Orthodontic treatment planning based on reinforcement learning

Abstract

A method comprising receiving, by a processing device, clinical data of a first state of one or more dental arches of a patient. The method further comprises determining, by the processing device, one or more stages of an orthodontic treatment plan for correction of the one or more dental arches based on processing of the clinical data. Determining the stages of the orthodontic treatment plan comprises determining one or more actions to be performed on the one or more dental arches and determining a target state of the one or more dental arches predicted to result, at least in part, from the one or more actions.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of dentistry, and more particularly to a system and method for automated orthodontic treatment planning. Background Technology

[0002] Orthodontic treatment can be performed using a series of appliances and / or other orthodontic instruments. Typically, an orthodontist determines an orthodontic treatment plan and orders a series of appliances. The patient wears the appliances in a predetermined sequence to adjust the patient's teeth from the initial configuration to the final target configuration. The orthodontist usually makes one or more clinical decisions associated with the orthodontic treatment plan, which have a significant impact on the success or failure of the plan. Furthermore, in many cases, teeth do not move as planned in the orthodontic treatment plan. However, the orthodontist may find it difficult to identify which teeth might move according to the plan and which might lag or not move. Additionally, the orthodontist may also find it difficult to identify other deviations from the orthodontic treatment plan during treatment, determine why teeth are not moving, and determine how to remedy these deviations. Determining the optimal orthodontic treatment plan for a patient, determining which clinical actions to perform on the patient in one or more treatment phases, and determining the optimal number of treatment phases are also challenging for the orthodontist. Summary of the Invention

[0003] In a first aspect of this disclosure, a method includes: receiving clinical data of a first state of a patient's dentition (e.g., one or more dental arches) by a processing device; and determining one or more stages of an orthodontic treatment plan for correcting the dentition (e.g., one or more dental arches) by the processing device (with or without user input) based on processing of the clinical data, wherein the stage of determining the orthodontic treatment plan includes: determining one or more actions to be performed on the dentition (e.g., one or more dental arches); and determining a target state of the dentition (e.g., one or more dental arches) predicted to be at least partially caused by the one or more actions.

[0004] The second aspect of this disclosure may further extend the first aspect of this disclosure. In the second aspect of this disclosure, the method further includes: determining an orthodontic treatment plan comprising one or more stages; determining one or more additional orthodontic treatment plans; outputting the orthodontic treatment plan and the one or more additional orthodontic treatment plans to a display; receiving a selection of the orthodontic treatment plan; and implementing the selected orthodontic treatment plan.

[0005] The third aspect of this disclosure may further extend the second aspect of this disclosure. In the third aspect of this disclosure, the method further includes: determining, for an orthodontic treatment plan, a first score associated with the predicted completion of one or more target conditions by the orthodontic treatment plan; and determining, for each of one or more additional orthodontic treatment plans, an additional score associated with the completion of one or more target conditions by the additional orthodontic treatment plan.

[0006] The fourth aspect of this disclosure may further extend the third aspect of this disclosure. In the fourth aspect of this disclosure, a first score and additional scores from one or more additional orthodontic treatment plans are output to a display.

[0007] The fifth aspect of this disclosure may further extend the first through fourth aspects of this disclosure. In the fifth aspect of this disclosure, a trained machine learning model is used to perform processing of clinical data, wherein the trained machine learning model outputs one or more actions and a target state of one or more dental arches.

[0008] The sixth aspect of this disclosure may further extend the fifth aspect of this disclosure. In the sixth aspect of this disclosure, the first state of one or more dental arches is an intermediate state achieved by a previously determined orthodontic treatment plan, and the method further includes: receiving information associated with the previously determined orthodontic treatment plan; determining a cost value or a reward value based on the similarity between the first state of one or more dental arches and a predicted state of one or more dental arches from the previously determined orthodontic treatment plan; and updating the training of a trained machine learning model based on the cost value or reward value, wherein processing of clinical data using the trained machine learning model is performed after the training of the trained machine learning model has been updated.

[0009] The seventh aspect of this disclosure may further extend any of the first through sixth aspects of this disclosure. In the seventh aspect of this disclosure, the receipt of clinical data and the determination of one or more phases of orthodontic treatment planning are performed before the commencement of orthodontic treatment for one or more dental arches, the method further comprising: receiving new clinical data on the new state of one or more dental arches by a processing device during an intermediate phase of orthodontic treatment planning; and determining one or more updated phases of orthodontic treatment planning by the processing device based on the processing of the new clinical data, with or without user input.

[0010] The eighth aspect of this disclosure may further extend the seventh aspect of this disclosure. In the eighth aspect of this disclosure, one or more updated stages include a final stage having a new target state with one or more dental arches.

[0011] The ninth aspect of this disclosure may further extend the seventh or eighth aspects of this disclosure. In the ninth aspect of this disclosure, the method also includes adding one or more new intermediate stages to orthodontic treatment planning based on new clinical data.

[0012] The tenth aspect of this disclosure may further extend any one of the first to ninth aspects of this disclosure. In the tenth aspect of this disclosure, the clinical data includes at least one of the following: a color two-dimensional (2D) image of one or more dental arches, a three-dimensional (3D) model of each of the one or more dental arches, an intraoral scan of one or more dental arches, or an X-ray image (e.g., a radiograph) of one or more dental arches.

[0013] The 11th aspect of this disclosure may further extend any one of the 1st to 10th aspects of this disclosure. In the 11th aspect of this disclosure, one or more actions include at least one of the following actions: widening one or more dental arches, adding one or more attachments to one or more dental arches, extracting one or more teeth from one or more dental arches, performing proximal enamel removal on one or more teeth of one or more dental arches, or using elastic elements to fix one or more dental arches to each other.

[0014] The 12th aspect of this disclosure may further extend any one of the 1st to 11th aspects of this disclosure. In the 12th aspect of this disclosure, one or more actions include adjusting at least one of the positions or orientations of one or more teeth on one or more dental arches.

[0015] The 13th aspect of this disclosure may further extend any of the 1st to 12th aspects of this disclosure. In the 13th aspect of this disclosure, one or more actions and target states are determined to minimize at least one of the number of stages in orthodontic treatment planning or the duration of orthodontic treatment performed according to the orthodontic treatment plan.

[0016] The 14th aspect of this disclosure may further extend any one of the 1st to 13th aspects of this disclosure. In the 14th aspect of this disclosure, the method further includes: determining the occlusal type of one or more dental arches based on processing of clinical data, wherein the occlusal type includes at least one of open bite, reverse bite, or deep overbite; wherein one or more actions are determined at least in part based on the occlusal type.

[0017] The 15th aspect of this disclosure may further extend any one of the 1st to 14th aspects of this disclosure. In the 15th aspect of this disclosure, the method further includes: determining the type of malocclusion of one or more dental arches based on processing of clinical data; wherein one or more actions are determined at least in part based on the type of malocclusion.

[0018] The 16th aspect of this disclosure may further extend any one of the 1st to 15th aspects of this disclosure. In the 16th aspect of this disclosure, one or more dental arches include the patient's upper dental arch, and the method further includes: receiving clinical data of a first state of the patient's lower dental arch by a processing device; and determining one or more stages of orthodontic treatment planning for correcting the lower dental arch by the processing device based on processing of the clinical data, wherein the stage of determining the orthodontic treatment planning for correcting the lower dental arch includes: determining one or more additional actions to be performed on the lower dental arch; and determining a target state of the lower dental arch predicted to be at least partially caused by the one or more additional actions.

[0019] The 17th aspect of this disclosure may further extend any one of the 1st to 16th aspects of this disclosure. In the 17th aspect of this disclosure, the method further includes: determining a current relationship between the upper dental arch and the lower dental arch in one or more dental arches, wherein determining a target state of one or more dental arches predicted to be at least partially caused by one or more actions includes determining the target relationship between the upper and lower dental arches.

[0020] The 18th aspect of this disclosure may further extend any of the 1st to 17th aspects of this disclosure. In the 18th aspect of the invention, determining one or more actions to be performed on one or more dental arches includes determining one or more tooth restoration actions to be performed, the one or more tooth restoration actions being associated with one or more restoration options.

[0021] The 19th aspect of this disclosure may further extend the 18th aspect of this disclosure. In the 19th aspect of this disclosure, one or more restoration options include at least one of a) one or more sizes that can be used as a patch, b) a type of patch used, or c) a patch thickness; and one or more restoration tooth actions to be performed include an amount of tooth body removed from the tooth to which the patch is to be received.

[0022] The 20th aspect of this disclosure may further extend the 18th aspect of this disclosure. In the 20th aspect of this disclosure, the space for placing a dental implant is compressed by the roots of one or more teeth adjacent to the space, and wherein one or more tooth restoration actions to be performed include opening the roots of the one or more teeth to allow placement of the dental implant in the space.

[0023] In a 21st aspect of this disclosure, a method includes: receiving training data items by a processing device, the training data items including clinical data on the state of one or more dental arches (and optionally, the spatial relationship or occlusal relationship between the upper and lower dental arches) and information regarding the execution of orthodontic treatment for the one or more dental arches; determining one or more phases of an orthodontic treatment plan for correcting the one or more dental arches by the processing device based on processing the clinical data using a machine learning model, wherein determining the phases of the orthodontic treatment plan includes: determining one or more actions to be performed on the one or more dental arches; and determining a target state of the one or more dental arches at least in part as a result of the one or more actions; determining a cost value or reward value associated with the one or more phases of the orthodontic treatment plan using a cost function; and updating one or more nodes of the machine learning model based on the cost value or reward value.

[0024] The 22nd aspect of this disclosure may further extend the 21st aspect of this disclosure. In the 22nd aspect of this disclosure, determining the cost value or reward value and updating one or more nodes of the machine learning model are performed according to a reinforcement learning algorithm.

[0025] The 23rd aspect of this disclosure can further extend the 27th aspect of this disclosure. In the 23rd aspect of this disclosure, the reinforcement learning algorithm is one of Q-learning, deep Q-learning, or a dual deep Q-network.

[0026] The 24th aspect of this disclosure may further extend any of the 21st to 23rd aspects of this disclosure. In the 24th aspect of this disclosure, the method further includes: determining the number of stages in the orthodontic treatment plan; determining the amount of variation between the number of stages in the orthodontic treatment plan and the target maximum number of stages in the orthodontic treatment plan; and determining a cost value or a reward value based on the amount of variation.

[0027] The 25th aspect of this disclosure may further extend any one of the 21st to 24th aspects of this disclosure. In the 25th aspect of this disclosure, the method further includes: determining a predicted amount of time for completing the orthodontic treatment plan; determining a change between the predicted amount of time and a target maximum amount of time for completing the orthodontic treatment plan; and determining a cost value or a reward value based on the change.

[0028] The 26th aspect of this disclosure may further extend any one of aspects 21 to 25 of this disclosure. In the 26th aspect of this disclosure, the method further includes: receiving input for completing at least one of a target number of stages or a target amount of time for orthodontic treatment planning; and setting a cost function based on the input.

[0029] The 27th aspect of this disclosure may further extend the 26th aspect of this disclosure. In the 27th aspect of this disclosure, a machine learning model is trained to generate an orthodontic treatment plan that achieves a target state of the dental arch while minimizing at least one of the number of stages or the amount of time used to complete the orthodontic treatment plan.

[0030] The 28th aspect of this disclosure may further extend any of the 21 to 27 aspects of this disclosure. In the 28th aspect of this disclosure, clinical data are marked using occlusal type, wherein the occlusal type is one of open bite, deep overbite, or reverse bite.

[0031] The 29th aspect of this disclosure may further extend any of the 21st to 28th aspects of this disclosure. In the 29th aspect of this disclosure, clinical data are labeled using treatment problems, wherein the treatment problems include at least one of vertical problems, sagittal problems, or transverse problems.

[0032] The 30th aspect of this disclosure may further extend any of the 21st to 29th aspects of this disclosure. In the 30th aspect of this disclosure, clinical data are characterized by the number of treatment phases, actions performed in one or more treatment phases, and the degree of success of orthodontic treatment.

[0033] The 31st aspect of this disclosure may further extend any of the 21st to 30th aspects of this disclosure. In the 31st aspect of this disclosure, a machine learning model is trained to receive new clinical data on a patient's dental arch and output multiple treatment planning options for orthodontic treatment of the patient's dental arch.

[0034] The 32nd aspect of this disclosure may further extend any of the 21st to 31st aspects of this disclosure. In the 32nd aspect of this disclosure, the training data item includes multiple states of one or more dental arches, each of which is associated with a different stage of orthodontic treatment.

[0035] The 33rd aspect of this disclosure may further extend any of the 21st to 32nd aspects of this disclosure. In the 33rd aspect of this disclosure, clinical data includes at least one of the following: one or more color two-dimensional (2D) images of one or more dental arches, one or more three-dimensional (3D) models of one or more dental arches, one or more intraoral scans of one or more dental arches, or one or more X-ray images (e.g., radiographs) of one or more dental arches.

[0036] The 34th aspect of this disclosure may further extend any of the 21 to 33 aspects of this disclosure. In the 34th aspect of this disclosure, one or more actions include at least one of the following actions: widening one or more dental arches, adding one or more attachments to a dental arch, extracting one or more teeth from a dental arch, performing proximal enamel removal on one or more teeth of a dental arch, or using an elastic member to secure a dental arch to an opposing dental arch.

[0037] The 35th aspect of this disclosure may further extend any of the 21st to 34th aspects of this disclosure. In the 35th aspect of this disclosure, the training data item also includes clinical data on the occlusal relationship between the maxillary and mandibular arches in one or more dental arches.

[0038] The 36th aspect of this disclosure may further extend any one of the first to 35 aspects of this disclosure. In the 36th aspect of this disclosure, a computer-readable medium includes instructions that, when executed by a processing apparatus, cause the processing apparatus to perform the method of any one of the first to 35 aspects of this disclosure.

[0039] The 37th aspect of this disclosure may further extend any one of the 1st to 35th aspects of this disclosure. In the 37th aspect of this disclosure, a system includes: a computing device including a memory; and a processing device, wherein the processing device is configured to execute instructions from the memory to perform the method of any one of the 1st to 35th aspects of this disclosure.

[0040] The 38th aspect of this disclosure may further extend the 37th aspect of this disclosure. In the 38th aspect of this disclosure, the computing device is configured to manufacture one or more orthodontic appliances according to one or more stages of an orthodontic treatment plan.

[0041] The 39th aspect of this disclosure may further extend the 37th or 38th aspects of this disclosure. In the 39th aspect of this disclosure, the system further includes: an additional computing device configured to transmit clinical data to the computing device; and a storage device configured to store orthodontic treatment plans. Attached Figure Description

[0042] In the various figures of the accompanying drawings, embodiments of the invention are shown by way of example and not limitation.

[0043] Figure 1 An embodiment of a system for orthodontic treatment planning and implementation according to an embodiment of the present disclosure is shown.

[0044] Figure 2 The present disclosure illustrates a model training workflow and a model application workflow for treatment planning applications according to embodiments of the present disclosure.

