MESH

TECHNICAL FIELD

The disclosure relates generally to manipulation of elements that are represented by a three-dimensional mesh and more

particularly to methods and apparatus for tooth crown surface

characterization in a surface contour image that has been obtained using reflectance imaging.

BACKGROUND

Three-dimensional (3-D) imaging and 3-D image processing are of growing interest to dental/orthodontic practitioners for computer-aided diagnosis, for prosthesis design and fabrication, and for overall improved patient care. For cephalometric measurement and analysis, 3-D imaging and 3-D image processing offer significant advantages in terms of flexibility, accuracy, and repeatability. 3- D cephalometric analysis overcomes some of the shortcomings associated with conventional methods of two-dimensional (2-D) cephalometric analysis, such as 2-D geometric errors of perspective projection, magnification, and head positioning in projection, for example. 3-D cephalometrics has been shown to yield objective data that is more accurate, since it is based on calculation rather than being largely dependent upon discrete measurements, as is the case with 2-D cephalometrics.

Early research using 3-D cephalometrics methods employed 3-D imaging and parametric analysis of maxillo-facial anatomical structures using cone beam computed tomography (CBCT) of a patient's head. Using CBCT methods, a significant role of the 3-D cephalometric analysis was to define mathematical models of maxillary and mandibular arches for which the axes of inertia were calculated for each tooth or group of teeth. This, in turn, required the segmentation of individual teeth from the acquired CBCT head volume of a patient.

Conventionally, during an orthodontic treatment procedure, multiple 2-D X-ray cephalogram acquisitions are used to assess treatment progress. Conventional 3-D cephalometric analysis can also be used for this purpose, requiring multiple CBCT scans. However, both 2-D and 3-D

radiographic imaging methods expose the patient to ionizing radiation. Reducing overall patient exposure to radiation is desirable, particularly for younger patients.

Optical intraoral scans, in general, produce contours of dentition objects and have been helpful in improving visualization of teeth, gums, and other intra-oral structures. Surface contour characterization using visible or near- visible light can be particularly useful for assessment of tooth condition and has recognized value for various types of dental procedures, such as for restorative dentistry. This can provide a valuable tool to assist the dental practitioner in identifying various problems and in validating other measurements and observations related to the patient's teeth and supporting structures. Surface contour information can also be used to generate 3-D models of dentition components such as individual teeth; the position and orientation information related to individual teeth can then be used in assessing orthodontic treatment progress. With proper use of surface contour imaging, the need: for multiple 2-D or 3-D X-ray acquisitions of a patient's dentition can be avoided.

A number of techniques have been developed for obtaining surface contour information from various types of objects in medical, industrial, and other applications. Optical 3 -dimensional (3-D) measurement methods provide shape and spatial information using light directed onto a surface in various ways.

Among types of imaging methods used for contour imaging are fringe or structured light projection devices. Structured light projection imaging uses patterned or structured light and camera/sensor triangulation to obtain surface contour information for structures of various types. Once the structured light projection images are processed, a point cloud can be generated. A mesh can then be formed from the point cloud or a plurality of point clouds, in; order to reconstruct at least a planar approximation to the surface.

Mesh representation can be particularly useful for showing surface structure of teeth and gums and can be obtained using a handheld camera and without requiring harmful radiation levels. However, when using conventional image processing approaches, mesh representation has been found to lack some of the inherent versatility and utility that is available using cone-beam computed tomography (CBCT) or other techniques that expose the patient to radiation. One area in which mesh representation has yielded only disappointing results relates to segmentation. Segmentation allows the practitioner to identify and isolate the crown and other visible portions of the tooth from gums and related supporting structure. Conventional methods for segmentation of mesh images can often be inaccurate and may fail to distinguish tooth structure from supporting tissues.

Various approaches for addressing the segmentation problem for mesh images have been proposed, such as the following:

(i) A method described in the article "Snake-Based Segmentation of Teeth from Virtual Dental Casts" by Thomas Kronfeld et al. (in Computer-Aided Design & applications, 7(a), 2010) employs an active contour

segmentation method that attempts to separate every tooth and gum surface in a single processing iteration. The approach that is described, however, is not a topology-independent method and can fail, particularly where there are missing teeth in the jaw mesh.

(ii) An article entitled "Perception-based 3D Triangle Mesh Segmentation Using Fast Marching Watershed" by Page, D.L. et al. (in Proc. CVPIwo II 2003) describes using a Fast Marching Watershed method for mesh segmentation. The Fast Marching Watershed method that is described requires the user to enter seed points manually. The seed points must be placed at both sides of the contours of the regions under segmentation. The method then attempts to segment all regions in one step, using seed information. For jaw mesh segmentation, this type of method segments each tooth as well as the gum at the same time. This makes the method less desirable, because segmenting teeth from the gum region typically requires parameters and processing that differ from those needed for the task of segmenting teeth from each other. Using different segmentation strategies for different types of dentition components with alternate segmentation requirements would provide better performance.

(iii) For support of his thesis, "Evaluation of software developed for automated segmentation of digital dental models", J.M. Moon used a software tool that decomposed the segmentation process into two steps: separation of teeth from gingival structure and segmentation of whole arch structure into individual tooth objects. The software tool used in Moon's thesis finds maximum curvature in the mesh and requires the user to manually choose a curvature threshold to obtain margin vertices that are used for segmenting the tooth. The software also requires the user to manually edit margins in order to remove erroneous segmentation results. Directed to analysis of shape and positional characteristics, this software tool does not employ color information in the separation of teeth regions from the gum regions.

(iv) U.S. Patent application 20030039389 Al entitled "Manipulating a digital dentition model to form models of individual dentition

components" by Jones, T. N. et al. discloses a method of separating portions of the dentition model representing the adjacent teeth.

While conventional methods for tooth segmentation exhibit some level of success in a limited set of test cases, none of these methods appears to be robust and commercially viable. In addition, conventional methods do not appear to be able to properly segment orthodontic braces and brackets that frequently appear in scanned dentition mesh models.

At different intervals during the orthodontic treatment process, it is desirable to remove the physical bracket braces from the teeth before performing intraoral scanning in order to obtain a clear 3D view of the teeth mesh model for progress assessment. However, due to factors such as de-bonding, staining, and plaque accumulation on rough surfaces of the tooth, the enamel can be damaged by removing the braces. The enamel thickness lost during bracket removal has been estimated to be approximately 150 micron. To prevent damage and enamel loss, there would be advantages in forgoing removal of the brace features if possible. One solution is to scan the dentition/dental arch without removing the physical braces from the teeth, and clean up the dental arch mesh by mesh manipulation.

US 8,738,165 to Cinader Jr. et al., entitled "Methods of preparing a virtual dentition model and fabricating a dental retainer therefrom", discloses a virtual model of a dental patient's dentition provided by obtaining a digital data file of the patient's teeth and orthodontic appliances connected to the teeth, and combined with data from the data file with other data that represents surfaces of the teeth underlying the appliances. In the Ί65 disclosure, the virtual model is used in preparing a physical model of the patient's current dentition that can be used to make a dental retainer. The Ί65 disclosure also notes editing tools used in image manipulating software to remove the data representing the orthodontic appliances. Image manipulating software described in the Ί65 disclosure is "Geomagic Studio" (from Geomagic, Inc. of Research Triangle Park, N.C.), in which portions of an image are identified and deleted by a technician using a computer mouse or other input device. The US 8,738,165 disclosure further mentions software known as "ZBrush" (from Pixologic, Inc. of Los Angeles, Calif.) used to digitally/manually fine-tune and sculpt the combined data. These methods can require considerable operator skill and results can be highly subjective.

There is, then, a need for improved methods and/or apparatus, preferably with little or no human intervention, for segmentation of mesh representations of tooth and gum structures including bracket removal for tooth/crown surface reconstruction. SUMMARY

An aspect of this application is to advance the art of tooth segmentation and/or manipulation in relation to volume imaging and visualization used in medical and dental applications.

Another aspect of this application is to address, in whole or in part, at least the foregoing and other deficiencies in the related art. It is another aspect of this application to provide, in whole or in part, at least the advantages described herein.

