KR102542016B1 - Method and apparatus for pseudonymizing medical image - Google Patents

Abstract

The present invention relates to a method and apparatus for pseudonymizing medical images. A pseudonymization method according to an embodiment of the present invention is a method of pseudonymizing a medical image performed by an electronic device, comprising the steps of distinguishing an identified part from a non-identified part of a target body part in a medical image of a target body part of a subject. ; and generating a pseudonymized image by mapping the non-identified part to an image (processed image) of a corresponding target body part of another target person.

Description

Medical image pseudonymization method and apparatus {METHOD AND APPARATUS FOR PSEUDONYMIZING MEDICAL IMAGE}

The present invention relates to medical image processing, and more particularly, to a method and apparatus for pseudonymizing medical images.

Recently, a health care big-data analysis technology using medical information (or health information) is in the limelight. To implement these healthcare big data analysis technologies, a large amount of medical information must first be collected. However, in relation to the collection of medical information, a new problem of conflict with laws related to personal information protection is emerging.

Accordingly, various pseudonymization techniques for de-identification processing by pseudonymizing personal information in medical information have been proposed. At this time, pseudonymization refers to the process of assigning a random code or number (pseudonym) to personal information so that it is impossible to identify a specific individual without combining the use of additional information to restore it to its original state. . In addition, the pseudonymized information is referred to as "pseudonym information".

Conventional pseudonymization technology for medical information was only a pseudonymization technology for text-type personal information related to personal information such as name, age, gender, and phone number. In addition to personal information in the form of tests, medical images in the form of images may also be recognized as personal information in some cases. However, until now, there has been no pseudonymization technology for such medical images, that is, technology for pseudonymizing medical images to generate pseudonymized images, which are pseudonymous information.

Meanwhile, in order to pseudonymize the medical image, a technique of applying a simple image processing technique such as masking or blurring to a specific region in the medical image may be proposed. However, in the case of this proposed technology, there is a problem in that the reality of the pseudonymized image itself obtained as a result of image processing is poor. For example, when a medical image of a human face is pseudonymized using the proposed technology, the pseudonymized image generated is recognized by medical personnel (such as doctors) as not being an image of a human face, or is significantly deformed so that it is unlikely to actually exist. can be recognized as an image of a human face.

However, the matters described as the background art above are only for the purpose of improving understanding of the background of the present invention, and should not be taken as acknowledging that they correspond to prior art already known to those skilled in the art. It won't.

JP 2011-508332 A

In order to solve the above problems, the present invention proposes a new technology for pseudonymizing medical images, thereby pseudonymizing medical images that can prevent issues related to laws related to personal information protection when using medical images. Its purpose is to provide a method and apparatus.

Another object of the present invention is to provide a pseudonymization method and apparatus for generating a pseudonymized image that maintains reality, that is, that can be recognized as actually existing by a medical person (doctor, etc.).

However, the problem to be solved by the present invention is not limited to the above-mentioned problems, and other problems not mentioned can be clearly understood by those skilled in the art from the description below. There will be.

To solve the above problems, a pseudonymization method according to an embodiment of the present invention is a method for pseudonymizing a medical image performed by an electronic device, and includes an identification part of a target body part in a medical image of a target body part of a subject. and separating the non-identified part; and generating a pseudonymized image by mapping the non-identified part to an image (processed image) of a corresponding target body part of another target person.

The processed image may be a fake image that has a body structure of the corresponding target body part and is recognized as a corresponding target body part of another target person.

The processed image may be a synthesized image using medical images of corresponding target body parts of various other subjects.

The processed image or the pseudonymized image may be generated using a pre-learned machine learning model.

The machine learning model may be learned through machine learning of a generative adversarial network (GAN) algorithm.

In the classifying step, the identification part may be adjusted according to a user's selection.

In the classifying step, the target body part of the subject is divided into a plurality of subregions according to weight ranges of importance, and types of subregions included in the identification part may be adjusted according to the applied weight.