[0045] Figure 3A An application of reinforcement learning for training one or more machine learning models to generate orthodontic treatment plans, according to an embodiment of this disclosure, is illustrated.

[0046] Figure 3B An example machine learning model architecture according to an embodiment of this disclosure is shown.

[0047] Figure 3C An example autoencoder according to an embodiment of this disclosure is shown.

[0048] Figure 4 A flowchart is shown, according to an embodiment, of a method for performing reinforcement learning to train a machine learning model to perform treatment planning.

[0049] Figure 5 A flowchart of a method for training a machine learning model to generate orthodontic treatment plans, according to an embodiment, is shown.

[0050] Figure 6 A flowchart is shown for a method for determining cost or reward values ​​for one or more treatment phases applied to a generated treatment plan.

[0051] Figure 7 A flowchart illustrating a method for generating orthodontic treatment plans using a trained machine learning model, according to an embodiment, is shown.

[0052] Figure 8 A flowchart illustrating a method for selecting a trained machine learning model to generate orthodontic treatment plans according to an embodiment is shown.

[0053] Figure 9 A flowchart is shown of a method for determining one or more stages of orthodontic treatment planning according to an embodiment.

[0054] Figure 10A A tooth repositioning system according to certain embodiments is shown.

[0055] Figure 10B A method of orthodontic treatment using multiple instruments according to certain embodiments is shown.

[0056] Figure 11 A method for designing orthodontic appliances according to certain embodiments is shown.

[0057] Figure 12 A method for digitally planning orthodontic treatment, according to certain embodiments, is shown.

[0058] Figure 13 A block diagram of an example computing device according to an embodiment of the present invention is shown. Detailed Implementation

[0059] This document describes methods and apparatus for using machine learning to determine one or more aspects of orthodontic treatment planning. In embodiments, clinical data representing the current state of one or more patient dental arches and / or one or more past states of one or more patient dental arches (e.g., including spatial or occlusal relationships between the patient's upper and lower dental arches) are input into a trained machine learning model. The trained machine learning model can then output a complete treatment plan (e.g., a complete orthodontic treatment plan), one or more phases of the treatment plan, recommendations for the treatment plan, etc. In embodiments, the machine learning model is trained to learn strategies for achieving clinical (e.g., orthodontic and / or rehabilitation) goals while maximizing one or more defined rewards or objectives (e.g., minimizing the number of orthodontic appliances). The machine learning model can be trained, for example, to perform clinical judgment regarding the generation of new treatment plans (e.g., orthodontic treatment plans, rehabilitation treatment plans, or orthodontic-rehabilitation treatment plans) or the updating of existing treatment plans. In some embodiments, a rehabilitation treatment plan or orthodontic treatment plan can be updated to an orthodontic-rehabilitation treatment plan. For one or more phases of orthodontic treatment planning or orthodontic-restoration treatment planning, a trained machine learning model can output one or more actions to be performed, such as interproximal enamel removal, tooth extraction, adding attachments to one or more teeth, using elastomers, widening one or more dental arches (e.g., via palatal expanders), increasing space between two or more teeth, repositioning one or more teeth in one or more directions, rotating one or more teeth about one or more axes, etc. In embodiments, the trained machine learning model can suggest actions that do not require physician intervention and / or can suggest clinical actions that can be performed by a physician, and can use these suggested actions to supplement the physician's decisions. In embodiments, reinforcement learning can be used to train the machine learning model based on a library of previous successful and / or unsuccessful treatments of past patients. The machine learning model may have been trained to optimize for one or more objectives, such as minimizing treatment time while still achieving the target end state of the patient's dentition, minimizing the number of treatment phases while still achieving the target end state of the patient's dentition, etc. In embodiments, the machine learning model can be trained to satisfy both clinical constraints (e.g., avoiding tooth and / or root collisions, tooth movement constraints, etc.) and business constraints (e.g., the number of treatment phases, the length of treatment, etc.).

[0060] Orthodontic treatment is typically performed in stages, and the treatment plan may require the patient's teeth to move a specific amount in each stage. However, dentists may find it difficult to determine the number of stages needed to achieve the goal, which actions to perform on the patient's teeth, when to perform those actions, whether the treatment is progressing according to the plan, and / or what actions to perform if it is determined that the treatment is not progressing according to the plan. For example, dentists may find it difficult to determine whether teeth have moved according to the treatment plan, whether some teeth have not moved according to the treatment plan, whether the planned occlusal level will be achieved, whether the planned arch expansion will be achieved, whether any teeth will be extracted, whether interproximal enamel reduction (IPR) will be performed, whether an elastomer will be used between the upper and lower arches, where attachments will be placed on the teeth, etc. Some examples of orthodontic treatment have been discussed. However, it should be understood that such examples also apply to orthodontic-restorative treatment, which includes both orthodontic treatment and restorative treatment.

[0061] The embodiments provide a method and system for using machine learning to assist in treatment planning, particularly using machine learning to suggest actions associated with the treatment plan and / or suggest actions to be taken by a dentist in one or more phases of orthodontic treatment to achieve one or more objectives, such as minimizing the number of treatment phases, minimizing possible treatment time, minimizing patient discomfort, etc. In the embodiments, the system and method are additionally capable of evaluating the actual progress of an orthodontic treatment plan with a target final position (e.g., evaluating patient teeth during an intermediate phase of a multi-phase orthodontic treatment plan) and updating the treatment plan and / or suggesting one or more actions for the treatment plan based on such evaluation. Suggested actions may include corrective actions, such as modifying the final treatment plan (e.g., modifying the final tooth position) and / or phasing the tooth positions in the treatment plan (if the treatment plan is a multi-phase treatment plan). Phasing refers to the sequence of movements from the current or initial tooth position to the new tooth position. Phasing includes determining which tooth movements will be performed at different treatment phases. Some corrective actions may require the dentist to perform one or more actions or operations, such as placing attachments, IPR, applying palatal expanders, using temporary anchoring devices, etc.

[0062] In the field of orthodontics, the embodiments offer significant advantages over traditional treatment techniques and can improve orthodontic treatment outcomes while shortening treatment time and reducing the number of stages used. The embodiments can provide a system that suggests stages of orthodontic treatment, instructs dentists on recommended actions to be performed in one or more treatment stages, etc. Therefore, treatment planning is improved in the embodiments. This improvement in treatment planning, by reducing the number of continuous refinements to the treatment plan (and the associated sequence of additional appliances) during treatment, can lead to increased patient satisfaction and reduced costs.

[0063] This document discusses embodiments of multi-stage treatment planning. However, such embodiments are also applicable to single-stage orthodontic treatment planning with a target final position. For example, image data may be generated at some point after the start of single-stage orthodontic treatment planning. If the image data shows that the progress of the single-stage treatment plan is not as expected, the target final position of the single-stage treatment plan may be adjusted and / or one or more treatment parameters used to achieve the target final position may be adjusted. Therefore, it should be understood that all discussions of multi-stage treatment planning herein also apply to single-stage treatment planning with a target final position and / or condition.

[0064] Physicians have noted for many years that controlling the upper teeth is very difficult. When the embodiments of this disclosure are applied to this common clinical problem, they can help improve control over these upper teeth.

[0065] Furthermore, this document discusses some embodiments with reference to the creation and use of orthodontic appliances. As used herein, an orthodontic appliance is an orthodontic instrument used for repositioning teeth. It should be noted that the embodiments are also applicable to other types of orthodontic instruments, including but not limited to frameworks and wires, retainers, or functional instruments. For example, actions defined using frameworks and wires at one or more stages of orthodontic treatment may include attaching the framework to a specific location on the teeth, selecting archwires, selecting ligature types, changing archwire selection, repositioning one or more frameworks, bending archwires at specific times, etc. Therefore, it should be understood that any discussion of orthodontic appliances herein also applies to other types of orthodontic instruments.

[0066] Figure 1 An embodiment of a system 100 for performing orthodontic treatment or orthodontic-rehabilitation treatment is shown. In this embodiment, system 100 performs one or more operations described below in methods 200-1200. System 100 may include only components located at a single location (e.g., dental clinic 108 or dental laboratory), or may include components located at multiple different locations (e.g., components at dental clinic 108 or dental laboratory and components at a second location remote from the dental clinic, such as a server farm providing cloud computing services 109). System 100 may also optionally include a manufacturing facility 110, which includes one or more three-dimensional (3D) printers 185 and / or other additive manufacturing machines that can be used to manufacture orthodontic appliances. Dental clinic 108 includes a computing device 105. In embodiments incorporating the use of cloud computing services 109 or other remote servers, system 100 also includes a remote server computing device 109 connected to computing device 105 via network 170. Network 170 can be a local area network (LAN), a public wide area network (WAN) (e.g., the Internet), a private WAN (e.g., an intranet), or a combination thereof.

[0067] Computing device 105 may be connected to and / or include data storage 110. Computing device 106 may also be connected to and / or include data storage (not shown). Data storage may be local data storage and / or remote data storage. Computing device 105 and computing device 106 may each include one or more processing devices, memory, auxiliary storage, one or more input devices (e.g., such as a keyboard, mouse, tablet computer, etc.), one or more output devices (e.g., a display, printer, etc.) and / or other hardware components. In some embodiments, computing device 105 and / or computing device 106 may not include input and / or output devices (e.g., not connected to a keyboard, mouse, display, etc.). In some embodiments, computing device 105 may be integrated into scanner 150 or image capture device 160 to improve mobility.

[0068] In some embodiments, a scanner 150 for obtaining three-dimensional (3D) data of tooth locations in a patient's oral cavity is operatively connected to a computing device 105. In embodiments, one or more handheld intraoral scanners 150 (also referred to as intraoral scanners or simply scanners) are wirelessly connected to the computing device 105. In one embodiment, the scanner 150 is wirelessly connected to the computing device 105 via a direct wireless connection. In another embodiment, the scanner 150 is wirelessly connected to the computing device 105 via a wireless network. Alternatively, the scanner 150 may be connected to the computing device 105 via a wired connection.

[0069] Scanner 150 may include a probe (e.g., a handheld probe) for optically capturing three-dimensional structures. An example of such scanner 150 is the iTero® intraoral digital scanner manufactured by Align Technology, Inc. In some embodiments, scanner 150 corresponds to the intraoral scanner described in U.S. Application No. 2019 / 0388193, filed June 19, 2019, entitled “Intraoral 3D Scanner Employing Multiple Miniature Cameras and Multiple Miniature Pattern Projectors,” which is incorporated herein by reference. In some embodiments, scanner 150 corresponds to the intraoral scanner described in U.S. Application No. 2019 / 0388193, filed June 23, 2020, entitled “Intraoral 3D Scanner Employing Multiple Miniature Cameras and Multiple Miniature Pattern Projectors,” which is incorporated herein by reference. In some embodiments, scanner 150 corresponds to the intraoral scanner described in U.S. Patent No. 10,835,128, published November 17, 2020, which is incorporated herein by reference. In some embodiments, scanner 150 corresponds to the intraoral scanner described in U.S. Patent No. 10,918,286, published February 21, 2021, which is incorporated herein by reference.

[0070] Intraoral scanner 150 can generate an intraoral scan, which may be or include color or monochrome 3D information, and send the intraoral scan to computing device 105. In some embodiments, the intraoral scan includes a height map. Additionally or alternatively, intraoral scanner 150 can generate a color two-dimensional (2D) image (e.g., a viewfinder image) and send the color 2D image to local server computing device 105. Additionally or alternatively, scanner 150 can generate 2D or 3D images under specific lighting conditions, such as infrared or near-infrared (NIRI) light and / or ultraviolet light, and can send such 2D or 3D images to server computing device 105. Intraoral scans, color images, and images under specific lighting conditions (e.g., NIRI images, infrared images, ultraviolet images, etc.) are collectively referred to as intraoral scan data. Oral scan data and other data on the dental arches, such as X-ray images (also referred to herein as radiographs), 2D or 3D images generated by devices other than oral scanners, cone-beam computed tomography (CBCT) scan data, etc., are collectively referred to as clinical data 135. Clinical data may include images of a single dental arch or both the upper and lower dental arches of a patient. For clinical data that includes information on both the upper and lower dental arches, the clinical data may include occlusal relationship (also referred to as spatial relationship) data between the upper and lower dental arches. This information can indicate how the upper and lower dental arches occlude relative to each other and / or occlusal contact information between the teeth in the upper and lower dental arches. An operator may begin recording a scan when the scanner 150 is in a first position in the oral cavity, move the scanner 150 to a second position within the oral cavity while the scan is being performed, and then stop recording the scan. In some embodiments, recording may begin automatically when the scanner 150 identifies teeth and / or other objects.

[0071] An intraoral scanning application 108 running on computing device 105 can communicate with scanner 150 to perform intraoral scanning. The results of the intraoral scanning can be clinical data 135, which may include one or more sets of intraoral scans, one or more sets of viewfinder images (e.g., color 2D images showing the field of view of the intraoral scanner), one or more sets of NIRI images, etc. Each intraoral scan can be a two-dimensional (2D) or 3D image that includes height information (e.g., a height map) of a portion of tooth locations and therefore may include x, y, and z information. In one embodiment, each intraoral scan is a point cloud. In one embodiment, intraoral scanner 150 generates a plurality of discrete (i.e., individual) intraoral scans and / or additional images. In some embodiments, the set of discrete intraoral scans may be combined into a smaller set of hybrid intraoral scans, wherein each hybrid scan is a combination of multiple discrete intraoral scans.

[0072] In one embodiment, scanner 150 generates a stream of intraoral scan data and sends it to computing device 105. In some embodiments, the stream of intraoral scan data may include separate streams of intraoral scans, color images, and / or NIRI images (and / or other images under specific lighting conditions). In one embodiment, a stream of mixed intraoral scans is sent to computing device 105.

[0073] The computing device 105 receives intraoral scan data from the scanner 150 and then stores the intraoral scan data as clinical data 135 in the data storage 110. According to an example, a user (e.g., a doctor) can perform an intraoral scan on a patient. In doing so, the user can apply the scanner 150 to one or more intraoral locations of the patient. The scan can be divided into one or more segments. For example, a segment may include the patient's lower cheek region, the patient's lower tongue region, the patient's upper cheek region, the patient's upper tongue region, one or more of the patient's prepared teeth (e.g., patient teeth with dental appliances such as crowns or orthodontic alignment devices applied), one or more teeth in contact with the prepared teeth (e.g., teeth that are not fitted with dental appliances themselves but are located next to one or more such teeth or abut against one or more such teeth when the mouth is closed), and / or the patient's occlusion (e.g., a scan performed with the patient's mouth closed, where the scan is directed towards the interface region between the patient's upper and lower teeth). Through such scanner application, the scanner 150 can provide intraoral scan data to the computing device 105. Intraoral scan data can be provided in the form of intraoral scan / image datasets, each of which may include 2D and / or 3D intraoral scans / images of regions of specific teeth and / or intraoral points. In one embodiment, separate scan / image datasets are created for the maxillary arch, mandibular arch, patient occlusion, and each prepared tooth. Alternatively, a single large intraoral scan / image dataset (e.g., for the mandibular arch and / or maxillary arch) can be generated. Such scans / images can be provided from the scanner to the computing device 105 in the form of one or more points (e.g., one or more pixels and / or groups of pixels). For example, the scanner 150 may provide such 3D scans / images as one or more point clouds.