Certain exemplary method and/or apparatus embodiments according to the present disclosure can address particular needs for improved visualization and assessment of 3D dentition models, where brace representations have been removed or reduced and tooth surfaces added or restored for clarity. Restored 3D dentition models can be used with internal structures obtained using CBCT and other radiographic volume imaging methods or can be correlated to reflectance image data obtained from the patient.

These objects are. given only by way of illustrative example, and such objects may be exemplary of one or more embodiments of the application. Other desirable objectives and advantages inherently achieved by exemplary method and/or apparatus embodiments may occur or become apparent to those skilled in the art. The invention is defined by the appended claims.

According to one aspect of the disclosure, there is provided a method for generating a digital model of reconstructed dentition that can include obtaining a 3-D digital mesh model of the patient's dentition including braces, teeth, and gingival, modifying the 3-D digital mesh dentition model by removing wire portions of the braces therefrom, modifying the 3-D digital mesh dentition model by removing bracket portions of the braces therefrom, approximating teeth surfaces of the modified 3-D digital mesh dentition model previously covered by the wire portions and the bracket portions of the braces, and displaying, storing, or transmitting over a network to another computer, the reconstructed 3-D digital mesh dentition model.

According to another aspect of the disclosure, there is provided a method for generating a digital model of a patient's dentition, the method executed at least in part by a computer that can include acquiring a 3-D digital mesh that is representative of the patient's dentition along a dental arch, wherein the digital mesh includes braces, teeth, and gingival tissue; modifying the 3-D digital mesh to generate a digital mesh dentition model by: (i) processing the digital mesh and automatically detecting one or more initial bracket positions from the acquired mesh; (ii) processing the initial bracket positions to identify bracket areas for braces that lie against tooth surfaces; (iii) identifying one or more brace wires extending between brackets; (iv) removing one or more brackets and one or more wires from the dentition model; (v) forming a reconstructed tooth surface within the digital mesh dentition model where the one or more brackets have been removed; and displaying, storing, or transmitting over a network to another computer, the modified 3-D digital mesh dentition model. BRIEF DESCRIPTION OF THE DRAWINGS

The foregoing and other objects, features, and advantages of the invention will be apparent from the following more particular description of the embodiments of the invention, as illustrated in the accompanying drawings.

Elements of the drawings are not necessarily to scale relative to each other.

Figure 1 is a schematic diagram that shows components of an imaging apparatus for surface contour imaging of a patient's teeth and related structures.

Figure 2 shows schematically how patterned light is used for obtaining surface contour information using a handheld camera or other portable imaging device.

Figure 3 shows an example of surface imaging using a pattern with multiple lines of light.

Figure 4 shows a point cloud generated from structured light imaging, such as that shown in Figure 3.

Figure 5 shows a polygon mesh in the simple form of a triangular mesh.

Figure 6A is a logic flow diagram that shows a hybrid sequence for mesh segmentation according to an embodiment of the present disclosure.

Figure 6B is a logic flow diagram that shows a workflow sequence for hybrid segmentation of the tooth according to an embodiment of the present disclosure.

Figure 7A shows an example of a poorly segmented tooth.

Figure 7B shows an example of an improved segmentation.

Figure 8A shows an example of a seed line trace pattern.

Figure 8B shows an example of a block line trace pattern.

Figures 9A, 9B and 9C show operator interface screens for review and entry of markup instructions for refining tooth mesh segmentation processing according to certain embodiments of the present disclosure.

Figure 10 is a logic flow diagram that shows a sequence for bracket removal from the tooth mesh surface according to an exemplary embodiment of the application. Figure 11 shows an example of a dentition mesh containing teeth, brackets and gingival.

Figure 12 is a diagram that shows exemplary resultant separated teeth from a 3D dentition mesh according to an exemplary embodiment of the application.

Figures 13A-13C shows an example of removing a bracket from a tooth surface of a 3D dentition mesh and reconstructing the tooth surface afterwards.

Figures 13D and 13E are diagrams that show a hole on tooth mesh surface where a bracket is removed and an initial patch approximated to fill the hole.

Figure 13F is a diagram that shows an initial arrangement of triangles in a tooth surface mesh patch and a modified arrangement of triangles for a tooth surface mesh patch.

Figure 13G shows an exemplary corrected 3D dentition mesh.

Figure 14 is a logic flow diagram that shows an exemplary sequence for automatic braces and brackets detection and removal by processing logic according to an embodiment of the present disclosure.

Figure 15 is a logic flow diagram showing a process for bracket area detection.

Figure 16 shows images for a sequence that follows steps given in

Figure 15.

Figure 17 shows exemplary coarse brackets obtained using the described sequence.

Figure 18 shows braces wire detection.

Figure 19 shows an arrangement of vertices for mask generation.

Figure 20 shows the pruning operation for masks that can inaccurately extend to the opposite side in schematic representation.

Figure 21 shows a post-processing sequence.

Figure 22 shows an exemplary Fast Marching process.

Figure 23 shows exemplary Fast March computation for arrival time from different seed-points along mask boundaries. Figure 24 shows results of using a sequence of different approaches for refinement of bracket regions according to an embodiment of the present disclosure.

Figure 25 shows steps of an optional refinement of bracket regions using a convex hull computation.

Figure 26 shows fine tuned bracket regions obtained using the described sequence.

Figure 27 shows the recovered tooth surface following bracket removal.

Figure 28 is a logic flow diagram that shows a sequence for bracket removal from the tooth mesh surface according to another exemplary embodiment of the application.

Figure 29 shows an operator interface screen embodiment for review and entry of delineation instructions for separating brackets from tooth mesh according to one exemplary embodiment of the application. Figure 29 also shows an example of a closed contour or snake encircling a bracket.

Figure 30 shows an example of highlighted mesh vertices within a closed contour.

Figure 31 shows an example of a reconstructed tooth surface after the bracket is removed.

Figures 32-34 are diagrams that shows respectively, an example of dentition model with brackets, the same dentition model with brackets identified, and reconstructed teeth after brackets are removed according to one exemplary embodiment of the application.

Figure 35A is a diagram that shows an example of a dentition mesh containing teeth, bridged brackets and gingival tissue.

Figure 35B is a diagram that shows an example dentition mesh with bridges (e.g., wires) between brackets broken according to exemplary embodiments of the application.

Figure 35C is a diagram that shows an example dentition mesh illustrating detection of bridges (e.g., wires). Figure 36 shows example results of bracket removal and surface reconstruction after breaking the bridge wires according to exemplary

embodiments of the application.

Figure 37 is a logic flow diagram that shows a sequence for bridged bracket removal from the tooth mesh surface according to an embodiment of the present disclosure.

Figure 38 is a logic flow diagram that shows a sequence for bridged bracket removal from the tooth mesh surface according to another embodiment of the present disclosure.

DETAILED DESCRIPTION OF EXEMPLARY EMBODIMENTS

The following is a detailed description of exemplary embodiments, reference being made to the drawings in which the same reference numerals identify the same elements of structure in each of the several figures.

Where they are used, the terms "first", "second", and so on, do not necessarily denote any ordinal or priority relation, but may be used for more clearly distinguishing one element or time interval from another.

The term "exemplary" indicates that the description is used as an example, rather than implying that it is ideal or preferred.

The term "in signal communication" as used in the application means that two or more devices and/or components are capable of communicating with each other via signals that travel over some type of signal path. Signal communication may be wired or wireless. The signals may be communication, power, data, or energy signals which may communicate information, power, and/or energy from a first device and/or component to a second device and/or component along a signal path between the first device and/or component and second device and/or component. The signal paths may include physical, electrical, magnetic, electromagnetic, optical, wired, and/or wireless connections between the first device and/or component and second device and/or component. The signal paths may also include additional devices and/or components between the first device and/or component and second device and/or component. In the context of the present disclosure, the terms "pixel" and "voxel" may be used interchangeably to describe an individual digital image data element, that is, a single value representing a measured image signal intensity. Conventionally an individual digital image data element is referred to as a voxel for 3 -dimensional or volume images and a pixel for 2-dimensional (2-D) images. For the purposes of the description herein, the terms voxel and pixel can generally be considered equivalent, describing an image elemental datum that is capable of having a range of numerical values. Voxels and pixels have attributes of both spatial location and image data code value.