In the classifying, each partial area in the medical image may be displayed in various colors.

The subregion may include a first subregion having the highest weight range, a second subregion having a lower weight range than the first subregion, and a third subregion having the lowest weight range.

In the distinguishing, the identification part may be selected from at least a first partial area, and the non-identified part may be selected from at least a third partial area.

In the classifying step, as the applied weight decreases, the second partial area in addition to the first partial area may also be included in the identification part.

In the dividing step, a plurality of spaced apart contour points among the contour points constituting the contour of the target body part may be divided into identification parts.

The pseudonymization method according to an embodiment of the present invention may further include storing information on the identification part as meta information.

The pseudonymization method according to an embodiment of the present invention may further include selecting whether or not to transmit the meta information when transmitting the pseudonymized video to another electronic device.

A pseudonymization device according to an embodiment of the present invention includes a memory storing a medical image of a target body part of a subject; and a controller for controlling pseudonymization of the medical image, wherein the controller distinguishes between an identified part and a non-identified part among the target body parts of the medical image, and the non-identified part is classified as a corresponding part of another subject. A pseudonymized image is generated by mapping to an image (processed image) of a target body part.

A pseudonymization server according to an embodiment of the present invention is a server for pseudonymizing a medical image of a target body part of a subject received from a terminal, receives the medical image from the terminal, and transmits the pseudonymized image to the terminal. a communication unit that transmits; and a control unit controlling the pseudonymization process, wherein the control unit distinguishes between an identified part and a non-identified part among the target body parts of the medical image, and identifies the non-identified part as a corresponding target body part of another subject. A pseudonymized image is generated by mapping to an image (processed image).

The present invention configured as described above has an advantage of preventing issues that violate laws related to personal information protection when using medical images by proposing a new technology for pseudonymizing medical images.

In addition, the present invention has an advantage of being able to generate a pseudonymized image that maintains reality, that is, can be recognized as actually existing by a medical person (doctor, etc.).

In addition, the present invention has an advantage in that the learning time and the pseudonymization processing time can be significantly reduced when a pseudonymized image is generated by applying the GAN algorithm.

The effects obtainable in the present invention are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below. will be.

1 shows a block diagram of a pseudonymization device 100 according to an embodiment of the present invention.
2 shows a flowchart of a pseudonymization method according to an embodiment of the present invention.
3 to 11 show examples of various medical images.
12 shows an example of a medical image from which a shape of an occlusal surface of a tooth is extracted.
13 and 14 show conceptual diagrams of machine learning models of GAN algorithms.
15 is a flowchart of a pseudonymization method according to another embodiment of the present invention.

The above objects and means of the present invention and the effects thereof will become clearer through the following detailed description in conjunction with the accompanying drawings, and accordingly, those skilled in the art to which the present invention belongs can easily understand the technical idea of the present invention. will be able to carry out. In addition, in describing the present invention, if it is determined that a detailed description of a known technology related to the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted.

Terms used in this specification are for describing the embodiments and are not intended to limit the present invention. In this specification, singular forms also include plural forms in some cases unless otherwise specified in the text. In this specification, terms such as "comprise", "have", "provide" or "have" do not exclude the presence or addition of one or more other elements other than the mentioned elements.

In this specification, terms such as “or” and “at least one” may represent one of the words listed together, or a combination of two or more. For example, "or B" and "at least one of B" may include only one of A or B, or may include both A and B.

In this specification, descriptions following "for example" may not exactly match the information presented, such as cited characteristics, variables, or values, and tolerances, measurement errors, limits of measurement accuracy and other commonly known factors It should not be limited to the embodiments of the present invention according to various embodiments of the present invention with effects such as modifications including.

In this specification, when a component is described as being 'connected' or 'connected' to another component, it may be directly connected or connected to the other component, but there may be other components in the middle. It should be understood that it may be On the other hand, when a component is referred to as 'directly connected' or 'directly connected' to another component, it should be understood that no other component exists in the middle.

