturing both local and global features that define the geometric structure of the dental preparation. The feature cloud 114 may refer to a transformed representation of the original point cloud, where each point is enriched with extracted features. Instead of merely representing spatial coordinates, the featurecloud 114 may include additional dimensions that encode the learned attributes, providing a more comprehensive representation of the data. The enriched dataset serves as an input for further processing and analysis, enabling more accurate predictions and evaluations. Hence, computing 1210 the feature cloud 114 may involve the application of the trained DGCNN to the relevant point cloud 100, wherewith the network processes the input data to generate the feature cloud 114.

[0087] A trained transformer-based encoder-decoder is a neural network architecture initially popularized in natural language processing but increasingly applied to various domains, including point cloud processing. Transformers rely on attention mechanisms to model relationships within the data, allowing them to capture complex dependencies between elements. In the context of point clouds, a transformer-based encoder-decoder may encode the input data into a latent representation through the encoder, then decode the latent representation to predict desired outputs, such as an inference cloud 102. The encoder may focus on capturing the essential features and patterns from the input, while the decoder may reconstruct or generate the output based on the learned latent representation.

[0088] The transformer-based encoder-decoder may process the feature cloud 114 by first encoding it into an intermediate representation through the encoder component. The encoder may leverage attention mechanisms to capture complex dependencies and interactions between the extracted features, allowing for modeling of both local and global patterns. The encoded representation, conceived as a condensed and structured form of the input feature cloud 114, may then be passed to the decoder component. The decoder may use the encoded information to generate and predict 1220 the inference cloud 102 with inference points 103 that reflect the predicted properties or features of the dental preparation.

[0089] Using a DGCNN and a transformer-based encoder-decoder may be advantageous for handling complex relationships within the data. The DGCNN may focus on extracting features from the relevant point cloud 100 through dynamic graphbased operations, while the transformer-based encoder-decoder may further process these features to generate predictions informed by a deeper understanding of the structure and dependencies of the data. A layered approach may enhance the accuracy and robustness of the predictions.

[0090] In one embodiment, predicting 1200 the inference cloud 102 may include discarding 1230 inference outliers 120 from the inference cloud 102 to improve the accuracy of the inference cloud 102. For example, removing outliers may involve assessing the local point density 121 around each candidate point 122 within the inference cloud 102. Discarding 1230 the inference outliers 120 may include evaluating 1240 a local point density 121 for a candidate point 122 from the inference points 103 by calculating an average distance to one or more nearest inference points 103. An alignment of the candidate point 122 with a main trend of one or more nearby inference points 103 may then be assessed 1250. An outlier score 123 from the local point density 121 and the alignment may be computed 1260. Then, the candidate point 122 may be discarded 1270 when the outlier score 123 is above a maximum outlier score threshold 124.

[0091] The local density may be evaluated 1240 using the average distance from the candidate point 122 to its nearest neighboring inference points 103. Points situated in regions of low density compared to their surroundings may be flagged as potential outliers, as they might not conform to the expected distribution of the data. Another criterion for identifying outliers may be to assess 1250 the alignment of a candidate point 122 with the main trend of nearby inference points 103. The assessment 1250 of the alignment may involve evaluating whether the candidate point 122 fits within the general orientation or pattern established by its neighboring points. Points that deviate significantly from the established trend may be considered outliers. An outlier score 123 may be computed 1260 for each candidate point 122, based on factors such as the local point density 121 and alignment assessments. The outlier score 123 may quantify the likelihood that a point is an outlier, with higher scores indicating a greater probability of being an outlier. Using the computed outlier scores 123, a threshold may be established to determine which points should be discarded. Candidate points 122 with outlier scores 123 exceeding the established threshold may be removed from the inference cloud 102. The threshold may be set to balance the removal of true outliers with the retention of legitimate data points, minimizing the risk of erroneously discarding relevant information. The threshold may be fixed or adjusted according to specific situations. For example, a threshold may be chosen such that the number of remaining points is above a certain value, such as 250 points. The threshold may also be adjusted in different areas of the cloud points. For example, when features have been identifiedor density is uneven, it may be advantageous to provide different thresholds for different regions of the cloud.

[0092] The dental preparation score 105 may include a confidence score 125 computed 1280 from a percentage of the inference outliers 120. Once outliers have been identified and discarded based on criteria such as local point density 121 and alignment with the main trend, the percentage of outliers relative to the total number of points in the inference cloud 102 may be calculated. The calculated percentage reflects the extent to which the initial predictions contained deviations or anomalies. The confidence score 125 may then be derived from this calculated outlier percentage. A lower percentage of outliers generally indicates a higher confidence score 125, suggesting that the inference cloud 102 is more reliable and accurate. Conversely, a higher percentage of outliers may result in a lower confidence score 125, indicating potential issues with the accuracy or reliability of the predictions. In situations where the threshold to qualify a point is adaptive, such as according to regions of the cloud, the computation 1280 of the confidence score 125 may be performed based on a fixed threshold rather than the number of discarded outliers. The confidence score 125 may serve as an indicator of the quality and trustworthiness of the predictions of the inference cloud 102 and may be provided with the final report to the dental practitioner to nuance the assessment of the results.

[0093] A visual representation 126 of the dental preparation may be displayed 1910 along with the confidence score 125. Actions to improve the confidence score 125 may also be suggested 1920. The visual representation 126 may refer to a graphical depiction of the three-dimensional structure of the prepared tooth or dental area, as captured and processed from the point cloud data. The depiction may include the detailed geometry of the tooth, highlighting contours, edges, and any significant features pertinent to the dental preparation. The visual representation 126 may assist dental practitioners by providing an intuitive and comprehensive view of the preparation, enabling assessment and decision-making.

[0094] Areas communicating the confidence score 125 within the visual representation 126 may be specific regions of the dental preparation highlighted or annotated to reflect the reliability of the predictions made regarding those areas. The visual representation 126 may be achieved through visual cues including color coding, shading, grayscale intensity, stippling, hatching patterns, contour lines, labels, and / or other annotations, indicating different levels of confidence. For example, areas withhigh confidence may be displayed using a first visual indicator, while areas with lower confidence may be displayed using a second visual indicator distinct from the first, thereby signaling potential inaccuracies or areas requiring further inspection or refinement (e.g., in green, indicating reliability and consistency with the expectations of the model, while areas with lower confidence may be shown in red or another distinct color).

[0095] Displaying 1910 the information may be achieved through the use of a printed report or visualization software that processes the point cloud and associated data to render a three-dimensional model of the dental preparation. The software may integrate the confidence score 125 data to dynamically adjust the visual representation 126, ensuring that users can easily interpret the reliability of different areas. Technologies such as 3D rendering engines, graphical user interfaces (GUIs), and augmented reality (AR) systems may be employed to create an interactive and informative display.

[0096] Suggesting 1920 actions to improve the confidence score 125 may include providing specific recommendations or interventions that can enhance the reliability and accuracy of the predictions made by the inference model. Suggested actions may address factors contributing to lower confidence scores 125, thereby improving the overall quality and certainty of the dental preparation analysis. For example, suggestions may include acquiring additional scan data or improving the quality of existing scans to fill in gaps or increase point density in areas with lower confidence. Rescanning the dental preparation with higher resolution or adjusting the scanning technique to capture more comprehensive data may also be suggested 1920.

[0097] Recommendations might involve tuning the parameters of the inference model 104 to better accommodate the specific characteristics of the dental preparation. Tuning the inference model 104 may include adjusting thresholds for outlier detection or modifying the parameters used in the fitting of splines. In areas where the predictions of the inference model 104 are uncertain, manual inspection and correction by dental professionals may be advised. If certain regions of the dental preparation consistently exhibit low confidence, targeted data acquisition strategies may be suggested, such as focusing scanning efforts on problematic areas or using complementary imaging techniques to capture additional details.

[0098] From the inference cloud 102, the evaluation of the dental preparation may be achieved by computing 1300 a dental preparation score 105. Computing 1300 the dental preparation score 105 from the inference cloud 102 may involve deriving a comprehensive metric that evaluates the readiness and suitability of a dental preparation for restoration. The dental preparation score 105 may be calculated using the inference points 103 within the inference cloud 102. The dental preparation score 105 may include various sub-scores or components that assess different aspects of the preparation. For example, the dental preparation score 105 may include a margin line score 141 , an undercut score 143, a thickness score 147, and an alignment and fit score. Each sub-score may assess how well the preparation aligns with the intended restoration and how effectively it fits, considering occlusal and contact points. The dental preparation score 105, as a whole, may be used to provide a holistic assessment of the suitability of the preparation for restoration, highlighting areas that may require adjustment or further preparation.

[0099] In embodiments, the dental preparation score 105 comprises a first-stage feasibility outcome computed without generating a dental restoration mesh. The first- stage feasibility outcome may be computed from one or more clearance distances and / or exclusion criteria and may provide an early indication that a restoration cannot be fabricated or seated satisfactorily without additional preparation.

[0100] In embodiments, the dental preparation score 105 further comprises a second-stage outcome computed using an obtained or generated dental restoration mesh 127. The second-stage outcome may provide a more precise evaluation than the first-stage feasibility outcome, including by measuring alignment, gaps, contact areas, and / or occlusal interactions between the dental restoration mesh 127 and the dental preparation. By way of example, Figure 7D illustrates an exploded view in which a restoration geometry is positioned relative to a prepared tooth and an insertion direction is indicated for seating and alignment.

[0101] In embodiments, the dental preparation score 105 may include a scan quality score 109. Broadly, a dental preparation mesh 110 of the dental preparation may be computed 1010 from a dental preparation scan 111 , the dental preparation mesh 110 may be converted 1020 into the dental preparation point cloud 101 , and the scan quality score 109 may be computed 1030 from the dental preparation scan 111.

[0102] The dental preparation mesh 110 may be generated from the raw scan data and then converted into a dental preparation point cloud 101. The transformation may be achieved by applying various computational techniques such as decimation and sampling. The scan quality score 109 may include several components, such as a completeness score 112 and a registration score 113. The completeness score 112 may assess how thoroughly the scan captures the entire surface of the dental preparation, evaluating whether there are any missing areas or gaps in the scan that could affect the accuracy of the restoration design. The registration score 113 may evaluate how well the scan aligns with predetermined reference points or models, ensuring that the scan accurately represents the actual geometry of the dental preparation without distortion or misalignment.

[0103] In one embodiment, the completeness score 112 may be computed by performing an analysis of point density across the scanned surface. The analysis may involve assessing the distribution of points in the point cloud to identify areas that are sparsely populated or missing. A uniform point density across the surface may suggest a complete scan, while areas with lower density may indicate missing or incomplete data. Another approach might involve comparing the scanned surface with a reference model or expected geometry by aligning the scanned point cloud with a digital model representing the ideal or expected shape of the dental preparation. Differences in surface coverage, such as gaps or missing regions, are quantified to determine completeness. The score may also be computed by detecting and analyzing the boundaries of the scanned area. If the boundaries of the scan align well with the expected limits of the dental preparation, it suggests completeness. Conversely, if boundaries are irregular or incomplete, it may indicate that some parts of the preparation were not captured. Specific quantitative metrics, such as the percentage of surface area covered or the number of missing regions beyond a certain size threshold, could be used to numerically express completeness. These metrics may be derived from surface comparisons or point density analyses. Various algorithms may be employed to automatically detect and flag gaps or holes in the point cloud. These algorithms might use techniques such as surface triangulation to identify discontinuities in the mesh derived from the point cloud. The completeness score 112 may also involve validating the scan against known anatomical features of the tooth, where missing or partially captured key features may negatively impact the score.