"Patterned light" is used to indicate light that has a predetermined spatial pattern, such that the light has one or more features such as one or more discernable parallel lines, curves, a grid or checkerboard pattern, or other features having areas of light separated by areas without illumination. In the context of the present disclosure, the phrases "patterned light" and "structured light" are considered to be equivalent, both used to identify the light that is projected onto the head of the patient in order to derive contour image data.

In the context of the present disclosure, the terms "viewer", "operator", and "user" are considered to be equivalent and refer to the viewing practitioner, technician, or other person who views and manipulates a contour image that is formed from a combination of multiple structured light images on a display monitor.

A "viewer instruction", "operator instruction", or "operator command" can be obtained from explicit commands entered by the viewer or may be implicitly obtained or derived based on some other user action, such as making an equipment setting, for example. With respect to entries entered on an operator interface, such as an interface using a display monitor and keyboard, for example, the terms "command" and "instruction" may be used interchangeably to refer to an operator entry.

In the context of the present disclosure, a single projected line of light is considered a "one dimensional" pattern, since the line has an almost negligible width, such as when projected from a line laser, and has a length that is its predominant dimension. Two or more of such lines projected side by side, either simultaneously or in a scanned arrangement, provide a two-dimensional pattern. In exemplary embodiments, lines of light can be linear, curved or three- dimensional.

The terms "3-D model", "point cloud", "3-D surface", and "mesh" may be used synonymously in the context of the present disclosure. The dense point cloud is formed using techniques familiar to those skilled in the volume imaging arts for forming a point cloud and relates generally to methods that identify, from the point cloud, vertex points corresponding to surface features. The dense point cloud is thus generated using the reconstructed contour data from one or more reflectance images. Dense point cloud information serves as the basis for a polygon model at high density for the teeth and gum surface.

According to the present disclosure, the phrase "geometric primitive" refers to basic 2-D geometric shapes that can be entered by the operator in order to indicate areas of an image. By way of example, and not limitation, geometric primitives can include lines, curves, points, and other open shapes, as well as closed shapes that can be formed by the operator, such as circles, closed curves, rectangles and squares, polygons, and the like.

Embodiments of the present disclosure pro vide exemplary methods and/or apparatus that can help to eliminate the need for multiple CBCT scans for visualization of tooth and jaw structures. Exemplary methods and/or apparatus embodiments can be used to combine a single CBCT volume with optical intraoral scans that have the capability of tracking the root position at various stages of orthodontic treatment, for example. To achieve this, the intraoral scans are segmented so that exposed portions, such as individual tooth crowns, from the intraoral scan can be aligned with the individual tooth and root structure segmented from the CBCT volume.

Figure 1 is a schematic diagram showing an imaging apparatus 70 for projecting and imaging using structured light patterns 46. Imaging apparatus 70 uses a handheld camera 24 for image acquisition according to an embodiment of the present disclosure. A control logic processor 80, or other type of computer that may be part of camera 24 controls the operation of an illumination array 10 that generates the structured light and controls operation of an imaging sensor array 30. Image data from surface 20, such as from a tooth 22, is obtained from imaging sensor array 30 and stored in a memory 72. Control logic processor 80, in signal communication with camera 24 components that acquire the image, processes the received image data and stores the mapping in memory 72. The resulting image from memory 72 is then optionally rendered and displayed on a display 74. Memory 72 may also include a display buffer for temporarily storing display 74 image content.

In structured light projection imaging of a surface, a pattern of lines is projected from illumination array 10 toward the surface of an object from a given angle. The projected pattern from the surface is then viewed from another angle as a contour image, taking advantage of triangulation in order to analyze surface information based on the appearance of contour lines. Phase shifting, in which the projected pattern is incrementally shifted spatially for obtaining additional measurements at the new locations, is typically applied as part of structured light projection imaging, used in order to complete the contour mapping of the surface and to increase overall resolution in the contour image.

The schematic diagram of Figure 2 shows, with the example of a single line of light L, how patterned light is used for obtaining surface contour information using a handheld camera or other portable imaging device. A mapping is obtained as an illumination array 10 directs a pattern of light onto a surface 20 and a corresponding image of a line L' is formed on an imaging sensor array 30. Each pixel 32 on imaging sensor array 30 maps to a corresponding pixel 12 on illumination array 10 according to modulation by surface 20. Shifts in pixel position, as represented in Figure 2, yield useful information about the contour of surface 20. It can be appreciated that the basic pattern shown in Figure 2 can be implemented in a number of ways, using a variety of illumination sources and sequences and using one or more different types of sensor arrays 30. Illumination array 10 can utilize any of a number of types of arrays used for light modulation, such as a liquid crystal array or digital micromirror array, such as that provided using the Digital Light Processor or DLP device from Texas Instruments, Dallas, TX. This type of spatial light modulator is used in the illumination path to change the light pattern as needed for the mapping sequence. By projecting and capturing images that show structured light patterns that duplicate the arrangement shown in Figures 1 and 2 multiple times, the image of the contour line on the camera simultaneously locates a number of surface points of the imaged object. This can speed the process of gathering many sample points, while the plane of light (and usually also the receiving camera) is laterally moved in order to "paint" some or all of the exterior surface of the object with the plane of light.

Figure 3 shows surface imaging using a pattern with multiple lines of light. Incremental shifting of the line pattern and other techniques help to compensate for inaccuracies and confusion that can result from abrupt transitions along the surface, whereby it can be difficult to positively identify the segments that correspond to each projected line. In Figure 3, for example, it can be difficult to determine whether line segment 16 is from the same line of illumination as line segment 18 or adjacent line segment 19.

By knowing the instantaneous position of the camera and the instantaneous position of the line of light within an object-relative coordinate system when the image was acquired, a computer and software can use triangulation methods to compute the coordinates of numerous illuminated surface points. As the plane is moved to intersect eventually with some or all of the surface of the object, the coordinates of an increasing number of points are accumulated. As a result of this image acquisition, a point cloud of vertex points or vertices can be identified and used to represent the extent of a surface within a volume. By way of example, Figure 4 shows a dense point cloud 50 generated from a structured light imaging apparatus, CS 3500 3-D camera made by

Carestream Heath, Inc., Rochester NY, USA, using results from patterned illumination such as that shown in Figure 3. The point cloud 50 models physical location of sampled points on tooth surfaces and other intraoral surfaces or, more generally, of surfaces of a real-world object. Variable resolution can be obtained. The example of Fig. 4 shows an exemplary 100 micron resolution. The points in the point cloud represent actual, measured points on the three dimensional surface of an object. The surface structure can be approximated from the point cloud representation by forming a polygon mesh, in which adjacent vertices are connected by line segments. For a vertex, its adjacent vertices are those vertices closest to the vertex in terms of Euclidean distance.

By way of example, Figure 5 shows a 3-D polygon mesh model 60 in the simple form of a triangular mesh. A triangular mesh forms a basic mesh structure that can be generated from a point cloud and used as a digital model to represent a 3-D object by its approximate surface shape, in the form of triangular plane segments sharing adjacent boundaries. Methods and apparatus for forming a polygon mesh model, such as a triangular mesh or more complex mesh structure, are well known to those skilled in the contour imaging arts. The polygon unit of the mesh model, and relationships between neighboring polygons, can be used in embodiments of the present disclosure to extract features (e.g., curvatures, minimum curvatures, edges, spatial relations, etc.) at teeth boundaries.

In intra-oral imaging, segmentation of individual components of the image content from a digital model can be of value to the dental practitioner in various procedures, including orthodontic treatment and preparation of crowns, implants, and other prosthetic devices, for example. Various methods have been proposed and demonstrated for mesh-based segmentation of teeth from gums and of teeth from each other. However, drawbacks of conventional segmentation solutions include requirements for a significant level of operator skill and a high degree of computational complexity. Conventional approaches to the problem of segmenting tooth components and other dentition features have yielded disappointing results in many cases. Exemplary method and apparatus embodiments according to the present disclosure address such problems with segmentation that can utilize the polygonal mesh data as a type of source digital model and can operate in more than one stage: e.g., first, performing an automated segmentation that can provide at least a close or coarse approximation of the needed segmentation of the digital model; and second, allowing operator interactions to improve, correct or clean up observed errors and inconsistencies in the automated results, which can yield highly accurate results that are difficult to achieve in a purely automated manner without significant requirements on operator time or skill level or on needed computer resources. This hybrid approach in exemplary method and apparatus embodiments can help to combine computing and image processing power with operator perception to check, correct, and refine results of automated processing.