[0074] The method of scanning a patient's oral cavity can depend on the procedure to be applied. For example, if a maxillary or mandibular denture is to be created, a full scan of the edentulous arch of the mandible or maxilla may be performed. Conversely, if a dental bridge is to be created, only a portion of the entire arch may be scanned, including the edentulous area, adjacent prepared teeth (e.g., adjacent teeth), and the opposing arch and dentition. Furthermore, the method of scanning the oral cavity can depend on the dentist's scanning preferences and / or the patient's condition.

[0075] By way of non-limiting examples, dental procedures can be broadly categorized into restorative (restorative) and orthodontic procedures, which are then further subdivided into specific forms of these procedures. Furthermore, dental procedures can include the identification and treatment of gingival diseases, sleep apnea, and intraoral conditions. The term "restorative procedure" specifically refers to any procedure involving the oral cavity, involving the design, fabrication, or installation of dental restorations at intraoral sites or on real or virtual models of teeth, or involving the design and preparation of intraoral sites to receive such restorations. Restorations can include any restoration, such as crowns, veneers, inlays, onlays, implants, and bridges, as well as any other artificial partial or complete dentures. The term "orthodontic procedure" specifically refers to any procedure involving the oral cavity, involving the design, fabrication, or installation of orthodontic elements at intraoral sites or on real or virtual models of teeth, or involving the design and preparation of intraoral sites to receive such orthodontic elements. These elements can be instruments, including but not limited to frameworks and wires, retainers, clear aligners, or functional appliances.

[0076] During the intraoral scanning process, the intraoral scanning application 108 receives and processes intraoral scanning data (e.g., intraoral scans) and generates a 3D surface of the scanned area of ​​the oral cavity (e.g., tooth locations) based on this processing. To generate the 3D surface, the intraoral scanning application 108 can register and “stitch” or merge the intraoral scans generated from the intraoral scanning process in real-time or near real-time during the scan execution. In one embodiment, performing registration includes capturing 3D data of individual points of the surface in multiple scans (views from a camera) and registering the scans by calculating transformations between scans. The 3D data can be projected into 3D space for transformation and stitching. By applying appropriate transformations to the points of each registered scan and projecting each scan into 3D space, the scans can be integrated into a common reference frame.

[0077] In one embodiment, registration is performed on adjacent or overlapping intraoral scans (e.g., each consecutive frame of an intraoral video). In another embodiment, registration is performed using mixed scans and / or reduced or cropped scans. A registration algorithm is performed to register two or more adjacent intraoral scans and / or to register intraoral scans with an already generated 3D surface, which essentially involves determining a transformation to align one scan with another scan and / or with a 3D surface. Registration may involve identifying multiple points in each scan (e.g., a point cloud) of a scan pair (or scan and 3D model), performing surface fitting on the points, and using a local search around the points to match the points of the two scans (or scan and 3D model). For example, intraoral scan application 115 may match a point of one scan with the nearest point interpolated on the surface of another image and iteratively minimize the distance between the matched points. Other registration techniques may also be used. Intraoral scan application 115 may repeat registration and stitching for all scans in an intraoral scan sequence and update the 3D surface upon receiving a scan.

[0078] When the scanning process is complete (e.g., all scans for intraoral or tooth locations have been captured), the intraoral scanning application 108 or the treatment planning application 115 can generate a virtual 3D model (also referred to as a digital 3D model) of one or more scanned tooth locations. The virtual 3D model includes 3D surfaces of one or more scanned tooth locations, but with a higher degree of accuracy than the 3D surfaces generated during the scanning process. To generate the virtual 3D model, the intraoral scanning application 108 or the treatment planning application 115 can register and “stitch” or merge the intraoral scans generated from the intraoral scanning process. In one embodiment, registration is performed on adjacent and / or overlapping intraoral scans (e.g., each consecutive frame of an intraoral video). In one embodiment, registration is performed using blended scans and / or reduced or cropped scans. Registration algorithms can be performed to register two or more adjacent intraoral scans and / or register intraoral scans with a 3D model, which essentially involves determining transformations to align one scan with another scan and / or with a 3D model. Registration can involve identifying multiple points in each scan (e.g., a point cloud) of a scan pair (or scan and 3D model), performing surface fitting on the points, and using a local search around the points to match the points between the two scans (or scan and 3D model). For example, an intraoral scanning application 108 or a treatment planning application 115 might match points from one scan with the closest interpolated points on the surface of another scan, iteratively minimizing the distance between the matched points. Other registration techniques can also be used. Performing registration and stitching for generating a 3D model can be more accurate than performing registration and stitching for generating a 3D surface that is displayed in real-time or near real-time during the scanning process.

[0079] The intraoral scanning application 108 or treatment planning application 115 can repeatedly register all scans in the intraoral scanning sequence to obtain a transformation for each scan, thereby registering each scan with the previous and / or with a common reference frame (e.g., with a 3D model). The intraoral scanning application 108 or treatment planning application 115 integrates all scans into a single virtual 3D model by applying appropriate, defined transformations to each scan. Each transformation may include rotation about one to three axes and translation in one to three planes.

[0080] In addition to clinical data 135, which includes data captured by scanner 150 and / or data generated from such captured data (e.g., virtual 3D models), clinical data 135 may also, or alternatively, include data from one or more additional image capture devices 160. Additional image capture devices 160 may include X-ray devices capable of generating standard X-rays (e.g., apical X-rays), panoramic X-rays, cephalometric X-rays, etc. Additionally or alternatively, additional image capture devices 160 may include X-ray devices capable of generating cone-beam computed tomography (CBCT) scans. Additionally or alternatively, additional image capture devices 160 may include standard optical image capture devices (e.g., cameras) that generate two-dimensional or three-dimensional images or videos of a patient's oral cavity and dental arches. For example, additional image capture devices 160 may be mobile phones, laptop computers, image capture attachments attached to laptops or desktop computers (e.g., devices using Intel® RealSense™ 3D image capture technology), etc. This additional image capture device 160 can be operated by the patient or a friend or family member, and can generate 2D or 3D images that are transmitted via network 170 to computing device 105 or computing device 109. Therefore, clinical data 135 can include 2D optical images, 3D optical images, virtual 2D models, virtual 3D models, intraoral scans, 2D X-ray images, 3D X-ray images, etc.

[0081] Once the intraoral scan is complete, the treatment planning application 115 can receive the intraoral scan data and / or a 3D model generated based on the intraoral scan data. If 3D models of the upper and lower dental arches have not yet been generated, the treatment planning application 115 can generate one or more 3D models as described above. The treatment planning application 115 can then input clinical data 135 into a trained machine learning model that has been trained to facilitate the generation of orthodontic treatment planning. This clinical data may include one or more 3D models, intraoral scans, 2D images, 3D images, projections of one or more 3D models onto one or more planes, CBCT scans, X-ray images, and / or other data. In some embodiments, a final target state of the patient's dentition is generated (e.g., a 3D model of one or more of the patient's dental arches after treatment), and the clinical data 135 input into the trained machine learning model also includes information about the final target state of the patient's dentition. For example, a physician can determine the final target dentition and can generate 3D models of the patient's upper and / or lower dental arches and input them, along with the current state of the patient's dentition, into the trained machine learning model. In an embodiment, the trained machine learning model outputs a complete orthodontic treatment plan or one or more stages of orthodontic treatment plan 186. The machine learning model may additionally output one or more actions to be performed at one or more treatment stages, such as tooth extraction, IPR, palatal expansion, adding attachments, using elastomers, moving one or more teeth in the sagittal, lateral, and / or vertical planes, rotating one or more teeth about one or more axes, changing the thickness and / or shape of the appliance, adding reinforcing features (e.g., indentations) to the appliance, etc. In an embodiment, treatment plan 186 may be stored in data storage 110.

[0082] In one embodiment, an intraoral scan is performed during an intermediate phase of the multi-stage orthodontic treatment plan 186. Alternatively or additionally, other clinical data 135 may be generated at the beginning of the multi-stage orthodontic treatment plan (e.g., before the treatment plan has been generated). In one embodiment, one or more 2D images of the patient's dentition are received during an intermediate phase of treatment. These images may be generated, for example, by a user device, such as a mobile phone belonging to the patient or another person associated with the patient.

[0083] Multi-stage orthodontic treatment planning186 can be used for multi-stage orthodontic treatment or procedures (or multi-stage orthodontic-rehabilitation treatment). The term orthodontic procedure specifically refers to any procedure involving the oral cavity, involving the design, fabrication, or installation of orthodontic elements at tooth locations within the oral cavity or at actual or virtual models, or involving the design and preparation of tooth locations to receive such orthodontic elements. These elements can be instruments, including but not limited to frameworks and wires, retainers, appliances, or functional appliances. Different appliances can be formed for each treatment stage to provide forces for moving the patient's teeth. The shape of each appliance is unique and customized for a specific patient and a specific treatment stage. Each appliance has a tooth-receiving cavity that accommodates the tooth and flexibly repositions it according to the specific treatment stage.

[0084] Restorative dentistry refers to dental procedures and practices aimed at restoring the function, integrity, and appearance of damaged, decayed, or missing teeth. The primary goal of restorative dentistry is to repair or replace teeth to improve oral health and function, allowing patients to chew, speak, and smile confidently. Common restorative dental procedures include applying fillings, applying crowns or pedicles, applying bridges, applying dentures, applying dental implants (e.g., metal posts or frames surgically placed into the jawbone to act as a base for replacing teeth), applying inlays and onlays, applying veneers, and root canal treatment.

[0085] In some cases, machine learning models are used to generate part or all of a new orthodontic treatment plan or orthodontic-rehabilitation treatment plan. In some cases, a multi-stage orthodontic treatment plan 186 for a patient may be initially generated by a dentist (e.g., an orthodontist) or by a treatment planning application 115 after performing an initial pre-treatment scan of the patient's dental arches, and updates to the treatment plan may be determined. The treatment plan 186 may also begin at home (based on the patient's own scan) or at a scanning center. The treatment plan 186 may be created automatically (e.g., via the treatment planning application 115) or by professionals (including orthodontists) in a remote service center. Intraoral scans can provide surface topography data of the patient's oral cavity (including teeth, gingival tissue, etc.). Surface topography data can be generated by directly scanning the oral cavity using a suitable scanning device (e.g., a handheld scanner, a desktop scanner, etc.), a physical model (positive or negative) of the oral cavity, or an impression of the oral cavity. Clinical data from the initial intraoral scan can be used to generate a virtual three-dimensional (3D) model or other digital representation of the initial or starting condition of the patient's upper and / or lower dental arches.

[0086] The dentist or treatment planning application 115 can then determine the desired final state of the patient's dental arch. The final state of the patient's dental arch may include the final arrangement, position, orientation, etc., of the patient's teeth, and may additionally include the final occlusal position, final occlusal surface, final arch length, etc. In an embodiment, the final state is the state of the patient's upper and lower dental arches, and includes the final occlusal relationship of the upper and lower dental arches. Subsequently, the treatment planning application 115 can calculate or determine the motion paths of some or all of the patient's teeth and the changes in the patient's occlusion from the initial position to the planned final position. In some embodiments, one or more suitable computer programs are used to calculate the motion paths, which can take numerical representations of the initial and final positions as input and provide a numerical representation of the motion paths as output. In some embodiments, a trained machine learning model is used to determine the motion paths for one or more treatment phases. The motion path of any given tooth can be calculated based on the position and / or motion paths of other teeth in the patient's dentition. For example, the motion path can be optimized based on minimizing the total distance of movement, preventing collisions between teeth, avoiding more difficult tooth movements, or any other suitable criteria. In some cases, the motion path can be provided as a series of incremental tooth movements, which, when executed sequentially, result in the repositioning of the patient's teeth from their initial to final positions, and / or in the final occlusal relationship between the patient's upper and lower dental arches.

[0087] Multiple treatment phases can be generated based on the determined motion paths using a treatment planning application 115 employing machine learning. Each treatment phase can be an incremental repositioning phase of an orthodontic procedure designed to move one or more of the patient's teeth from the initial tooth arrangement for one or two dental arches to the target arrangement for that phase. The treatment planning application 115 can generate different 3D models of the target condition for each treatment phase and for each dental arch.

[0088] Subsequently, at manufacturing facility 110, one or a set of orthodontic appliances (e.g., aligners) are manufactured based on the generated treatment phases (e.g., based on 3D models of the target condition for each treatment phase). For example, a set of appliances may be manufactured, each individually shaped to fit the tooth arrangement specified by one of the treatment phases, allowing the patient to wear the appliances sequentially to incrementally reposition the teeth from the initial arrangement to the target arrangement. The configuration of the aligners can be selected to induce tooth movements specified by the respective treatment phase. In some embodiments, 3D printer 185 prints molds for the dental arch based on 3D models across multiple treatment phases, and polymer sheets are thermoformed on the molds to form the aligners. In some embodiments, 3D printer 185 directly prints aligners based on 3D models of the aligners determined according to 3D models of the patient's dental arch at multiple treatment phases.

[0089] Sometimes, it can be difficult to determine the final treatment plan at the start of orthodontic treatment. This can be due to a variety of reasons. For example, in some cases, teeth may not move as expected due to the specific biology of a particular patient. Furthermore, sometimes patient compliance with the treatment plan is not ideal (e.g., the patient does not wear his or her appliances as instructed). Therefore, in some embodiments, it may be beneficial to generate more than one treatment plan before treatment begins and / or to generate multiple treatment plan update options during treatment. In some embodiments, a set of multi-stage orthodontic treatment plans is generated before treatment begins. This set of treatment plans may include a first treatment plan with the optimal target condition. The first treatment plan may be the most aggressive treatment plan in the set (e.g., the one that will likely produce the greatest movement of the teeth relative to any other treatment plan in the set). The set of treatment plans may additionally include less aggressive treatment plans (e.g., with alternative target conditions involving less arch change) and / or other, more aggressive treatment plans. Additionally or alternatively, the first treatment plan may involve fewer physical manipulations or procedures in the patient's mouth, while other treatment plans may require more physical manipulations or procedures. For example, the first treatment plan might generate a 5 mm increase in arch length by moving the patient's molars distally, while the second treatment plan might generate a 4 mm increase in arch length by moving the patient's molars distally, and additionally require proximal enamel removal or extraction.

