Mouthpiece brush size selection

Abstract

The present invention relates to a computer-implemented method of selecting a size of a mouthpiece brush (10) for an individual. The computer-implemented method comprises: receiving (500) data associated with the individual; generating (502), using a plurality of machine learning models, a vector of latent parameters of the individual's jaw (310) based on the received data; measuring (504) a dimension of the jaw based on the vector of latent parameters of the jaw; and selecting (506) a size of the mouthpiece brush based on the measured dimension of the jaw.

Description

MOUTHPIECE BRUSH SIZE SELECTIONFIELD

[0001] The subject-matter of the present disclosure relates to computer implemented methods of selecting a size of a mouthpiece brush for an individual, computer- implemented methods of training machine learning models to generate a vector of latent parameters of a jaw of an individual from a face from an image and training machine learning models to generate a vector of latent parameters of a head of an individual from a face from an image, and transitory, or non-transitory, computer-readable media.BACKGROUND

[0002] The making of successful oral health care (OHC) products, e.g. single-mouth-piece toothbrushes, implies that the product perfectly fit the jaw shape of the individual. However, the jaw shapes and sizes for different people vary a lot. For example, the jaw length in adults can vary in range of about 20mm. When ‘one size fits all’ concept does net work, multiple sizes, e.g. ‘small’, ‘medium’ and ‘large’ can be envisioned. This creates a challenge for the end-user to decide whether ‘one size fits all’ will fit them, or which ‘small’, ‘medium’ and ‘large’ size to choose. Moreover, it would be difficult for the end user to try different mouthpieces in a shop (in a similar way as clothes for instance) for hygiene reasons. The problem can be amplified with a premium price for (technically complex) single-mouth-piece toothbrushes.

[0003] It is an aim of the subject-matter of the present disclosure to improve on the prior art.SUMMARY

[0004] According to a first aspect of the present invention, there is provided a computer- implemented method of selecting a size of a mouthpiece brush for an individual, the computer-implemented method comprising: receiving data associated with the individual; generating, using a plurality of machine learning models, a vector of latent parameters of the individual’s jaw based on the received data; measuring a dimension of the jaw based on the vector of latent parameters of the jaw; and selecting a size of the mouthpiece brush based on the measured dimension of the jaw. Using the plurality of machine learning models in this way enables a mouthpiece to be selected easily without having to measure the jaw intraoral ly.

[0005] In an embodiment, the data comprises an image of the individual’s face, and wherein the generating, using the plurality of machine learning models, the vector of latent parameters of the jaw comprises: generating, using a first machine learning model of the plurality of machine learning models, a vector of latent parameters of a head of the individual using the image of the individual’s face as an input; and transforming, using a second machine learning model of the plurality of machine learning models, the vector of latent parameters of the head into a vector of latent parameters of the jaw.

[0006] In an embodiment, the data comprises an image of the individual’s face, and the generating, using the plurality of machine learning models, the vector of latent parameters of the jaw comprises: generating, using a third machine learning model of the plurality of machine learning models, a vector of latent parameters of a jaw of the individual using the image of the individual’s face as an input.

[0007] In an embodiment, the data comprises metadata and wherein the generating, using one or more machine learning models, a vector of latent parameters of a jaw of the individual based on the face in the received image, comprises: generating, using a fourth machine learning model of the plurality of machine learning models, a vector of latent parameters of the jaw by inputting the metadata to the third machine learning model.

[0008] In an embodiment, the data comprises metadata and wherein the generating, using one or more machine learning models, a vector of latent parameters of the jaw of the individual based on the face in the received image, comprises: generating, using a fifth machine learning model of the one or more machine learning models, a vector of latent parameters of a head of the individual using the metadata as an input; and transforming, using a second machine learning model of the one or more machine learning models, the vector of latent parameters of the head into a vector of latent parameters of the jaw.

[0009] In an embodiment, the generating, using the plurality of machine learning models, the vector of latent parameters of the jaw comprises: fusing, with an aggregator machine learning model of the plurality of machine learning models, two or more vectors of latent parameters of the jaw; each vector corresponding to the respective output from one of the machine learning models of the plurality of machine learning models.

[0010] In an embodiment, the measuring a dimension of the jaw based on the vector of latent parameters of the jaw comprises: inputting, to a sixth machine learning model, the vector of latent parameters of the jaw to obtain directly the dimensions of the jaw.