In the present specification, when an element is described as being 'on' or 'in contact with' another element, it may be in direct contact with or connected to the other element, but another element may be present in the middle. It should be understood that On the other hand, if an element is described as being 'directly on' or 'directly in contact with' another element, it may be understood that another element in the middle does not exist. Other expressions describing the relationship between components, such as 'between' and 'directly between', can be interpreted similarly.

In this specification, terms such as 'first' and 'second' may be used to describe various elements, but the elements should not be limited by the above terms. In addition, the above terms should not be interpreted as limiting the order of each component, and may be used for the purpose of distinguishing one component from another. For example, a 'first element' may be named a 'second element', and similarly, a 'second element' may also be named a 'first element'.

Unless otherwise defined, all terms used in this specification may be used in a meaning commonly understood by those of ordinary skill in the art to which the present invention belongs. In addition, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless explicitly specifically defined.

Hereinafter, a preferred embodiment according to the present invention will be described in detail with reference to the accompanying drawings.

1 shows a block diagram of a pseudonymization device 100 according to an embodiment of the present invention.

The pseudonymization device 100 according to an embodiment of the present invention is a device that pseudonyms pre-stored medical images, and may be an electronic device capable of computing or a computing network. The pseudonymization device 100 may operate as a server or the like for pseudonymization processing. That is, in the case of a server, the pseudonymization apparatus 100 may receive a medical image from a terminal storing the medical image, pseudonymize the received medical image, and generate a pseudonymized image (hereinafter referred to as “pseudonymized image”). ”) may be operated as a server or the like that transmits data to a corresponding terminal or other terminals. In this case, the terminal may be an electronic device that measures a medical image or an electronic device that previously stores medical images measured by the corresponding measuring device.

On the other hand, a medical image is an image of a target body part (treatment target part, etc.) of a subject (patient to be treated, etc.), hard tissue (bone, cartilage, etc.) or soft tissue (muscle, fascia, tendon, ligament, joint capsule, skin, fat) etc.) can be included. For example, a medical image is a 2-dimensional or 3-dimensional image using ultrasound or radiation to identify the inside of a body, such as an ultrasound image, an X-ray image, a computed tomography (CT) image, or a magnetic resonance image (MRI) image. can In addition, the medical image may be a 2D or 3D optical image for recognizing internal and external surfaces of the body, such as a clinical image of the surface of an external part of the body (face, etc.) or an oral scan image of the internal surface of the oral cavity. Such an optical image is an image obtained by photographing the surface of a part of a subject to be treated with an optical camera or the like, and may be an image used to check the surface of a part of a subject to be treated in plastic surgery or dentistry. Since it is an image for , it may include an image for the skin surface among soft tissues, but is not limited thereto. Also, the medical image may be an image of a target body part in various directions (front, back, plane, side, etc.). However, the medical image is not limited to the above-mentioned types, and may include various images of the body used during treatment (or medical practice) such as diagnosis, correction, and surgery.

For example, the electronic device includes a desktop personal computer, a laptop personal computer, a tablet personal computer, a netbook computer, a workstation, and a personal digital assistant (PDA). , a smart phone (smartphone), a smart pad (smartpad), or a mobile phone (mobile phone), etc., but is not limited thereto.

As shown in FIG. 1 , the pseudonymization device 100 may include an input unit 110, a communication unit 120, a display 130, a memory 140, a controller 150, and the like.

The input unit 110 generates input data in response to various user inputs and may include various input means. For example, the input unit 110 may receive input from the user to distinguish an identified part from a non-identified part among target body parts of a medical image or to adjust an identified part. For example, the input unit 110 includes a keyboard, a key pad, a dome switch, a touch panel, a touch key, a touch pad, and a mouse. (mouse), menu button (menu button), etc. may be included, but is not limited thereto.