[0104] The registration score 113 may be computed from an initial alignment of the scanned point cloud with a reference model or standard geometry. The initial alignment may use common features or anatomical landmarks to establish a rough positioning, providing a starting point for more refined registration. In one embodiment, an Iterative Closest Point (ICP) algorithm may be used to refine the alignment, iteratively refining the alignment by minimizing the distance between corresponding points in the scanned point cloud and the reference model. The registration score 113 may include error metrics that quantify the deviation between the aligned scan and the reference model. Metrics such as the root mean square error (RMSE) or mean absolute error (MAE) of point-to-point or point-to-surface distances may provide a numerical measure of registration accuracy. The score may assess both global alignment, which considers the overall fit of the scan to the reference model, and local alignment, which may focus on specific regions or features. Discrepancies in either may affect the registration score 113. The score may also involve a feature-based analysis, where specific anatomical landmarks or features are identified in both the scan and the reference model. The alignment of these features may be evaluated to ensure they match accurately, contributing to the overall registration score 113. In embodiments, visual inspection tools may be used to assess the quality of the alignment, providing qualitative feedback that complements quantitative metrics. Color maps or overlays may highlight areas of misalignment, offering insights into the registration quality.

[0105] In one embodiment, the dental preparation score 105 may be computed 1300 using an inference spline 116. Computing 1300 the dental preparation score 105 may include ordering 1310 the inference points 103 into an ordered point cloud 117 and generating 1320 the inference spline 116 from the ordered point cloud 117. Ordering 1310 the inference points 103 into an ordered point cloud 117 may involve organizing the individual inference points 103 in a systematic sequence or structure. An ordered point cloud 117 may refer to a point cloud where the points are arranged in a coherent order, often based on spatial proximity or connectivity, rather than being randomly distributed. The ordering 1310 may, for example, involve algorithms that determine the closest neighboring points and establish a path or sequence that connects these points logically. Ordering 1310 the inference points 103 using a nearest neighbor algorithm may be achieved by arranging the points in the inference cloud 102 based on their spatial proximity to one another. The nearest neighbor algorithm systematically selects and sequences points by continually choosing the point closestto the last selected point. A path or order through the point cloud that reflects the natural structure and continuity of the surface being represented may be created. Employing a nearest neighbor algorithm may ensure that the inference points 103 are organized in a way that facilitates the generation of continuous structures, such as splines.

[0106] Although a nearest neighbor approach may be straightforward, computationally efficient, and generally effective, other ordering 1310 methods may also be considered. For example, Traveling Salesman Problem (TSP) algorithms may provide an optimal ordering by minimizing the total path length, though they are computationally intensive and may not be practical for large point clouds typical in dental applications. Greedy algorithms may also be considered, as they are generally simple and can be effective. Graph-based approaches may be employed to capture complex relationships within the point cloud, such as in cases with non-uniform point distribution. Clustering techniques may be used in conjunction with ordering 1310 to organize points in a preprocessing step or iteratively as ordering 1310 is being performed, organizing the points into coherent groups.

[0107] Ordering 1310 may facilitate the generation of continuous structures, such as splines, by ensuring that the points are arranged in a manner that reflects the underlying geometry of the dental preparation. Generating 1320 the inference spline 116 from the ordered point cloud 117 may involve the creation of a smooth, continuous curve or surface that approximates the shape represented by the ordered point cloud 117. A spline, in a broad sense, is a mathematical function used to create smooth and flexible curves or surfaces through a set of given points.

[0108] Splines are often used in computer graphics, data interpolation, and geometric modeling due to their ability to produce visually appealing and mathematically sound representations of complex shapes. An inference spline 116 may be a spline generated from the inference points 103 that have been ordered into an ordered point cloud 117. The inference spline 116 serves as a refined representation of the predicted features of the dental preparation, capturing the essential geometric characteristics in a smooth and continuous form. The inference spline 116 may then be used to compute 1300 the dental preparation score 105, as it provides a clear and precise depiction of important features such as margin lines or contours that are relevant for evaluating the readiness and suitability of the dental preparation for restoration.

[0109] In one embodiment, the inference spline 116 may be a B-spline. A B-spline, or basis spline, is a type of spline composed of a series of polynomial segments that are smoothly connected to form a continuous curve. B-splines are defined by a set of control points and a degree that determines the polynomial order of the spline segments and a knot sequence. One of the key characteristics of B-splines may be their flexibility and ability to accurately model complex shapes and curves with a high degree of smoothness. The control points may influence the shape of the spline, but unlike some other types of splines, a B-spline may not necessarily pass through these control points, allowing for greater control over the smoothness and shape of the curve. Modeling the points using a B-spline may be appropriate for several reasons. Firstly, the ability of a B-spline to create smooth and continuous curves may be advantageous for modeling the intricate surfaces and contours of dental structures. Smoothness may be particularly significant in accurately representing features such as margin lines, thereby ensuring the proper fit and function of dental restorations. Additionally, the flexibility of B-splines may allow for adjustments to be made easily by manipulating the control points, providing dental professionals with the ability to refine the model as needed.

[0110] Other examples of splines, such as NURBS (Non-Uniform Rational B- Splines), Catmull-Rom splines, or Bezier splines, are each defined by a set of control points and offer varying degrees of adjustment and control over the geometry and smoothness of the modeled segments. The choice of spline may depend on the specific requirements of the modeling task, such as the need for smoothness, the ability to adjust the curve, and the complexity of the dental surface being represented.

[0111] In one embodiment, a predicted inference spline 118 may be computed by fitting the predicted inference spline 118 to the point cloud, and one or more control points 119 of the predicted inference spline 118 may be adjusted to generate the inference spline 116. Fitting the predicted inference spline 118 to the point cloud may refer to the process of creating a spline that conforms to the shape and structure represented by the ordered point cloud 117 or a subset thereof. Fitting may involve using mathematical techniques to determine a spline curve that best approximates or represents the distribution and arrangement of points in the point cloud. The fitting process may capture the essential geometric features of the data, such as contours and boundaries, by minimizing the distance between the spline and the points in the cloud. Once a predicted spline has been fitted to a subset of points from the point cloud, the spline may be manipulated by adjusting its control points. Adjusting one or morecontrol points 119 of the predicted inference spline 118 involves modifying the position or influence of specific points that define the shape of the spline. Control points determine the curvature and path of the spline. By adjusting the control points, the shape of the spline may be refined to better fit the desired features or contours of the dental preparation. The adjustment may be performed according to known features or manually by dental professionals.

[0112] Dental professionals may interactively modify the spline to achieve an optimal representation of the dental preparation, enhancing the accuracy and reliability of the subsequent evaluation and restoration processes. The adjustment may be performed freely by the dental professionals, or with some assistance or constraints from the point cloud and / or known dental features. For example, as the spline is being adjusted, control points may snap to ideal positions, or other control points may be refitted from the point cloud to improve the overall fitting of the spline. In embodiments, dental professionals may add or remove control points to adjust the resolution of the spline or have it conform to intricate features.

[0113] In one embodiment, a dental restoration mesh 127 prepared for the dental preparation may be obtained 1410 and aligned 1420 with the dental preparation point cloud 101. Computing 1300 the dental preparation score 105 may then be performed by measuring alignment between the inference cloud 102 and the dental restoration mesh 127.

[0114] The dental restoration mesh 127 is a three-dimensional digital model representing the intended dental restoration, such as a crown, bridge, or implant, designed to fit the prepared tooth or dental structure. The dental restoration mesh 127 may be generated using computer-aided design (CAD) software, based on the specifications and geometry of the dental preparation, to ensure a precise fit and optimal function of the restoration. In some cases, when the dental restoration is available, the dental restoration mesh 127 may be obtained from a 3D scan. Aligning 1420 the dental restoration mesh 127 with the dental preparation point cloud 101 involves positioning and orienting the mesh so that it corresponds accurately to the scanned data of the dental preparation. The alignment process may use techniques such as Iterative Closest Point (ICP) algorithms or other registration methods to ensure that the dental restoration mesh 127 is correctly overlaid on the dental preparation point cloud 101. The dental preparation score 105 may then be computed 1300 by measuring alignment between the inference cloud 102 and the dental restoration mesh127. The alignment measurement may involve evaluating how well the predicted features and contours from the inference cloud 102 correspond to the designed dental restoration mesh 127. The alignment measurement may include metrics such as point- to-surface distances, overlap, and fit accuracy, which collectively inform the dental preparation score 105. Once aligned, the dental restoration mesh 127 may be evaluated in a seated configuration relative to the prepared tooth to assess one or more of gap, contact, clearance, and occlusal interactions. By way of example, Figure 7G illustrates a dental arch representation in which a restoration is positioned on the prepared tooth.

[0115] In one embodiment, the dental restoration mesh 127 may be a generated dental restoration mesh 128 created from the preparation point cloud. The generation may be achieved without necessitating a separate dental restoration model. The generation of the generated dental restoration mesh 128 may utilize the detailed three- dimensional spatial information inherent in the preparation point cloud to design a restoration that closely fits the specific contours and features of the prepared tooth or dental structure, without the need for a pre-existing model of the restoration. By relying solely on the preparation point cloud, the method may enable the generation of a dental restoration mesh 127 on site, in near real-time. The ability to directly process point cloud data allows advanced algorithms and CAD software to convert the data into a cohesive mesh model of the restoration. The generated dental restoration mesh 128 may accurately reflect the unique geometry of the dental preparation and enable adjustment without having to wait for the actual dental preparation. A near real-time on-site generation of the dental restoration mesh 127 — understood as being within minutes or any reasonable delay for which a patient might prefer to wait rather than leave and come back — streamlines the workflow, potentially reducing the need for adjustments and enhancing the efficiency and success of the dental procedure. By eliminating the dependency on pre-designed restoration models, dental professionals may respond quickly to patient needs and offer a more immediate and customized restorative solution.

[0116] The dental preparation score 105 may further include a gap analysis score 129 computed 1340 from an average deviation and a maximum deviation between a dental preparation surface and the dental restoration mesh 127. The gap analysis score 129 may, for example, be computed 1340 by assessing both the average deviation and the maximum deviation between these two surfaces. Theaverage deviation may refer to the mean distance between corresponding points on the dental preparation surface and the dental restoration mesh 127. The average deviation may help in understanding the general alignment and fit consistency across the entire surface, indicating how well the restoration conforms to the prepared tooth or dental structure. A lower average deviation may suggest a more uniform and accurate fit, indicative of the efficiency and durability of the dental restoration. The maximum deviation may identify the largest disparity at any point between the dental preparation surface and the restoration mesh. The maximum deviation may help pinpoint specific areas where the fit may be inadequate or problematic, potentially leading to issues such as gaps or misalignments that could compromise the effectiveness of the restoration.

[0117] The gap analysis score 129 may also include a restoration distance 130 calculated 1330 between the dental restoration mesh 127 and the inference cloud 102. The restoration distance 130 may refer to a metric that quantifies the spatial differences between the dental restoration mesh 127 and the inference cloud 102. The restoration distance 130 may provide a measure of how well the predicted features of the dental preparation, as represented by the inference cloud 102, align with the designed contours and surfaces of the dental restoration mesh 127.