[0090] Treatment planning application 115 (e.g., a machine learning model trained by treatment planning application 115) can process clinical data 135 received during intermediate stages of a multi-stage orthodontic treatment plan 186 or an orthodontic-restoration treatment plan, as well as prior clinical data generated during one or more prior stages of the orthodontic treatment plan 186 and / or the stored state of the orthodontic treatment plan. Based on this processing, treatment planning application 115 can determine one or more updated treatment stages and / or one or more actions associated with one or more treatment stages. For example, treatment planning application 115 can determine one or more corrective actions, such as performing IPR, tooth extraction, adding attachments to one or more teeth, updating the planned position and / or orientation of one or more teeth, adding an additional treatment stage to the treatment plan, changing the shape of one or more appliances, etc. Additionally or alternatively, the treatment planning application can add one or more restorative tooth actions to one or more treatment stages. For example, a restorative workflow can be added at a stage of orthodontic treatment or between two stages of orthodontic treatment.

[0091] In some embodiments, computing device 105 offloads one or more operations to remote server computing device 106. Remote server computing device 106 may be provided by a cloud computing service 109 such as Amazon Web Services (AWS). Compared to local server computing device 105, remote server computing device 106 may have increased resources, such as storage resources, processing resources, etc. Remote server computing device 106 may include a version of treatment planning application 115, which may perform some or all of the operations described above. In embodiments, computing device 105 may send clinical data 135 to remote computing device 106, which may execute treatment planning application 115. Once a treatment plan has been generated, the remote computing device may send instructions to manufacturing facility 110 to manufacture one or more orthodontic appliances associated with a stage of the orthodontic treatment plan. Additionally or alternatively, remote server computing device 106 may store the treatment plan in data storage 110 and / or another data storage.

[0092] Figure 2 A model training workflow 205 and a model application workflow 217 for a treatment planning application according to embodiments of the present disclosure are illustrated. In embodiments, the model training workflow 205 may be executed at a server, which may or may not include the treatment planning application, and the trained model is provided (e.g., in...) Figure 1A treatment planning application (on computing device 105 or 106) can execute a model application workflow 217. The model training workflow 205 and the model application workflow 217 can be executed by processing logic executed by the processor of the computing device. One or more of these workflows 205, 217 can be implemented, for example, through one or more machine learning modules implemented in the treatment planning application 115 or in… Figure 13 This is achieved by other software and / or firmware executed on the processing unit of the computing device 1300 shown.

[0093] Model training workflow 205 is used to train one or more machine learning models (e.g., deep learning models) to perform one or more predictions, recommendations, or task generation regarding orthodontic treatment planning or orthodontic-rehabilitation treatment planning based on clinical data (e.g., 3D scans, height maps, 2D color images, NIRI images, 3D surfaces generated from intraoral scan data, 3D models of dental arches, etc.). Model application workflow 217 applies one or more trained machine learning models to perform predictions, recommendations, or task generation regarding orthodontic treatment planning based on clinical data. One or more of the machine learning models can receive and process 3D data (e.g., 3D point clouds, 3D surfaces, portions of 3D models, etc.). One or more of the machine learning models can receive and process 2D data (e.g., 2D images, height maps, projections of 3D surfaces onto a plane, etc.), which can be used for a single dental arch or for both the upper and lower dental arches.

[0094] One type of machine learning model that can be used to perform some or all of the above requirements is an artificial neural network, such as a deep neural network. Artificial neural networks typically include feature representation components with classifier or regression layers that map features to a desired output space. For example, a convolutional neural network (CNN) hosts multiple layers of convolutional filters. Pooling is performed, and nonlinearity can be processed at lower layers, typically topped by a multilayer perceptron that maps the top-level features extracted by the convolutional layers to a decision (e.g., a classification output). Deep learning is a class of machine learning algorithms that use cascaded, multi-layered nonlinear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. Deep neural networks can learn in a supervised (e.g., classification) and / or unsupervised (e.g., pattern analysis) manner. Deep neural networks consist of hierarchical layers, where different layers learn different levels of representation corresponding to different levels of abstraction. In deep learning, each level learns to transform its input data into a slightly more abstract and comprehensive representation. In image recognition applications, for example, the raw input might be a pixel matrix; the first representation layer abstracts the pixels and encodes the edges; the second layer synthesizes and encodes the arrangement of edges; the third layer encodes higher-level shapes (e.g., teeth, lips, gums, etc.); and the fourth layer identifies the scanned object. It's worth noting that the deep learning process can learn on its own which features are best placed at which level. The "depth" in "deep learning" refers to the number of layers through which the data is transformed. More precisely, deep learning systems have a basic credit allocation path (CAP) depth. CAP is a transformation chain from input to output. CAP describes the underlying causal connection between the input and output. For feedforward neural networks, the CAP depth can be the network depth, or the number of hidden layers plus one. For recurrent neural networks, where signals can propagate through layers more than once, the CAP depth can be infinite.

[0095] Training neural networks can be achieved through supervised learning, which involves feeding the network a training dataset consisting of labeled inputs, observing its outputs, defining the error (by measuring the difference between the output and the labeled values), and using techniques such as deep gradient descent and backpropagation to adjust the network's weights across all its layers and nodes to minimize the error. In many applications, repeating this process on numerous labeled inputs from the training dataset yields a network that can produce correct outputs when presented with inputs different from those present in the training dataset. This generalization is achieved in high-dimensional settings such as large images when sufficiently large and diverse training datasets are available.

[0096] For model training workflow 205, the training dataset contains hundreds, thousands, tens of thousands, hundreds of thousands, or more clinical datasets (e.g., including intraoral scans, images, and / or 3D models, etc.) that should be used to form the training dataset. In an embodiment, up to millions of cases of patients' dentition that may have undergone orthodontic procedures can be used to form the training dataset, wherein each case can include various labels of one or more types of useful information. Each case can include, for example, data showing one or more tooth locations, such as 3D models, intraoral scans, height maps, color images, NIRI images, etc.; data showing pixel-level segmentation of the data (e.g., 3D models, intraoral scans, height maps, color images, NIRI images, etc.) into various tooth categories (e.g., teeth, restoration objects, gingiva, moving tissue, palate, etc.); data showing one or more specified classifications of the data; data indicating the number of stages of orthodontic treatment used; data indicating the length of time of orthodontic treatment; data indicating the success rate of orthodontic treatment, etc. This data can be processed to generate one or more training datasets 236 for training one or more machine learning models. Machine learning models can be trained to, for example, generate one or more recommendations for treatment planning, generate treatment plans, generate one or more phases of treatment plans, and suggest actions to be performed at one or more phases of treatment plans. Such trained machine learning models can be added to treatment planning applications and can be used to facilitate the generation of orthodontic treatment plans.

[0097] In one embodiment, generating one or more training datasets 236 includes collecting one or more clinical data (e.g., labeled) from historical orthodontic treatments 210 and / or one or more 3D models. Processing logic may collect the training dataset 236, which includes clinical data from historical orthodontic treatments 210. To perform training, the processing logic feeds the training dataset(s) 236 into one or more untrained machine learning models. The machine learning models may be initialized before the first input is fed into them. The processing logic trains the untrained machine learning models(s) based on the training dataset(s) to generate one or more trained machine learning models that perform the various operations described above.

[0098] Training can be performed by feeding one or more of an image, scan, or 3D surface (or data from an image, scan, or 3D surface) into a machine learning model (one at a time). Each input can include clinical data, which may include data from an image, an intraoral scan, or a 3D surface from a training data item in the training dataset. As mentioned above, training data items can also include other types of data, such as color images, images generated under specific lighting conditions (e.g., UV or IR radiation), etc. In an embodiment, training data items include a 2D or 3D representation of the current or initial state of one or more of the patient's dental arches and a target final state of the patient's dental arches.

[0099] Machine learning models process inputs to generate outputs. Artificial neural networks consist of an input layer composed of values ​​from data points (e.g., intensity and / or height values ​​of pixels in a height map). The next layer is called a hidden layer, and nodes in the hidden layers each receive one or more input values. Each node contains parameters (e.g., weights) applied to the input values. Thus, each node primarily feeds the input values ​​into a multivariable function (e.g., a nonlinear mathematical transformation) to produce an output value. The next layer can be another hidden layer or an output layer. In each case, nodes in the next layer receive output values ​​from nodes in the previous layer, and each node applies weights to these values ​​and then generates its own output value. This can be performed at each layer. The final layer is the output layer, where a node exists for each class, prediction, and / or output that the machine learning model can produce. Therefore, the output can include one or more predictions, actions, recommendations, treatment phases, complete treatment plans, etc.

[0100] The processing logic can then compare the generated output with one or more values ​​(e.g., one or more labels included in the training data items). In some embodiments, reinforcement learning is used to train a machine learning model based on a learning strategy for treating orthodontic cases, as described in more detail below. Therefore, the output can be processed according to a reward function or cost function to determine the cost or reward associated with that output. The determined cost or reward can be used to adjust the weights of one or more nodes in the machine learning model. An error term or amount of change for each node in the artificial neural network can be determined based on the calculated cost or reward value. Based on this error, the artificial neural network adjusts one or more parameters (weights of one or more inputs to the node) for one or more of its nodes. Parameters can be updated in a backpropagation manner, such that the nodes at the highest layer are updated first, followed by the nodes at the next layer, and so on. An artificial neural network contains multiple layers of "neurons," where each layer receives values ​​from neurons in the previous layer as input. The parameters of each neuron include weights associated with the values ​​received from each neuron in the previous layer. Therefore, adjusting the parameters can include adjusting the weights of each input to one or more neurons in one or more layers of the artificial neural network.

[0101] Once the model parameters have been optimized, model validation can be performed to determine if the model has improved and to determine the current accuracy of the deep learning model. After one or more rounds of training, the processing logic can determine whether a stopping criterion has been met. The stopping criterion can be a target accuracy level, the target number of processed images in the training dataset, the target change in parameters on one or more previous data points, a combination of the above criteria, and / or other criteria. In one embodiment, the stopping criterion is met when at least a minimum number of data points have been processed and at least a threshold accuracy has been reached. The threshold accuracy can be, for example, 70%, 80%, or 90% accuracy. In one embodiment, the stopping criterion is met if the accuracy of the machine learning model has stopped improving. If the stopping criterion is not yet met, further training is performed. If the stopping criterion has been met, training can be completed. Once the machine learning model has been trained, a reserved portion of the training dataset can be used to test the model.

[0102] Once one or more trained ML models 238 are generated, these models can be stored in model storage 245 and added to a treatment planning application (e.g., treatment planning application 115). The treatment planning application 115 can then use one or more trained ML models 238, along with additional processing logic, to automate and / or assist physicians in designing treatment plans and / or updating treatment plans that are not progressing according to the plan.

[0103] In one embodiment, the model application workflow 217 includes one or more trained machine learning models serving as a treatment plan generator 276. Alternatively or additionally, the one or more trained machine learning models may serve as a treatment action recommender, a partial treatment plan generator, a treatment plan updater, and / or a treatment plan evaluator. For example, the treatment plan generator 276 may be a deep neural network trained to generate treatment plans and / or information associated with treatment plans (e.g., actions, phases, recommendations, etc. for the treatment plan).

[0104] For model application workflow 217, according to one embodiment, an intraoral scanner generates a sequence of intraoral scans 248 for one or two dental arches. A 3D surface generator 255 can perform registration between these intraoral scans, stitching the intraoral scans together, and generating 3D surfaces or models 260 for the upper and / or lower dental arches from the intraoral scans. The intraoral scans 248, the 3D surfaces 260, and / or other information, such as 2D images 250 (e.g., generated by the intraoral scanner or other device), 3D images, X-ray images, CBCT scans, panoramic X-ray images, patient case details 252 (e.g., indicating the level of malocclusion, indicating one or more missing teeth, indicating occlusal relationships, indicating the type of treatment to be performed (e.g., appliance or bracket / framework), patient age, patient sex, etc.), can constitute clinical data 262.

[0105] Clinical data 262 may also include previously generated intraoral scans and / or 3D surfaces / models, previously generated treatment plans, indications of the current treatment stage associated with the previously generated treatment plans, the saved state of the treatment plans (e.g., as previously output by the treatment plan generator 276), the target final state of the patient's upper and / or lower dental arches (e.g., in the form of one or more 3D models), indications of whether the patient is experiencing pain, and / or other information. Some or all of the clinical data 262 may be input into the treatment plan generator 276, which may include a trained neural network. Based on the input clinical data, the machine learning model may identify, for example, one or more clinical problems (e.g., the category or type of malocclusion), the amount of crowding in one or more areas, the presence of clinical problems in the vertical, sagittal, and / or transverse planes, inappropriate or undesirable occlusal relationships, etc. Based on clinical data 262, the treatment plan generator 276 outputs information associated with the patient's orthodontic treatment, the patient's rehabilitation treatment, and / or orthodontic-rehabilitation treatment, such as one or more treatment plans 278.

[0106] In an embodiment, the generated treatment plan may include a sequence of phases and may include one or more actions in one or more treatment phases (e.g., some of which will be performed by a physician). Examples of actions include slowing down tooth movement (e.g., reducing the amount of tooth movement between phases in a pre-existing treatment plan), adjusting the final position of teeth (e.g., for a pre-existing treatment plan), adding overcompensation to one or more specific tooth movements, applying a more conservative phased pattern (e.g., phased phases that allow space for problematic contacts in a pre-existing treatment plan), adding elastics to an appliance, adding occlusal ramps to an appliance, adding compliance indicators for an appliance, adding attachments to one or more patient teeth, adjusting one or more attachments previously applied to patient teeth (e.g., for a pre-existing treatment plan), adding or removing one or more temporary anchoring devices, reattaching attachments to teeth, performing interproximal enamel removal, tooth extraction, performing palatal expansion, adjusting the position of one or more teeth in the sagittal, vertical, and / or transverse planes, rotating one or more teeth about one or more axes, increasing the spacing between two or more teeth, changing the shape of an appliance, adding implants, adding restorations (e.g., veneers, bridges, crowns, etc.), etc. Additionally or alternatively, the treatment planning generator 276 may output suggestions for one or more phases of treatment planning, one or more recommendations for actions to be performed at the start of treatment or at one or more phases of treatment, etc.

[0107] In some embodiments, the treatment planning generator 276 includes multiple machine learning models trained using reinforcement learning with different cost or reward functions. For example, one model may be trained with a reward function that minimizes the number of stages for orthodontic treatment while still achieving the target dentition; another model may be trained with a reward function that minimizes the amount of time required to perform orthodontic treatment; and yet another model may be trained with a reward function that combines minimizing treatment stages and treatment time. Other models may also be trained using reward functions that reward other objectives. In embodiments, one or more objectives of the treatment plan are provided to the treatment planning generator 276, and the treatment planning generator 276 selects models trained using reward functions associated with the selected one or more objectives. In some embodiments, a single machine learning model is trained such that the objective is provided as input to the model and generates an output optimized for the input objective.

[0108] The physician can implement the output treatment plan or modify the treatment plan and then implement the modified treatment plan. In some embodiments, the treatment plan generator 276 outputs multiple treatment plans, which may have different numbers of stages, different tooth movements within the stages, different actions performed in one or more stages, etc., and the physician can select the treatment plan to use from the multiple treatment plans.