The communication unit 120 is a component that performs communication with other devices such as terminals. For example, the communication unit 120 may receive a medical image from a terminal or transmit a pseudonymized image generated by the controller 150 to the corresponding terminal or another terminal. Also, the communication unit 120 may receive information on a pre-learned machine learning model from another device. For example, the communication unit 120 may perform 5th generation communication (5G), long term evolution-advanced (LTE-A), long term evolution (LTE), Bluetooth, bluetooth low energy (BLE), near field communication (NFC), Wireless communication such as WiFi communication or wired communication such as cable communication may be performed, but is not limited thereto.

The display 130 displays various image data on a screen. For example, the display 130 may display a medical image, a pseudonymization process for the medical image, and a pseudonymized image. The display 130 may be composed of a non-emissive panel or a light emitting panel. For example, the display 130 may include a liquid crystal display (LCD), a light emitting diode (LED) display, an organic LED (OLED) display, and a micro electromechanical system (MEMS). mechanical systems) display, or electronic paper (electronic paper) display, etc. may be included, but is not limited thereto. Also, the display 130 may be combined with the input unit 120 and implemented as a touch screen or the like.

The memory 140 stores various types of information necessary for the operation of the pseudonymization device 100 . For example, stored information may include medical images, pseudonymized images, machine learning models, program information related to a pseudonymization method to be described later, and the like. For example, the memory 140 may be of a hard disk type, a magnetic media type, a compact disc read only memory (CD-ROM), or an optical media type according to its type. ), Sagneto-optical media type, Sultimedia card micro type, flash memory type, read only memory type, or RAM type (random access memory type), etc., but is not limited thereto. In addition, the memory 140 may be a cache, a buffer, a main memory device, an auxiliary memory device, or a separately provided storage system depending on its purpose/location, but is not limited thereto.

The controller 150 may perform various control operations of the pseudonymization device 100 . That is, the control unit 150 may control the execution of a pseudonymization method to be described later, and the remaining components of the pseudonymization device 100, that is, the input unit 110, the communication unit 120, the display 130, and the memory 140 etc. can be controlled. For example, the control unit 150 may include, but is not limited to, a processor that is hardware or a process that is software that is executed in the corresponding processor.

2 shows a flowchart of a pseudonymization method according to an embodiment of the present invention.

Referring to FIG. 2 , the pseudonymization method according to an embodiment of the present invention may include steps S101 and S102. At this time, the execution of S101 and S102 may be controlled through hardware or software of the control unit 150, respectively.

First, S101 is a step of distinguishing an identified part and a non-identified part among target body parts in a medical image.

In this case, the identification part refers to a feature part of a target body part of a corresponding subject to be provided to a medical person (doctor, etc.) as a minimum for treatment of the subject, among target body parts of the medical image. That is, by grasping the corresponding identification part, the medical person can grasp the subject's condition for treatment such as diagnosis, correction, surgery, and the like. Of course, since the identification part is only a very small part of the entire medical image, it is difficult to identify a specific individual only with the identification part. On the other hand, the non-identified part refers to a part of the target body part of the medical image that does not need to be provided to the medical person because the medical person does not need to refer to it when treating the subject. That is, the non-identified part may be the remaining part of the target body part of the medical image that is not set as an identification part. In addition, even if the non-identified part is not a body part in the medical image, it may be an object part, such as a prosthesis, that affects the target body part among the remaining parts not set as identification parts.

3 to 11 show examples of various medical images. That is, FIG. 3 is an X-ray image (cephalo image) of the lateral hair head, FIG. 4 is an X-ray image of the hairy head, FIG. 5 is a clinical image of the lateral hair head, and FIG. 6 is a clinical image of the hairy head. 7 is a clinical image of the frontal mouth area during a smile, FIG. 8 is an X-ray image (panoramic image) of the frontal area around the mouth, FIG. 9 is a frontal intraoral surface image, and FIG. A lateral intraoral surface image, FIG. 11 shows an intraoral surface image of the other lateral surface, respectively. Also, FIG. 12 shows an example of a medical image from which the shape of an occlusal surface of a tooth is extracted. However, in FIGS. 3 to 12 , dots of various colors are anatomical landmarks for various body parts (ie, one or more dots, lines, or planes representing positions or shapes of body parts).