[0118] Calculating 1330 the restoration distance 130 may involve several steps. The restoration distance 130 may be calculated by measuring the shortest distances between corresponding points on the inference cloud 102 and the surface of the dental restoration mesh 127. The calculation of the shortest distances may involve determining the closest point on the mesh for each point in the inference cloud 102 and computing the Euclidean distance between them. Several other mathematical metrics may be used to calculate the restoration distance 130. For example, the Chamfer Distance (CD) metric may calculate the average of the distances from each point in one set to the nearest point in the other set, and vice versa, providing a symmetric measure of similarity between the two sets. The Hausdorff Distance metric may consider the maximum distance from a point in one set to the nearest point in the other set, providing an indication of the worst-case scenario for alignment. The Root Mean Square Error (RMSE) metric computes the square root of the average squared differences between corresponding points, offering a measure of the overall deviation.

[0119] Other metrics may also be considered, such as the Earth Mover’s Distance (EMD), also known as the Wasserstein distance. The EMD is a metric used to quantify the dissimilarity between two probability distributions over a region. EMD may beapplied to measure the “effort” required to transform one distribution of points (e.g., the inference cloud 102) into another distribution (e.g., the dental restoration mesh 127). To compute 1330 a restoration distance 130 using EMD, the inference cloud 102 and the dental restoration mesh 127 may be treated as discrete distributions of points in a three-dimensional space. Each point may be assigned a “mass” or weight, representing its significance or contribution to the overall structure. EMD may then find the minimal cost required to transport the “mass” from the points in the inference cloud 102 to the points in the restoration mesh. The EMD approach may be conceptually similar to finding the least costly way to move a pile of earth (the inference cloud 102) to fill a hole (the restoration mesh), where the cost is defined by the distance the earth must be moved. Computation of EMD may involve solving an optimization problem that seeks to minimize the total transportation cost. The optimization problem may be achieved by determining a flow that specifies how much “mass” should be moved from each point in the inference cloud 102 to each point in the restoration mesh, subject to constraints that ensure the total mass is preserved and all points are accounted for. When used as part of the gap analysis score 129, EMD provides a measure of the overall dissimilarity between the predicted dental preparation features and the designed restoration. Unlike metrics like Chamfer or Hausdorff distances, EMD may consider the entire distribution of points, making it sensitive to both local and global differences in shape and alignment.

[0120] The calculated restoration distances 130 are then used to compute the gap analysis score 129, in addition to or as an alternative to the average and maximum deviations. The restoration distance 130 score may provide a comprehensive evaluation of the alignment quality, highlighting both the overall fit and specific areas where the restoration may not conform perfectly to the predicted preparation features.

[0121] The dental preparation score 105 may also include a contact area score 131 computed 1350 from the percentage of the inference cloud 102 that is in contact with the dental restoration mesh 127.

[0122] The contact area score 131 may provide a quantitative measure of how effectively the surfaces of the dental preparation, as predicted by the inference cloud 102, may contact the designed dental restoration mesh 127. Computing 1350 the contact area score 131 may involve determining the percentage of the inference cloud 102 in contact with the dental restoration mesh 127. The first step may involve identifying points within the inference cloud 102 that are close enough to the dentalrestoration mesh 127 to be considered in contact. Proximity may typically be defined by a threshold distance, which may account for acceptable tolerances in the fit between the preparation and the restoration. Once the contact points are identified, the contact area score 131 may be computed by determining the percentage of the total points in the inference cloud 102 that fall within the threshold distance of the dental restoration mesh 127. The calculated percentage reflects the extent to which the predicted features of the dental preparation align with and support the restoration. A higher contact area score 131 may indicate a greater proportion of the inference cloud 102 in contact with the restoration mesh, suggesting a better fit and more reliable support for the restoration. Conversely, a lower contact area score 131 may highlight areas where the preparation does not adequately support the restoration, potentially indicating gaps or misalignments that could affect the stability and effectiveness of the restoration.

[0123] The dental preparation score 105 may also include an occlusal fit score 132 computed 1370 from a simulated 1360 dental occlusion 133 of the dental restoration mesh 127 aligned onto the preparation point cloud. The dental occlusion 133 may refer to the contact between teeth in opposing dental arches when the jaws are closed. Proper occlusion may be essential for effective mastication, speech, and overall oral health. The occlusion may involve the alignment and contact points of the upper and lower teeth, ensuring that the teeth meet correctly to distribute biting forces evenly across the dentition. Simulating 1360 a dental occlusion 133 may involve creating a virtual model of how the dental restoration mesh 127 interacts with the opposing teeth when the mouth is closed. The simulation may use the alignment of the dental restoration mesh 127 with the preparation point cloud to evaluate how the restoration fits into the existing occlusal scheme. Techniques such as finite element analysis or other computational modeling approaches may be used to predict the dynamics of the occlusion, including contact points, pressure distribution, and potential interference with the natural bite.

[0124] The occlusal fit score 132 may be computed 1370 by analyzing the results of the dental occlusion 133 simulation to derive a quantitative measure of the fit and functionality of the restoration within the occlusal context. The occlusal fit score 132 may consider various factors, such as the number and distribution of contact points, the balance of occlusal forces, and the presence of any premature contacts or interferences. A high occlusal fit score 132 may indicate that the restoration meshes well with the opposing teeth, ensuring proper function and comfort. Conversely, a lowocclusal fit score 132 may suggest misalignment or excessive forces on certain teeth, potentially leading to discomfort or damage to the restoration or natural teeth.

[0125] The occlusal fit score 132 may include an occlusal clearance score 134, a contact points score 135, and / or an alignment score 136. The occlusal clearance score 134 may assess the vertical space, or clearance, between the dental restoration and the opposing dentition when the jaws are not in contact for premolars and molars. Adequate occlusal clearance may ensure that there is sufficient space to accommodate the restoration material without interfering with the natural movement of the jaws. In the case of incisors, the assessment of occlusal clearance may focus on horizontal or diagonal contact due to the orientation and nature of occlusion in the anterior region. Proper clearance may prevent excessive contact that may lead to premature wear or discomfort. The occlusal clearance score 134 may quantify whether the occlusal space is adequate or whether adjustments are needed to avoid excessive or insufficient clearance.

[0126] The contact points score 135 may evaluate the number, distribution, and quality of contact points between the dental restoration and the opposing teeth when the jaws are closed. Ideally, contact points should be evenly distributed to achieve a balance of occlusal forces, preventing localized pressure that could lead to damage or discomfort. The contact points score 135 may quantify how well the restoration achieves the desired balance of occlusal forces, identifying areas where contact may be too light or too heavy, potentially requiring adjustments to ensure even distribution.

[0127] The alignment score 136 may assess how well the dental restoration aligns with the existing occlusal plane and the overall dental arch. Proper alignment may ensure that the restoration integrates seamlessly with the natural dentition, supporting functional efficiency and aesthetic harmony. Misalignment may lead to occlusal interferences, discomfort, or inefficiencies in bite force distribution. The alignment score 136 may quantify the degree to which the restoration conforms to the ideal alignment, guiding necessary adjustments to improve fit and function.

[0128] In one embodiment, actions to improve the dental preparation score 105 may be suggested 1930. Examples of suggestion 1930 may include a tooth reduction 137, a tooth contouring 138, a margin definition 139, and / or a core buildup 140. The suggestions may be based on the assessment results and are intended to optimize the preparation for successful restoration outcomes. Suggestionsmay relate to a tooth being restored or to any other teeth. By identifying specific areas of the preparation that may benefit from improvement, guidance may be provided to dental professionals in making informed adjustments that can lead to better fit, function, and longevity of the dental restoration. For example, tooth reduction 137 may involve removing excess tooth structure either from the tooth being prepared or from neighboring teeth to achieve the desired shape and size for accommodating the dental restoration. Proper reduction may ensure that the restoration fits snugly and functions effectively without causing interference with opposing teeth or affecting adjacent teeth.

[0129] T ooth contouring 138 may refer to the shaping of the surface of the tooth to create smooth transitions and appropriate angles that facilitate the placement and retention of the restoration. Tooth contouring 138 may involve refining the contours of the prepared tooth or adjusting the contours of neighboring teeth to improve aesthetics and functionality, ensuring that the contours do not impede the restoration.

[0130] Margin definition 139 actions may focus on clearly delineating the margin line, which is the boundary where the restoration meets the prepared tooth. A well- defined margin may ensure a secure fit and prevent gaps that could lead to leakage or decay. In some cases, adjustments to neighboring teeth may be necessary to ensure that the margin is accessible and properly defined.

[0131] In cases where significant tooth structure is missing or compromised, a core buildup 140 may be recommended to provide a solid foundation for the restoration. While core buildup 140 may typically apply to the tooth being restored, adjustments to neighboring teeth may also be considered to ensure they provide adequate occlusion and support or do not interfere with the restoration.

[0132] The dental preparation score 105 may include additional scores, such as a margin line score 141 , a gingival proximity score 142, an undercut score 143, an occlusal reduction score 144, an axial reduction score 145, a smoothness score, a thickness score 147, an interproximal clearance score 148, an alignment score 136, a path of insertion score 149, a contact area score 131 , and an abutment spacing score 150.

[0133] The margin line score 141 may evaluate the precision and clarity of the margin line, which is the boundary where the dental restoration meets the tooth preparation. The margin line score 141 may be computed by evaluating the precision and clarity of the margin line predicted by the trained inference model 104. The marginline may be identified within the inference cloud 102 derived from the relevant point cloud 100 of the dental preparation. A comparison may be made between the predicted margin line and a reference margin line, which may be determined from additional imaging data or expert annotations, to assess alignment accuracy. Precision may involve calculating deviations between the predicted margin line and the reference, utilizing metrics such as point-to-point distance measurements or a Chamfer Distance metric, wherewith smaller values may indicate higher precision. Clarity may be assessed by analyzing the continuity and definition of the margin line, possibly by using a spline fitting technique, such as a B-spline, to represent the smoothness of the line. The margin line score 141 may incorporate aspects such as the smoothness of the fitted spline, where deviations or abrupt changes may indicate lower clarity and precision. Sampling points along the margin line may aid in determining if the line is well-defined and free from unwanted interruptions or inconsistencies. In computing the margin line score 141 , a set of threshold parameters may be established to quantify what constitutes adequate precision and clarity, with results potentially categorized into levels such as outstanding, satisfactory, or warning.

[0134] In embodiments, the margin line is displayed as a curve or set of points around the prepared tooth to facilitate review and scoring. By way of example, Figure 7F illustrates a close-up representation of a prepared tooth with a margin line indicated in relation to surrounding gingival anatomy, which may support evaluation of margin line quality and gingival proximity.

[0135] When the dental preparation score 105 includes a margin line score 141 , the trained inference model 104 may be trained for detecting margin line points located on a margin line of the dental preparation. The trained inference model 104, when trained for detecting margin line points on the margin line of the dental preparation, may be useful for computing the margin line score 141 by ensuring precise and accurate identification of the margin line location. During training, the inference model may be exposed to a dataset comprising known margin line points across various dental preparations, allowing it to learn patterns that distinguish margin line features from other tooth structures. The model may employ neural network architectures capable of capturing complex spatial relationships within point cloud data, thereby improving predictive accuracy.

