Method for generating images of expected results of medical cosmetic procedures on human anatomical features from images of the anatomical features prior to these medical cosmetic procedures - Patents.com

Abstract

[Solution] A method for imaging a predicted outcome of a medical cosmetic procedure includes converting an input image of an anatomical feature into an input image vector. A directional vector corresponding to the medical cosmetic procedure is determined. An amplitude of the directional vector is determined. The directional vector is multiplied with the determined amplitude to obtain a product vector. The product vector is a vector that is added to the input image vector. An output image corresponding to a predicted visual appearance of the human anatomical feature is generated from the vector-added product vector and the input image vector. A computer program stored on a non-transitory computer readable medium causes a computer to perform the imaging method.

Description

[Technical field]

[0001] The present disclosure relates to the field of processing images of human anatomical features, such as faces, to predict the expected facial appearance as a result of proposed procedures, including cosmetic medical procedures, surgical procedures, or medical procedures such as injection of fillers and other agents. More specifically, the present disclosure relates to a method for creating predicted images based on machine learning processing of actual appearance changes resulting from facial procedures on real people. [Background technology]

[0002] Cosmetic procedures, such as medical cosmetic procedures, surgical procedures, and injections (e.g., but not limited to, Botox and fillers), are performed to modify the appearance of human features, such as the face. Although physicians, particularly plastic surgeons, have detailed knowledge of how such procedures are performed and have an informed understanding of how such procedures change a person's appearance, it is difficult for them to convey to the individual what can reasonably be expected as a result of such procedures. Furthermore, while physicians may have the ability to select the amount of feature change with reference to the strength of the procedure and the products used, they may still not be able to convey to the individual how different amounts of feature change will affect the overall appearance when the treatment is completed. Summary of the Invention [Problem to be solved by the invention]

[0003] Therefore, there is a need for a tool that can visualize how a facial treatment will change the appearance of a recipient based on their actual pre-treatment appearance. [Means for solving the problem]

[0004] One aspect of the present disclosure is a method of generating an image of a predicted appearance of a human anatomical feature resulting from a medical cosmetic procedure from an image of the human anatomical feature before the procedure. The method according to this aspect of the present disclosure includes: sending an image of the human anatomical feature as an input to a computer before performing the medical cosmetic procedure; converting, in the computer, the input image into an input image vector; selecting, in the computer, a direction vector corresponding to the medical cosmetic procedure; determining, in the computer, an amplitude of the direction vector; multiplying, in the computer, the direction vector by the determined amplitude to obtain a product vector; vector-adding, in the computer, the product vector to the input image vector; and generating, in the computer, an output image corresponding to the predicted visual appearance from the added vector, the product vector, and the input image vector.

[0005] In another aspect of the present disclosure, a computer program is stored on a non-transitory computer readable medium, the program having logic operable to cause a programmable computer to perform the operations of the methods according to the present disclosure, as described below, in various embodiments.

[0006] Some embodiments further include optimizing, in a computer, the transformation of the input image by: (a) generating an optimized image from the input image vector; (b) comparing the optimized image to the input image to calculate a loss function; (c) modifying the input image vector and repeating the generation of the optimized image; and (d) repeating (a), (b), and (c) until either (i) the value of the loss function falls below a predetermined threshold or (ii) (a), (b), and (c) have been repeated a selected number of times.

[0007] In some embodiments, determining the upper amplitude limit includes, in a computer, (a) converting pre-treatment images and corresponding post-treatment images of corresponding human anatomical features into vectors; (b) calculating a difference vector between each converted pre-treatment image and each corresponding post-treatment converted image; (c) calculating an amplitude for each image vector pair of a pre-treatment image vector and a post-treatment image vector using the difference vector and the previously identified direction vector; and (d) determining an upper amplitude limit using the distribution of amplitude values ​​calculated in (c).