These identification parts and non-identification parts may be classified in various ways according to the type of target body part, the type (purpose) of treatment, and the type of medical image. In particular, the identification part and the non-identification part may be automatically extracted (hereinafter referred to as “automatic extraction processing”) through various algorithms. In addition, the input unit 110 may receive input from the user about which part of the target body part of the medical image to be set as the identification part, and the controller 150 may adjust the identification part according to the corresponding input. The adjustment of the identification part according to the user's selection may be additionally performed after the automatic extraction process.

3 to 12, the control unit 150 may use landmarks for various body parts in medical images or landmarks for object parts such as prosthesis (ie, locations of object parts) during automatic extraction processing. Alternatively, one or more points, lines, or planes to represent the shape may be displayed (hereinafter, landmarks for these body parts or object parts are simply referred to as “landmarks”). Of course, these landmarks can be set by the user through the input unit 110 when the automatic extraction process is not performed or after the automatic extraction process. At this time, it may be displayed in various colors according to the automatic extraction process or the user's selection according to the body part or the importance of the body part.

Referring to FIG. 3, in the X-ray image of the temporal head, each point is a landmark that can be referred to during dental treatment, red and orange dots represent hard tissue landmarks, and green dots represent soft tissue landmarks. indicate In this case, red dots may be landmarks of high importance in dental treatment, orange dots may be landmarks of medium importance in dental treatment, and green dots may be landmarks of low importance in dental treatment. That is, a weight is set for each landmark according to its importance, and various colors may be set according to the weight range.

As described above, the control unit 150 divides the subject's target body part into a plurality of partial regions (any region of the body parts of the medical image) according to the weight range of importance through automatic extraction processing or user setting, and each part Areas can be displayed in different colors. Also, the controller 150 may adjust the type of partial area included in the identification part according to the applied weight. That is, the user may select the applied weight through the input unit 110 so that a partial area having a weight range of a certain range or more is included in the identification part, and in this case, the controller 150 identifies a landmark included in the corresponding weight range. It can be set in parts.

For example, in FIG. 3 , a first subregion having the highest weight range is red dots, a second subregion having a lower weight range than the first subregion is orange dots, and a third subregion having the lowest weight range. may be the area of all other body parts including the green dots. In this case, as the applied weight selected by the user increases, only the first partial area may be set as the identification portion.

Conversely, as the applied weight decreases, the second partial area in addition to the first partial area may be set as the identification part. However, since the green dot is a landmark for the soft tissue of the outer surface of the side hair, even if the applied weight is set to the lowest, it may not be set as an identification part but a non-identification part according to the purpose of dental treatment. In this case, at least red dots may be selected as identification parts, and at least green dots may be selected as non-identification parts. However, the present invention is not limited to the above examples.

Meanwhile, in FIGS. 4 to 6 , red dots represent landmarks for each body part in the top or side hairs. In Fig. 7, red dots represent landmarks for the teeth from the front, and yellow dots represent landmarks for the lips from the front. 8 to 11, red dots represent landmarks for the upper teeth from the front or side, and yellow dots represent landmarks for the lower teeth from the front or side. That is, each landmark is not only displayed in various colors according to the weight of its importance according to automatic extraction process or user setting, but also various colors according to the type of the body part as shown in FIGS. 7 to 11 can also be displayed as Also, among body parts, landmarks for hard tissues and landmarks for soft tissues may be displayed in various colors different from each other. In addition, landmarks for body parts and landmarks for object parts may also be displayed in various colors different from each other. According to these various color displays, a user of a medical person can more easily select an identification part according to the type of target body part or the type (purpose) of treatment.