[0136] The gingival proximity score 142 may measure the closeness of the dental preparation to the gingival tissue. The gingival proximity score 142 may be computedby measuring the spatial relationship between the dental preparation and the surrounding gingival tissue. Measurements may involve the analysis of the relevant point cloud 100, which may include points representing both the dental preparation and the gingival tissue, potentially obtained from detailed imaging techniques such as CBCT scans. The gingival proximity score 142 may be computed by identifying and delineating points within the point cloud that define the boundary of the gingival tissue. Identification of the gingival tissue may involve segmentation algorithms or predefined anatomical markers that distinguish gingival areas from tooth structures. Subsequently, the minimum distances between the dental preparation surface and the gingival boundary may be calculated for various points along the margin line of the dental preparation. These distances may provide insight into how close the preparation is to the gingival tissue, with smaller distances suggesting closer proximity. Thresholds for what constitutes optimal gingival proximity may be established, accounting for clinical guidelines regarding space necessary to avoid gingival irritation or damage during the placement of a dental restoration. In one embodiment, the computed distances may be averaged or weighted to reflect areas of concern where minimal clearance occurs. The gingival proximity score 142 may then be derived by categorizing the computed proximity data into levels such as outstanding, satisfactory, or warning, based on the degree to which the measured distances meet or exceed the established thresholds. The gingival proximity score 142 provides a quantitative measure that may guide adjustments in the dental preparation process to maintain a safe and effective distance from the gingival tissue. In some embodiments, gingival anatomy is segmented or otherwise identified to support margin-related evaluation and gingival proximity evaluation. By way of example, Figure 7E illustrates a three-dimensional representation in which gingival anatomy 7060 surrounding a prepared tooth 7020 and a margin line 7070 are identified, which may be used to evaluate gingival proximity and / or clearance.

[0137] When the dental preparation score 105 includes a gingival proximity score 142, the trained inference model 104 may be trained for detecting points located on a gingival margin of the dental preparation. The trained inference model 104, when trained for detecting gingival margin points, may be useful for computing the gingival proximity score 142 by ensuring precise and accurate identification of the gingival margin location. During training, the inference model may be exposed to a dataset comprising known gingival margin points in various dental preparations, allowing it to learn patterns that distinguish gingival margin features from other areas of the toothstructure. The model may employ neural network architectures capable of capturing complex spatial relationships within point cloud data, thereby enhancing predictive accuracy.

[0138] The undercut score 143 may assess the presence and extent of undercuts, which are areas where the tooth structure overhangs in a way that could impede the placement or retention of the restoration. The undercut score 143 may be computed by evaluating the geometry of the dental preparation to identify and quantify areas where the tooth structure creates overhangs, known as undercuts, that may interfere with the placement or retention of a dental restoration. The undercut score 143 may be computed by identifying the path of insertion for the restoration, potentially using principal component analysis (PCA) on the relevant point cloud 100 to determine the optimal insertion direction. The path of insertion may serve as a reference for evaluating the geometry of the tooth, indicating the intended direction along which the restoration will engage the preparation. Once the path of insertion is established, the geometry of the preparation may be analyzed to detect areas where the tooth structure overhangs relative to the established path of insertion. Techniques such as computational geometry algorithms or point cloud slicing may be employed to assess the contour of the tooth and detect deviations that indicate undercuts. The extent of each identified undercut may be quantified by measuring the depth and area of the overhangs. The measurements of depth and area of the overhangs may be standardized against defined clinical thresholds to determine their significance, with deeper or larger undercuts potentially posing greater challenges to restoration placement. The undercut score 143 may then be computed by categorizing the presence and extent of these features. The score may express results as categories such as outstanding, satisfactory, or warning, depending on how well the detected undercuts align with acceptable preparation standards. Results may be integrated into the dental preparation score 105, and corrective measures may be suggested to address identified undercut issues.

[0139] When the dental preparation score 105 includes an undercut score 143, the trained inference model 104 may be trained for predicting undercut points 151 lying within an undercut from the relevant point cloud 100 and the margin line 7070 points. The trained inference model 104, when trained for predicting undercut points 151 , may be useful for computing the undercut score 143 by ensuring precise and accurate identification of undercut locations. During training, the inference model may beexposed to a dataset comprising known undercut points 151 in various dental preparations, allowing it to learn patterns that distinguish undercut features from other areas of the tooth structure. The model may employ neural network architectures capable of capturing complex spatial relationships within point cloud data, thereby enhancing predictive accuracy.

[0140] In embodiments, an undercut area may be generated from the undercut points 151 , an undercut profile 152 may be computed, with an undercut height measured from the undercut area, and the undercut score 143 may be computed from the undercut profile 152. After the trained inference model 104 predicts undercut points 151 , the undercut points 151 may be used to define an undercut area. The undercut area may represent the spatial region where the tooth structure creates an overhang relative to the margin line 7070 or path of insertion 7050. The generation of the undercut area may involve grouping the predicted undercut points 151 into a coherent geometric form that captures the extent of the undercut within the dental preparation. From the generated undercut area, an undercut profile 152 may be computed, which includes specific measurements such as the undercut height. The undercut height may indicate how far the undercut extends into the dental preparation. The measurements provide insight into how significant the undercut is and whether it may present a challenge to restoration placement. The profile may encompass relevant metrics to describe the geometry and impact of the undercut. Using the computed undercut profile 152, the undercut score 143 may be derived. The score may reflect the suitability of the preparation for restoration, considering the presence and extent of undercuts. Factors such as the depth, width, and area of the undercut, as captured in the profile, determine the score. A higher score may indicate minimal undercut issues, while a lower score may highlight significant undercuts that require attention.

[0141] The occlusal reduction score 144 may evaluate the amount of tooth structure removed from the occlusal surface during preparation. The occlusal reduction score 144 may be computed by evaluating the extent of dental structure removal from the occlusal surface during preparation. The computation may involve comparing the current geometry of the occlusal surface, represented in the relevant point cloud 100, with a pre-preparation model or a standardized anatomical reference. The prepreparation model may be obtained through prior scans or standardized dental anatomy databases, providing a baseline for comparison. The amount of occlusal reduction may be determined by calculating differences in height or volume betweenthe occlusal surface in the relevant point cloud 100 and the pre-preparation model. For premolars and molars, spatial analysis techniques may be used to measure the vertical displacement of occlusal points, while volume-based methods may quantify material removal. For incisors, spatial analysis may focus on horizontal or angular changes. The computed extent of reduction may then be compared against clinical guidelines or targets for occlusal reduction, which may represent preparation standards ensuring proper alignment and restoration fitting while maintaining tooth structure and vitality. Deviations from the targets may be quantified to assess whether the reduction is appropriate, excessive, or insufficient. The occlusal reduction score 144 may be derived by categorizing the computed reduction into levels such as outstanding, satisfactory, or warning, based on alignment with predefined targets. The occlusal reduction score 144 may provide feedback on the adequacy of occlusal surface preparation for supporting a successful restoration.

[0142] The axial reduction score 145 may assess the reduction of tooth structure along the axial (vertical) walls of the preparation. The axial reduction score 145 may be computed by assessing the reduction of tooth structure along the axial walls during preparation. The computation may involve evaluating the vertical tooth surfaces represented in the relevant point cloud 100 against a pre-preparation model or an established reference for ideal preparation. The axial walls of the prepared tooth in the relevant point cloud 100 may be identified, potentially referencing prior scans or standardized anatomical data to define the original tooth contours. The original contours may serve as a baseline to quantify changes. The reduction may be determined by calculating the differences in axial wall thickness between the prepared surfaces in the relevant point cloud 100 and the pre-preparation model. These differences may be measured using spatial analysis techniques to determine the displacement or material removal along the vertical surfaces. The measured axial reduction may then be compared against clinical guidelines or targets for axial reduction, indicating the optimal amount to support a restoration without compromising the structural integrity and vitality of the tooth. Deviations from these guidelines may be quantified to assess the adequacy of the reduction. The axial reduction score 145 may be derived by categorizing the computed reduction into levels such as outstanding, satisfactory, or warning, according to alignment with the predefined targets. The axial reduction score 145 may inform the readiness of the axial walls to accommodate a dental restoration effectively.

[0143] The smoothness score 146 may measure the smoothness of the prepared tooth surfaces. The smoothness score 146 may be computed by evaluating the surface texture of the prepared tooth surfaces represented in the relevant point cloud 100. The measurement may involve analyzing geometrical properties such as curvature, gradient, or surface normals to determine deviations from an ideally smooth surface. Initially, prepared tooth surfaces in the relevant point cloud 100 may be segmented to focus on specific areas, such as occlusal and axial surfaces. Using techniques such as curvature analysis, variations in surface curvature across these areas may be quantified, as larger variations may indicate surface roughness. Surface normal consistency may additionally be assessed by evaluating the orientation of normals across the segmented surfaces. A consistent normal orientation may suggest smoothness, whereas variations may denote irregularities. The computed measurements may then be compared against reference values or standards that represent ideal smoothness, perhaps derived from clinical guidelines or idealized models. Deviations from these standards may be quantified for smoothness assessment. The smoothness score 146 may be characterized by categorizing the computed smoothness into levels such as outstanding, satisfactory, or warning, according to adherence to smoothness references. The smoothness score 146 may provide feedback on the quality of the preparation surface for supporting a dental restoration.

[0144] In embodiments, the dental preparation score 105 comprises a cornerrounding score. The corner-rounding score may be computed by identifying one or more transition regions of the preparation, including transition regions between axial walls and occlusal surfaces, and evaluating a sharpness metric at the transition regions. The sharpness metric may include a local radius of curvature, a curvature magnitude, an angle between estimated surface normals, and / or a feature derived from a fitted curve or spline.

[0145] In embodiments, the corner-rounding score is computed by comparing the sharpness metric to a minimum rounding threshold. The minimum rounding threshold may be selected based on manufacturability constraints associated with fabrication of the restoration, including a characteristic dimension of a milling tool, such as a tool diameter or tool radius. In embodiments, transition regions that do not satisfy the minimum rounding threshold are flagged as presenting a manufacturability risk and / or a durability risk associated with stress concentrations.

[0146] The thickness score 147 may evaluate the thickness of the remaining tooth structure after preparation. The thickness score 147 may be computed by evaluating the thickness of the remaining tooth structure after preparation, as represented in the relevant point cloud 100. The evaluation may involve analyzing cross-sectional profiles or point-to-point distances within the tooth structure to determine whether adequate thickness is maintained. The process may begin by segmenting the relevant point cloud 100 to isolate prepared areas requiring thickness evaluation. Cross-sectional slices may be generated perpendicular to the long axis of the tooth to measure wall thickness at various locations. Thickness measurements may be computed by determining distances between opposing surfaces within these cross-sectional slices. Spatial analysis techniques may be employed to verify that the measured thickness meets or exceeds predefined clinical thresholds for structural integrity. The computed thickness values may then be compared against clinical guidelines or material-specific requirements for dental restorations, ensuring that the remaining structure can support a restoration without risk of fracture or failure. The thickness score 147 may be derived by categorizing the computed thickness into levels such as outstanding, satisfactory, or warning, based on the adherence of the structure to thickness requirements. The thickness score 147 may provide insight into the adequacy of the preparation for successful restoration placement.

[0147] When the dental preparation score 105 includes a thickness score 147, the trained inference model 104 may be a thickness model 153 trained to predict thin points exhibiting insufficient thickness from the relevant point cloud 100. The trained inference model 104, when configured to detect thin points, may be useful for computing the thickness score 147 by ensuring accurate identification of areas where the tooth structure is too thin. During training, the thickness model 153 may be exposed to a dataset comprising known regions of insufficient thickness in various dental preparations, allowing it to learn patterns that identify thin areas distinct from adequately thick structures. The model may employ neural network architectures capable of capturing complex spatial relationships within point cloud data, thereby improving predictive accuracy.

[0148] In one embodiment, a thin surface 154 from the thin points may be generated, a thickness profile 155 from the thin surface 154 may be computed, and the thickness score 147 may be computed from the thickness profile 155. Once the thickness model 153 predicts thin points from the relevant point cloud 100, a thinsurface 154 may be generated from these points. The thin surface 154 may represent the contiguous areas of the dental preparation where the tooth structure exhibits insufficient thickness. The generation of the thin surface 154 may involve connecting the identified thin points into a coherent representation that visually and geometrically highlights the regions of concern. From the generated thin surface 154, a thickness profile 155 may be computed. The thickness profile 155 may provide a detailed characterization of the thin areas, potentially including metrics such as the minimum, average, and maximum thickness values across the thin surface 154. The profile may also describe the spatial distribution and extent of thin sections, offering a comprehensive view of the structural integrity of the tooth. Using the computed thickness profile 155, the thickness score 147 may be derived.