[0011] In an embodiment, the measuring a dimension of the jaw based on the vector of latent parameters of the jaw comprises: inputting, to a seventh machine learning model, the vector of latent parameters of the jaw to obtain a three-dimensional model of the jaw; and measuring the dimensions of the jaw from the obtained three-dimensional model of the jaw.

[0012] In an embodiment, the selecting a size of the mouthpiece brush based on the measured dimensions of the jaw comprises: inputting, to an eight machine learning model, the measured dimensions to obtain a selection of the size of the mouthpiece.

[0013] In an embodiment, the selecting the size of the mouthpiece brush based on the measured dimension of the jaw comprises: inputting, to a rules-based mathematical model, the measured dimension to obtain a selection of the size of the mouthpiece.

[0014] In an embodiment, the dimension is at least one of: a length of the jaw; an angle between a molar, an incisor, and a canine; and a molar thickness.

[0015] According to an aspect of the present invention, there is provided a computer- implemented method of training a machine learning model, MLM, to generate a vector of latent parameters of a jaw of an individual from a face from an image, the computer- implemented method comprising: receiving a plurality of images and a plurality of scans of jaws; registering a three-dimensional jaw mesh with each scan of the plurality of scans of jaws; deconstructing the registered meshes using principal component analysis to obtain a vector of latent parameters of a jaw; inputting the plurality of images of faces to a machine learning model to generate respective vectors of latent parameters of a jaw; calculating an error between the generated vectors of latent parameters and the obtained vectors of latent parameters; and updating parameters of the machine learning model to minimise the error. The vector of latent parameters of a jaw can be used as an input to a machine learning model trained to derive one or more measurements of the jaw, for example.

[0016] According to an aspect of the present invention, there is provided a computer- implemented method of training a machine learning model, MLM, to generate a vector of latent parameters of a head of an individual from a face from an image, the computer- implemented method comprising: receiving a plurality of images and a plurality of scans of faces; registering a three-dimensional head mesh with each scan of the plurality of scans of faces; deconstructing the registered meshes using principal component analysis to obtain a vector of latent parameters of a head; inputting the plurality of images of facesto a machine learning model to generate respective vectors of latent parameters of a head; calculating an error between the generated vectors of latent parameters and the obtained vector of latent parameters; and updating parameters of the machine learning model to minimise the error.

[0017] According to an aspect of the present invention, there is provided a transitory, or non-transitory, computer-readable medium, having instructions stored thereon that when executed by one or more processors, cause the one or more processors to perform the computer-implemented method of any preceding aspect or embodiment.

[0018] These and other aspects of the present invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter.BRIEF DESCRIPTION OF DRAWINGS

[0019] The embodiments of the present inventions may be best understood with reference to the accompanying figures, in which:

[0020] Figure 1 shows a mouthpiece brush;

[0021] Figure 2 shows a mobile device and a remote device used to implemented the computer-implemented methods described herein;

[0022] Figure 3 shows a block diagram covering a computer-implemented method of selecting a size of a mouthpiece brush for an individual according to one or more embodiments;

[0023] Figure 4 shows a block diagram of a computer-implemented method of training a plurality of machine learning models according to one or more embodiments;

[0024] Figure 5 shows a flow chart summarising a computer-implemented method of selecting a size of a mouthpiece brush for an individual;

[0025] Figure 6 shows a flow chart summarising a computer-implemented method of training a machine learning model, MLM, to generate a vector of latent parameters of a jaw of an individual from a face from an image; and

[0026] Figure 7 shows a flow chart summarising a computer-implemented method of training a machine learning model, MLM, to generate a vector of latent parameters of a head of an individual from a face from an image.DESCRIPTION OF EMBODIMENTS

[0027] At least some of the example embodiments described herein may be constructed, partially or wholly, using dedicated special-purpose hardware. Terms such as ‘component’, ‘module’ or ‘unit’ used herein may include, but are not limited to, a hardware device, such as circuitry in the form of discrete or integrated components, a Field Programmable Gate Array (FPGA) or Application Specific Integrated Circuit (ASIC), which performs certain tasks or provides the associated functionality. In some embodiments, the described elements may be configured to reside on a tangible, persistent, addressable storage medium and may be configured to execute on one or more processors. These functional elements may in some embodiments include, by way of example, components, such as software components, object-oriented software components, class components and task components, processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. Although the example embodiments have been described with reference to the components, modules and units discussed herein, such functional elements may be combined into fewer elements or separated into additional elements. Various combinations of optional features have been described herein, and it will be appreciated that described features may be combined in any suitable combination. In particular, the features of any one example embodiment may be combined with features of any other embodiment, as appropriate, except where such combinations are mutually exclusive. Throughout this specification, the term “comprising” or “comprises” means including the component(s) specified but not to the exclusion of the presence of others.