[0109] The following Figures 3A to 12 This document describes an example application for automated orthodontic treatment planning and / or orthodontic-rehabilitation treatment planning generation and recommendation. The reference description includes flowcharts illustrating some examples of the process for performing automated orthodontic treatment planning generation and / or modification. The flowcharts provide examples that can be generated by… Figure 1 Example processes executed by system 100 (e.g., by computing devices 105 and / or 106). Figures 3 to Figure 12 The methods described herein can be executed by processing logic, which may include hardware (e.g., circuits, special-purpose logic, programmable logic, microcode, etc.), software (e.g., instructions running on a processing device to execute hardware emulation), or a combination thereof.

[0110] Figure 3A An application of reinforcement learning, according to embodiments of the present disclosure, for training one or more machine learning models to generate orthodontic treatment plans and / or orthodontic-rehabilitation treatment plans. In some embodiments, reinforcement learning methods can be used for training. Figure 2The treatment planning generator 276, or one or more machine learning models of the treatment planning generator 276, employs a reinforcement learning method that enables the treatment planning generator 276 to select actions and / or phases of orthodontic treatment to maximize the cumulative reward on the treatment planning from the current observable state to the target observable state. Therefore, the training process may involve running one or more agents 302, such that each agent 302 is assigned a specific objective (e.g., assigning specific values ​​to one or more treatment conditions or treatment parameters, such as achieving a specific observable state of the patient's dentition or treatment plan within a target number of phases (e.g., with a minimum required number of phases) or a target amount of treatment time (e.g., with a minimum amount of treatment time)). In embodiments, the objective may include clinical and / or commercial objectives. In embodiments, a set of clinical measurements is used to define the state. In some embodiments, the state is defined as a combination of an initial or current clinical measurement (e.g., an initial 3D model of the dental arch) and the next clinical measurement (e.g., a 3D model of the dental arch after the application of one or more actions). In some embodiments, a state is defined as a combination of an initial or current clinical measurement (e.g., an initial 3D model of the dental arch) and changes in the clinical measurement (e.g., a flow diagram indicating the magnitude and direction of changes in position on the 3D model). In some embodiments, the clinical measurement is embedded in a learned lower-dimensional representation. For such embodiments, a machine learning model can generate a lower-dimensional representation of the current state and / or next state of a patient(s)'s dental arch(s) based on input clinical data. Figure 4 The diagram schematically illustrates the operation of an agent 302 implemented according to various aspects of this disclosure, which can cause one or more trained machine learning (ML) models 310 (e.g., neural networks) 310 to iteratively generate phases of orthodontic treatment and / or actions to be performed during those phases. At each iteration, the agent 302 can feed a vector identifying the values ​​of an observable state 312 to one or more trained ML models 310. The observable state 312 can be represented, for example, by a predicted state of the patient's dentition. Each action can correspond to a clinical approach to treatment and can include adjusting one or more teeth, adding attachments, performing an IPR, widening the dental arch, tilting or tilting incisors or other teeth, and / or any other actions described herein. In embodiments, actions can include orthodontic actions, tooth restoration actions, or a combination of orthodontic actions and tooth restoration actions.

[0111] Upon receiving a representation of observable state 312, one or more trained ML models 310 process the observable state to generate a set of possible actions 315A-315N and their corresponding scores, such that the score associated with a particular action 315 indicates the likelihood that the action will trigger a transition in the observable state that falls within the shortest path from the current observable state to the desired observable state (i.e., the shortest number of orthodontic treatment stages required to progress from the patient's current dentition to the patient's target dentition). In some embodiments, only orthodontic actions are output. In some embodiments, only actions to restore teeth are output. In some embodiments, both orthodontic actions and actions to restore teeth are output.

[0112] Agent 302 selects a random action or an action 315 associated with the highest score from candidate actions generated by the neural network with a known probability ε. This action may be the generation of a new treatment phase and / or may include one or more actions performed by a physician in association with a treatment phase, such as IPR, tooth extraction, or any other action discussed herein. The probability ε may be chosen as a monotonically decreasing function of the number of training iterations, such that the probability ε is close to one at the initial iterations (thus forcing the agent to favor random user interface actions relative to actions generated by an untrained agent), and subsequently decreases with iterations to asymptotically approach a predetermined low value, thus giving the neural network output more bias as training progresses.

[0113] Agent 302 transmits the selected action 315Q to environment 320. Environment 320 may represent the current state of the patient's dental arch. Environment 320 applies action 315Q to the current state of the patient's dental arch and returns a new observable state 312 and an optional reward 322 to agent 302. In embodiments, the reinforcement learning techniques applied include Q-learning, deep Q-learning networks, dual deep Q-learning networks, or other reinforcement learning techniques.

[0114] It can continue to iterate until the target observable state is reached (e.g., the final target state of the patient's dentition) or until an error condition is detected (e.g., exceeding the predetermined number threshold of orthodontic treatment stages or the neural network not returning a valid action for the current observable state).

[0115] During reinforcement learning training of one or more ML models 310, actions can be taken from historical clinical data. Favorable actions (e.g., shorter than expected treatment time) can be rewarded, while unfavorable actions (e.g., longer treatment time) can be penalized. The model parameters are updated to make favorable actions more likely and unfavorable actions less likely for a given state.

[0116] Upon completion of training, the processing logic can validate the trained model by running it multiple times with added noise, thereby forcing agent 302 to select a random action or the action associated with the highest score from among the candidate actions generated by the one or more ML models 310 with a known small probability γ. The validated model can be stored... Figure 2 The model is stored in 245. Once the system is trained, during inference, the selected action 315A-N can be the action with the highest output probability.

[0117] Figure 3B An example machine learning model architecture 323 according to an embodiment of the present disclosure is shown. In one embodiment, the machine learning model architecture 323 corresponds to Figure 3A One or more machine learning models 310. For example... Figure 3A As shown, agent 302 can receive information about the current state 312 from environment 320, which can be input into one or more ML models 310. In embodiments, the details of the machine learning architecture can vary considerably depending on the input type. Inputs may include patient case details 340 (e.g., age, gender, potential health problems, etc.), 2D data 324 (e.g., teeth, 2D images of a patient's smile, etc.), and / or 3D data 334 (e.g., 3D intraoral scans, 3D models generated from intraoral scans, etc.). One method of combining these different data types includes a dedicated network (e.g., an autoencoder) that transforms 2D data 324 and / or 3D data 334 into lower-dimensional embeddings (e.g., 2D embeddings 328 and / or 3D embeddings 338). As shown, in an embodiment, 2D autoencoder 326 can transform 2D data 324 into 2D embeddings 328, while 3D autoencoder 338 can transform 3D data 334 into 3D embeddings 338. Embeddings can be lower-dimensional representations of the input data. In some embodiments, the 2D embedding and the 3D embedding have the same dimensions and / or data format. In one embodiment, one or more convolutional layers and / or transformer layers of a neural network (e.g., a 2D autoencoder 326) are used to generate the 2D embedding 328. In one embodiment, one or more layers of a graph neural network or a 3D convolutional network are used to generate the 3D embedding 338 from the 3D data 334.

[0118] In some embodiments, 2D embeddings 328 and / or 3D embeddings 338 are combined with patient case details 340 to generate combined clinical data 342. The combined clinical data 342 can be fed into an output algorithm 346. The output algorithm can be another machine learning model (or an additional layer of machine learning models), which can form a neural network or another architecture such as a decision tree.

[0119] Output algorithm 346 (e.g., a trained ML model) may output one or more categories or types of actions, such as orthodontic action 348 and / or recovery action 350. In an embodiment, output algorithm 346 may determine whether orthodontic action 348 and / or recovery action 350 is appropriate. Output algorithm 346 may then determine which action is appropriate among many possibilities of the determined type(s) of actions(s). For example, output algorithm 346 may determine which of orthodontic action 1 352A, orthodontic action 2 352B through orthodontic action N 352N is the best orthodontic action to be performed. Similarly, output algorithm 346 may determine which of recovery action 1 354A, recovery action 2 354B through recovery action N 354N is the best recovery action to be performed. In an embodiment, the output action may include probability values ​​and / or confidence values.

[0120] During reinforcement learning, dedicated 2D and / or 3D networks (e.g., 2D autoencoders 326 and / or 3D autoencoders 336) can be trained directly using other parameters. Alternatively, the 2D and / or 3D networks can be trained separately from reinforcement learning. Separate training can be performed on different, related tasks, such as numbering teeth from 2D images or 3D scans. Separate training can also come from autoencoders or similar unsupervised learning frameworks that learn how to compress data with high fidelity.

[0121] Figure 3C An example autoencoder architecture 380 according to an embodiment of this disclosure is illustrated. An autoencoder is a type of neural network designed to learn efficient encodings (embeddings) of input data. As shown, the autoencoder architecture 380 may include an encoder 382 and a decoder 384. Each of the encoder 382 and decoder 384 may include one or more layers. The encoder 382 is part of a network that transforms the input data into a compressed representation (referred to as a latent space or embedding 388). The encoder 382 reduces the dimensionality of the data while preserving its essential features. The decoder 384 generates an output 390 from the embedding 388, which is a reconstruction of the original input 386. The goal is to make the output 390 as close as possible to the input 386. In some embodiments, the decoder 384 may be discarded once the autoencoder has been fully trained. For example, in embodiments, the decoder may be omitted from 2D autoencoder 326 and / or 3D autoencoder 336.

[0122] Figure 4 An example method for generating orthodontic treatment plans or orthodontic-rehabilitation treatment plans using a trainable agent, implemented according to one or more aspects of this disclosure, is described. As described above, a trainable agent can be employed to generate treatment plans, generate recommendations for actions to be performed by a physician for orthodontic treatment planning, etc. In some embodiments, the agent corresponds to... Figure 2 Treatment planning generator 276. In the illustrative example, the generation of the optimal treatment plan may involve employing multiple trainable agents to realize various target observable states (e.g., target tooth alignment).

[0123] In an embodiment, method 400 can be provided by Figure 2 The treatment planning generator 276 is used to implement this. Each of method 400 and / or its individual functions, routines, subroutines, or operations can be implemented by a computing device (e.g., Figure 13 The method 400 may be executed by one or more processors of a computing device 1300. In some embodiments, method 400 may be executed by a single processing thread. Alternatively, method 400 may be executed by two or more processing threads, each thread performing the operation of one or more individual functions, routines, subroutines, or methods. In the illustrative example, the processing threads implementing method 400 may be synchronized (e.g., using semaphores, critical sections, and / or other thread synchronization mechanisms). Alternatively, the processing threads implementing method 400 may execute asynchronously relative to each other. Therefore, although Figure 4 The associated description lists the operations of method 400 in a specific order, but various implementations of the method may perform at least some of the described operations in parallel and / or in any chosen order.

[0124] like Figure 4 Schematably illustrated, at box 410, the processing logic identifies the current observable state of the patient's dental arch. The current observable state can be identified based on the processing of clinical data on the patient's dental arch, such as image data, scan data, 3D models, etc. In some embodiments, the current observable state of the patient's dental arch can be represented by a vector of values ​​characterizing one or more parameters of the patient's dental arch.

[0125] The method terminates in response to a match between the current observable state and the target observable state determined at box 420; otherwise, processing continues at box 430.

[0126] At box 430, the computing device feeds the current observable state (e.g., a numerical vector representing the current observable state) into a neural network or other trained ML model, which can generate multiple actions available in the current observable state (e.g., multiple different options for the next stage of orthodontic treatment and / or actions to be performed by the physician to achieve the next stage of orthodontic treatment) and their corresponding action scores. The action scores can be represented by positive integers or real values.

[0127] At box 440, the processing logic selects the next stage of orthodontic treatment or orthodontic-rehabilitation treatment and / or one or more associated actions from multiple possible next treatment stages based on the action score. In an illustrative example, the computing device selects the next stage of orthodontic treatment associated with the best (e.g., maximum or minimum) score among the scores associated with the next stage of orthodontic treatment generated by the neural network. In another illustrative example, for example, to train the neural network, the processing logic selects a random action or the action associated with the highest score from the user interface actions generated by the neural network with a known probability ε, as described in more detail above.

[0128] At box 450, the processing logic selects the next stage of the chosen orthodontic treatment and / or applies the selected action to the current observable state of the patient's dentition, as described in more detail above. At box 460, the next state of the patient's dentition (e.g., the upper and / or lower dental arches) associated with the determined next treatment stage is generated and made the new observable state of the patient's dentition.

[0129] The operations of boxes 410 to 460 are iteratively repeated until the target observable state of the patient's dentition is achieved. Therefore, in response to the completion of the operation of box 460, the method loops back to box 410. In some embodiments, in response to the failure to achieve the desired observable state of the patient's dentition within a predetermined number of iterations, the processing logic may initiate retraining of the neural network to modify one or more parameters of the neural network, as described in more detail above.

[0130] Figure 5 A flowchart is shown of a method 500 for training a machine learning model to generate orthodontic treatment plans or orthodontic-rehabilitation treatment plans, according to an embodiment. In some embodiments, the machine learning model corresponds to or includes... Figure 2In the treatment planning generator 276, at box 505 of method 500, the processing logic receives training data items including clinical data on the state of the dental arch and information regarding the execution of orthodontic treatment for the dental arch. At box 510, the processing logic determines one or more orthodontic treatment plans for the dental arch treatment. This may include determining one or more phases of orthodontic treatment at box 515. Determining the phases of orthodontic treatment may include determining one or more actions to be performed on the patient's dental arch at box 520, such as moving one or more teeth by a determined amount in one or more directions, rotating one or more teeth by a determined amount about one or more axes, performing an IPR, performing palatal expansion, performing tooth extraction, adding attachments to one or more teeth, removing attachments from one or more teeth, etc. Actions may include actions to be performed by the dentist (e.g., such as an IPR or tooth extraction) and actions that do not need to be performed by the dentist (e.g., such as designing an appliance to apply forces to move and / or rotate one or more teeth). Determining the phases of orthodontic treatment may additionally include determining a predicted target state of the upper and / or lower dental arches based on the results of the actions. Treatment plans may include a sequence of treatment phases, each of which may include associated actions and associated predicted states resulting from treatment in that phase.

[0131] At box 530, the processing logic determines a cost or reward value associated with one or more phases of the determined orthodontic treatment. The cost or reward value can be calculated by applying the determined actions, treatment phases, and / or states associated with those phases to a cost or reward function. The cost / reward function can be adjusted to reward and / or penalize specific outcomes. For example, the cost / reward function can be configured to reward treatment planning for a target dentition with fewer treatment phases and penalize treatment planning for a target dentition with a larger number of treatment phases. At box 535, the processing logic updates one or more nodes of the machine learning model based on the cost or reward value (e.g., based on backpropagation).

[0132] At box 540, the processing logic determines whether training is complete. If training is complete, at box 545, the trained machine learning model can be deployed to the treatment planning application. If training is not complete, the method can return to box 505 and can receive new training data items for further training of the machine learning model.