For reference, in FIG. 5, details of each landmark are as follows.

number detail part number detail part One Glabella 12 Upper embrasure 2 Soft tissue Nasion 13 Lower embrasure 3 Nasal bridge 14 Stomion inferius 4 Dorsum of Nose 15 Lower lip 5 Pronasale 16 Labrale inferius 6 Columella 17 Soft tissue B point 7 Subnasale 18 Soft tissue pogonion 8 Soft tissue A point 19 Soft tissue gnathion 9 Labrale superius 20 Soft tissue Menton 10 Upper lip 21 Submandibular point 11 Stomion superius 22 Cervical point

Meanwhile, among the contour points constituting the contour of the target body part, a plurality of spaced apart contour points may be set as an identification part according to automatic extraction process or user's setting. Of course, the rest of the target body part may be set as the tooth non-identification part.

That is, referring to FIGS. 7 to 11 , when a target body part of a medical image includes a tooth for dental treatment, a plurality of spaced apart contour points forming the contour of the corresponding tooth are automatically extracted or set by the user. It can be set as an identification part according to. In addition, the remaining part of the corresponding tooth may be set as a tooth non-identification part. For example, the identification part for one tooth may include four spaced contour points, and each point may be a point representing the top, bottom, left, and right positions of the corresponding tooth. However, the present invention is not limited to the above examples.

Meanwhile, for automatic extraction processing, the controller 150 may include various algorithms for setting and displaying landmarks according to the target body part type, treatment type (purpose), medical image type, and the like.

In particular, the controller 150 may set a corresponding landmark through a machine learning model. In this case, the machine learning model may be a model learned according to a machine learning technique of supervised learning through training data of input data and output data pairs (datasets). That is, the machine learning model may be learned using a training data set including input data of a medical image and output data in which landmarks are set in the corresponding medical image. A corresponding landmark of the output data may be set by a medical person or the like according to the type of target body part, type of treatment (purpose), type of medical image, and the like.

Accordingly, the machine learning model has a function for a relationship between a medical image, which is input data, and a landmark (or a medical image in which a landmark is set), which is output data, and is expressed using various parameters. For example, a machine learning model may express a relationship between a medical image and a landmark using parameters such as weights and biases. That is, when input data of a medical image is input to a pre-learned machine learning model, output data of a landmark (or a medical image in which a landmark is set) according to a corresponding function may be output.

For example, machine learning techniques may include Artificial neural network, Boosting, Bayesian statistics, Decision tree, Gaussian process regression, Nearest neighbor algorithm, Support vector machine, Random forests, Symbolic machine learning, Ensembles of classifiers, or Deep learning. It may, but is not limited thereto.

In particular, deep learning models learned by deep learning techniques express the relationship between input data and output data as multiple layers (layers), and these multiple expression layers are referred to as “neural networks.” do it too Such a deep learning model may have impressive performance in the field of image processing such as the present invention.

For example, deep learning techniques include Deep Neural Network (DNN), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Restricted Boltzmann Machine (RBM), Deep Belief Network (DBN), Deep Q-Networks, etc. It can be done, but is not limited thereto.

Thereafter, in S102, a pseudonymized image is generated by mapping the non-identified part to an image of a corresponding target body part of another subject (hereinafter, referred to as “processed image”) and setting a pseudonymized region. That is, in the pseudonymized image, the pseudonymized region refers to the region of the non-identified portion replaced with the processed image. In this case, the identification part may not be subjected to image processing, and the original value of the medical image may be maintained.