[0028] With reference to Figure 1 , a mouthpiece toothbrush 100 (also referred to herein for the sake of concise terminology as a “mouthpiece brush”) is shaped to conform to a jaw of an individual. The term jaw is used herein to mean the teeth of a dental arch and the gum, where an individual has an upper jaw and a lower jaw.

[0029] With reference to Figure 2, the computer-implemented methods described herein may be embodied using a user device 200 and a remote device 202. The user device may be a mobile device. The mobile device includes a user interface 204 which comprises display and user input functionality. The mobile device also includes an imaging device 206. In the mobile device shown in Figure 2 the imaging device is a 2D camera. In otheruser devices, the imaging device may be a 3D scanning device for capturing 3D scans of an individual.

[0030] The remote device 202 is configured to receive data from the mobile device 200. The remote device includes storage 208 and one or more processors 210. The remote device 202 may be used to perform the computer-implemented methods described herein and transfer a mouthpiece size back to the mobile device 200 or the computer- implemented methods can be performed directly on the mobile device 200. More specifically, the computer-implemented method described herein may be stored as instructions as electronic data on the storage 208 and executed by the one or more processors 210 such that the one or more processors 210 performed the computer- implemented methods described herein. In this way, the storage may be non-transitory computer-readable media. In addition, transitory computer-readable media may include the instructions and downloaded to the storage 208.

[0031] With reference to Figure 3, a computer-implemented method of selecting a size of a mouthpiece brush for an individual is shown and includes a first step of receiving data associated with the individual.

[0032] The data includes images of an individual’s face. The images include a smiling image 300 and a neutral expression image 302. The smiling image 300 shows the individual showing a smile which is a happy smile and one which shows the teeth of the individual. The neutral expression is of the user where the teeth of not visible.

[0033] The data also includes metadata 304. The metadata 304 includes parameters about the individual such as weight, height, age, sex, etc. The metadata 304 may include other information such as dental records.

[0034] The method may include generating, using a plurality of machine learning models, MLMs, a vector of latent parameters of the individual’s jaw based on the received data. The machine learning models may be neural networks, for example.

[0035] There are various ways to obtain the vector of latent parameters of the jaw of the individual, which are described below.

[0036] When the data is either of the images 300, 302, a first machine learning model 306 of the plurality of machine learning models generates a vector of latent parameters of a head of the individual from the image of the individual’s face which is used as an input. A second machine learning model 308, of the plurality of machine learning models, is thenused to transform the vector of latent parameters of the head into a vector of latent parameters of the jaw 310.

[0037] In other embodiments, the vector of latent parameters of the jaw may be obtained directly without the need for transformation using the second machine learning model. This can be achieved by generating, using a third machine learning model 312, of the plurality of machine learning models, a vector of latent parameters of a jaw of the individual using the individual’s face as an input. In other words, either image 300, 302, may be input to the third machine learning model 312 to obtain directly the vector of latent parameters of the jaw.

[0038] When the data is, or comprises, metadata, a fourth machine learning model 314, of the plurality of machine learning models, may be used. The vector of latent parameters of the jaw may be obtained by inputting the metadata to the fourth machine learning model 314.

[0039] Similarly, the vector of latent parameters of the jaw may be obtained by inputting the meta data into a fifth machine learning model 316 of the plurality of machine learning models. The fifth machine learning model 316 may be trained to generate a vector of latent parameters of a head of the individual using the metadata as an input. The latent vector of the head is then used as an input to the second machine learning model 308 to transform it into a vector of latent parameters of the jaw.

[0040] Next, an aggregator machine learning model of the plurality of machine learning models, is used to fuse two or more of the vectors of latent parameters of the jaw, where each vector corresponds to the respective output from one of the machine learnings of the plurality of machine learning models. More specifically, the vectors of the latent parameters of the jaw obtained based on the smiling image 300, the neutral expression image 302, and the metadata 304, are fused using the aggregator machine learning model 305. Numerically, the aggregator machine learning model may be a ninth machine learning model.

[0041] The output from the aggregator machine learning model will be one vector of latent parameters of the jaw of the individual.