[0008] In some embodiments, determining the amplitude includes, in a computer, (a) converting a pre-treatment image of the corresponding human anatomical feature into an input vector; (b) first estimating the amplitude and generating an intermediate product vector with a previously estimated direction vector; (c) adding the intermediate product vector to the converted input vector; (d) generating an output image from the intermediate product vector; (e) determining a change in landmark position, color, and texture between the input image and the output image; and (f) adjusting the amplitude and repeating (b), (c), (d), and (e) until a change in at least one of the landmarks, color, and texture reaches a predefined threshold or the amplitude reaches a calculated predefined upper limit value.

[0009] In some embodiments, determining the direction vector comprises, for a cosmetic procedure: converting at least one image of the anatomical feature before the medical cosmetic procedure into a first input image vector; converting at least one image of the anatomical feature after the medical cosmetic procedure into a second input image vector; treating the pre-treatment image vector and the post-treatment image vector as a binary classification, with the pre-treatment image and the post-treatment image representing negative and positive classes, respectively; setting a linear decision boundary; and determining a vector orthogonal to the linear decision boundary.

[0010] In some embodiments, determining the directional vector includes, for a medical cosmetic procedure, converting at least one image of the anatomical feature before the medical cosmetic procedure to a first input image vector, converting at least one image of the anatomical feature after the medical cosmetic procedure to a second input image vector, calculating a difference vector by subtracting the first input image vector from the second input image vector, and normalizing the difference vector to a unit vector. If multiple difference vectors are calculated from multiple input image pairs, the directional vector is calculated by statistical aggregation.

[0011] In some embodiments, converting the input image into an input image vector comprises processing the image through a pre-trained encoder in a computer.

[0012] In some embodiments, generating the output image consists of inputting the summed vector, the product vector, and the input image vector into a generative adversarial network.

[0013] Other aspects and possible advantages will become apparent from the following description and claims. [Brief description of the drawings]

[0014] [Figure 1] 1 is a flow chart of an exemplary process according to the present disclosure. [Diagram 2] 2 shows a flowchart of an example process for determining hard-coded limits on amplitudes for scaling directional vectors in the process of FIG. 1. [Diagram 3] 2 is a flow chart illustrating an exemplary process for determining amplitude in the process of FIG. 1 . [Figure 4] 1 is a flowchart illustrating a process for determining a directional vector corresponding to a particular cosmetic treatment. [Diagram 5] 1 illustrates an exemplary embodiment of a computer system that may be used to perform methods according to the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] The method according to the present disclosure begins with an actual image of an external anatomical feature, such as the face, of a human being who is intended to have a medical cosmetic procedure performed on that feature. As the term "medical cosmetic procedure" is used herein, it may include any surgical procedure for changing the structural appearance of an anatomical feature, injection of fillers or other agents to change the structural appearance or to change the appearance of the skin, and any other medical procedure intended to change the color and / or textural appearance of a human anatomical feature.

[0016] The image may be generated using any optical imaging device known in the art to generate a printed image for subsequent digitization, or to generate an initial digitized representation of the color and intensity distribution within an image field or frame. The image in digitized form may be input to a computer or computer system for processing according to the present disclosure. Such an image is referred to for convenience as the "input image."

[0017] Another input to the method according to the present disclosure is the specific medical cosmetic procedure(s) performed on any one or more parts of a human anatomical feature (e.g., face). When the term "aesthetic" is used herein, it is intended to mean any kind of procedure that changes the visual appearance of a human anatomical feature, however administered, including, but not limited to, literal aesthetic procedures as well as medical cosmetic procedures. As explained above, any form of medical procedure intended to change the visual appearance of a human anatomical feature is within the scope of the present disclosure.

[0018] The output of the method according to the present disclosure is an image of the same human anatomical feature (e.g., face) as it would be expected to visually appear after the cosmetic procedure has been performed. In some embodiments, the cosmetic procedure may be an aesthetic surgical procedure. In some embodiments, the cosmetic procedure is an aesthetic filler injection.

[0019] An exemplary embodiment of a process according to the present disclosure may be better understood with reference to the flow chart of FIG.