The logic flow diagram of Figure 6 A shows a hybrid sequence for tooth mesh segmentation and generation of a digital model to identify individual features or intraoral components such as teeth from within the mouth according to an exemplary embodiment of the present disclosure. In an image acquisition step SI 00, a plurality of structured light images of the patient's dentition are captured, providing a set of contour images for processing. A point cloud generation step SI 10 then generates a point cloud of the patient's dentition using the set of contour images. A polygon mesh generation step S 120 forms a polygon mesh by connecting adjacent points from point cloud results. A triangular mesh provides one type of polygon mesh that can be readily generated for approximating a surface contour; more complex polygon mesh configurations can alternately be used.

Continuing with the Figure 6 A sequence, given the polygon mesh, a segmentation step SI 30 can be executed. For a dental contour image, for example, segmentation step SI 30 can distinguish teeth from gum tissue, as well as distinguishing one tooth from another. Segmentation results can then be displayed, showing the results of this initial, automated segmentation processing. The automated segmentation step SI 30 can provide an intermediate image. Thus automated step SI 30 can perform the bulk of segmentation processing, but can further benefit from operator review and refinements of results. For its automatic processing, segmentation step S130 can use any of a number of known

segmentation techniques, such as fast-marching watershed algorithms, so-called snake-based segmentation, and other methods known to those skilled in the imaging arts, as noted earlier.

Figure 6A also shows an optional repeat loop that can enable viewer interaction with the intermediate image for refining the results of the automated segmentation processing, for example, using the basic apparatus shown in Figure 1. An accept operator instructions step SI 40 can be executed, during which the viewer indicates, on the displayed results, seed points, seed lines, block lines, boundary features, or other markings that identify one or more distinct features of the segmentation results to allow further segmentation refinement and processing. Viewer markup instructions cause segmentation step SI 30 to be executed at least a second time, this second time using input markup(s) from entered viewer instructions. It can be appreciated that different segmentation algorithms can be applied at various stages of automated or manual processing. Final results of segmentation processing can be displayed, stored, and transmitted between computers, such as over a wired or wireless network, for example.

The process shown in Figure 6A can thus allow automated segmentation to perform the coarse segmentation (e.g., first segmentation) that can be more easily accomplished, such as segmentation of teeth from gum tissue, for example. Thus, for example, tooth and gum partitioning can be automated. In one embodiment, tooth and gum partitioning can use an automated curvature- based method that computes curvature of vertices in the mesh, and then uses a thresholding algorithm to identify margin vertices having large negative curvature. Alternately, color-based segmentation can be used for tooth segmentation from the gums. This type of method can obtain average hue values from regions of the image and calculate threshold values that partition image content.

An exemplary embodiment of workflow for the hybrid tooth segmentation system is depicted in the logic flow diagram of Figure 6B. Upon receiving a dentition mesh such as the one described in Step SI 20 and shown in Figs. 4 and 5, the control logic processor 80 (Figure 1) initiates an automated segmentation step S202 in which a fully automatic tooth segmentation tool is evoked to delineate teeth and gum regions and delineate individual teeth regions. The fully automatic tooth segmentation tool employs exemplary algorithms such as active contour models published in the literature or otherwise well-known to those skilled in the image processing arts. The delineation of teeth effectively produces individually segmented teeth; however, these generated teeth may contain poorly segmented intraoral components. A first checking step S204 then checks for poorly segmented intraoral components. Checking for incorrect or incomplete segmentation in step S204 can be accomplished either computationally, such as by applying trained artificial intelligence algorithms to the segmentation results, or by viewer interaction, such as following visual inspection by the viewer. By way of example, Figure 7 A shows an exemplary poorly segmented or mis-segmented tooth 302. As shown in Figure 7A, a segmented tooth boundary 306 is not aligned with an actual tooth boundary 308.

Still referring to the workflow process in Figure 6B, if checking Step S204 identifies one or more poorly segmented teeth, either computationally or visually, a primary assisted segmentation step S206 executes, activating a segmentation procedure that is also automated, but allows some level of operator adjustment. Primary assisted segmentation step S206 applies an algorithm for segmentation that allows operator adjustment of one or more parameters in a parameter adjustment step S210. Another checking step S208 executes to determine if additional segmentation processing is needed. The adjustable parameter can be altered computationally or explicitly by an operator instruction in step S210. Subsequent figures show an exemplary operator interface for parameter adjustment.

An exemplary algorithm employed in primary assisted segmentation Step S206 can be a well-known technique, such as the mesh minimum curvature-based segmentation method. The adjustable parameter can be the threshold value of the curvature. With the help of the parameter adjustment in step S210, a correction of the poorly segmented tooth can be made. Figure 7B shows an image of tooth 312 that, by comparison with Figure 7 A, shows a segmented tooth boundary 316 now well aligned with the actual boundary.

However, as is clear from the exemplary workflow embodiment shown in Figure 6B, the delineation of teeth performed in Step S206 may still produce poorly segmented intraoral components or features, so that a repeated segmentation process is helpful. The checking of poor segmentation in step S208 can be accomplished either computationally, such as by applying artificial intelligence algorithms to the segmentation results, or more directly, by visual inspection performed by the user. In addition to the adjustable parameter adjusted in Step S210, the hybrid tooth segmentation system optionally allows the user to add exemplary geometric primitives such as seed lines on the tooth region and add blocking lines between the teeth or between the teeth and gum to aid the tooth segmentation process. Figure 8 A shows an exemplary seed line 406 for marking a tooth, added to a mesh image 62. Figure 8B shows an exemplary block line 408 for indicating space between two teeth, added to a mesh image 62.

The three basic steps, Step S206, Step S208 and Step S210 in the

Figure 6B sequence constitute an exemplary primary segmentation loop 54 that follows the fully automatic segmentation of step S202 and checking step S204. This exemplary primary segmentation loop 54 is intended to correct segmentation errors from the fully automated segmentation of automated segmentation step S202, as identified in step S204. Exemplary primary segmentation loop 54 can be executed one or more times, as needed. When exemplary primary segmentation loop 54 is successful, segmentation can be complete.

In some cases, however, additional segmentation processing beyond what is provided by primary segmentation loop 54 is needed.

Segmentation processing can be complicated by various factors, such as tooth crowding, irregular tooth shapes, artifacts from scanning, indistinct tooth contours, and undistinguishable interstices among others. Where additional segmentation is needed, an exemplary secondary segmentation loop 56 can be used to provide more interactive segmentation approaches. The secondary segmentation loop 56 can include an interactive segmentation step S212, another checking step S214, and an operator markup step S216. Interactive segmentation step S212 can activate a segmentation process that works with the operator for indicating areas of the image to be segmented from other areas. Interactive segmentation step S212 can have an automated sequence, implemented by an exemplary algorithm such as a "fast march" method known to those skilled in the image segmentation arts. Step S212 may require population of the tooth region images by operator-entered seeds or seed lines or other types of geometric primitives before activation or during processing. In certain exemplary embodiments, seed lines or other features can be automatically generated in Step S100, S110 and S 120 when the dentition mesh is entered into the system for optional operator adjustment (e.g., subsequent operations such as secondary segmentation loop 56 or Step 212). In addition, the features, seeds or seed lines can be added to the segmentation process in operator markup Step S216 by the user. The results from Step S212 are subject to inspection by the user in Step S216. Results from the hybrid automated/interactive segmentation processing can then be displayed in a display step S220, as well as stored and transmitted to another computer. . ...

Following the sequence of Figure 6B, some exemplary methods and apparatus of the present disclosure provide a hybrid tooth segmentation that provides the benefits of interactive segmentation with human-machine synergy.