[0133] Figure 6A flowchart of a method 600 for determining a cost or reward value applied to one or more treatment phases in a generated treatment plan, according to an embodiment, is shown. In some embodiments, method 600 is performed at block 530 of method 500. At block 602, processing logic may receive inputs for a target maximum number of phases for orthodontic treatment and / or a target maximum amount of time to complete orthodontic treatment. Additionally or alternatively, processing logic may receive objectives that minimize the number of phases and / or minimize the amount of time to perform treatment. Alternatively, processing logic may determine one or more objectives and / or may apply a cost or reward function that has been optimized for one or more predetermined objectives.

[0134] At box 605, the processing logic determines the number of stages in the generated orthodontic treatment plan. At box 610, the processing logic determines the predicted time amount for completing the orthodontic treatment plan. At box 615, the processing logic determines a first amount of variation between the number of stages and the target maximum number of stages. At box 620, the processing logic determines a second amount of variation between the predicted time amount and the target maximum time amount. At box 625, the processing logic determines a cost value or a reward value based on the first and / or second amount of variation. In this embodiment, a larger amount of variation results in a lower reward and / or a higher cost.

[0135] Figure 7 A flowchart of a method 700 for generating an orthodontic treatment plan or orthodontic-rehabilitation treatment plan using a trained machine learning model according to an embodiment is shown. At block 705 of method 700, processing logic receives clinical data of a first state of the patient's dental arch and / or a target final state of the patient's dental arch. At block 710, the processing logic determines one or more orthodontic treatment plans and / or orthodontic-rehabilitation treatment plans for treating the patient's dental arch (e.g., for correcting one or more malocclusions in the patient's dental arch) based on the processed clinical data. In one embodiment, this includes determining one or more phases of orthodontic treatment and / or orthodontic-rehabilitation treatment at block 715 using a trained machine learning model (e.g., which may have been trained using reinforcement learning). Generating treatment phases may include determining one or more actions to be performed on the patient's dental arch at block 720 and determining a predicted state of the patient's dental arch after applying one or more actions. In an embodiment, the predicted state of the patient's dentition (e.g., the upper and / or lower dental arches) may include a generated 3D model of the patient's upper and / or lower dental arches. In some embodiments, the 3D model of the dental arch can be automatically generated by a machine learning model. Actions may include movement, translation, rotation, etc., of one or more teeth; extraction of one or more teeth; interdental retraction (IPR), etc. In some embodiments, the processing logic iteratively determines the sequence of stages of orthodontic treatment and / orthodontic-restoration treatment until the target state of the patient's dental arch is achieved.

[0136] At box 730, the processing logic determines a score for each of one or more defined orthodontic treatment plans. At box 735, the processing logic can output one or more generated treatment plans and their associated scores to a display.

[0137] In one embodiment, at box 740, the processing logic determines whether multiple treatment plans have been generated. If multiple treatment plans have been generated, at box 745, the user may be prompted to select one of the orthodontic treatment plans. Multiple treatment plans may include one or more orthodontic-rejuvenation treatment plans, which include rejuvenation options such as allowing the use of different types and thicknesses of veneer sizes, allowing minimal or reasonable removal of enamel and tooth structure. Some treatment plans allow for greater tooth movement, thus reducing the need for further tooth removal. Some treatment plans allow for additional steps in treatment, considering that the target tooth movement has not been completed as planned in previous phases of treatment. At this point in the treatment, more than one option may exist. One possibility is to update the treatment plan. Another possibility is to change the dental instrument (e.g., orthodontic appliance) to achieve a different movement. Another example of an orthodontic-rejuvenation treatment plan is the need for space to place a dental implant, where this space is compressed by the roots of teeth adjacent to the space where the implant will be placed. In this case, it is necessary to open up the roots of the adjacent teeth to allow for implant placement. In another example, an orthodontic-restoration treatment plan might require the insertion of a specific type of implant in the middle of treatment (e.g., after teeth have been moved to create space for implant placement). However, during the implant placement phase, insufficient space may not have been created for the implant. In response, the processing logic could either add an additional treatment phase to allow more time to create space for implant insertion, or update the treatment plan to use a different type of implant that requires less space. In some embodiments, different orthodontic-restoration treatment plans can be generated, one for each possible option.

[0138] At box 750, the processing logic may subsequently receive a selection from one of the automatically generated treatment plans. Alternatively, the processing logic may automatically select the orthodontic treatment plan or orthodontic-rehabilitation treatment plan associated with the highest score (e.g., which may be associated with the highest reward or the lowest cost). At box 755, the selected orthodontic treatment plan may be implemented.

[0139] Method 700 can be implemented before or at any intermediate stage during orthodontic treatment. Patient teeth may not respond to treatment as predicted. For example, teeth may move faster or slower than expected. If method 700 is implemented during an intermediate stage of treatment, one or more previous states of the patient's dentition can be fed into the machine learning model along with the current state of the model. In some embodiments, previously generated treatment plans are stored, and / or the current implementation state of previously generated treatment plans is stored. In embodiments, the current implementation state of previously generated treatment plans and / or approved treatment plans can be fed into the machine learning model. The machine learning model can output an updated treatment plan that can replace one or more remaining stages of the previously generated treatment plan.

[0140] Figure 8 A flowchart of method 800 for selecting a trained machine learning model to generate an orthodontic treatment plan, according to an embodiment, is shown. At block 805 of method 800, processing logic receives a selection of one or more objectives for orthodontic treatment, such as minimizing the number of stages and / or minimizing the amount of treatment time. At block 810, the processing logic can select a trained machine learning model associated with the selected one or more objectives. This machine learning model is one that can be applied at method 700 in an embodiment.

[0141] Figure 9 A flowchart of method 900 according to an embodiment is shown. Method 900 determines information that can be used to assist in determining one or more stages of orthodontic treatment planning. In one embodiment, method 900 is performed using a trained machine learning model, which is trained to receive input clinical data (e.g., intraoral scans, 3D models, 2D images, etc.) and output one or more classifications regarding the patient's dentition. At box 905, clinical data is input into the trained machine learning model, and the machine learning model outputs the occlusal type of the patient's dental arch. Clinical data may include the current state of the patient's dental arch (e.g., a currently existing 3D model of the patient's dental arch) and / or the predicted future state of the patient's dental arch associated with a treatment stage (e.g., a synthetically generated 3D model of the upper and lower dental arches). Occlusal types may include, for example, open bite, cross bite, overbite, underbite, deep bite, etc.

[0142] At box 910, clinical data is fed into a trained machine learning model, which outputs one or more classifications of malocclusion, such as Class I, Class II, Class III, etc. Additionally or alternatively, the trained machine learning model may output the current occlusal relationship between the upper and lower dental arches. Clinical data may include the current state of the patient's dental arches (e.g., a currently existing 3D model of the patient's dental arches, including the current occlusal relationship between the upper and lower dental arches), and / or a predicted future state of the patient's dental arches associated with a treatment phase (e.g., a synthetically generated 3D model of the upper and lower dental arches, and optionally a defined occlusal relationship between the upper and lower dental arches). In some embodiments, the operations at boxes 905 and 910 are performed together.

[0143] In an embodiment, method 900 may be performed, for example, at block 710 of method 700.

[0144] Figure 10AA tooth repositioning system 1010 comprising multiple instruments 1012, 1014, and 1016 is illustrated. Instruments 1012, 1014, and 1016 can be designed based on the generation of a sequence of 3D models of a dental arch, which can be generated according to the techniques discussed above. Any instrument described herein can be designed and / or provided as part of a collection of multiple instruments for use in a tooth repositioning system, and can be designed based on an orthodontic treatment plan generated according to embodiments of this disclosure. Each instrument can be configured such that the tooth-accommodating cavity has a geometry corresponding to the intermediate or final tooth arrangement intended by the instrument. By placing a series of incremental position-adjusting instruments on the patient's teeth, the patient's teeth can be progressively repositioned from an initial tooth arrangement to a target tooth arrangement. For example, the tooth repositioning system 1010 may include a first instrument 1012 corresponding to the initial tooth arrangement, one or more intermediate instruments 1014 corresponding to one or more intermediate arrangements, and a final instrument 1016 corresponding to the target arrangement. The target tooth arrangement can be the final planned tooth arrangement selected for the patient's teeth at the end of all planned orthodontic treatment, as an optional output using a trained machine learning model. Alternatively, the target arrangement can be one of several intermediate arrangements used for the patient's teeth during orthodontic treatment, which can include a variety of different treatment scenarios, including but not limited to cases where surgery is recommended, cases suitable for interproximal enamel reduction (IPR), cases for scheduling progress checks, cases optimal for anchor placement, cases where palatal expansion is desired, cases involving restorative dentistry (e.g., inlays, high onlays, crowns, bridges, implants, veneers, etc.). Thus, it should be understood that the target tooth arrangement can be any planned arrangement of the patient's teeth after one or more incremental repositioning phases. Similarly, the initial tooth arrangement can be any initial arrangement of the patient's teeth followed by one or more incremental repositioning phases.

[0145] In some embodiments, devices 1012, 1014, 1016 (or portions thereof) may be manufactured using indirect manufacturing techniques, such as by thermoforming on a male or female mold. Indirect manufacturing of orthodontic devices may involve producing a male or female mold for a patient’s teeth to be arranged in a targeted manner (e.g., by rapid prototyping, milling, etc.), and thermoforming one or more sheets of material on a mold to generate an device housing.

[0146] In the example of indirect manufacturing, the mold for the patient's dental arch can be manufactured from a digital model of the dental arch generated by a machine learning model trained as described above, and the shell can be formed on the mold (e.g., by thermoforming a polymer sheet on the mold of the dental arch, followed by trimming the thermoformed polymer sheet). Mold manufacturing can be performed using a rapid prototyping machine (e.g., a stereolithography (SLA) 3D printer). The digital models of instruments 1012, 1014, and 1016 have been computed by a computing device (e.g., Figure 13 After processing by the processing logic of the computing device (in the machine), the rapid prototyping machine can receive digital models of dental arch molds and / or digital models of instruments 1012, 1014, 1016. The processing logic may include hardware (e.g., circuits, dedicated logic, programmable logic, microcode, etc.), software (e.g., instructions executed by the processing device), firmware, or a combination thereof. For example, one or more operations can be performed by an execution device... Figure 2 The treatment plan generator 276 is used by the processing device to execute the treatment plan.

[0147] To create a mold, the shape of the patient's dental arch at the stage of treatment is determined based on the treatment plan. In orthodontic examples, the treatment plan can be generated based on an intraoral scan of the dental arch to be modeled. An intraoral scan of the patient's dental arch can be performed to generate a three-dimensional (3D) virtual model of the patient's dental arch (mold). For example, a full scan of the patient's mandibular and / or maxillary arch can be performed to generate its 3D virtual model. Intraoral scanning can be performed by creating multiple overlapping intraoral images from different scanning stations and then stitching the intraoral images or scans together to provide a synthetic 3D virtual model. In other applications, a virtual 3D model can also be generated based on a scan of the object to be modeled or based on the use of computer-aided drawing techniques (e.g., designing a virtual 3D mold). Alternatively, an initial negative mold can be generated from the actual object to be modeled (e.g., a dental impression, etc.). The negative mold can then be scanned to determine the shape of the positive mold to be produced.

[0148] Once a virtual 3D model of the patient's dental arch is generated, the dentist can determine the desired treatment outcome, including the final position and orientation of the patient's teeth. In one embodiment, the treatment planning generator 276 outputs the desired treatment outcome based on processing the virtual 3D model of the patient's dental arch (or other dental arch data associated with the virtual 3D model). The processing logic can then determine the number of treatment stages to allow the teeth to progress from their initial position and orientation to their target final position and orientation. The shapes of the final virtual 3D model and each intermediate virtual 3D model can be determined by calculating the progression of tooth movement throughout the entire orthodontic treatment, from the initial tooth placement and orientation to the final corrected tooth placement and orientation. For each treatment stage, a separate virtual 3D model of the patient's dental arch for that treatment stage can be generated. In one embodiment, for each treatment stage, the treatment planning generator 276 outputs a different 3D model of the dental arch. The shape of each virtual 3D model will be different. The original virtual 3D model, the final virtual 3D model, and each intermediate virtual 3D model are unique and customized for the patient.

[0149] Therefore, multiple different virtual 3D models (digital designs) of the dental arch can be generated for a single patient. The first virtual 3D model can be a unique model of the patient's currently existing dental arch and / or teeth, and the final virtual 3D model can be a model of the patient's dental arch and / or teeth after correction of one or more teeth and / or jaws. Multiple intermediate virtual 3D models can be modeled, each of which can differ from the previous virtual 3D model in an incremental manner.

[0150] Each virtual 3D model of a patient's dental arch can be used to generate a unique, customized physical mold of the arch for a specific treatment phase. The shape of the mold can be at least partially based on the shape of the virtual 3D model used for that treatment phase. The virtual 3D model can be represented in a file such as a computer-aided design (CAD) file or a 3D printable file such as a stereolithography (STL) file. The virtual 3D model of the mold can be sent to a third party (e.g., a clinician's office, laboratory, manufacturing facility, or other entity). The virtual 3D model may include instructions that will control the manufacturing system or apparatus to produce a mold with a specific geometry.

[0151] Clinicians' offices, laboratories, manufacturing facilities, or other entities can receive virtual 3D models of the mold, digital models already created as described above. Entities can then input the digital model into a 3D printer. 3D printing encompasses any layer-based additive manufacturing process. 3D printing can be achieved using additive processes, where continuous layers of material are formed into a prescribed shape. 3D printing can be performed using extrusion deposition, particulate material bonding, lamination, photopolymerization, continuous liquid interface production (CLIP), or other techniques. 3D printing can also be achieved using subtractive processes such as milling.

[0152] In some cases, stereolithography (SLA), also known as optical fabrication solid-state imaging, is used to create SLA molds. In SLA, a mold is created by sequentially printing thin layers of a photocurable material (e.g., a polymeric resin) on top of each other. A platform rests in a bath of liquid photopolymer or resin, just below the surface of the bath. A light source (e.g., an ultraviolet laser) tracks the pattern on the platform to cure the photopolymer where the light source is directed, thus forming the first layer of the mold. The platform is incrementally lowered, and the light source tracks the new pattern on the platform to form another layer of the mold at each increment. This process is repeated until the mold is completely fabricated. Once all the layers of the mold have been formed, the mold can be cleaned and cured.

[0153] The following materials can be used to directly form molds: such as polyesters, copolyesters, polycarbonates, thermopolymerized polyurethanes, polypropylene, polyethylene, polypropylene and polyethylene copolymers, acrylics, cyclic block copolymers, polyetheretherketones, polyamides, polyethylene terephthalate, polybutylene terephthalate, polyetherimide, polyethersulfone, polypropylene terephthalate, styrene-based block copolymers (SBCs), silicone rubber, elastomer alloys, thermopolymerized elastomers (TPEs), thermopolymerized vulcanized rubber (TPV) elastomers, polyurethane elastomers, block copolymer elastomers, polyolefin blend elastomers, thermopolymerized copolyester elastomers, thermopolymerized polyamide elastomers, or combinations thereof. Materials used to manufacture molds can be provided in uncured form (e.g., as liquids, resins, powders, etc.) and can be cured (e.g., by photopolymerization, light curing, gas curing, laser curing, crosslinking, etc.). The properties of the material before curing may differ from those after curing.