[0149] The thickness score 147 may be computed according to a property of the material used for the dental restoration. Different dental materials, such as porcelain, metal, or composite, may have unique mechanical properties, including strength, flexibility, and wear resistance. The thickness score 147 computation may incorporate these properties to determine the minimum thickness required for effective support and durability of the restoration made from these materials. For example, porcelain may be known for its translucency and aesthetic similarity to natural enamel but requires a certain thickness to prevent chipping or cracking. When computing the thickness score 147 for a porcelain veneer, the model might require a minimum thickness threshold to ensure functional strength while maintaining aesthetic appearance, with the score potentially being lower if the tooth structure is too thin to meet the minimum thickness threshold. Metals such as gold or metal alloys offer high strength and durability but may be placed with slightly thinner profiles compared to ceramic materials. When computing the thickness score 147 for a metal crown, the required thickness may be less stringent, allowing for a higher score even if the remaining tooth structure is less substantial compared to what would be acceptable for ceramic restorations. Composites, on the other hand, have versatile applications but may require specific thickness profiles 155 to optimize bonding strength and wear resistance. For a composite filling, the thickness score 147 computation may favor configurations that provide enough material for effective bonding without excessive removal of tooth structure, allowing for tailored evaluations based on different product formulations.

[0150] The interproximal clearance score 148 may measure the space available between adjacent teeth for the dental restoration. The interproximal clearance score 148 may be computed by measuring the space available between adjacent teeth as represented in the relevant point cloud 100, specifically in areas intended for restoration placement. The computation may involve analyzing the distance between proximal surfaces of the prepared tooth and its neighboring teeth. The process may commence by segmenting the relevant point cloud 100 to identify the proximal surfaces of both the tooth undergoing restoration and the adjacent teeth. Spatial analysis techniques may be applied to calculate the distances between these identified surfaces to determine available clearance. Measured interproximal distances may be compared against clinical guidelines or established standards for minimum clearance, as adequate interproximal clearance is crucial for ensuring that the dental restoration fits snugly without impinging on neighboring teeth. The interproximal clearance score 148 may then be derived by categorizing the measured clearance into levels such as outstanding, satisfactory, or warning, based on alignment with clearance requirements. The interproximal clearance score 148 may provide insights into the ease of restoration placement and the likelihood of achieving optimal dental function and aesthetics.

[0151] The alignment score 136 may assess how well the preparation aligns with the rest of the dental arch. The alignment score 136 may be computed by evaluating how well the dental preparation aligns with the rest of the dental arch as represented in the relevant point cloud 100. The computation may involve analyzing the spatial orientation and positioning of the prepared tooth in relation to contiguous dental structures within the arch. The process may begin by registering the relevant point cloud 100 of the prepared tooth with the point cloud of the entire dental arch. The registration may establish a unified coordinate framework, enabling assessment of discrepancies in alignment between the prepared tooth and its neighboring teeth. Alignment may be assessed by examining how the occlusal plane and axial orientations of the prepared tooth conform to those of adjacent teeth. Techniques such as fitting curves or splines along the arch line may be employed to determine deviations or rotations that indicate misalignment. Measured deviations may be compared against established standards or clinical guidelines for proper alignment, ensuring the preparation seamlessly integrates into the arch for optimal restoration function and aesthetics. The alignment score 136 may then be derived by categorizing the computedalignment into levels such as outstanding, satisfactory, or warning, based on adherence to alignment metrics.

[0152] The path of insertion score 149 may evaluate the path along which the restoration may be placed onto the prepared tooth. The path of insertion score 149 may be computed by evaluating the anticipated insertion trajectory of the dental restoration onto the prepared tooth, as represented in the relevant point cloud 100. The assessment may involve determining whether the path of insertion is unobstructed and aligned with the dental preparation.

[0153] The process may begin by analyzing the relevant point cloud 100 to establish the optimal path of insertion, potentially using techniques such as principal component analysis (PCA) to identify the primary axis along which the restoration should be placed. Angles and directions that minimize interference during insertion may be revealed. The established path may be examined to identify potential obstructions or geometric constraints that could impede the restoration process. Areas of interference, such as undercuts or misalignments, may be flagged as potential challenges. Measured characteristics of the path of insertion, including angles and clearance, may be compared against clinical guidelines or standards to ensure the insertion path accommodates the dental restoration efficiently and without disruption. The path of insertion score 149 may then be derived by categorizing the evaluated path into levels such as outstanding, satisfactory, or warning, based on adherence to path criteria. Insights gained from evaluating the path of insertion may assist in identifying adjustments required to optimize the restoration process, thereby facilitating a stable and precise fit onto the prepared tooth.

[0154] In embodiments, the dental preparation score 105 comprises a taper score. The taper score may be computed by measuring one or more taper angles of axial walls of the prepared tooth relative to the path of insertion and / or a computed preparation axis. The measured taper angle may be evaluated against a target taper value or a target taper range derived from one or more preparation guidelines. In embodiments, the taper score expresses whether the taper is outside an acceptable range (out-of-bounds) or within an acceptable range (safe zone), and optionally whether the taper is within the acceptable range while still improvable.

[0155] In embodiments, the taper score and / or the corner-rounding score are provided in a training mode. In embodiments, the taper score and / or the corner-rounding score are optionally activatable in a practitioner mode. In embodiments, user interface elements may be configured such that the practitioner mode emphasizes safe-zone / out-of-bounds indications and the training mode further provides additional metrics and guidance.

[0156] The contact area score 131 may evaluate the extent and distribution of contact between the restoration and the opposing teeth or other structures. The contact area score 131 may be computed by evaluating the extent and distribution of contact between the dental restoration and opposing teeth or other structures, as represented in the simulated occlusion of the relevant point cloud 100. The assessment may involve determining how the restoration interfaces with adjacent entities to ensure functional and balanced contact. The process may start by simulating the occlusion of the dental restoration with the opposing arch, using computational models to replicate dynamic interactions during normal dental function. Simulation may reveal contact points and pressure distributions across the surfaces involved. The extent of contact may be evaluated by identifying areas where the restoration closely interfaces with opposing structures. Contact patterns may be mapped and quantified to determine the overall distribution and uniformity of contact, aiming to achieve balance and prevent localized pressure that might cause wear or discomfort. Measured contact area and distribution may be compared against clinical guidelines or standards to ensure the restoration achieves optimal functional contact without adversely affecting dental health. The contact area score 131 may be derived by categorizing the evaluated contact into levels such as outstanding, satisfactory, or warning, based on adherence to contact criteria. Evaluating the contact area may provide insights into necessary adjustments to optimize restoration fit and function within the oral environment.

[0157] The abutment spacing score 150 may assess the space available around abutment teeth for the placement of a dental bridge or similar restoration. The abutment spacing score 150 may be computed by assessing the space available around abutment teeth for the placement of a dental bridge or similar restoration as represented in the relevant point cloud 100. The assessment may involve evaluating the spatial relationships between abutment teeth and surrounding structures to ensure adequate room for restoration placement. The process may begin by identifying and segmenting the abutment teeth within the relevant point cloud 100, focusing on areas where the dental bridge will anchor. Spatial analysis techniques may be applied to measure distances between abutment teeth and any neighboring dental or anatomicalstructures. Measured spacing around abutment teeth may be compared against clinical guidelines or standards for minimal spacing required to accommodate a dental bridge. Ensuring sufficient spacing is crucial for proper placement and stabilization of the restoration without interference. The abutment spacing score 150 may be derived by categorizing the measured spacing into levels such as outstanding, satisfactory, or warning, based on alignment with spacing requirements. Insights gained from assessing abutment spacing may guide necessary adjustments to facilitate successful restoration placement and function.

[0158] In embodiments, when the dental restoration is a dental bridge, the trained inference model 104 may be an abutment alignment model 156 trained to detect alignment points describing the alignment of abutment teeth, and the dental preparation score 105 may include an alignment score 136. The trained inference model 104, when configured for detecting alignment points, may be useful for computing the alignment score 136 by ensuring the precise and accurate identification of the spatial positioning and orientation of abutment teeth relative to each other and the surrounding dental arch. During training, the abutment alignment model 156 may be exposed to a dataset comprising known configurations of aligned abutment teeth in various dental bridge preparations. The model may learn to recognize patterns that reflect optimal alignment, capturing how abutment teeth should be positioned to provide effective support for a dental bridge. The model may employ neural network architectures capable of capturing complex spatial relationships within point cloud data, enhancing predictive accuracy. By detecting alignment points, the model may ensure that the evaluation focuses on key geometrical aspects that inform the proper fit and function of the dental bridge. The alignment score 136 derived from the model output may assess how well the detected alignment points adhere to clinical guidelines or standards for optimal abutment alignment. The evaluation informs dental practitioners about necessary adjustments to the abutment preparation to facilitate the successful integration and stability of the dental bridge.

[0159] In one embodiment, an alignment spline 157 may be fitted through the alignment points. A curvature of the alignment spline 157 may be evaluated against an alignment threshold 158, and the alignment score 136 may be computed from the curvature of the alignment spline 157, such that the alignment spline 157 may be the inference spline 116 when the method presented herein is applicable. The alignment spline 157 may be fit through the alignment points detected by the abutment alignmentmodel 156. The spline may serve as a continuous geometric curve that represents the trajectory of alignment points along the abutment teeth. In this context, the alignment spline 157 may also refer to the inference spline 116, generated from an ordered point cloud 117 to capture geometrical features accurately. The evaluation of the curvature of the fitted alignment spline 157 is performed to determine how well it conforms to an ideal alignment, which involves assessing the smoothness and consistency of the curvature along the spline. Comparison against an alignment threshold 158 helps identify any deviations that might indicate misalignments. The threshold represents the maximum permissible deviation from a defined alignment standard and is used as a benchmark for quality assessment. The alignment score 136 is computed based on the curvature evaluation of the alignment spline 157. By quantifying how closely the curvature aligns with expectations, the score reflects the adequacy of abutment tooth alignment for supporting a dental bridge. Lower scores may highlight significant deviations from ideal alignment, while higher scores indicate proper alignment that supports stable and functional bridge placement.

[0160] When the dental preparation score 105 includes an abutment spacing score 150, the trained inference model 104 may be an abutment spacing model 159 trained to detect abutment spacing points located between abutment teeth. The trained inference model 104, when configured to detect spacing points, may be useful for computing the abutment spacing score 150 by ensuring accurate identification of the spatial gaps and spaces that exist between abutment teeth in preparation for the placement of a dental bridge. During training, the abutment spacing model 159 may be exposed to a dataset comprising known configurations of abutment teeth and their corresponding spacing parameters across various dental bridges. The model may learn to identify patterns that denote sufficient and optimal spacing, recognizing how spacing impacts the fitting and support of a dental bridge. The model may employ neural network architectures capable of capturing detailed spatial relationships within point cloud data, thereby enhancing its ability to predict spacing points accurately. By detecting these points, the model may ensure a comprehensive evaluation of the spacing available for dental bridge placement. The abutment spacing score 150 derived from the model may assess how well the detected spacing points align with clinical guidelines or standards for adequate spacing.