Figures 9A-9C show operator interface screens 52 for portions of a sequence for review and entry of markup instructions for refining mesh segmentation processing according to certain exemplary embodiments of the present disclosure. Interim mesh segmentation results are shown in a display area 86 on screen 52. A number of controls 90 for adjustment of the segmentation process are available, such as an adjustment control 84 for setting a level for overall aggressiveness or other parameter or characteristic of the segmentation processing algorithm. Optional selection controls 88 allow the viewer to specify one or more segmentation algorithms to be applied. This gives the operator an opportunity to assess whether one particular type of segmentation algorithm or another appear to be more successful in performing the segmentation task for the given mesh digital model. The operator can compare results against the original and adjust parameters to view results of successive segmentation attempts, with and without operator markup.

Figure 9A also shows a trace pattern 96 that is entered as an operator seed line instruction for correcting or refining segmentation processing, as was shown previously with respect to Figure 8A. According to an embodiment of the present disclosure, an operator mark in the form of trace pattern 96 or other arbitrary marking/geometric can be used to provide seed points that indicate a specific feature for segmentation, such as a molar or other tooth feature that may be difficult to process for conventional segmentation routines. Seed marks can then be used as input to a fast marching algorithm or other algorithm type, as described previously. In some cases, for example, adjacent teeth may not be accurately segmented with respect to each other; operator markup can provide useful guidance for segmentation processing where standard segmentation logic does not perform well. As Figure 9A shows, the operator can have controls 90 available that allow the entered markup to be cleared or provided to the segmentation processor. As Figure 9B shows, color or shading can be used to differentiate various teeth or other structures identified by segmentation.

Additional controls 90 can also be used to display individual segmented elements, such as individual teeth, for example. As Figure 9C highlights, in some exemplary embodiments, individual controls 90 can be used individually or in combination.

In one embodiment, segmentation of individual teeth from each other can use curvature thresholds to compute margin and border vertices, then use various growth techniques to define the bounds of each tooth relative to margin detection.

In some exemplary embodiments, controls 90 can include, but are not limited to enter/adjust seed or boundary geometries, enter/adjust selected segmentation procedures, enter/adjust number of objects to segment, subdivide selected object, modify segmented object display, etc.

Bracket and wires removal with reconstruction

The logic flow diagram of Figure 10 shows an exemplary embodiment of a workflow for bracket removal from a dentition 3D mesh according to an embodiment of the present disclosure. As shown in Figure 10, a virtual or digital 3D dentition mesh model is obtained in an acquisition step SI 002. For example, a digital 3D dentition mesh model can be obtained by using an intraoral scanner that employs structured light.

Figure 11 is a diagram that shows an exemplary 3D dentition mesh that can be acquired in step SI 002 of Figure 10. As shown in Figure 11, 3D dentition mesh 1100 can include brackets 1102, gingival tissue 1104 and teeth 1106. Preferably, a result from the exemplary workflow process of Figure 10 will be a 3D dentition mesh including the teeth 1106 and gingival tissue 1104 from the 3D dentition mesh 1100, but without the brackets 1102 and tooth surfaces previously covered by brackets 1102 and with the tooth surfaces accurately reconstructed.

As shown in Figure 10, separation steps 1004 and 1006 constitute a tooth segmentation method for an obtained dentition 3D mesh. As described herein, in one embodiment steps Si 004 and SI 006 can be implemented by similar steps of a hybrid sequence for tooth mesh segmentation depicted in Figure 6 A. Alternatively in another embodiment, steps SI 004 and SI 006 can be implemented by similar steps of a hybrid tooth segmentation method or system depicted in Figure 6B. Segmentation distinguishes each tooth from its neighboring teeth and from adjacent gingival tissue.

Continuing with the workflow in Figure 10 and with reference to Figures 11, 12, and 13A, 13B, and 13C, brackets 1102 are automatically removed from the 3D dentition mesh 1100 (e.g., tooth surfaces) in a removal step S1008. In one exemplary embodiment, the separated (or segmented) teeth resulting from step S 1006 can individually undergo bracket removal and surface reconstruction described hereafter. Figure 12 is a diagram that shows exemplary resultant separated teeth 1202 contained within the 3D dentition mesh 1100.

In removal step SI 008, to automatically remove the brackets from surfaces of the separated teeth 1202, each individually segmented tooth (or crown) is examined and processed. An exemplary segmented tooth 1202 with bracket 1302 to be removed is shown in Figure 13 A. In one exemplary embodiment, an automatic bracket removal algorithm first detects boundaries of the bracket 1302. Various approaches known to one skilled in the imaging arts can be used to detect bracket boundaries in the 3D dentition mesh 1100. In one exemplary

embodiment, bracket boundary detection can use an automated curvature-based algorithm that detects and computes the curvatures of vertices in the mesh of tooth surfaces, and then uses a thresholding algorithm to identify margin vertices that have large negative curvature values, indicative of a high degree of curvature.

As shown in Figure 13 A, these identified margin vertices form a closed 3D curve or bracket boundary 1303 (or the boundary vertices of the bracket) that surrounds the bracket 1302. Then, mesh vertices within the closed 3D boundary are removed in the 3D dentition mesh 1100. As Figure 13B shows, this results in a gap or hole 1304 on the tooth surface. Figure 13B is a diagram that shows an exemplary segmented tooth 1202 with bracket 1302 removed. As shown in Figure 13B, small white patches can be present in the bracket hole 1304; these white patches do not belong to the bracket 1302 itself, but can be other artifacts behind the original bracket. These artifacts can become visible after the bracket 1302 has been removed from the 3D dentition mesh 1100 by an automatic bracket removal algorithm.

Referring again to the flow diagram of Figure 10, in a reconstruction step SI 010, tooth surfaces of the segmented tooth 1202 having the bracket removed are automatically reconstructed. Various approaches known to those skilled in the imaging arts can be used to fill holes in the 3D dentition mesh 1100. An exemplary segmented tooth 1202 having automatically reconstructed tooth surface 1306 is shown in Figure 13C. In exemplary embodiments, hole- filling procedures (e.g., tooth or crown surface reconstruction) can include a first step to generate an initial patch to fill the hole and a second step to smooth the reconstructed mesh to obtain improved quality polygons (e.g., triangles) therein.

Figure 13D schematically shows a part of the 3D dentition mesh 1100 forming a 3D crown mesh surface after mesh portions representing a bracket are removed. A closed polygon 1303' represents a boundary of the (removed) bracket. A region 1308 enclosed by the closed polygon 1303' is the gap or hole left by bracket removal. First in step S 1010, an initial patch is generated to fill the tooth surface or hole of region 1308 (e.g., within the closed polygon 1303 '). In one embodiment, the initial patch contains a plurality of triangles 1310 arranged in an exemplary prescribed pattern such as one formed by connecting vertices in the closed polygon 1303' to form the pattern shown in Figure 13E. Then, in reconstruction step S1010, polygons such as triangles 1310 of the initial patch can be further modified or optimized. One exemplary procedure of modifying or optimally arranging the triangles 1310 is illustrated in Figure 13F where four points A, B, C, and D form two triangles ABD and CDB in the triangles 1310, which are rearranged to become triangles ABC and CD A in an improved set of triangles 1310'. An improved triangle arrangement can reduce or avoid long, thin triangles. In a second part of reconstruction step S1010 of the Figure 10 sequence, the 3D mesh with the initial patch can be smoothed to obtain better quality. In one embodiment, the second part of step S1010 can correct positions of points created in the initial patch using local information globally. Thus, the 3D mesh including the initial patch (e.g., triangles 1310, 1310' within the hole of polygon 1303') and the surrounding regions, such as triangles 1312 surrounding (or nearby) the hole 1308' in Figure 13D can be smoothed using a Laplacian smoothing method that adjusts the location of each mesh vertex to the geometric center of its neighbor vertices. For example, an implementation of mesh smoothing is described by Wei Zhao et al. in "A robust hole-filling algorithm for triangular mesh" in The Visual Computer (2007) December 2007, Volume 23, Issue 12. pp 987-997, that can implement a patch refinement algorithm using the Poisson equation with Dirichlet boundary conditions. The Poisson equation is formulated as wherein /is an unknown scalar function; d ² d ² d ²

Af =— +— _T +—

dx ² dy ² dz ² is a Laplacian operator; h is the guidance vector field; div(h) is the divergence of h; and/* is a known scalar function providing the boundary condition. The guidance vector field on a discrete triangle mesh as used in Wei Zhao's method is defined as a piecewise constant vector function whose domain is the set of all points on the mesh surface. The constant vector is defined for each triangle and this vector is coplanar with the triangle.