[0154] The instruments can be formed from each mold, and when applied to a patient's teeth, the instruments can provide force to move the patient's teeth as instructed in the treatment plan. The shape of each instrument is unique and customized for a specific patient and a specific treatment stage. In the example, instruments 1012, 1014, and 1016 can be pressure-formed or thermoformed on a mold. Each mold can be used to manufacture the instrument that applies force to the patient's teeth at a specific stage of orthodontic treatment. Instruments 1012, 1014, and 1016 each have a tooth-receiving cavity that receives and resiliently repositions the teeth according to the specific treatment stage.

[0155] In one embodiment, a sheet of material is pressure-formed or thermoformed on a mold. This sheet can be, for example, a polymer sheet (e.g., a sheet of an elastic thermopolymer, a sheet of a polymer material, etc.). To thermoform the shell on the mold, the sheet of material can be heated to a temperature at which the sheet becomes pliable. Pressure can be applied simultaneously to the sheet to form the now pliable sheet around the mold. Once the sheet cools, it will have a shape consistent with the mold. In one embodiment, a release agent (e.g., a non-sticky material) is applied to the mold before forming the shell. This facilitates the subsequent removal of the mold from the shell. Force can be applied to lift the device from the mold. In some cases, the removal force may cause breakage, warping, or deformation. Therefore, the embodiments disclosed herein can identify one or more potential points of failure in the digital design of the device prior to manufacturing and can perform corrective actions.

[0156] Additional information can be added to the device. This additional information can be any information about the device. Examples of such additional information include part number identifiers, patient names, patient identifiers, case numbers, sequence identifiers (e.g., indicating a specific liner of the device in a treatment sequence), manufacturing dates, clinician names, tissue identifiers, etc. For example, after identifying potential damage points in the device's digital design, an indicator can be inserted into the device's digital design. In some embodiments, the indicator may indicate a recommended location to begin removing the polymer device to prevent damage points from occurring during removal. In embodiments, additional information can be automatically added to the generated 3D model by the treatment planning generator 276 during the generation of the 3D model.

[0157] After the instrument for the treatment phase is formed on the mold, it is removed from the mold (e.g., automatically removed from the mold), and subsequently trimmed along the cutting line (also called the trimming line). Processing logic can determine the cutting line of the instrument. In one embodiment, treatment planning generator 276 outputs the cutting line of the instrument associated with a 3D model output by dental arch generator 268. One or more cutting lines can be determined based on a virtual 3D model of the dental arch at a particular treatment phase, a virtual 3D model of the instrument to be formed on the dental arch, or a combination of a virtual 3D model of the dental arch and a virtual 3D model of the instrument. The location and shape of the cutting line can be important for the functionality of the instrument (e.g., the ability of the instrument to exert desired forces on the patient's teeth) and for the fit and comfort of the instrument. For housings such as orthodontic instruments, orthodontic retainers, and orthodontic splints, the trimming of the housing can play a role in the effectiveness of the housing for its intended purpose (e.g., aligning, retaining, or positioning one or more of the patient's teeth) and in the fit of the housing on the patient's dental arch. For example, if the shell is over-trimmed, it may lose rigidity, and its ability to apply force to the patient's teeth may be impaired. When the shell is over-trimmed, it may become weak at that location and could become a point of damage when the patient removes the shell from their teeth or when the shell is removed from the mold. In some embodiments, when a potential point of damage is identified in the digital design of the device, the cutting line may be modified in the digital design of the device as one of the corrective actions to be taken.

[0158] On the other hand, if the outer shell is under-trimmed, certain portions of the shell may affect the patient's gums and cause discomfort, swelling, and / or other dental problems. Furthermore, if the shell is under-trimmed at a particular location, the shell may be too rigid at that location. In some embodiments, the cutting line may be a straight line passing through the instrument at, below, or above the gingival line. In some embodiments, the cutting line may be a gingival cutting line, which represents the interface between the instrument and the patient's gums. In such embodiments, the cutting line controls the distance between the edge of the instrument and the patient's gingival line or gingival surface.

[0159] Each patient has a unique dental arch, and each dental arch has a unique gingiva. Therefore, the shape and position of the incision line can be unique and can be customized for each patient and each stage of treatment. For example, the incision line is customized to run along the gingival line (also known as the gum line). In some embodiments, the incision line may be away from the gingival line in some areas and on the gingival line in other areas. For example, in some cases, it may be desirable for the incision line to be away from the gingival line (e.g., not touching the gum), where the shell will touch the teeth and be on the gingival line in the interproximal region between teeth (e.g., touching the gum). Therefore, it is important to trim the shell along the predetermined incision line.

[0160] Figure 10B A method 1050 for orthodontic treatment using multiple instruments according to an embodiment is illustrated. Method 1050 can be practiced using any of the instruments or groups of instruments described herein. In block 1060, a first orthodontic instrument is applied to the patient's teeth to reposition the teeth from a first dental arrangement to a second dental arrangement. In block 1070, a second orthodontic instrument is applied to the patient's teeth to reposition the teeth from the second dental arrangement to a third dental arrangement. Method 1050 can be repeated as needed, using any appropriate number of instruments in sequence and combinations thereof, to incrementally reposition the patient's teeth from an initial arrangement to a target arrangement. Instruments can be manufactured all at the same stage or in groups or batches (e.g., at the start of a treatment stage), or instruments can be manufactured one at a time, and the patient can wear each instrument until they can no longer feel pressure from each instrument on their teeth or until the maximum amount of tooth movement expected for that given stage has been achieved. Multiple different instruments (e.g., a set of instruments) can be designed or even manufactured before the patient wears any of the multiple instruments. After an appropriate period of wearing the instruments, the patient can replace the current instrument with the next instrument in the series until no instruments remain. Instruments are typically not fixed to the teeth, and patients can place and replace instruments at any time during the procedure (e.g., patient-removable instruments). The final instrument or several instruments in a series may have one or more geometries selected for overcorrecting tooth arrangement. For example, one or more instruments may have geometries that, if treatment is fully achieved, move a single tooth beyond a geometry that has been selected as the “final” tooth arrangement. This overcorrection may be desirable to counteract potential regression after the repositioning method has been terminated (e.g., allowing movement of individual teeth toward their pre-corrected position). Overcorrection can also be beneficial to accelerate the correction rate (e.g., an instrument with a geometry that positions a single tooth toward a desired intermediate or final position can move it toward that position at a greater rate). In this case, instrument use may be terminated before the tooth reaches the position defined by the instrument. Furthermore, overcorrection may be intentionally applied to compensate for any inaccuracies or limitations of the instrument.

[0161] Figure 11 A method 1100 for designing an orthodontic device to be manufactured directly or indirectly, according to an embodiment, is illustrated. Method 1100 can be applied to any embodiment of the orthodontic device described herein, and in embodiments, it can be performed using one or more trained machine learning models. Some or all blocks of method 1100 can be performed by any suitable data processing system or apparatus, such as one or more processors configured with suitable instructions.

[0162] At box 1105, the target arrangement of one or more teeth of the patient can be determined. The target arrangement of teeth can be received from a clinician in the form of a prescription (e.g., expectations and anticipated end results of orthodontic treatment), calculated from basic orthodontic principles, extrapolated from a clinical prescription, and / or determined by methods such as... Figure 2 The treatment planning generator 276 is generated using a trained machine learning model. Utilizing the desired final position of each specified tooth and its own digital representation, the final position and surface geometry of each tooth can be specified to form a complete model of the tooth arrangement at the end of the desired treatment.

[0163] In box 1110, a motion path is determined to move one or more teeth from an initial arrangement to a target arrangement. The initial arrangement can be determined from a model or scan of the patient's teeth or oral tissues, such as using wax bite, direct contact scanning, X-ray imaging, tomography, ultrasound imaging, and other techniques for obtaining information about the position and structure of teeth, jaws, gingiva, and other orthodontic-related tissues. From the obtained data, a digital dataset, such as a 3D model of one or more dental arches of the patient, representing the initial (e.g., pre-treatment) arrangement of the patient's teeth and other tissues, can be derived. Optionally, the initial digital dataset is processed to segment the tissue components relative to each other. For example, a data structure digitally representing the crown of a single tooth can be generated. Advantageously, a digital model of the entire tooth can be generated, optionally including measured or extrapolated hidden surfaces and root structures, as well as surrounding bone and soft tissue.

[0164] Given the initial and target positions of each tooth, a motion path can be defined for the movement of each tooth. Determining the motion path for one or more teeth may include identifying multiple incremental arrangements of the one or more teeth to achieve the motion path. In some embodiments, the motion path implements one or more force systems on one or more teeth (e.g., as described below). In some embodiments, the motion path is determined by a trained machine learning model, such as treatment planning generator 276. In some embodiments, the motion path is configured to move the teeth in the fastest way possible, with minimal round trips, to move the teeth from their initial positions to their desired target positions. The tooth path may optionally be segmented, and the segmentation may be computed such that the movement of each tooth within a segment remains within threshold limits of linearity and rotational translation. Thus, the endpoints of each path segment can constitute a clinically feasible repositioning, and the set of segment endpoints can constitute a clinically feasible sequence of tooth positions such that moving from one point in the sequence to the next does not cause tooth collisions.

[0165] In some embodiments, a force system for generating movement of the one or more teeth along a movement path is determined. In one embodiment, the force system is determined by a trained machine learning model, such as a data arch generator 268. The force system may include one or more forces and / or one or more torques. Different force systems can cause different types of tooth movement, such as flipping, translation, rotation, extraction, intrusion, root movement, etc. Biomechanical principles, modeling techniques, force calculation / measurement techniques, etc., including knowledge and methods commonly used in orthodontics, can be used to determine the appropriate force system to be applied to the teeth to complete the tooth movement. In determining the force system to be applied, sources may be considered including literature, force systems determined through experimental or virtual modeling, computer-based modeling, clinical experience, minimization of undesirable forces, etc.

[0166] Determining the force system can include constraints on permissible forces, such as the allowed direction and magnitude, and the desired movement resulting from the applied forces. For example, different patients may require different movement strategies when fabricating a palatal expander. For instance, the amount of force required to separate the palate can depend on the patient's age, as very young patients may not have fully formed sutures. Therefore, palatal expansion can be achieved with lower forces in young patients and others whose palatal sutures are not fully closed. Slower palatal movements can also aid bone growth to fill the expanded sutures. For other patients, faster expansion may be required, which can be achieved by applying greater forces. These requirements can be combined as needed to select the instrument's structure and materials; for example, by selecting a palatal expander capable of applying greater forces to break palatal sutures and / or cause rapid palatal expansion. Subsequent instrumentation stages can be designed to apply varying amounts of force, such as initially applying greater forces to break the sutures, followed by smaller forces to maintain suture separation or gradually expand the palate and / or dental arch.

[0167] Determining the force system may also involve modeling the patient's facial structures, such as the skeletal structure of the jaw and palate. Scanning data of the palate and dental arches (such as X-ray data or 3D optical scan data) can, for example, be used to determine parameters of the patient's oral skeletal and muscular systems in order to determine sufficient forces to provide the desired palatal and / or dental arch expansion. In some embodiments, the thickness and / or density of the palatal sutures may be considered. In other embodiments, the treatment professional may select appropriate treatment based on the patient's physiological characteristics. For example, the nature of the palate may also be estimated based on factors such as the patient's age; for instance, younger adolescent patients typically require less force to expand the sutures than older patients because the sutures have not yet fully formed.

[0168] In box 1130, the design of one or more dental instruments shaped to implement a motion path is determined. In one embodiment, the one or more dental instruments are shaped to move one or more teeth toward a corresponding incremental arrangement. In one embodiment, orthodontic application is determined by treatment planning generator 276. A treatment or force application simulation environment can be used to perform the determination of one or more dental or orthodontic instruments, instrument geometry, material composition, and / or properties. The simulation environment may include, for example, a computer modeling system, biomechanical system, or device. Optionally, digital models of the instruments and / or teeth, such as finite element models, can be generated. Finite element models can be created using computer program application software available from various vendors. To create a solid geometry model, computer-aided engineering (CAE) or computer-aided design (CAD) programs can be used, such as AutoCAD® software products available from Autodesk Ltd., San Lafrey, California. To create and analyze finite element models, program products from multiple vendors can be used, including finite element analysis packages from ANSYS Ltd., Fort Cannons, Pennsylvania, and SIMULIA (Abaqus) software products from Dassault Systèmes, Waltham, Massachusetts.

[0169] In box 1140, instructions for manufacturing one or more dental instruments are identified or defined. In some embodiments, the instructions identify one or more geometries of the one or more dental instruments. In some embodiments, the instructions identify slices of layers for fabricating one or more dental instruments using a 3D printer. In some embodiments, the instructions identify one or more geometries for use in indirectly manufacturing the one or more dental instruments (e.g., by thermoforming plastic sheets on a 3D-printed mold). Dental instruments may include one or more of appliances (e.g., orthodontic appliances), retainers, incremental palatal expanders, attachment templates, etc.

[0170] In one embodiment, instructions for manufacturing one or more dental appliances are generated by a treatment planning generator 276. The instructions may be configured to control a manufacturing system or equipment to produce orthodontic appliances having a specific orthodontic function. In some embodiments, the instructions are configured to manufacture the orthodontic appliance using direct manufacturing methods (e.g., stereolithography, selective laser sintering, fused deposition modeling, 3D printing, continuous direct manufacturing, multimaterial direct manufacturing, etc.) according to the various methods presented herein. In alternative embodiments, the instructions may be configured to manufacture the appliance indirectly, for example, by 3D printing a mold and thermoforming a plastic sheet on the mold.

[0171] Method 1100 may include additional boxes: 1) intraorally scanning the patient’s maxillary arch and palate to generate three-dimensional data of the palate and maxillary arch; 2) determining the three-dimensional shape profile of the instrument to provide the gap and tooth engagement structure as described herein.

[0172] Although the boxes above illustrate a method 1100 for designing orthodontic appliances according to some embodiments, those skilled in the art will recognize some variations based on the teachings described herein. Some boxes may include sub-boxes. Some boxes may be repeated frequently as needed. One or more boxes of method 1100 can be performed using any suitable manufacturing system or apparatus, such as the embodiments described herein. Some boxes may be optional, and the order of the boxes may be changed as needed.

[0173] Figure 12 A method 1200 for digitally planning orthodontic treatment and / or designing or manufacturing instruments is illustrated according to an embodiment. Method 1200 can be applied to any treatment procedure described herein and can be performed by any suitable data processing system.