[0020] In some embodiments, the input image may be pre-processed (block 10 processing). Pre-processing may include: 1. The input image may be converted into a Red, Green, Blue (RGB) intensity distribution (block 12 processing). 2. A human face present in the image frame may be detected (block 14 process), for example, using a common face detection algorithm. Such algorithms are available, for example, as open source code. A predetermined number of facial "landmarks" (68 in this exemplary embodiment) may then be extracted from within the image boundary of the detected face. Landmarks may be understood as characteristics of certain anatomical parts of the human face. Landmarks are described in U.S. Patent Application Publication No. 2021 / 0012097, filed by Velthuis et al. The landmarks may be used to align the image such that the face is geometrically moved to the center of the image frame. 3. The image may be resized (block 16 process) to a selected resolution, e.g., 1024x1024. If necessary, the image is padded to fit the selected resolution. The image may be trimmed or cropped.

[0021] In some embodiments, after preprocessing, the preprocessed image may be encoded (block 18 process). The encoding may be performed using a pre-trained image-to-vector encoder trained to project images, i.e., vector encoding (or transformation), into the latent space of a machine learning algorithm known as a generative adversarial network (GAN). Since the process disclosed herein has as its intended output a manipulated version of an input (actual) image of a human face (i.e., a predicted visual appearance image after processing), one goal of the encoding is to project the actual image (i.e., the input image before processing) into an N×D vector space, which is the GAN latent space. Exemplary embodiments of the encoding process include the following: 1. Train an encoder (neural network) that takes a sample image as input and generates an N×D vector representation as output. This encoder is trained prior to actual processing, but is trained only once and then used for each encoding flow. The preprocessed input image is then encoded using the trained encoder to generate a vector representation in the GAN latent space. The parameters represented by N and D, and the values ​​of N and D, are at the discretion of the method designer and are not intended to limit the scope of the present disclosure. Exemplary parameters are those related to the resolution of the input image and the resolution of the intended output image. In this specification, the N×D vector generated from the input image may be referred to as the "input image vector." 2. Optimization of the encoded image may include fine-tuning the previously encoded vector representation (input image vector) to better reflect the original image (block 20 process). The optimization may be iterative and based on a loss function. In some embodiments, such a loss function may be the mean squared loss between the actual pixel values ​​or extracted higher level features of the preprocessed input image and the image generated from the current vector representation (current input image vector) using a GAN to transform the vector representation back to image space. GANs are described further below. In this exemplary embodiment, a stochastic gradient descent method may be used to direct changes to the current vector representation at each iteration of the optimization in order to minimize the loss function in a relatively small number of iterations. 3. Optimization as in (2) above is performed (a) for a predetermined number of iterations, or (b) until the loss function falls below a predetermined threshold.

[0022] Image manipulation and generation of expected image: After an optimized N×D (input image) vector representation of the input image is generated, manipulation of the optimized input image vector can be performed by simple vector addition of a directional vector to the optimized input image (N×D) vector representation, which may be referred to as a “manipulated image vector” (process block 22). The output of the vector addition, i.e., the manipulated image vector, may be used to generate an output image that emulates the effect of a medical cosmetic procedure. The output image may be generated by inputting the manipulated image vector, as described above, into a machine learning algorithm known as a generative adversarial network (GAN). The GAN generates as its output a manipulated image, which may be stored and / or displayed in any known manner for storing and displaying images. The GAN may also be used in the process of optimizing the generation of the vector representation of the input image, as described above.

[0023] It should be noted that the direction vector only defines a direction in the latent space and functions as a unit vector. An additional input may be the amplitude of the direction vector, which directly changes the magnitude of the image manipulation. The amplitude may be modified (processing block 24) and a new manipulated image may be generated using the GAN (processing block 26). When the amplitude is the final value of the amplitude, as determined below (processing block 28), the process The process stops (block 30 process) and the GAN output image existing at that point is used as the final output image.