In a display step SI 012 of Figure 10, the exemplary segmented tooth 1202 having automatically reconstructed tooth surface 1306 (see Figure 13C) can be displayed. Although described for one exemplary segmented tooth 1202, steps S1008, S1010 and S1012 can be repeatedly performed until all brackets are removed from 3D dentition mesh 1100. In this manner, the resultant corrected 3D dentition mesh 1100 can be displayed in step S1012 after each additional segmented tooth surface is corrected. Alternatively, steps SI 008 and S1010 can be performed for all teeth in the 3D dentition mesh 1100, before the resultant corrected 3D dentition mesh 1100 is displayed in step 1012. Figure 13G shows an exemplary corrected 3D dentition mesh 1316.

Braces and brackets detection and removal

Certain exemplary method and/or apparatus embodiments can provide automatic braces detection and removal by initial (e.g., coarse) bracket detection, subsequent wire detection, and refinement of detected (e.g., separated) initial brackets, which can then be removed from the initial 3D mesh and subsequently filled by various surface reconstruction techniques.

The logic flow diagram of Figure 14 shows an exemplary sequence for automatic braces and brackets detection and removal, without tooth

segmentation as a pre-step, by processing logic according to an embodiment of the present disclosure. A coarse bracket detection step SI 302 provides estimated positions of brackets, using an approach such as that described subsequently. A brace wires detection step SI 304 then detects connecting wires that extend across the bracket region. A masks generation step SI 308 generates masks for the brackets; these masks narrow the search area for detection. A processing step S 1310 provides pruning and other morphological operations for further defining the masks. A Fast March application step SI 320 executes a fast march algorithm according to the defined mask region. A refinement step SI 330 performs the necessary refinement of detected bracket areas or regions using morphological operators. A fine tuning step SI 340 generates the fine-tuned bracket regions that are then used for removal steps.

Coarse bracket detection

Coarse bracket detection in step SI 302 can proceed as described in the flow diagram of Figure 15 and as shown visually in the sequence of Figure 16. In a computation step SI 306, the system computes a parabola 502 that is a suitable fit for the imaged dentition. This is typically executed from image content in a top view image 500 as shown, using curvature detection logic.

Parabola 502 can be traced along the edges of teeth in the arch using the imaged content. Given this processing, in a side detection step S1312, a buccal side 504 or, alternately, the opposite lingual side of the arch is then identified.

With the lingual or buccal side 504 of the arch and parabola 502 located, one or more bracket areas or regions 506 disposed on that side can then be identified in a bracket areas detection step SI 324. According to an

embodiment of the present disclosure, bracket area 506 detection uses the following general sequence, repeated for points along parabola 502:

(i) Extend a normal outward toward the side from the generated parabola 502;

(ii) Detect a maximum length of the extended normals within a local neighborhood, such as within a predetermined number of pixels or calculated measurement;

(iii) Select nearby points on the mesh that lie within a predetermined distance from the detected maximum.

These substeps identify candidate bracket areas or regions 508 as shown in the example of Figure 16. These candidate areas 508 can be processed in order to more accurately identify bracket features that lie against the teeth and to distinguish each bracket from the corresponding tooth surface.

Once areas 508 have been identified, a decision step SI 328 determines whether or not post treatment is needed in order to correct for processing errors. If post treatment is not required, bracket areas have been satisfactorily defined. If post treatment is required, further processing is applied in a false detection correction step SI 332 to remove false positives and in a clustering step SI 344 to effect further clustering of bracket areas that are in proximity and that can be assumed to belong to the same bracket 510. Figure 17 shows exemplary coarse brackets 510 obtained using the described coarse bracket detection sequence of Figure 15.

Brace wires detection

Brace wires detection step SI 304 from the Figure 14 sequence can proceed as shown in Figure 18 and as described following. Coarse brackets 510 may be connected by brace wires 512.

Processing can detect wire extending from each bracket region. It is useful to remove these wires in order to obtain improved bracket removal.

For each vertex V in the bracket region as shown in Figures 18 and 19, processing can perform a nearest neighbor search within a suitable radius, such as within an exemplary 5mm radius, resulting in a set of neighbor vertices VN. Processing then checks the normal of each of the vertices in VN.

The detected wires can facilitate identification of the individual brackets. If it is determined that the normal for at least one vertex in VN points to the opposite direction of the normal of the vertex V (e.g. if the dot product of the two normal vectors < -0.9), then V is considered a candidate vertex on the wire (or bridge). This can be measured, for example, because there is space between the wire feature and the tooth. This procedure can be applied to the entire mesh, resulting in a set that has a number of candidate vertices.

The set of candidate vertices is used to compute a plurality of connected regions. Each of the connected regions can be analyzed using a shape detection algorithm, such as principal component analysis PCA, familiar to those skilled in the imaging arts and used for shape detection, such as wire detection.

Figure 18 shows results of wire detection for wires 512 extending between brackets. These detected wires can then be used to algorithmically identify and separate connected brackets.

Generating initial masks

With separated coarse brackets detected in some manner (either originally detected using step SI 302 or using the results from wire detection step SI 304), an initial mask can be generated for each individual coarse bracket.

These initial masks can be helpful for narrowing the search area in Fast Marching brackets detection. In practice, a proper initial mask should be adequately large enough to cover all the components (base, pad, slots, hook, band, etc.) that belong to a bracket.

Generating and processing initial masks from steps SI 308 and

S 1310 in Figure 14 can be executed as follows. Referring to the schematic diagram of Figure 19, this processing can generate a mask for each bracket. The mask is used to define the region of interest (ROI) for subsequent fast march bracket detection.

Processing for mask generation can use the following sequence, with reference to Figure 19:

(i) Jaw mesh orientation. The z axis is orthogonal to the bite plane.

(ii) Sorting. Brackets, separated by wire detection, in each dental arch are sorted and center, normal, and bi-normal features are computed for each bracket.

(iii) Identification. Each bracket type is identified as either lingual or buccal, on back molar or on other teeth. A suitable radius is set for mask generation for each bracket.

(iv) Radius search. A radius search is executed from the center of each initial bracket in order to generate an initial mask 520 for each bracket. The mask should be large enough to contain the bracket.

The centroid of each mask 520 is connected to each neighbor along the arch, as represented in flattened form in Figure 19.

Processing initial masks

Processing initial masks in step S1310 of Figure 14 performs a pruning operation that shapes the mask correctly for its corresponding bracket and removes areas where the initial mask is inaccurate and extends to opposite sides of the same tooth or extends between teeth. Figure 20 shows the pruning operation for masks that inaccurately extend to the opposite side in schematic representation at images 530. Pruning results are shown in the example of image 532. Pruning for masks that inaccurately extend across teeth, as shown in image 540, is shown in image 542.

Starting from one end of the dental arch, the bi-normal bn can be defined as the vector from a bracket's own center to that of the next bracket in the series that is formed by sorting all brackets that lie along the dental arch from one side to another. The cross product of the z-axis and bi-normal can be used to generate its normal as depicted in Figure 19, showing the z-axis, normal n, and bi- normal bn of each bracket. For pruning where masks extend to the opposite side as shown in schematic representation in Figure 20 at image 530 with pruning at image 532, the following processing can be executed at each vertex in the mask:

(i) Compute D _norma the dot product of the normal and bracket normal for each vertex:

normal ~ ^ Nv/ > Nbracket ^

wherein N _vi is the normal of vertex Vj, Nbracket is the bracket normal. (The notation < a,b > indicates dot product and can alternately be expressed as a * b. )

(ii) Remove the vertices whose D _norma i value is below a predetermined threshold value (for example, below -0.1). This dot product value indicates vectors tending towards opposite directions.