[0174] In box 1210, a digital representation of the patient's teeth is received. The digital representation may include surface topography data for the patient's oral cavity (including teeth, gingival tissue, etc.). The surface topography data can be generated by directly scanning the oral cavity, a physical model (positive or negative) of the oral cavity, or an impression of the oral cavity using a suitable scanning device (e.g., a handheld scanner, a desktop scanner, etc.).

[0175] In box 1220, one or more treatment phases are generated based on a digital representation of the teeth. In some embodiments, the one or more treatment phases are generated based on processing input dental arch data by a machine learning model trained by a treatment planning generator 276. Each treatment phase may include a 3D model of the dental arch generated in that treatment phase. A treatment phase may be an incremental repositioning phase of an orthodontic treatment procedure designed to move one or more of the patient's teeth from an initial tooth arrangement to a target tooth arrangement. For example, a treatment phase may be generated by determining an initial tooth arrangement indicated by the digital representation, determining the target tooth arrangement, and determining the motion paths necessary for achieving the target tooth arrangement for one or more teeth in the initial arrangement. The motion paths may be optimized based on minimizing the total distance of movement, preventing collisions between teeth, avoiding more difficult tooth movements, or any other suitable criteria.

[0176] In box 1230, at least one orthodontic appliance is manufactured based on the generated treatment phase. For example, a set of appliances can be manufactured, each shaped according to a tooth arrangement specified by a treatment phase, allowing the patient to wear the appliances sequentially to incrementally reposition the teeth from the initial arrangement to the target arrangement. The appliance set may include one or more orthodontic appliances described herein. The manufacture of the appliances may involve creating a digital model of the appliance to serve as input to a computer-controlled manufacturing system. Direct manufacturing methods, indirect manufacturing methods, or combinations thereof may be used to form the appliances, as needed.

[0177] In some cases, the phased arrangement or treatment phases may not be necessary for the design and / or manufacture of the device. For example... Figure 12 As shown by the dashed lines, the design and / or manufacture of orthodontic appliances, and possibly specific orthodontic treatments, may include using a representation of the patient's teeth (e.g., receiving a digital representation of the patient's teeth at box 1210), and subsequently designing and / or manufacturing orthodontic appliances based on the representation of the patient's teeth in the arrangement represented by the received representation.

[0178] Figure 13 An illustration of a machine in an example form of computing device 1300 is shown, within which a set of instructions can be executed to cause the machine to perform any one or more of the methods discussed herein. In alternative embodiments, the machine may be connected (e.g., networked) to other machines in a local area network (LAN), intranet, extranet, or the Internet. The machine may operate as a server or client machine in a client-server network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The machine may be a personal computer (PC), tablet computer, set-top box (STB), personal digital assistant (PDA), cellular phone, web device, server, network router, switch, or bridge, or any machine capable of executing a specified set of instructions (sequential or otherwise) of actions to be taken by the machine. Furthermore, although only a single machine is shown, the term "machine" should also be understood to include any collection of machines (e.g., computers) that individually or jointly execute a set (or more) of instructions to perform any one or more of the methods discussed herein. In one embodiment, computing device 1300 corresponds to Figure 1 The computing device 105.

[0179] Example computing device 1300 includes processing device 1302, main memory 1304 (e.g., read-only memory (ROM), flash memory, dynamic random access memory (DRAM) such as synchronous DRAM (SDRAM), etc.), static memory 1306 (e.g., flash memory, static random access memory (SRAM), etc.), and secondary memory (e.g., data storage device 1328), which communicate with each other via bus 1308.

[0180] Processing device 1302 represents one or more general-purpose processors, such as microprocessors, central processing units, etc. More specifically, processing device 1302 may be a Complex Instruction Set Computing (CISC) microprocessor, a Reduced Instruction Set Computing (RISC) microprocessor, a Very Long Instruction Word (VLIW) microprocessor, a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. Processing device 1302 may also be one or more special-purpose processing devices, such as application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), network processors, etc. Processing device 1302 is configured to execute processing logic (instruction 1326) for performing the operations and steps discussed herein.

[0181] The computing device 1300 may also include a network interface device 1322 for communicating with the network 1364. The computing device 1300 may also include a video display unit 1310 (e.g., a liquid crystal display (LCD) or a cathode ray tube (CRT)), an alphanumeric input device 1312 (e.g., a keyboard), a cursor control device 1314 (e.g., a mouse), and a signal generation device 1320 (e.g., a speaker).

[0182] Data storage device 1328 may include machine-readable storage medium (or more specifically, non-transitory computer-readable storage medium) 1324 thereon storing one or more sets of instructions 1326 embodying any one or more of the methods or functions described herein, such as instructions for treatment planning application 1350. Non-transitory storage medium refers to storage medium other than a carrier wave. Instructions 1326 may also reside wholly or at least partially within main memory 1304 and / or processing device 1302 during execution by computer device 1300, which also constitute computer-readable storage media.

[0183] Computer-readable storage medium 1324 can also be used to store treatment planning applications 1350, which can correspond to Figure 1Similar named components. The computer-readable storage medium 1324 may also store a software library containing methods for the treatment planning application 1350. Although the computer-readable storage medium 1324 is shown as a single medium in the example embodiment, the term "computer-readable storage medium" should be understood to include a single medium or multiple media (e.g., a centralized or distributed database and / or associated caches and servers) storing one or more sets of instructions. The term "computer-readable storage medium" should also be understood to include any medium other than a carrier wave capable of storing or encoding a set of instructions for machine execution and causing the machine to perform any one or more of the methods of the present invention. The term "computer-readable storage medium" should therefore be understood to include, but is not limited to, solid-state memory as well as optical and magnetic media.

[0184] It should be understood that the above description is illustrative and not restrictive. Many other embodiments will become apparent upon reading and understanding the above description. Although embodiments of the invention have been described with reference to specific exemplary embodiments, it will be appreciated that the invention is not limited to the described embodiments but can be practiced with modifications and changes within the spirit and scope of the appended claims. Therefore, the specification and drawings should be considered illustrative and not restrictive. Accordingly, the scope of the invention should be determined by reference to the appended claims and the full scope of their equivalents.

Claims

1. A method comprising: The processing device receives clinical data on the initial state of one or more dental arches of the patient; as well as The processing device determines one or more stages of orthodontic treatment planning for correcting one or more dental arches based on the processing of clinical data, wherein the stages of determining the orthodontic treatment planning include: Determine one or more actions to be performed on one or more dental arches; and Determine the target state of one or more dental arches that is predicted to be at least partially caused by the one or more actions.

2. The method according to claim 1, further comprising: Determine the orthodontic treatment plan, which includes one or more phases; Determine one or more additional orthodontic treatment plans; Output the orthodontic treatment plan and one or more additional orthodontic treatment plans to the display; Receive the choice of orthodontic treatment plan; as well as Implement the selected orthodontic treatment plan.

3. The method according to claim 2, further comprising: For orthodontic treatment planning, a first score is determined that is associated with the predicted achievement of one or more target conditions by the orthodontic treatment plan; as well as For each of the one or more additional orthodontic treatment plans, determine an additional score associated with the completion of one or more target conditions by the additional orthodontic treatment plan.

4. The method according to claim 3, wherein, Output the first score and additional scores from one or more additional orthodontic treatment plans to the display.

5. The method according to claim 1, wherein, The processing of clinical data is performed using a trained machine learning model, which outputs one or more actions and the target state of one or more dental arches.

6. The method according to claim 5, wherein, The first state of one or more dental arches is an intermediate state achieved through a previously determined orthodontic treatment plan, and the method further includes: Receive information associated with the previously determined orthodontic treatment plan; Cost or bonus values ​​are determined based on the similarity between the initial state of one or more dental arches and the predicted state of one or more dental arches from a previously determined orthodontic treatment plan; and The training of the trained machine learning model is updated based on cost or reward values, wherein processing of clinical data using the trained machine learning model is performed after the training of the trained machine learning model is updated.

7. The method according to claim 1, wherein, The receipt of clinical data and the determination of one or more phases of orthodontic treatment planning are performed before the commencement of orthodontic treatment for one or more dental arches, the method further comprising: The processing device receives new clinical data on the new state of one or more dental arches during the intermediate stage of orthodontic treatment planning; and The processing device determines one or more updated phases of orthodontic treatment planning based on the processing of new clinical data, with or without user input.

8. The method according to claim 7, wherein, One or more updated phases include the final phase with a new target state of one or more dental arches.

9. The method according to claim 7, further comprising: One or more new intermediate stages can be added to the orthodontic treatment plan based on new clinical data.

10. The method according to claim 1, wherein, Clinical data include at least one of the following: color two-dimensional (2D) images of one or more dental arches, three-dimensional (3D) models of each of one or more dental arches, intraoral scans of one or more dental arches, or X-ray images of one or more dental arches.

11. The method according to claim 1, wherein, One or more actions include at least one of the following: widening one or more dental arches, adding one or more attachments to one or more dental arches, extracting one or more teeth from one or more dental arches, performing proximal enamel removal on one or more teeth of one or more dental arches, or using elastic elements to fix one or more dental arches to each other.

12. The method according to claim 1, wherein, One or more actions include adjusting the position or orientation of at least one of one or more teeth on one or more dental arches.

13. The method according to claim 1, wherein, One or more actions and target states are determined to minimize at least one of the number of phases in the orthodontic treatment plan or the duration of orthodontic treatment performed according to the orthodontic treatment plan.

14. The method according to claim 1, further comprising: The occlusion type of one or more dental arches is determined based on the processing of clinical data, wherein the occlusion type includes at least one of open bite, reverse bite, or deep overbite; One or more actions are determined at least in part based on the type of bite.

15. The method according to claim 1, further comprising: The type of malocclusion in one or more dental arches is determined by processing clinical data; One or more actions are determined at least in part based on the type of malocclusion.

16. The method according to claim 1, wherein, One or more dental arches include the patient's upper dental arch, and the method further includes: The processing device receives clinical data on the initial state of the patient's mandibular arch; and The processing device determines one or more stages for orthodontic treatment planning to correct the lower dental arch based on the processing of clinical data, wherein the stages for determining orthodontic treatment planning to correct the lower dental arch include: Determine one or more additional actions to be performed on the lower dental arch; and Determine the target state of the inferior dental arch that is predicted to be at least partially caused by one or more additional actions.

17. The method according to claim 1, further comprising: Determining the current relationship between the upper dental arch and the lower dental arch in one or more dental arches, wherein determining the target state of one or more dental arches predicted to be at least partially caused by one or more actions includes determining the target relationship between the upper and lower dental arches.

18. The method according to claim 1, wherein, Determining one or more actions to be performed on one or more dental arches includes determining one or more tooth restoration actions to be performed, which are associated with one or more restoration options.

19. The method of claim 18, wherein: One or more recovery options include at least one of the following: a) one or more sizes that can be used as patches, b) the type of patch used, or c) patch thickness; as well as One or more dental restoration procedures to be performed include the removal of a amount of tooth body from the tooth to which the veneer will be received.

20. The method according to claim 18, wherein, The space for placing a dental implant is compressed by the roots of one or more teeth adjacent to the space, and wherein one or more tooth restoration actions to be performed include opening the roots of one or more teeth to allow placement of a dental implant in the space.

21. A method comprising: The processing device receives training data items, which include clinical data on the state of one or more dental arches and information on the execution of orthodontic treatment for one or more dental arches; The processing device determines one or more stages of orthodontic treatment planning for correcting one or more dental arches based on the processing of clinical data using a machine learning model. The stages of determining the orthodontic treatment planning include: determining one or more actions to be performed on one or more dental arches; and determining the target state of one or more dental arches, which is at least partially caused by the one or more actions. Use cost functions to determine the cost or reward values ​​associated with one or more phases of orthodontic treatment planning; and Update one or more nodes of the machine learning model based on cost or reward values.

22. The method according to claim 21, wherein, Determining cost or reward values ​​and updating one or more nodes in the machine learning model are performed according to a reinforcement learning algorithm.

23. The method according to claim 22, wherein, Reinforcement learning algorithms are one of Q-learning, deep Q-learning, or dual deep Q-networks.

24. The method of claim 21, further comprising: Determine the number of stages in the orthodontic treatment plan; Determine the amount of change between the number of stages in the orthodontic treatment plan and the target maximum number of stages in the orthodontic treatment plan; as well as The cost value or reward value is determined based on the amount of change.

25. The method of claim 21, further comprising: Determine the predicted timeframe for completing the orthodontic treatment plan; Determine the amount of change between the predicted time and the target maximum time for completing the orthodontic treatment plan; as well as The cost value or reward value is determined based on the amount of change.

26. The method of claim 21, further comprising: Receives input for at least one of the target number of stages or target time for completing orthodontic treatment planning; as well as The cost function is set based on the input.

27. The method according to claim 26, wherein, The machine learning model is trained to generate orthodontic treatment plans that achieve a target state of the dental arch while minimizing at least one of the number of stages or the amount of time required to complete the orthodontic treatment plan.

28. The method according to claim 21, wherein, Clinical data are labeled using occlusal type, which is one of open bite, deep overbite, or reverse bite.

29. The method according to claim 21, wherein, Clinical data are labeled using treatment problems, which include at least one of vertical, sagittal, or lateral problems.

30. The method according to claim 21, wherein, Clinical data are categorized by the number of treatment phases, actions performed in one or more treatment phases, and the degree of orthodontic success.

31. The method according to claim 21, wherein, The machine learning model is trained to receive new clinical data on the patient's dental arch and output multiple treatment planning options for orthodontic treatment of the patient's dental arch.

32. The method according to claim 21, wherein, The training data items include multiple states of one or more dental arches, each of which is associated with a different stage of orthodontic treatment.

33. The method according to claim 21, wherein, Clinical data include at least one of the following: one or more color two-dimensional (2D) images of one or more dental arches, one or more three-dimensional (3D) models of one or more dental arches, one or more intraoral scans of one or more dental arches, or one or more X-ray images of one or more dental arches.

34. The method according to claim 21, wherein, One or more actions include at least one of the following: widening the dental arch in one or more dental arches, adding one or more attachments to the dental arch, extracting one or more teeth from the dental arch, performing proximal enamel removal on one or more teeth of the dental arch, or using an elastic element to fix the dental arch to the opposing dental arch.

35. The method according to claim 21, wherein, The training data also includes clinical data on the occlusal relationship between the upper and lower dental arches in one or more dental arches.

36. A computer-readable medium comprising instructions that, when executed by a processing means, cause the processing means to perform the method according to any one of claims 1 to 35.

37. A system comprising: A computing device, comprising a memory and a processing device, wherein the processing device is configured to execute instructions from the memory to perform the method according to any one of claims 1 to 35.

38. The system according to claim 37, wherein, The computing device is configured to manufacture one or more orthodontic appliances according to one or more stages of an orthodontic treatment plan.

39. The system of claim 37, further comprising: An additional computing device is configured to send clinical data to the computing device; as well as The storage device is configured to store orthodontic treatment plans.

Images