[0161] When the dental preparation score 105 includes a gingival proximity score 142, the trained inference model 104 may be a gingival clearance model 161trained to detect gingival clearance points located in the space between the edge of the dental restoration and the gingival margin. The trained inference model 104, when configured to detect gingival clearance points, may be useful for computing the gingival proximity score 142 by ensuring precise identification of the spatial separation necessary to avoid impingement on the gingival tissue. During training, the gingival clearance model 161 may be exposed to a dataset comprising examples of dental restorations and the required clearance from the gingival tissue in various clinical scenarios. The model may learn to recognize patterns that indicate sufficient gingival clearance, understanding the balance needed to achieve optimal restoration support without compromising gingival health. The model may employ neural network architectures capable of capturing intricate spatial relationships within point cloud data, thereby improving its predictive accuracy in identifying gingival clearance points. Through reliable detection, the model ensures a focused evaluation of the space that might affect restoration placement. The gingival proximity score 142 derived from the model may assess how well the detected clearance points meet clinical guidelines or standards for maintaining appropriate distances from the gingival margin. The evaluation may provide insights to dental practitioners regarding necessary modifications to achieve a balance between restoration fit and gingival health, supporting successful restoration outcomes.

[0162] In embodiments, when the dental restoration is a dental bridge, a gingival clearance spline 162 may be fitted through the gingival clearance points. A gingival clearance distance may be calculated along the gingival clearance spline 162 to determine an available clearance. The gingival proximity score 142 may be computed from the gingival clearance distance. A gingival clearance spline 162 may be fitted through the gingival clearance points, which are detected by the trained gingival clearance model 161. The spline may act as a continuous curve that represents the trajectory of these points, capturing the contour and variability of the clearance space between the edge of the dental restoration and the gingival margin. Along the fitted gingival clearance spline 162, the gingival clearance distance is calculated to determine the available space. The distance of the spline from the gingival margin may be measured at various points, providing a comprehensive profile of how much clearance is maintained. The calculation considers both the minimum required spacing and the variability across different sections of the restoration, ensuring that there are no segments where the clearance is insufficient. Using the calculated gingival clearancedistance, the gingival proximity score 142 is computed. The gingival proximity score 142 may evaluate the adequacy of clearance maintained between the restoration and the gingiva, ensuring that the restoration does not impinge on or irritate the gingival tissue. The clearance measurements are compared against clinical standards or guidelines to determine if the clearance is within acceptable limits. A higher score may indicate sufficient clearance, whereas a lower score may suggest the need for adjustments to improve the clearance.

[0163] In embodiments, when the dental restoration is a dental bridge, the dental preparation score 105 may include a load distribution score 163. Occlusal forces may be simulated on the dental restoration mesh 127 aligned onto the preparation point cloud. A stress distribution assessment 164 may be computed from the occlusal forces across abutment teeth and the dental bridge. The load distribution score 163 may be computed from the stress distribution assessment 164. Occlusal forces, which refer to the forces exerted during biting and chewing, may be simulated on the dental restoration mesh 127 aligned onto the preparation point cloud. The simulation may replicate the dynamic interactions the dental bridge would experience in the mouth of the patient, providing insight into how these forces distribute across the restoration and the supporting abutment teeth. Using the simulated occlusal forces, a stress distribution assessment 164 may be computed. The assessment may evaluate how the forces are transmitted through the dental bridge and the abutment teeth. Computational models, such as finite element analysis (FEA), may be employed to visualize and quantify stress concentrations, identifying areas where forces might be excessive or unevenly distributed, which could lead to mechanical failure or discomfort. Based on the stress distribution assessment 164, the load distribution score 163 may be computed. The score may reflect the efficacy of the dental preparation and bridge design in evenly distributing occlusal forces to prevent overloading of specific areas. The score may be derived by comparing the computed stress distributions against idealized or clinical standards that ensure long-term functionality and durability of the dental bridge. A higher load distribution score 163 indicates a favorable stress distribution, while a lower score suggests the need for alterations in the preparation or bridge design to achieve balance and structural integrity.

[0164] In one embodiment, the stress distribution assessment 164 may take into account specific properties of the material used in constructing the dental bridge. By incorporating material properties into the assessment, the evaluation of how occlusalforces are distributed across the dental restoration becomes more precise and reflective of real-world performance. Different dental bridge materials, such as metals, ceramics, or composites, have unique mechanical properties that influence their response to stress. These properties include tensile and compressive strength, modulus of elasticity, and fracture toughness. The consideration of such properties ensures that the stress distribution assessment 164 accurately reflects how these materials will perform under actual occlusal forces. Including material properties in the assessment may alter how forces are predicted to distribute across the dental bridge and abutment teeth. For example, metal bridges, made from materials such as alloys, typically have high strength and can withstand significant stress without deformation. Stress distribution assessments 164 for metal bridges account for their ability to distribute forces over larger areas effectively, reducing localized stress concentrations. Ceramic materials are often more brittle than metals, so the assessment for ceramic bridges might emphasize the need to avoid stress concentrations that could lead to cracking or chipping. The brittle nature of ceramics demands a more even stress distribution to prevent material failure. Composite bridges offer a balance between flexibility and strength. Assessments for composites may focus on ensuring that the flexible nature of the material does not lead to excessive movement or instability under occlusal forces.

[0165] In one embodiment, the dental preparation score 105 may include an alignment score 136. An alignment spline 157 may be computed from the dental restoration mesh 127 aligned onto the preparation point cloud, and the alignment score 136 may be computed from the alignment spline 157. An alignment spline 157 is computed based on the dental restoration mesh 127 as it aligns with the preparation point cloud. The alignment spline 157 serves as a continuous geometric curve that represents the trajectory of alignment points along the interface between the restoration and the prepared dental structures. The alignment spline 157 may capture the nuances of spatial orientation and positioning that are useful for proper restoration fit. The alignment spline 157 provides a detailed representation of how well the restoration mesh conforms to the prepared tooth surfaces. The geometric model helps in visualizing any deviations, rotations, or offsets that may exist between the intended position of the restoration and its actual alignment when placed. The alignment score 136 is computed by analyzing the curvature and trajectory of the alignment spline 157, comparing these characteristics against established clinical guidelines orstandards that define optimal alignment parameters. Measurements such as spline smoothness, distance deviations, and angles are evaluated to determine how well the restoration achieves the desired alignment. A higher alignment score 136 may indicate that the restoration accurately follows the intended path, ensuring that it integrates seamlessly with the dental arch, while a lower score may highlight the need for adjustments.

[0166] In embodiments when the dental restoration is a dental bridge and the dental preparation score 105 includes an abutment space score 165, an available abutment space 166 may be determined between abutment teeth from the dental restoration mesh 127 aligned onto the preparation point cloud, and the abutment space score 165 may be computed from the available abutment space 166. The available abutment space 166 may be determined by aligning the dental restoration mesh 127 onto the preparation point cloud. The alignment may enable a precise assessment of the spatial relationship between abutment teeth, allowing accurate measurement of the gaps or spaces available for placing the dental bridge. Spatial analysis techniques may be employed to calculate the distance between abutment teeth, ensuring that the measured space reflects the true configuration of the dental arch. Adequate spacing between abutment teeth is crucial for the structural integrity and stability of the dental bridge. The available space must accommodate the physical dimensions of the bridge while maintaining proper margins and avoiding undue stress on the abutment teeth that could lead to periodontal issues or prosthetic failure. From the available abutment space 166, the abutment space score 165 may be computed by comparing measured distances against clinical guidelines or thresholds for minimal spacing required for dental bridges. The score evaluates whether the spacing is sufficient to support the dental bridge. Higher scores may indicate that the spacing meets or exceeds the necessary requirements, facilitating a successful fitting, while lower scores may highlight potential issues that need correction.

[0167] In embodiments, when the dental restoration is a dental bridge and the dental preparation score 105 comprises a gingival proximity score 142, gingival margin clearance 167 may be simulated from the dental restoration mesh 127 aligned onto the preparation point cloud, and the gingival proximity score 142 may be computed from the gingival margin clearance 167. Gingival margin clearance 167 is simulated by aligning the dental restoration mesh 127 onto the preparation point cloud. The alignment allows for an accurate analysis of the spatial relationship between the edgeof the dental bridge and the gingival margin, ensuring that the bridge does not impinge on or irritate the gingival tissue. Computational simulations may model how the bridge interacts with the gingiva, highlighting areas where clearance may be insufficient. Adequate clearance is necessary to prevent the dental bridge from causing irritation or damage to the gingival tissue. Proper clearance ensures that oral hygiene can be maintained, reducing the risk of inflammation or periodontal issues, while also allowing the restoration to integrate seamlessly with the existing dental structures. The gingival proximity score 142 is computed based on the simulated gingival margin clearance 167, comparing measured distances against clinical guidelines or thresholds that define acceptable clearance levels. The gingival proximity score 142 may evaluate whether the clearance is adequate to prevent gingival impingement. A higher score may indicate optimal clearance that protects gingival health, while a lower score could suggest areas where adjustments are needed to increase the space between the restoration and the gingiva.

[0168] A second aspect of the techniques described herein relates to a non- transitory computer-readable medium storing a set of instructions for executing the method described hereinabove. The set of instructions stored on the non-transitory computer-readable medium, when executed, may be used for evaluating a dental preparation for a dental restoration. The executed method may include assembling 1100 a relevant point cloud 100 from a dental preparation point cloud 101 of the dental preparation, predicting 1200 an inference cloud 102 comprising inference points 103 using a trained inference model 104 and the relevant point cloud 100; and computing 1300 a dental preparation score 105 from the inference cloud 102, thereby evaluating the dental preparation.

[0169] When executed by one or more processors of a device, the one or more instructions may cause the device to communicate a dental preparation score 105 and display color-coded risk indicators for at least one of an undercut, a thickness, and a margin line 7070. The executed instructions may cause the device to calculate and communicate a dental preparation score 105, which may be a comprehensive evaluation of the readiness of the dental preparation for supporting a restoration. The dental preparation score 105 may aggregate various individual metrics, such as undercut presence, thickness adequacy, and margin line precision, providing a clear and quantitative measure of the quality of the preparation. The instructions may also cause the device to display color-coded risk indicators for particular aspects of thepreparation, such as undercuts, thickness, and margin lines. These risk indicators may serve as visual tools, highlighting areas where the preparation may require attention or adjustment. For example, for undercuts, indicators may highlight regions where the tooth structure overhangs, potentially interfering with restoration placement. Color codes may range from green (no risk) to red (high risk), guiding practitioners to areas that may need modification. Indicators may show sections where the remaining tooth structure is too thin to meet restoration requirements, providing insights into necessary reinforcement. Indicators may illustrate the precision and clarity of the margin line, showing whether it adequately defines the restoration boundary without irregularities.

[0170] In one embodiment, an assessment comprising removal of additional material to improve the dental preparation score 105 may be provided. Areas where material should be added may be displayed. Areas where material should be removed may also be displayed, and these areas may include surrounding teeth. The assessment may identify areas on the dental preparation where excess material should be removed to improve the dental preparation score 105, which may include refining the contours or addressing bulkiness in the preparation that could hinder the proper seating or alignment of the dental restoration. Importantly, the assessment may include surrounding teeth where the removal of material may alleviate interferences or occlusal mismatches, contributing to a better fit and alignment with the dental arch. Areas identified for material reduction may be visually highlighted on a display, using techniques like color coding or annotations. The visualization aids dental professionals in easily identifying specific regions that require modifications, providing a clear and direct reference for executing changes during preparation adjustments. In conjunction with removal, the assessment may also identify regions where material should be added to the preparation. Such additions may enhance structural integrity or achieve adequate thickness for supporting the intended restoration, ensuring that preparation meets strength and stability requirements. Similar to areas of removal, regions where material addition is warranted may be displayed, guiding practitioners toward building up specific sections of the preparation. Strengthening thin areas or adjusting contours to achieve better conformity with the restoration design may involve specific changes to the preparation.