[0042] Next, the computer-implemented method comprises measuring a dimension of the jaw based on the vector of latent parameters of the jaw. The dimension of the jaw may be, for example, a length of a dental arch, a height of teeth, and a thickness of one ormore teeth. More specifically, the dimension may be at least one of a length of the jaw (e.g. a length of a dental arch), an angle between a molar an incisor and a canine, and a molar thickness. These dimensions are usually minimally required in order to derive a size of the mouthpiece brush. This may involve inputting, to a sixth machine learning model 318, the vector of latent parameters of the jaw to obtain directly the dimensions of the jaw.

[0043] In other embodiments, this may be done by first inputting the vector of latent parameters of the jaw into a seventh machine learning model 319 to obtain a three- dimensional model of the jaw, and then measuring the dimensional of the jaw from the obtained three-dimensional model of the jaw. The measuring may be done with a rules- based model 321. The rules-based model may be a first rules-based model.

[0044] As a result of these methods, a measured dimension of the jaw 320 is obtained.

[0045] Next, the computer-implemented method comprises selecting a size of the mouthpiece brush based on the measured dimension of the jaw. This may be achieved in one of a number of different ways.

[0046] In one embodiment, the measured dimension of the jaw 320 is input to an eighth machine learning model 324 to obtain a selection of the size of the mouthpiece 322. In other embodiments, the measured dimension may be input to a rules-based mathematical model to obtain a selection of the size of the mouthpiece. The rules-based mathematical model may be a second rules-based model, 326.

[0047] The training of the various machine learning models is described below.

[0048] With reference to Figure 4, there is a method of obtaining the vectors of the latent parameters of the 3D head and the 3D jaw for use in training the first, second, third, fourth, and fifth machine learning models.

[0049] First, a method of obtaining a vector of latent parameters of the jaw and head is described.

[0050] This method requires 3D scans of a head and 3D scans of a jaw of the individual. The scans are also captured together with an image of the individual for training the first, third, fourth, and fifth, machine learning models.

[0051] The number, A / , of scans of different individuals needs to be sufficiently large, e.g. greater than 100. The actual number of scans may exceed 1000. For each scan, we register a common 3D template Vi, i=1...N, where Vi is the template registered with the Ithscan. The common 3D template may comprise a collection of 2D landmarks, or a dense mesh of 3D landmarks. The common mesh is morphed to fit a 3D scan surface of each scan in order to register the common mesh with the scan. The registered mesh for each scan is then rigidly aligned to the average mesh AvgV. The principal component analysis, PCA, is applied to get a linear decomposition of the aligned models.

[0052] The linear decomposition results two things. First, there is a vector of PCA coefficients, i.e. latent parameters encoding the 3D scan of the head. This is the vector of latent parameters of the head. Second, there is a a PCA basis. The PCA basis conmprses principle 3D shape components. The PCA basis can be used to devoce the model of the 3D head from the vector of latent parameters of the head. It can also be used for encoding of new head scans.

[0053] This method can also be used for 3D scans of the jaw to obtain a vector of latent paramters of the jaw.

[0054] The method in Figure 4 is a method of training the second machine learning model 308. The vector of latent paramters of the jaw 402, and the vector of latent paramters of the head 404 are derived from the method described above. A jaw encoder 406 and a jaw decoder 408 may be neural networks. Similarly, a head encoder 410 and a head decoder 412 may be neural networks. The registered 3D model of the jaw 414 obtained above may be input to the jaw encoder 406 and the jaw decoder 408 may generate a corresponding registered 3D jaw model 415 from the resulting vector of latent parameters of the jaw 402. The jaw encoder and the jaw decoder may be trained to reduce a jaw loss function 416, which is a loss between the registered 3D jaw models.

[0055] It should be noted that the encoder-decoder neural network generalises the previously described PCA decomposition where the latent paramters in the “bottle-neck” generalised PCA coefficients. Thus, in the case of PCA, the encoder or decoder become linear matrix multiplications.

[0056] Similarly, the registered 3D head model 418 may be input to the head encoder 410, and the head decoder 412 may generate a corresponding registered 3D model of the head 420 from the resulting vector of latent paramters of the head 404. The head encoder 410 and the head decoder 412 may be trained to reduce a head loss function 422, which is a loss between the registered 3D head models.

[0057] The second machine learning model 308 is trained to transform the vector of latent parameters of the jaw into the vector of latent paramters of the head, and vice-versa using conventional machine learning techniques.