For pruning where masks extend to neighboring teeth, as shown in schematic representation at image 540 in Figure 20 with pruning at image 542, the following processing can be executed at each vertex in the mask:

(i) Compute Dbmormai for each vertex:

Dbmormai = < Nvi BN _br acket > * Sgtl( < Dir _vi , BN _b racket >) wherein N _vi is the normal of vertex v, ; BNbracket is the binormal of the bracket; Dir _vi is the direction from the bracket center to vertex v, ; and

Sgn(x) returns the +/- sign of (x) .

(ii) Remove vertices whose Dbmormai value is smaller than a threshold value (for example, smaller than -0.1).

After pruning, a post-processing procedure can be applied to each mask, as shown in the sequence of Figure 21. An image 550 shows encircled gaps 556 that can be filled in order to complete masked regions. The remaining vertices after pruning are dilated to connect discontinuous regions and to fill regions that may have been inaccurately pruned. Dilation can be followed by an erosion process to remove regions of the mask lying between the teeth, as shown in encircled areas 552, 554. An image 560 shows improvement to the encircled regions of image 550. There can be some small residual regions, as shown encircled in an area 572 in an image 570, other than the main bracket mask region. These can be redundant areas, for example; these small regions can be detected and removed and only the largest connected region retained as the resultant initial mask. An image 580 shows the completed mask following both dilation and erosion processing.

Fast March processing

Once well-pruned masks have been obtained, a Fast March algorithm can be applied within each mask, with boundaries of the mask used as seed vertices. Within the fast march algorithm, the arrival time for seed vertices can be set to 0. The arrival time for vertices within the mask can be computed with the common Fast Marching process, as shown schematically in Figure 22.

Fast March processing uses a weighting or cost function in order to determine the most likely path between vertices in each masked region. Figure 22 shows different computations that can apply for paths between given vertices using Fast March methods. Figure 23 shows exemplary Fast March computation for arrival time from different seed-points along mask boundaries using the Fast March method. Masked regions 590 are shown, with grayscale- or color-encoded arrival times used for comparison as shown in image 595 in Figure 23.

For Fast March processing, curvature values κ can be used. It should be noted that minimum κ values (for example, with negative values such as K = -10) indicate very high curvature. The boundary of a bracket is characterized by a high absolute value of curvature.

The Fast Marching algorithm applies a speed function in order to compute the weight assigned to each edge in the graph. For bracket removal, there is a need for reduced edge weights in flat regions and larger edge weight values in highly curved regions.

The speed function for Fast Marching execution is based on normal (s) - ds , where v ₀ and v are two neighbor vertices, the normal difference is equal to the integration of normal curvature κ _ηοηηαΙ in the geodesic line on the mesh surface from v ₀ to v, . The D _normal value is approximate to the averaged normal curvature of v ₀ and v, , times the distance S from v ₀ and v, :

normal ~ ^ In implementing the speed function, the mean curvature can be used. The mean curvature is readily computed (as compared against a normal curvature) and operates without concern for possible differences in estimation for the propagating front stop at regions that are highly curved. The speed function is therefore defined as:

w = w _nom A ^{Kmem{Vo) Kmea} " ^{Vl s , wherein f _mean is the mean curvature and w _nomal is a weight value.

The speed function used for processing with masked Fast Marching can be defined as a normal difference of two neighbor vertices along the edge of an area being processed. Where vertices v# and v; are two neighboring vertices (that is, within nearest proximity of each other relative to the display medium), the normal difference is equal to the integration of normal curvature K _norma i in the geodesic line from vertex v _/ to ¾. The normal difference is approximate to the average normal curvature of vo and v _/ , times a distance S from vo to v _/ .

Refining detected bracket regions

Morphological processing can be used for final refinement of detected bracket regions. Figure 24 shows results of using a sequence of different approaches for refinement of bracket regions according to an embodiment of the present disclosure. An image 600 shows fast marching results for a typical image having brackets and braces. An image 610 shows results of image thresholding, well known to those skilled in the imaging arts. An image 620 shows results of a dilation and fill process. An image 630 shows results following image erosion, using a maximum-sized region. Figure 25 shows steps of an optional refinement of bracket regions using a convex hull computation. The following sequence can be used for convex hull processing:

(i) compute the boundary of a bracket region in the mesh, as shown in an image 700; in the example shown, a large gap exists within the bracket region;

(ii) project the boundary vertices to the 2D PCA plane of the boundary as shown in an image 710;

(iii) compute the convex hull in the PCA plane, as shown in an image 720; (iv) detect and record pairs of points that are connected, such as non- neighbor vertices as shown in an image 730;

(v) connect paired vertices in the original 3D boundary with geodesic lines to form a 2-manifold convex hull, as shown in an image 740.

The resulting convex hull connects the gap that appears in image 700 and covers the full bracket.

Figure 26 shows the fine tuned bracket regions obtained using the described sequence.

Figure 27 shows the recovered tooth surface following bracket definition and removal by applying, to the results in Figure 26, the surface reconstruction process detailed in the preceding paragraphs and the sequence described with reference to Figures 13 A - 14.

Figure 28 is a logic flow diagram that shows a workflow of another exemplary embodiment of the present disclosure for bracket removal on a 3D dentition mesh. Unlike the workflow shown in Figure 10, the workflow shown in Figure 28 does not require tooth segmentation as a separate step. A 3D dentition mesh is received in an acquisition step S1402; the received mesh contains teeth, brackets, and gingival tissue. Then, in an instruction step S1404, instructions are received from an operator regarding brackets in the 3D dentition mesh.

Figure 29 is a diagram that displays an exemplary graphical user interface (GUI) that allows the user to input information to identify brackets in the 3D dentition mesh. As shown in Figure 29, one exemplary GUI interface 1500 enables nodes to be placed by the user for a 'snake' operation, which automatically encircles bracket 1502 boundaries, based on the entered nodes. An exemplary bracket boundary 1503 generated by the automated 'snake' operation is shown in Figure 29. The 'snake' is an active shape model that is frequently used in automatic object segmentation in image processing, for example by delineating an object outline from a possibly noisy 2D image. The active shape model of the snake is similar to that used in applications like object tracking, shape recognition, segmentation, edge detection and stereo matching. Methods using a snake active shape model or active contour model are well known to those skilled in the imaging arts.

Figure 30 shows vertices 1602 encircled by the boundary 1503 being highlighted in the 3D dentition mesh after the user presses the 'run' command button 1504 in Figure 29. Identified vertices 1602 are to be removed from the original 3D dentition mesh. In one exemplary embodiment, the GUI 1500 can let the user inspect the intermediate results for vertices 1602, and if satisfied, the user presses the 'cut' button 1506. The vertices 1602 change their highlight (e.g., color, texture, etc.) to indicate that these vertex features are to be removed from the original 3D dentition mesh. In one exemplary embodiment, pressing the 'cut' button 1506 causes processing to automatically remove the brackets from the teeth surface in a removal step SI 406 based on the operator instructions in step SI 404.

After bracket removal, the tooth surfaces are filled or reconstructed in a reconstruction step S1408. In one exemplary embodiment, step S1408 is performed when the user presses the 'fill' button 1506 in Figure 29 to reconstruct tooth surfaces and remove any holes or gaps caused by bracket removal. Step S1408 can be performed using known algorithms such as described herein with respect to Figure 10. Figure 31 shows an example of a reconstructed tooth surface 1702 after the bracket is removed.

The procedures shown in the Figure 28 sequence can be performed tooth by tooth, on a small group of adjacent teeth, or on all teeth simultaneously with respect to the 3D dentition mesh.

Figures 32-34 are diagrams that show sequential stages in the process leading to a complete, concurrent removal of all brackets from a 3D jaw mesh. Figure 32 is a diagram that shows a 3D dentition mesh 1800 with teeth, brackets and gingival tissue. Figure 33 is a diagram that shows the intermediate results of 'snake' cut operation with vertices 1802 that are to be removed shown in highlighted form. Figure 34 is a diagram that shows each of the final

reconstructed teeth surfaces 1806 after all brackets are removed and all fill operations are completed.

It is noted that the above described user actions such as pressing the 'cut' button, pressing the 'fill' button and pressing the 'run' button are illustrative. In actual applications, these separate actions may not necessarily be sequentially initiated and can be accomplished automatically by computer software.