[0024] An exemplary procedure for determining the amplitude of the direction vector to generate the manipulated image vector may include the following. 1. Iteratively increase the amplitude of the directional vector, vector-add it to an optimized NxD vector representation of the input image (input image vector), send the manipulated image vector representation to the GAN to generate the manipulated image, and repeat the above process with incrementally increasing amplitude until a final value of amplitude is reached. 2. In various embodiments, the final value of the amplitude can be determined using the following example: a. Referring to FIG. 2, the hard-coded upper amplitude limit can be determined as follows: During the selection of a direction vector for a particular medical cosmetic procedure, the process may use as input an actual image of a person taken before the medical cosmetic procedure (operation of block 32) and as input an actual image of a person taken after the same medical cosmetic procedure (operation of block 34) to form one or more image pairs. In operations of blocks 36 and 38, respectively, each of the pre-treatment and post-treatment images is encoded, for example using the trained encoder described above. A difference vector between the image vectors in each image pair is obtained by subtracting the pre-treatment vector representation from the post-treatment vector representation in each image pair. For each image pair of pre-treatment and post-treatment images, an amplitude value is calculated using the difference vector and the direction vector identified above. The final value of the hard-coded upper limit is then selected using the distribution of amplitude values ​​calculated as described above. (The upper amplitude limit may be determined using the distribution of amplitude values ​​calculated by calculating the mean and standard deviation, and then calculating the mean plus three standard deviations.) b. Referring to FIG. 3, a modification of a landmark may be used: using an input image before the aesthetic procedure is performed (process block 44) and using a manipulated image after the aesthetic procedure is performed (process block 46), the change of the facial landmarks associated with the facial area affected by the aesthetic procedure is evaluated both before and after the procedure is performed. A difference or ratio between the selected landmarks is determined (process block 48), and if the determined difference or ratio reaches a predefined limit (selection of block 50), further modification of the landmarks is terminated and the current amplitude value is used (process block 54). If the determined difference or ratio does not reach a predefined limit (selection of block 50), the landmark may be modified (process block 52) and the process returns to the act of determining the modification of the landmark (process block 48). The aforementioned ratios or differences may be understood to be related to the distances between various specific points of an anatomical feature, e.g., a human face. It will be understood by the medical practitioner that the distances or distance ratios between the specific anatomical features are within empirically determined limits. Such empirically determined limits may be used as limits in the processing of block 50 of FIG. c. Color and texture change: When the change in color and texture of the affected area reaches a predefined limit, the procedure is stopped and the amplitude value at that point is used as the final amplitude value. The predefined limits for landmarks and colors may be obtained, for example, from a facial image dataset using images of human anatomical features before and after treatment. 3. After determining the final value of the amplitude, the resulting manipulated image vector is fed as input to the GAN to generate the final manipulated image (output image). The resulting manipulated image vector is obtained by vector addition of the product of the orientation and amplitude with the original N×D representation of the input image (i.e., the input image vector) as explained above. The resulting manipulated image vector is then processed by the GAN to generate the manipulated image.

[0025] The identity of the direction vector for each type of aesthetic (or medical) procedure may be determined as described with reference to Figures 4A and 4B. , may be determined using vector representations of the actual pre-treatment image and the actual post-treatment image. The directional vectors of the various medical cosmetic procedures so determined may be stored in a database (e.g., as described with reference to FIG. 5 ) for later access when initializing the disclosed process for any new input image. An exemplary embodiment of a process for determining directional vectors may be performed according to one of the following methods: Method 1 (Figure 4A): 1. Each pre-treatment image input in block 56 and each post-treatment image input in block 58 are encoded (transformed, projected) into an NxD vector space in block 60 using the encoding process described above. 2. The directional vector determination is treated as a binary classification in the processing of block 62, with N×D latent vectors for the pre-treatment and post-treatment images representing the negative and positive classes, respectively. 3. A linear decision boundary is determined (step 64), and a vector orthogonal to the decision boundary is determined as the direction of the specific medical cosmetic procedure (step 66). Method 2 (Figure 4B): 1. Each pre-treatment image input in block 56A and each post-treatment image input in block 58A are converted into a vector, e.g., an NxD vector, in block 60A using the encoder described above. 2. A difference vector for each pair of before and after vectors is calculated by subtracting the before vector from the after vector (processing block 62A). 3. Each difference vector is normalized to a unit vector (processing block 64A). 4. A final directional vector is calculated as a statistical aggregation of the unit difference vectors, such as by taking the average or median of the unit difference vectors (processing block 66A).