That is, each of the identification part and the non-identification part includes one or more pixels, and for each corresponding pixel, a position value and a pixel value (color value, etc.) are stored as pixel information. At this time, pixels corresponding to the non-identification part may be replaced with corresponding pixels corresponding to the processed image, whereas pixels corresponding to the identification part are not replaced with corresponding pixels corresponding to the processed image, and the original landmark characteristics (eg, its position value, or its position value and pixel value). Information on the landmark of the identification part (in particular, information on the location value) may be separately stored in the memory 140 together with the pseudonymized image as meta information. When the pseudonymization device 100 operates as a server or the like, such meta information may be selectively transmitted when a pseudonymized image is transmitted to a terminal or the like.

Meanwhile, as a method of generating a pseudonymized image, various image processing techniques or machine learning models may be used. The pseudonymized image generated through this includes the synthesized processed region and the identification portion in which the characteristics of the original landmark are maintained, but must be an image recognized by medical personnel (such as a doctor) as an image of the target body part. That is, the pseudonymized image must be a fake image that has the body structure of the corresponding target body part and is recognized as the corresponding target body part of another target. For example, a pseudonymized image of an A-shaped tooth must be an image of a completely different B-shaped tooth, but must include the landmark characteristics of the identification portion of the A-shaped tooth as it is.

For example, when an image processing technique is used, corresponding target body parts are synthesized from previously stored medical images of other subjects, but only non-identified parts are synthesized while maintaining the landmark characteristics of the identified parts of the subject's medical images. can Such synthesis may be performed using a plurality of medical images of a subject other than one.

13 and 14 show conceptual diagrams of machine learning models of GAN algorithms.

In addition, when a machine learning model is used, a pseudonymized image may be generated using a machine learning model learned by an unsupervised machine learning technique in addition to a supervised method. In particular, as shown in FIGS. 13 and 14 , a pseudonymized image may be generated using a machine learning model of a generative adversarial network (GAN) algorithm.

This GAN algorithm is a method using two predefined network models, a generator (G) and a classifier (D). That is, the classifier (D) is trained first, then the generator (G) is trained, and the generator (G) and the classifier (D) compete with each other to learn little by little. . At this time, in the case of the classifier (D), after receiving actual medical images (real data) and learning to classify the corresponding medical images as real (real), on the contrary, synthetic medical images (fake data) generated by the generator (G) ) and can be learned to classify the corresponding medical image as synthetic (fake). In addition, in the case of the generator (G), a medical image and the landmark of the identification part are input, and the landmark characteristics of the identification part are left as they are, and the pseudonymization area is recognized as the corresponding target body part of another subject. Synthetic medical It can be trained to generate an image as a pseudonymized image. At this time, the fake data generated by the generator (G) is input to the discriminator (D), and the generator (G) can be trained to produce data similar to real data enough to classify the fake data as genuine. However, unlike the prior art, since the processed image generated by the generator G only needs to be recognized as an image of the target body part by a medical person (doctor, etc.), the learning level does not need to be too high. Therefore, when the GAN algorithm is applied to the present invention, there is an advantage in that the learning time and the pseudonymization processing time can be significantly reduced.

For example, GAN algorithms include DCGAN (Deep Convolutional GAN), LSGAN (Least Squares GAN), SGAN (Semi-Supervised GAN), ACGAN (Auxiliary Classifier GAN), WGAN (Wasserstein Generative Adversarial Networks, ConGAN (Continuous GAN), conditional GAN (cGAN), spectral normalization conditional GAN (SNcGAN), starGAN, etc., but is not limited thereto.

15 is a flowchart of a pseudonymization method according to another embodiment of the present invention.

Meanwhile, referring to FIG. 15 , the pseudonymization method according to another embodiment of the present invention may include steps S201 to S204. That is, steps S201 to S204 may be processes when the pseudonymization device 100 operates as a server or the like. At this time, the execution of S201 to S204 may be controlled through hardware or software of the control unit 150, respectively.

First, S201 is a step of receiving a medical image from a terminal or the like. That is, the terminal may transmit its own medical image to the pseudonymization device 100, which is a server that performs pseudonymization processing. In this case, the control unit 150 may control to receive the corresponding medical image through the communication unit 120 and control to store the received medical image in the memory 140 .