In some cases, 3D dentition models produced by an intraoral scanner may contain wires that bridge two neighboring brackets. In this situation, embodiments described previously may be insufficient for removal of the brackets and wires. Figure 35 A is a diagram that shows another exemplary dentition model. As shown in Figure 35A, dentition model 2100 includes brackets 2102, gingival tissue 2104, teeth 2106 and bridged brackets where a wire 2108 connects at least bracket 2110 and bracket 2112. Generally, wires 2108 will connect all brackets 2102. As shown in comparing Figures 35A, 35B, and 35C, the wire 2108 can, once identified, be erased automatically or interactively according to exemplary methods and apparatus of the present disclosure.

In Figure 36, an actual result 2204 for bridged brackets removal is shown. The surface reconstructed tooth 2210 and tooth 2212 in Figure 36 correspond to bracket 2110 and bracket 2112 in Figure 21 A before the brackets and wire 2108 are removed.

Figure 37 is a logic flow diagram that shows an exemplary sequence for bridged bracket removal from tooth mesh surfaces according to an embodiment of the present disclosure. As shown in Figure 37, a dentition model with bridged brackets is obtained in an acquisition step S2302, which is immediately followed by a cutting step S2304 that includes automatically

"breaking the bridge". One exemplary detection embodiment that can be used to automatically break the bridge (or wire) is described as follows. In a removal step S2306, given a vertex V in the dentition mesh model, processing logic performs a nearest neighbor search with an exemplary 5mm radius resulting in a set of neighbor vertices VN. As described in the preceding sections, the system checks the normal of each of the vertices V in set VN. If it is found that there is at least one vertex in VN whose normal points to the opposite direction of the normal of V (e.g. if these two normal vectors' dot product < -0.9), then vertex V is on the wire (or bridge). An exemplary bridge (wire) detection result 2118 resulting from step S2306 is shown in Figure 35C. These vertices of the 3D detention mesh detected in step S2306 (e.g., associated with the wires 2108) are excluded or removed from the 3D detention mesh in the subsequent removal step S2306 and reconstruction step S2308.

Removal step S2306 employs either exemplary automatic or interactive methods to remove the disconnected brackets. The bracket removed tooth surface is reconstructed automatically in a reconstruction step S2308 and the results are displayed for inspection in a display step S2310. For example, steps S2306 and S2308 can be performed as described above for Figures 10 and 28, respectively.

Figure 38 is a logic flow diagram that shows another exemplary method embodiment for bridged brackets removal. As shown in Figure 38, a dentition model with bridged brackets is acquired in an acquisition step S2402, which is immediately followed by an interaction step S2404 of interactively "breaking the bridge". In one exemplary embodiment, interactive operation effectively erases the thin wires with the assistance from a human by selecting and deleting mesh vertices that belong to the thin wires in step S2404. In one exemplary embodiment, step S2404 can use a GUI with selectable operator actions to "clear", "paint" (e.g., operator identify pixels showing wires), "auto paint", "approve" (e.g., paint or auto paint), and "clear" to interactively break the bridges or remove the wires from the 3D dentition mesh based on the operator instructions. Then, a removal step S2406 employs either automatic or interactive method to remove the disconnected brackets as previously described. The bracket removed tooth surfaces can be reconstructed automatically in a reconstruction step S2408 as previously described. Then, the results are displayed for inspection in a display step S2410.

As described herein, exemplary method and/or apparatus embodiments to remove bridged brackets and restore teeth surfaces in a 3D dentition model are intended to be illustrative examples and the application is not so limited. For example, in one exemplary embodiment, bridged brackets can be removed and teeth surfaces restored by automatically identifying parts of a bracket and/or wire without human intervention in an obtained 3D dentition model by growing the identified parts into a region that covers the brackets and/or wires entirely (e.g., and preferably slightly beyond the brackets and/or wires

boundaries), removing the region from the 3D dentition model surface, and restoring the removed region surfaces using hole filing techniques. In some exemplary embodiments, hole filling can fill portions of gingival tissue in addition to tooth surface portions. Surface data of the patient that were previously acquired, such as from a dentition mesh model obtained before braces were applied, can be used to generate the reconstructed tooth surface.

Consistent with one embodiment, the present disclosure can use a computer program with stored instructions that control system functions for image acquisition and image data processing for image data that is stored and accessed from an electronic memory. As can be appreciated by those skilled in the image processing arts, a computer program of an embodiment of the present invention can be utilized by a suitable, general-purpose computer system, such as a personal computer or workstation that acts as an image processor, when provided with a suitable software program so that the processor operates to acquire, process, transmit, store, and display data as described herein. Many other types of computer systems architectures can be used to execute the computer program of the present invention, including an arrangement of networked processors, for example.

The computer program for performing the method of the present invention may be stored in a computer readable storage medium. This medium may comprise, for example; magnetic storage media such as a magnetic disk such as a hard drive or removable device or magnetic tape; optical storage media such as an optical disc, optical tape, or machine readable optical encoding; solid state electronic storage devices such as random access memory (RAM), or read only memory (ROM); or any other physical device or medium employed to store a computer program. The computer program for performing the method of the present invention may also be stored on computer readable storage medium that is connected to the image processor by way of the internet or other network or communication medium. Those skilled in the image data processing arts will further readily recognize that the equivalent of such a computer program product may also be constructed in hardware.

It is noted that the term "memory", equivalent to "computer- accessible memory" in the context of the present disclosure, can refer to any type of temporary or more enduring data storage workspace used for storing and operating upon image data and accessible to a computer system, including a database. The memory could be non-volatile, using, for example, a long-term storage medium such as magnetic or optical storage. Alternately, the memory could be of a more volatile nature, using an electronic circuit, such as random- access memory (RAM) that is used as a temporary buffer or workspace by a microprocessor or other control logic processor device. Display data, for example, is typically stored in a temporary storage buffer that is directly associated with a display device and is periodically refreshed as needed in order to provide displayed data. This temporary storage buffer can also be considered to be a memory, as the term is used in the present disclosure. Memory is also used as the data workspace for executing and storing intermediate and final results of calculations and other processing. Computer-accessible memory can be volatile, non-volatile, or a hybrid combination of volatile and non-volatile types.

It is understood that the computer program product of the present disclosure may make use of various image manipulation algorithms and processes that are well known. It will be further understood that the computer program product embodiment of the present invention may embody algorithms and processes not specifically shown or described herein that are useful for implementation. Such algorithms and processes may include conventional utilities that are within the ordinary skill of the image processing arts. Additional aspects of such algorithms and systems, and hardware and/or software for producing and otherwise processing the images or co-operating with the computer program product of the present invention, are not specifically shown or described herein and may be selected from such algorithms, systems, hardware, components and elements known in the art.

In this document, the terms "a" or "an" are used, as is common in patent documents, to include one or more than one, independent of any other instances or usages of "at least one" or "one or more." In this document, the term "or" is used to refer to a nonexclusive or, such that "A or B" includes "A but not B," "B but not A," and "A and B," unless otherwise indicated. In this document, the terms "including" and "in which" are used as the plain-English equivalents of the respective terms "comprising" and "wherein." Also, in the following claims, the terms "including" and "comprising" are open-ended, that is, a system, device, article, or process that includes elements in addition to those listed after such a term in a claim are still deemed to fall within the scope of that claim.

Certain exemplary method and/or apparatus embodiments can provide automatic braces detection and removal by initial (e.g., coarse) bracket detection, subsequent wire detection, and refinement of detected (e.g., separated) initial brackets, which can then be removed from the initial 3D mesh. Exemplary embodiments according to the application can include various features described herein (individually or in combination).

While the invention has been illustrated with respect to one or more implementations, alterations and/or modifications can be made to the illustrated examples without departing from the spirit and scope of the appended claims. The presently disclosed embodiments are therefore considered in all respects to be illustrative and not restrictive. In addition, while a particular feature of the invention can have been disclosed with respect to one of several implementations, such feature can be combined with one or more other features of the other implementations as can be desired and advantageous for any given or particular function. The term "at least one of is used to mean one or more of the listed items can be selected. The term "about" indicates that the value listed can be somewhat altered, as long as the alteration does not result in nonconformance of the process or structure to the illustrated embodiment. Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims, and all changes that come within the meaning and range of equivalents thereof are intended to be embraced therein.