[0026] It should be noted that while the image encoding and processing described herein is in terms of the entire image, it is within the scope of the present disclosure to process only selected portions of an image to generate a visual representation of the expected appearance of the selected portion of the image. Thus, the term "image" as used herein may refer to all pixels in a frame of an object from which an image is created, or a selected subset thereof. Furthermore, while the present disclosure defines a process for generating a visual representation of an expected appearance corresponding to a single medical cosmetic procedure for a selected anatomical feature, it is within the scope of the present disclosure to process a single input image into multiple subset images, each having a different anatomical feature, and generate an expected visual appearance image for each subset image using a different selected medical cosmetic procedure. Such subset image processing may be performed sequentially or simultaneously, as long as it is not inconsistent with the scope of the present disclosure.

[0027] FIG. 5 illustrates an exemplary embodiment of a computing system 100 that may be used to perform the operations described with reference to FIGS. 1-4. The computing system 100 may be an individual computer system 101A or an arrangement of multiple distributed computer systems. The individual computer system 101A may include one or more analysis modules 102 that may be configured to perform various tasks according to some embodiments, such as the tasks described with reference to FIG. 5. To perform these various tasks, the analysis modules 102 may operate independently or in cooperation with one or more processors 104, which may be connected to one or more storage media 106. A display device, such as any known type of graphical user interface, may be in signal communication with the processor 104, which may enable a user to input commands and / or data and display the results of the execution of a sequence of instructions according to the present disclosure.

[0028] The processor(s) 104 are also connected to a network interface 108. data network 110, allowing individual computer system 101A to communicate with one or more additional individual computer systems and / or computing systems 101B, 101C, and / or 101D, etc. (note that computer systems 101B, 101C, and / or 101D may or may not share the same architecture as computer system 101A. For example, computer systems 101A and 101B may be at a well drilling site while communicating with one or more computer systems, such as 101C and / or 101D, that are located in one or more data centers and / or that are located in various countries on different continents).

[0029] A processor includes, but is not limited to, a microprocessor, a microcontroller, a processor module or subsystem, a programmable integrated circuit, a programmable gate array, or other control or computing device.

[0030] The storage medium 106 may be implemented as one or more computer-readable or machine-readable storage media, e.g., non-transitory storage media. The storage medium may include logic operable to cause a computer system to perform the operations described above with reference to Figures 1-4. It should be noted that while in the exemplary embodiment of Figure 5, the storage medium 106 is shown disposed within the individual computer system 101A, in some embodiments, the storage medium 106 may be distributed within and / or across multiple internal and / or external enclosures of the individual computing system 101A and / or additional computing systems, e.g., 101B, 101C, 101D. The storage medium 106 may include one or more different forms of memory. Examples include, but are not limited to, semiconductor memory devices (such as dynamic or static random access memory (DRAM or SRAM), erasable and programmable read-only memory (EPROM), electrically erasable and programmable read-only memory (EEPROM) and flash memory), magnetic disks (such as fixed disks, floppy disks and removable disks), other magnetic media (such as tapes), compact disks (CDs) or optical media (such as digital video disks (DVDs)), or other types of storage devices. It should be noted that the computer instructions for causing any individual computer system or computing system to perform the tasks described above may be provided on one computer-readable or machine-readable storage medium or may be provided on multiple computer-readable or machine-readable storage media distributed across a multiple component computing system having one or more nodes. Such computer-readable or machine-readable storage medium(s) are considered to be part of an article (or article of manufacture). An article or article of manufacture refers to a manufactured component or components.The storage medium(s) may be located on a machine that executes the machine-readable instructions, or may be located at a remote site from which the machine-readable instructions can be downloaded over a network for execution.

[0031] It should be understood that computing system 100 is but one example of a computing system, and that any other embodiment of a computing system may have more or fewer components than those shown, may combine additional components not shown in the exemplary embodiment of Figure 5, and / or computing system 100 may have a different configuration or arrangement of the components shown in Figure 5. The various components shown in Figure 5 may be implemented in hardware, software, or a combination of both hardware and software, including, for example, one or more signal processing circuits and / or application specific integrated circuits.