[0171] In one embodiment, the device may suggest a rescan of a highlighted document. The device may analyze the quality and completeness of the digital scans that serve as the basis for evaluating the dental preparation. If deficiencies orinaccuracies are detected in the scan data — such as missing details, artifacts, or inadequate resolution — the device may suggest performing a rescan. The recommendation is aimed at ensuring the highest accuracy and detail in the data used for generating the dental preparation score 105. The rescan suggestion may be triggered by specific issues highlighted in the document, such as areas where the scan data is incomplete or where discrepancies between expected and captured geometry are observed. By identifying these issues, the device helps practitioners pinpoint exactly where improvements in scan quality can enhance the overall evaluation. The device may communicate the need for a rescan through visual cues or alerts, possibly highlighting the affected areas on the document or providing guidance on achieving improved scan results. Recommendations on scanning techniques or changes in device settings to capture missing aspects of the dental preparation accurately may involve specific adjustments or techniques.

[0172] In one embodiment, areas of contact and a contact score may be displayed. The device may visually highlight the areas where the dental restoration makes contact with adjacent structures, such as opposing teeth or gingival tissue. These contact areas are crucial for understanding how the restoration will function within the oral environment, providing insight into its stability, alignment, and occlusion. By displaying contact areas, the device helps dental practitioners assess whether the restoration achieves the desired interactions with neighboring structures. The visualization can reveal potential issues, such as uneven contact distribution or excessive pressure points, that might compromise restoration performance or cause discomfort to the patient. Alongside the visualization, the device may compute a contact score that quantifies the quality and distribution of contact between the restoration and adjacent oral structures. The contact score provides a numerical measure indicating how well the contact areas align with predefined clinical standards or guidelines, assessing factors such as evenness, extent, and pressure distribution.

[0173] Reference is now made to the drawings in which Figure 6 shows a logical modular representation of an exemplary system 2000 comprising a network node 2100. The network node 2100 comprises a memory module 2160, a processor module 2120, an evaluation module 2130 and a network interface module 2170. The network node 2100 may also include a user interface module 2150.

[0174] The system 2000 may comprise a storage system 2300 for storing and accessing long-term (i.e., non-transitory) data and may further log data while thenetwork node 2100 is being used. Figure s shows examples of the storage system 2300 as a distinct database system 2300A, a distinct module 2300C of the network node 2100 ora sub-module 2300B of the memory module 2160 of the network node 2100. The storage system 2300 may be distributed over different systems A, B, C. The storage system 2300 may comprise one or more logical or physical as well as local or remote hard disk drive (HDD) (or an array thereof). The storage system 2300 may further comprise a local or remote database made accessible to the network node 2100 by a standardized or proprietary interface or via the network interface module 2170.

[0175] The network interface module 2170 represents at least one physical interface that can be used to communicate with other network nodes. The network interface module 2170 may be made visible to the other modules of the network node 2100 through one or more logical interfaces. The actual stacks of protocols used by the physical network interface(s) and / or logical network interface(s) 2172-2178 of the network interface module 2170 do not affect the teachings of the present invention.

[0176] The processor module 2120 may represent a single processor with one or more processor cores or an array of processors, each comprising one or more processor cores. The memory module 2160 may comprise various types of memory (different standardized or kinds of Random Access Memory (RAM) modules, memory cards, Read-Only Memory (ROM) modules, programmable ROM, etc.).

[0177] A bus 2180 is depicted as an example of means for exchanging data between the different modules of the network node 2100. The teachings presented herein are not affected by the way the different modules exchange information. For instance, the memory module 2160 and the processor module 2120 could be connected by a parallel bus, but could also be connected by a serial connection or involve an intermediate module (not shown) without affecting the teachings of the present invention.

[0178] The data acquisition module may serve within the network node 2100 for collecting and processing data necessary to evaluate a dental preparation. The module may assemble the relevant point cloud 100 from the dental preparation point cloud 101 by capturing the three-dimensional geometry and surface details of the tooth or teeth to be restored. Imaging technologies, such as intraoral scanners, CBCT scanners, or 3D cameras, may be interfaced with the module to obtain high-resolution scans of thedental preparation, capturing detailed representations of the dental structures. The module may preprocess the captured data, performing noise reduction, distortion correction, and alignment of the scans to ensure the data is ready for analysis. To manage data complexity, the dental preparation point cloud 101 may be decimated into a decimated point cloud 108, reducing the number of data points while preserving critical geometric features. Sampling techniques, such as Furthest Point Sampling FPS, may be used to accurately represent the geometry of the dental preparation within the point cloud, improving the accuracy of the inference cloud 102. The module may access historical or reference data from external databases or storage systems 2300 for comparison against standardized models or previous scans, enhancing the reliability of the evaluation. Real-time or near-real-time data processing may enable immediate feedback during the scanning process. Through these functionalities, the data acquisition module may provide a precise foundation for evaluating dental preparations, supporting effective restoration processes.

[0179] The evaluation module 2130 may be used to provide dental preparation evaluation-related services to the network node 2100, wherewith the data obtained from the relevant point cloud 100 and the inference cloud 102 may be processed to compute the dental preparation score 105. Algorithms and models, including the trained inference model 104, may be utilized by the evaluation module 2130 to assess various aspects of the dental preparation such as occlusal fit, margin line precision, and undercut presence. Through the processing capabilities thereof, the evaluation module 2130 may enable the system 2000 to deliver comprehensive feedback on the readiness and suitability of a dental preparation for restoration, thereby facilitating informed decision-making by dental professionals. Further functionalities and interactions of the evaluation module 2130 with other components of the network node 2100 may be elaborated upon in subsequent sections.

[0180] The variants of processor module 2120, memory module 2160 and network interface module 2170 usable in the context of the present invention will be readily apparent to persons skilled in the art. Likewise, even though explicit mentions of the evaluation module 2130, the memory module 2160, the user interface module 2150 and / or the processor module 2120 are not made throughout the description of the present examples, persons skilled in the art will readily recognize when such modules are used in conjunction with other modules of the network node 2100 to perform routine as well as innovative elements presented herein.

[0181] Various network links may be implicitly or explicitly used in the context of the present invention. While a link may be depicted as a wireless link, it could also be embodied as a wired link using a coaxial cable, an optical fiber, a category 5 cable, and the like. A wired or wireless access point (not shown) may be present on the link between. Likewise, any number of routers (not shown) may be present and part of the link, which may further pass through the Internet.

[0182] The present invention is not affected by the way the different modules exchange information between them. For instance, the memory module and the processor module could be connected by a parallel bus, but could also be connected by a serial connection or involve an intermediate module (not shown) without affecting the teachings of the present invention.

[0183] The invention described herein is not to be limited to the particular embodiments described hereinabove, as variations of these embodiments may be made and still fall within the scope of the appended claims. It is also to be understood that the terminology employed is for the purpose of describing particular embodiments; and is not intended to be limiting. Instead, the scope of the present invention will be established by the appended claims.

[0184] In order to provide a clear and consistent understanding of the terms used in the present specification, a number of definitions are provided below. Moreover, unless defined otherwise, all technical and scientific terms as used herein have the same meaning as commonly understood to one of ordinary skill in the art to which this disclosure pertains.

[0185] Use of the word “a” or “an” when used in conjunction with the term “comprising” in the claims and / or the specification may mean “one,” but it is also consistent with the meaning of “one or more”, “at least one”, and “one or more than one.” Similarly, the word “another” may mean at least a second or more.

[0186] Use of the expression “at least one of” followed by a set of elements suggests that any combination of the elements from the set is being considered, including a single element from the set, and all elements from the set. For clarity, “at least one of” followed by a set does not refer to having at least the whole set once, and possibly multiple times.

[0187] As used in this specification and claim(s), the words “comprising” (and any form of comprising, such as “comprise” and “comprises”), “having” (and any form ofhaving, such as “have” and “has”), “including” (and any form of including, such as “include” and “includes”) or “containing” (and any form of containing, such as “contain” and “contains”), are inclusive or open-ended and do not exclude additional, unrecited elements or process steps.

[0188] As will be understood by a skilled person, other variations and combinations may be made to the various embodiments of the invention as described herein above. The scope of the claims should not be limited by the preferred embodiments set forth; but should be given the broadest interpretation consistent with the description as a whole.

Claims

CLAIMSWhat is claimed is:

1. A method (1000) for evaluating a dental preparation for a dental restoration, comprising:- assembling (1100) a relevant point cloud (100) from a dental preparation point cloud (101 ) of the dental preparation;- predicting (1200) an inference cloud (102) comprising inference points (103) using a trained inference model (104) and the relevant point cloud (100); and- computing (1300) a dental preparation score (105) from the inference cloud (102), thereby evaluating the dental preparation.

2. The method of claim 1 , wherein the relevant point cloud (100) comprises at least one surface point (106) on a surface of a tooth and at least one internal point (107) internal to the tooth.

3. The method of claim 1 , wherein the relevant point cloud (100) comprises at least one surface point (106) on a surface of a tooth and at least one external point (207) external to the tooth.

4. The method of claim 1 , wherein the relevant point cloud (100) is a decimated point cloud (108), and wherein assembling (1100) the relevant point cloud (100) comprises:- decimating (1110) the dental preparation point cloud (101) into the decimated point cloud (108), thereby reducing a complexity of the relevant point cloud (100).

5. The method of claim 4, wherein the decimated point cloud (108) comprises between 250 and 15,000 points.

6. The method of claim 4, wherein the dental preparation score (105) comprises a scan quality score (109), the method further comprising:- computing (1010) a dental preparation mesh (110) of the dental preparation from a dental preparation scan (111);- converting (1020) the dental preparation mesh (110) into the dental preparation point cloud (101); and- computing (1030) the scan quality score (109) from the dental preparation scan (111).

7. The method of claim 6, wherein the scan quality score (109) comprises at least one of a completeness score (112), and a registration score (113).

8. The method of claim 1 , wherein assembling (1100) the relevant point cloud (100) comprises:- sampling (1120) the relevant point cloud (100) using Furthest Point Sampling (FPS), thereby improving accuracy of the inference cloud (102).

9. The method of claim 1 , wherein the trained inference model (104) comprises a trained Dynamic Graph Convolutional Neural Network (DGCNN) and a trained transformer-based encoder-decoder, and wherein predicting (1200) the inference cloud (102) comprises:- computing (1210) a feature cloud (114) by extracting a plurality of features (115) from the relevant point cloud (100) using the trained DGCNN; and- predicting (1220) the inference cloud (102) by processing the feature cloud (114) using the trained transformer-based encoder-decoder.

10. The method of claim 1 , wherein the dental preparation score (105) is computed using an inference spline (116), and wherein computing (1300) the dental preparation score (105) comprises:- ordering (1310) the inference points (103) into an ordered point cloud (117); and- generating (1320) the inference spline (116) from the ordered point cloud (117).

11. The method of claim 10, wherein ordering the inference points (103) is performed using a nearest neighbor algorithm.

12. The method of claim 10, wherein the inference spline (116) is a b-spline.

13. The method of claim 10, wherein generating (1320) the inference spline (116) comprises:- computing a predicted inference spline (118) by fitting the predicted inference spline (118) to the ordered point cloud (117); and- adjusting one or more control points (119) of the predicted inference spline (118), thereby generating the inference spline (116).