[0058] Since the image of the individuals is paired with the scans and metadata of the individuals, it is posisble then to train the first, third, fourth, and fifth machine learning models.

[0059] For instance, the first machine learning model is trained to generate a vector of latent parameters of a head of an individual from a face from an image. This method comprises using the method of obtaining the vector of latent parameters of the head from the scan of the head as described above. This may be summarised as receiving a plurality of images and a plurality of scans of faces; registering a three-dimensional head mesh with each scan of the plurality of scans of faces; and deconstructing the registered meshes using principal component analysis to obtain a vector of latent parameters of a head. Next, the plurality of faces are input to the first machine learning model to generate respective vectors of latent parameters of the head. Then, an error is calculated between the generated vectors of latent parmaters and the obtained vector of latent paramters. Finally, the paramters of the first machine learning model are updated to minimise the error.

[0060] Similarly, the third machine learning model is trained to generate a vector of latent parameters of a jaw of an individual from a face from an image. This method comprises using the method of obtaining the vector of latent parameters of the jaw from the scan of the jaw as described above. This may be summarised as receiving a plurality of images and a plurality of scans of faces (e.g. scans of the jaws); registering a three-dimensional head mesh with each scan of the plurality of scans of faces; and deconstructing the registered meshes using principal component analysis to obtain a vector of latent parameters of a jaw. Next, the plurality of faces are input to the first machine learning model to generate respective vectors of latent parameters of the jaw. Then, an error is calculated between the generated vectors of latent parmaters and the obtained vector of latent paramters. Finally, the paramters of the third machine learning model are updated to minimise the error.

[0061] Similarly, the fourth and fifth machine learning models can be trained in the same way but instead receiving metadata rather than images of the face and using those as inputs.

[0062] The other machine learning models described herein may be trained using conventional machine learning techniques using the associated inputs and outputs as described above.

[0063] With reference to Figure 5, the computer-implemented method of selecting a size of a mouthpiece brush for an individual may be summarised as comprising: receiving 500data associated with the individual; generating 502, using a plurality of machine learning models, a vector of latent parameters of the individual’s jaw based on the received data; measuring 504 a dimension of the jaw based on the vector of latent parameters of the jaw; and selecting 506 a size of the mouthpiece brush based on the measured dimension of the jaw.

[0064] With reference to Figure 6, the computer-implemented method of training a MLM to generate a vector of latent parameters of a jaw of an individual from a face from an image can be summarised as comprising: receiving 600 a plurality of images and a plurality of scans of faces; registering 602 a three-dimensional jaw mesh with each scan of the plurality of scans of faces; deconstructing 604 the registered meshes using principal component analysis to obtain a vector of latent parameters of a jaw; inputting 606 the plurality of images of faces to a machine learning model to generate respective vectors of latent parameters of a jaw; calculating 608 an error between the generated vectors of latent parameters and the obtained vectors of latent parameters; and updating 610 parameters of the machine learning model to minimise the error.

[0065] With reference to Figure 7, the computer-implemented method of training MLM to generate a vector of latent parameters of a head of an individual from a face from an image may be summarised as comprising: receiving 700 a plurality of images and a plurality of scans of faces; registering 702 a three-dimensional head mesh with each scan of the plurality of scans of faces; deconstructing 704 the registered meshes using principal component analysis to obtain a vector of latent parameters of a head; inputting 706 the plurality of images of faces to a machine learning model to generate respective vectors of latent parameters of a head; calculating 708 an error between the generated vectors of latent parameters and the obtained vector of latent parameters; and updating 710 parameters of the machine learning model to minimise the error.

[0066] While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments.

[0067] Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit may fulfil the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually differentdependent claims does not indicate that a combination of these measured cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.

Claims

CLAIMS1 . A computer-implemented method of selecting a size of a mouthpiece brush (10) for an individual, the computer-implemented method comprising: receiving (500) data associated with the individual; generating (502), using a plurality of machine learning models, a vector of latent parameters of the individual’s jaw (310) based on the received data; measuring (504) a dimension of the jaw based on the vector of latent parameters of the jaw; and selecting (506) a size of the mouthpiece brush based on the measured dimension of the jaw.

2. The computer-implemented method of Claim 1 , wherein the data comprises an image of the individual’s face, and wherein the generating, using the plurality of machine learning models, the vector of latent parameters of the jaw comprises: generating, using a first machine learning model (306) of the plurality of machine learning models, a vector of latent parameters of a head of the individual using the image of the individual’s face as an input (300, 302); and transforming, using a second machine learning model (308) of the plurality of machine learning models, the vector of latent parameters of the head into a vector of latent parameters of the jaw.