[0032] Additionally, the actions of the processing methods described above may be implemented by executing one or more functional modules in an information processing device, such as a general-purpose processor or an application-specific chip (e.g., ASIC, FPGA, PLD), or other suitable device. These modules, combinations of these modules, and / or combinations with common hardware are all within the scope of the present disclosure.

[0033] In light of the principles and exemplary embodiments described and illustrated herein, it will be recognized that the exemplary embodiments may be modified in arrangement and detail without departing from such principles. The foregoing discussion has focused on a particular embodiment, but other configurations are contemplated. In particular, even if expressions such as "in one embodiment" are used herein, these expressions are intended to generally refer to possible embodiments and are not intended to limit the disclosure to a particular embodiment configuration. When used herein, these terms can refer to the same embodiment or different embodiments that can be combined with other embodiments. In principle, any embodiment mentioned herein can be freely combined with any one or more of the other embodiments mentioned herein, and any number of features of different embodiments can be combined with each other, unless otherwise indicated. Although several examples have been described in detail above, those skilled in the art will easily understand that many modifications are possible within the scope of the described examples. Therefore, all such modifications are intended to be included within the scope of the present disclosure, as defined in the following claims.

Claims

1. 1. A method for generating an image of a human anatomical feature representative of a predicted visual appearance of the human anatomical feature after performance of a medical cosmetic procedure, the method comprising: transmitting the image of the human anatomical feature to a computer as an input image prior to performing a medical cosmetic procedure; converting, at the computer, the input image before the medical cosmetic procedure into an input image vector; selecting, on the computer, a direction vector corresponding to the medical cosmetic procedure, the direction vector being determined by: converting each of a plurality of corresponding images of the anatomical feature before the medical cosmetic procedure to a first input image vector; converting each of a plurality of corresponding images of the anatomical feature after the medical cosmetic procedure to a second input image vector; calculating a difference vector by subtracting the second input image vector from the first input image vector; normalizing the difference vector to a unit vector; and performing a statistical aggregation of the difference vectors for the plurality of corresponding images before the medical cosmetic procedure and after the medical cosmetic procedure; determining, in the computer, an amplitude of the direction vector; multiplying, at the computer, the direction vector by the determined amplitude to obtain a product vector; vector-adding, in the computer, the product vector to the input image vector to obtain a summed vector; generating, in the computer, an output image from the summed vector, the product vector, and the input image vector, the output image corresponding to the expected visual appearance; The method includes:

2. Optimizing the conversion on the computer includes: (a) generating an optimized image from the input image vector; (b) comparing the optimized image with the input image and calculating a loss function; (c) modifying the input image vector and repeating the generation of the optimized image; (d) repeating (a), (b), and (c) until (i) the value of the loss function falls below a predefined threshold or (ii) (a), (b), and (c) have been repeated a selected number of times; The method of claim 1 , further comprising:

3. The determining of the amplitude comprises: (a) converting pre-processing images of corresponding human anatomical features into input vectors; (b) initially estimating an amplitude and combining the initially estimated amplitude with the direction vector to generate an intermediate product vector; (c) adding the intermediate product vector to the transformed input vector; (d) generating an output image from the intermediate product vector; (e) determining changes in landmark position, color, and texture between the pre-treatment image and the output image; (f) adjusting the amplitude and repeating (b), (c), (d), and (e) until a change in at least one of the landmarks, the color, and the texture reaches a predefined threshold or the amplitude reaches a predefined upper limit; The method of claim 1 , comprising:

4. The determining of the amplitude comprises: (a) converting pre-treatment images and corresponding post-treatment images of corresponding human anatomical features into vectors; (b) calculating a difference vector between each transformed pre-processing image and each corresponding transformed post-processing image; (c) obtaining an amplitude value by calculating the amplitude of each difference vector; (d) selecting an upper amplitude limit using the distribution of amplitude values ​​calculated in (c); and The method of claim 1 , comprising:

5. The method of claim 1 , wherein generating the output image comprises inputting the summed vector, the product vector, and the input image vector into a generative adversarial network.