14. The method of claim 1 , wherein predicting (1200) the inference cloud (102) comprises:- discarding (1230) inference outliers (120) from the inference cloud (102), thereby improving an accuracy of the inference cloud (102).

15. The method of claim 14, wherein discarding (1230) the inference outliers (120) comprises:- evaluating (1240) a local point density (121) for a candidate point (122) from the inference points (103) by calculating an average distance therefrom to one or more nearest inference points (103);- assessing (1250) an alignment of the candidate point (122) with a main trend of one or more nearby inference points (103);- computing (1260) an outlier score (123) from the local point density (121) and the alignment; and- discarding (1270) the candidate point (122) when the outlier score (123) is above a maximum outlier score threshold (124).

16. The method of claim 15, wherein the dental preparation score (105) comprises a confidence score (125), the method further comprising:- computing (1280) the confidence score (125) from a percentage of the inference outliers (120).

17. The method of claim 16, further comprising:- displaying (1910) a visual representation (126) of the dental preparation, the visual representation (126) comprising areas communicating the confidence score (125).

18. The method of claim 17, further comprising- suggesting (1920) actions to improve the confidence score (125).

19. The method of claim 1 , further comprising:- obtaining (1410) a dental restoration mesh (127) prepared for the dental preparation;- aligning (1420) the dental restoration mesh (127) with the dental preparation point cloud (101); and wherein computing the dental preparation score (105) is performed by measuring an alignment between the inference cloud (102) and the dental restoration mesh (127).

20. The method of claim 18, wherein the dental restoration mesh (127) is a generated dental restoration mesh (128) generated from the preparation point cloud.

21. The method of claim 18, wherein the dental preparation score (105) comprises a gap analysis score (129), the method further comprising:- computing (1340) the gap analysis score (129) from an average deviation and a maximum deviation between a dental preparation surface and the dental restoration mesh (127).

22. The method of claim 21 , wherein computing the gap analysis score (129) comprises:- calculating (1330) a restoration distance (130) between the dental restoration mesh (127) and the inference cloud (102).

23. The method of claim 22, wherein the restoration distance (130) comprises at least one of a Chamfer Distance (CD), a Hausdorff Distance, and an Earth Mover’s Distance (EMD).

24. The method of claim 18, wherein the dental preparation score (105) comprises a contact area score (131), the method further comprising:- computing (1350) the contact area score (131) from a percentage of the inference cloud (102) in contact with the dental restoration mesh (127).

25. The method of claim 18, wherein the dental preparation score (105) comprises an occlusal fit score (132), the method further comprising:- simulating (1360) a dental occlusion (133) of the dental restoration mesh (127) aligned onto the preparation point cloud; and- computing (1370) the occlusal fit score (132) from the dental occlusion (133).

26. The method of claim 25, wherein the occlusal fit score (132) comprises at least one of an occlusal clearance score (134), a contact points score (135), and an alignment score (136).

27. The method of claim 1 , further comprising:- suggesting (1930) actions to improve the dental preparation score (105).

28. The method of claim 27, wherein the actions comprise at least one of a tooth reduction (137), a tooth contouring (138), a margin definition (139), and a core buildup (140).

29. The method of claim 1 , wherein the dental restoration comprises at least one of a crown, an implant, and a dental bridge.

30. The method of claim 1 , wherein the dental preparation score (105) comprises at least one of a margin line score (141), a gingival proximity score (142), an undercut score (143), an occlusal reduction score (144), an axial reduction score (145), a smoothness score, a thickness score (147), an interproximal clearance score (148),an alignment score (136), a path of insertion score (149), a contact area score (131 ), and an abutment spacing score (150).31 . The method of claim 1 , wherein:- the trained inference model (104) is trained for detecting margin line points located on a margin line of the dental preparation; and- the dental preparation score (105) comprises a margin line score (141).

32. The method of claim 31 , wherein:- the trained inference model (104) is trained to predict undercut points (151 ) lying within an undercut from the relevant point cloud (100) and the margin line points; and- the dental preparation score (105) comprises an undercut score (143).

33. The method of claim 32 further comprising:- generating an undercut area from the undercut points (151);- computing an undercut profile (152) comprising an undercut height measured from the undercut area; and- computing the undercut score (143) from the undercut profile (152).

34. The method of claim 1 , wherein:- the trained inference model (104) is trained to detect points located on a gingival margin of the dental preparation; and- the dental preparation score (105) comprises a gingival proximity score (142).

35. The method of claim 1 , wherein:- the trained inference model (104) is a thickness model (153) trained to predict thin points exhibiting insufficient thickness from the relevant point cloud (100); and- the dental preparation score (105) comprises a thickness score (147).

36. The method of claim 35, further comprising:- generating a thin surface (154) from the thin points;- computing a thickness profile (155) from the thin surface (154); and- computing the thickness score (147) from the thickness profile (155).

37. The method of claim 35, wherein the thickness score (147) is computed according to a property of a material used for the dental restoration.

38. The method of any one of claims 1 to 37, wherein the dental restoration is a dental bridge.

39. The method of claim 38, wherein:- the trained inference model (104) is an abutment alignment model (156) trained to detect alignment points describing an alignment of abutment teeth; and- the dental preparation score (105) comprises an alignment score (136).

40. The method of claim 39, further comprising:- fitting an alignment spline (157) through the alignment points;- evaluating a curvature of the alignment spline (157) against an alignment threshold (158);- computing the alignment score from the curvature of the alignment spline (157); and wherein the alignment spline (157) is the inference spline (116) when the method is dependent upon claim 10.41 . The method of claim 38, wherein:- the trained inference model (104) is an abutment spacing model (159) trained to detect abutment spacing points located between abutment teeth; and- the dental preparation score (105) comprises an abutment spacing score (150).

42. The method of claim 41 , further comprising:- fitting an abutment spacing spline (160) through the abutment spacing points;- calculating a distance along the abutment spacing spline (160) to determine an available abutment spacing between abutment teeth; and- computing the abutment spacing score from the available abutment spacing.

43. The method of claim 38, wherein:- the trained inference model (104) is a gingival clearance model (161) trained to detect gingival clearance points located in a space between an edge of the dental restoration and a gingival margin; and- the dental preparation score (105) comprises a gingival proximity score (142).

44. The method of claim 43, further comprising:- fitting a gingival clearance spline (162) through the gingival clearance points;- calculating a gingival clearance distance along the gingival clearance spline (162) to determine an available clearance; and- computing the gingival proximity score (142) from the gingival clearance distance.

45. The method of claim 38, wherein the dental preparation score (105) comprises a load distribution score (163), the method further comprising:- simulating occlusal forces on the dental restoration mesh (127) aligned onto the preparation point cloud;- computing a stress distribution assessment (164) from the occlusal forces across abutment teeth and the dental bridge; and- computing the load distribution score (163) from the stress distribution assessment (164).

46. The method of claim 45, wherein the stress distribution assessment (164) is computed in consideration of a property of a material of the dental bridge.

47. The method of claim 46, wherein the dental preparation score (105) comprises an alignment score (136), the method further comprising:- computing an alignment spline (157) from the dental restoration mesh (127) aligned onto the preparation point cloud; and- computing the alignment score from the alignment spline (157).

48. The method of claim 38, wherein the dental preparation score (105) comprises an abutment space score (165), the method further comprising:- determining an available abutment space (166) between abutment teeth from the dental restoration mesh (127) aligned onto the preparation point cloud; and- computing the abutment space score (165) from the available abutment space (166).

49. The method of claim 38, wherein the dental preparation score (105) comprises a gingival proximity score (142), the method further comprising:- simulating gingival margin clearance (167) from the dental restoration mesh (127) aligned onto the preparation point cloud; and- computing the gingival proximity score (142) from the gingival margin clearance (167).

50. The method of claim 1 , wherein computing (1300) the dental preparation score (105) further comprises:- determining at least one clearance distance between the dental preparation and a surrounding anatomy; and- evaluating the at least one clearance distance against at least one clearance threshold while generating an actual dental restoration mesh is unnecessary.

51. The method of claim 50, wherein evaluating the at least one clearance distance comprises determining that there exists at least one achievable dental restoration having at least a minimum prescribed thickness based on the at least one clearance distance.

52. The method of claim 1 , wherein computing (1300) the dental preparation score (105) comprises:- computing a first-stage feasibility outcome while generating an actual dental restoration mesh is unnecessary; and- computing a second-stage outcome, after the generating of the actual dental restoration mesh, wherein the actual dental restoration mesh is aligned with the dental preparation point cloud.

53. The method of claim 1 , wherein the dental preparation score (105) comprises a taper score, and wherein computing the taper score comprises:- determining a path of insertion for a dental restoration;- determining a taper angle of at least one axial wall of the dental preparation relative to the path of insertion; and- evaluating the taper angle against at least one taper threshold or taper range.

54. The method of claim 1 , wherein the dental preparation score (105) comprises a corner-rounding score, and wherein computing the corner-rounding score comprises identifying at least one transition region between an occlusal surface andan axial surface of the dental preparation and evaluating a sharpness metric at the at least one transition region against a minimum rounding threshold.

55. The method of claim 54, wherein the minimum rounding threshold is based on a characteristic dimension of a milling tool used to fabricate the dental restoration.

56. The method of claim 1 , further comprising classifying the dental preparation as being within an acceptable range or outside the acceptable range based on at least one component of the dental preparation score (105).

57. The method of claim 1 performed in at least one of a training mode and a practitioner mode, wherein at least one evaluation feature is enabled by default in the training mode and is user-activatable in the practitioner mode.

58. A non-transitory computer-readable medium storing a set of instructions for executing the method of any one of claims 1 to 57, the set of instructions comprising:- one or more instructions that, when executed by one or more processors of a device, cause the device to:- communicate a dental preparation score (105); and- display color-coded risk indicators for at least one of an undercut, a thickness, and a margin line.

59. The non-transitory computer-readable medium of claim 58, wherein the one or more instructions further cause the device to:- provides an assessment comprising removal of additional material to improve the dental preparation score (105);- display areas where material should be added; and- display areas where material should be removed;60. The non-transitory computer-readable medium of claim 58, wherein the one or more instructions further cause the device to suggest a rescan of a highlighted document.61 . The non-transitory computer-readable medium of claim 58 wherein the one or more instructions further cause the device to display areas of contact and a contact score.

62. The non-transitory computer-readable medium of claim 61 , wherein the display areas where material should be removed include surrounding teeth.

63. A system (2000) for evaluating a dental preparation for a dental restoration, comprising:- a data acquisition module configured to assemble a relevant point cloud (100) from a dental preparation point cloud (101 ) of the dental preparation;- a processing module (2120) configured to predict an inference cloud (102) comprising inference points (103) using a trained inference model (104) and the relevant point cloud (100); and- an evaluation module (2130) configured to compute a dental preparation score (105) from the inference cloud (102), thereby evaluating the dental preparation; and wherein the system is configured to perform the method of any one of claims 1 to 57.

64. A device (2100) for evaluating a dental preparation for a dental restoration, comprising:- one or more processors (2120) configured to execute a set of instructions for:- assembling a relevant point cloud (100) from a dental preparation point cloud (101 ) of the dental preparation;- predicting an inference cloud (102) comprising inference points (103) using a trained inference model (104) and the relevant point cloud (100); and- computing a dental preparation score (105) from the inference cloud (102), thereby evaluating the dental preparation;- a display (2150) configured to show color-coded risk indicators for at least one of an undercut, a thickness, and a margin line; and wherein the device is configured to perform the method of any one of claims 1 to 57.

Images