3. The computer-implemented method of Claim 1 , wherein the data comprises an image of the individual’s face, and the generating, using the plurality of machine learning models, the vector of latent parameters of the jaw comprises: generating, using a third machine learning model (312) of the plurality of machine learning models, a vector of latent parameters of a jaw of the individual using the image of the individual’s face as an input.

4. The computer-implemented method of any preceding claim, wherein the data comprises metadata and wherein the generating, using one or more machine learning models, a vector of latent parameters of a jaw of the individual based on the face in the received image, comprises:generating, using a fourth machine learning model (314) of the plurality of machine learning models, a vector of latent parameters of the jaw by inputting the metadata to the fourth machine learning model.

5. The computer-implemented method of any of Claims 1 to 3, wherein the data comprises metadata and wherein the generating, using one or more machine learning models, a vector of latent parameters of the jaw of the individual based on the face in the received image, comprises: generating, using a fifth machine learning model (316) of the one or more machine learning models, a vector of latent parameters of a head of the individual using the metadata as an input; and transforming, using a second machine learning model of the one or more machine learning models, the vector of latent parameters of the head into a vector of latent parameters of the jaw.

6. The computer-implemented method of every preceding claim, wherein the generating, using the plurality of machine learning models, the vector of latent parameters of the jaw comprises: fusing, with an aggregator machine learning model (305) of the plurality of machine learning models, two or more vectors of latent parameters of the jaw; each vector corresponding to the respective output from one of the machine learning models of the plurality of machine learning models.

7. The computer-implemented method of any preceding claim, wherein the measuring a dimension of the jaw based on the vector of latent parameters of the jaw comprises: inputting, to a sixth machine learning model (318), the vector of latent parameters of the jaw to obtain directly the dimensions of the jaw.

8. The computer-implemented method of any of Claims 1 to 6, wherein the measuring a dimension of the jaw based on the vector of latent parameters of the jaw comprises:inputting, to a seventh machine learning model (319), the vector of latent parameters of the jaw to obtain a three-dimensional model of the jaw; and measuring the dimensions of the jaw from the obtained three- dimensional model of the jaw.

9. The computer-implemented method of any preceding claim, wherein the selecting a size of the mouthpiece brush based on the measured dimensions of the jaw comprises: inputting, to an eighth machine learning model (324), the measured dimensions to obtain a selection of the size of the mouthpiece.

10. The computer-implemented method of any of Claims 1 to 8, wherein the selecting the size of the mouthpiece brush based on the measured dimension of the jaw comprises: inputting, to a rules-based mathematical model (326), the measured dimension to obtain a selection of the size of the mouthpiece.11 . The computer-implemented method of any preceding claim, wherein the dimension is at least one of: a length of the jaw; an angle between a molar, an incisor, and a canine; and a molar thickness.

12. A computer-implemented method of training a machine learning model, MLM, to generate a vector of latent parameters of a jaw of an individual from a face from an image, the computer-implemented method comprising: receiving (600) a plurality of images and a plurality of scans of faces; registering (602) a three-dimensional jaw mesh with each scan of the plurality of scans of faces; deconstructing (604) the registered meshes using principal component analysis to obtain a vector of latent parameters of a jaw; inputting (606) the plurality of images of faces to a machine learning model to generate respective vectors of latent parameters of a jaw;calculating (608) an error between the generated vectors of latent parameters and the obtained vectors of latent parameters; and updating (610) parameters of the machine learning model to minimise the error.

13. A computer-implemented method of training a machine learning model, MLM, to generate a vector of latent parameters of a head of an individual from a face from an image, the computer-implemented method comprising: receiving (700) a plurality of images and a plurality of scans of faces; registering (702) a three-dimensional head mesh with each scan of the plurality of scans of faces; deconstructing (704) the registered meshes using principal component analysis to obtain a vector of latent parameters of a head; inputting (706) the plurality of images of faces to a machine learning model to generate respective vectors of latent parameters of a head; calculating (708) an error between the generated vectors of latent parameters and the obtained vector of latent parameters; and updating (710) parameters of the machine learning model to minimise the error.

14. A transitory, or non-transitory, computer-readable medium, having instructions stored thereon that when executed by one or more processors, cause the one or more processors to perform the computer-implemented method of any preceding claim.

Images