6. 1. A computer program stored on a non-transitory computer readable medium for generating an image of a human anatomical feature representative of a predicted visual appearance of the human anatomical feature after performance of a medical cosmetic procedure, the program being configured to cause a programmable computer to: receiving an image of the human anatomical feature as an input image to the computer prior to performing the medical cosmetic procedure; converting, at the computer, the input image into an input image vector; selecting, on the computer, a direction vector corresponding to the medical cosmetic procedure, the direction vector being determined by: converting each of a plurality of corresponding images of the anatomical feature before the medical cosmetic procedure to a first input image vector; converting each of a plurality of corresponding images of the anatomical feature after the medical cosmetic procedure to a second input image vector; calculating a difference vector by subtracting the second input image vector from the first input image vector; normalizing the difference vector to a unit vector; and performing a statistical aggregation of the difference vectors for the plurality of corresponding images before the medical cosmetic procedure and after the medical cosmetic procedure; determining, in the computer, an amplitude of the direction vector; multiplying, at the computer, the direction vector by the determined amplitude to obtain a product vector; and in the computer, vector-adding the product vector to the input image vector to obtain a summed vector; generating, in the computer, an output image from the summed vector, the product vector, and the input image vector, the output image corresponding to the expected visual appearance; A computer program having logic operable to cause the computer to perform operations including:

7. (a) generating an optimized image from the input image vector; (b) comparing the optimized image with the input image and calculating a loss function; (c) modifying the input image vector and repeating the generation of the optimized image; (d) repeating (a), (b), and (c) until (i) the value of the loss function falls below a predefined threshold or (ii) (a), (b), and (c) have been repeated a selected number of times; The computer program product of claim 6 , further comprising logic operable to cause the computer to optimize the transformation by:

8. The determining of the amplitude comprises: (a) converting pre-processing images of corresponding human anatomical features into input vectors; (b) initially estimating amplitudes and combining the initially estimated amplitudes with the input vector to generate an intermediate product vector; (c) vector-adding the intermediate product vector to the transformed input vector; (d) generating an output image from the intermediate product vector; (e) determining changes in landmark position, color, and texture between the pre-treatment image and the output image; (f) adjusting the amplitude and repeating (b), (c), (d), and (e) until a change in at least one of the landmarks, the color, and the texture reaches a predefined threshold or the amplitude reaches a predefined upper limit; 7. The computer program of claim 6, comprising:

9. The determining the amplitude for the medical cosmetic procedure includes: (a) converting pre-treatment images and corresponding post-treatment images of corresponding human anatomical features into vectors; (b) calculating a difference vector between each transformed pre-processing image and each corresponding post-processing transformed image; (c) obtaining an amplitude value by calculating the amplitude of each difference vector; (d) selecting an upper amplitude limit using the distribution of amplitude values ​​calculated in (c); and 7. The computer program of claim 6, comprising:

10. 7. The computer program product of claim 6, wherein generating the output image comprises inputting the summed vector, the product vector, and the input image vector to a generative adversarial network.

11. 1. A method for generating an image of a human anatomical feature representative of a predicted visual appearance of the human anatomical feature after performance of a medical cosmetic procedure, the method comprising: transmitting the image of the human anatomical feature to a computer as an input image prior to performing a medical cosmetic procedure; converting, at the computer, the input image before the medical cosmetic procedure into an input image vector; selecting, at the computer, a directional vector corresponding to the medical cosmetic procedure; determining, in the computer, an amplitude of the direction vector; multiplying, at the computer, the direction vector by the determined amplitude to obtain a product vector; and in the computer, vector-adding the product vector to the input image vector to obtain a summed vector; generating, in the computer, an output image from the summed vector, the product vector, and the input image vector, the output image corresponding to the expected visual appearance; Including, The directional vector for the medical cosmetic procedure is: Transforming at least one image of the anatomical feature before the medical cosmetic procedure into a first input image vector; Transforming at least one image of the anatomical feature after the medical cosmetic procedure into a second input image vector; treating the first input image vector and the second input image vector as a binary classification, with at least one image of the anatomical feature before the medical cosmetic procedure and at least one image of the anatomical feature after the medical cosmetic procedure representing a negative class and a positive class, respectively; establishing a linear decision boundary between the negative and positive classes and determining a vector orthogonal to the linear decision boundary; The method is determined by:

12. The determining of the amplitude comprises: (a) converting pre-processing images of corresponding human anatomical features into input vectors; (b) initially estimating an amplitude and combining the initially estimated amplitude with the direction vector to generate an intermediate product vector; (c) adding the intermediate product vector to the transformed input vector; (d) generating an output image from the intermediate product vector; (e) determining changes in landmark position, color, and texture between the pre-treatment image and the output image; (f) adjusting the amplitude and repeating (b), (c), (d), and (e) until a change in at least one of the landmarks, the color, and the texture reaches a predefined threshold or the amplitude reaches a predefined upper limit; The method of claim 11 , comprising:

13. The determining of the amplitude comprises: (a) converting pre-treatment images and corresponding post-treatment images of corresponding human anatomical features into vectors; (b) calculating a difference vector between each transformed pre-image and each corresponding transformed post-image; (c) obtaining an amplitude value by calculating the amplitude of each difference vector; (d) selecting an upper amplitude limit using the distribution of amplitude values ​​calculated in (c); and The method of claim 11 , comprising:

14. 1. A computer program stored on a non-transitory computer readable medium for generating an image of a human anatomical feature representative of a predicted visual appearance of the human anatomical feature after performance of a medical cosmetic procedure, the program being configured to cause a programmable computer to: receiving an image of the human anatomical feature as an input image to the computer prior to performing the medical cosmetic procedure; converting, at the computer, the input image into an input image vector; selecting, at the computer, a directional vector corresponding to the medical cosmetic procedure; determining, in the computer, an amplitude of the direction vector; multiplying, at the computer, the direction vector by the determined amplitude to obtain a product vector; and in the computer, vector-adding the product vector to the input image vector to obtain a summed vector; generating, in the computer, an output image from the summed vector, the product vector, and the input image vector, the output image corresponding to the expected visual appearance; Including, The directional vector for the medical cosmetic procedure is: Transforming at least one image of the anatomical feature before the medical cosmetic procedure into a first input image vector; Transforming at least one image of the anatomical feature after the medical cosmetic procedure into a second input image vector; treating the first input image vector and the second input image vector as a binary classification, with at least one image of the anatomical feature before the medical cosmetic procedure and at least one image of the anatomical feature after the medical cosmetic procedure representing a negative class and a positive class, respectively; establishing a linear decision boundary between the negative and positive classes and determining a vector orthogonal to the linear decision boundary; A computer program having logic operable to cause the computer to perform operations including:

15. The determining of the amplitude comprises: (a) converting pre-processing images of corresponding human anatomical features into input vectors; (b) initially estimating amplitudes and combining the initially estimated amplitudes with the input vector to generate an intermediate product vector; (c) vector-adding the intermediate product vector to the transformed input vector; (d) generating an output image from the intermediate product vector; (e) determining changes in landmark position, color, and texture between the pre-treatment image and the output image; (f) adjusting the amplitude and repeating (b), (c), (d), and (e) until a change in at least one of the landmarks, the color, and the texture reaches a predefined threshold or the amplitude reaches a predefined upper limit; 15. The computer program of claim 14, comprising:

16. The determining the amplitude for the medical cosmetic procedure includes: (a) converting pre-treatment images and corresponding post-treatment images of corresponding human anatomical features into vectors; (b) calculating a difference vector between each transformed pre-processing image and each corresponding post-processing transformed image; (c) obtaining an amplitude value by calculating the amplitude of each difference vector; (d) selecting an upper amplitude limit using the distribution of amplitude values ​​calculated in (c); and 15. The computer program of claim 14, comprising:

Images