After that, S202 and S203 are performed. However, since S202 and S203 correspond to the above-described S101 and S102, respectively, a description thereof will be omitted.

Thereafter, S204 is a step of transmitting the pseudonymized image processed through S202 and S203 to a corresponding terminal or another terminal. At this time, the controller 150 may selectively transmit meta information, which is information about landmark characteristics of the identification part.

The present invention configured as described above has an advantage of preventing issues that violate laws related to personal information protection when using medical images by proposing a new technology for pseudonymizing medical images. In addition, the present invention has an advantage of being able to generate a pseudonymized image that maintains reality, that is, can be recognized as actually existing by a medical person (doctor, etc.). In addition, the present invention has an advantage in that the learning time and the pseudonymization processing time can be significantly reduced when a pseudonymized image is generated by applying the GAN algorithm.

In the detailed description of the present invention, specific embodiments have been described, but various modifications are possible without departing from the scope of the present invention. Therefore, the scope of the present invention is not limited to the described embodiments, and should be defined by the following claims and equivalents thereof.

100: pseudonymization device 110: input unit
120: communication unit 130: display
140: memory 150: control unit

Claims (15)

A pseudonymization method for medical images performed by an electronic device,
In a medical image including a target body part that is a body part of a subject's head, an identification part, which is a feature part to be provided to a medical person for treatment including diagnosis, correction, or surgery on the subject, and the medical person during the treatment Separating each non-identification part, which is a part that does not have to be provided to the user; and
generating a pseudonymized image each including a region of the identified portion and a pseudonymized region mapped using the processed image by mapping the non-identified portion using a processed image that is an image of a corresponding target body part of another subject Step; including,
The pseudonymized image is a fake image that has a body structure for a corresponding target body part and is recognized as a corresponding target body part of another target, and is an image of a corresponding target body part of another target, but is a corresponding target of the target An image including landmark characteristics of the identification part included in the body part,
In the step of classifying, the target body part of the subject is divided into a plurality of subregions according to a weight range of importance, and a type of the subregion included in the identification part according to an applied weight selected by a user from among the plurality of subregions. Pseudonymization method to control.
delete According to claim 1,
The pseudonymization method of claim 1 , wherein the processed image is an image synthesized using medical images of corresponding target body parts of various other subjects.
According to claim 1,
The pseudonymization method of claim 1 , wherein the processed image or the pseudonymized image is generated using a pre-learned machine learning model.
According to claim 4,
The machine learning model is a pseudonymization method learned by machine learning of a generative adversarial network (GAN) algorithm.
According to claim 1,
The distinguishing step is a pseudonymization method of adjusting the identification part according to a user's selection.
According to claim 6,
In the step of classifying, the identification part is selected from landmarks displayed according to body part type, hard tissue/soft tissue type, or object part type other than the body.
delete According to claim 1,
The step of classifying is a pseudonymization method of displaying each partial area in various colors in the medical image.
According to claim 1,
The subregion includes a first subregion having the highest weight range, a second subregion having a lower weight range than the first subregion, and a third subregion having the lowest weight range,
In the distinguishing step, the identification part is selected from at least a first partial area and the non-identified part is selected from at least a third partial area.
According to claim 10,
In the step of classifying, as the applied weight decreases, the pseudonymization method includes including the second partial area as well as the first partial area in the identification part.
According to claim 1,
In the step of distinguishing, a pseudonymization method of distinguishing a plurality of spaced apart from each other as an identification part among the contour points constituting the contour of the target body part.
According to claim 1,
storing information about the identification part as meta information; and
The pseudonymization method further comprising selecting whether or not to transmit the meta information when transmitting the pseudonymized video to another electronic device.
a memory storing a medical image including a body part of the subject, which is a part of the subject's head; and
A control unit controlling a pseudonymization process for the medical image;
The control unit determines an identification part, which is a feature part to be provided to a medical person for treatment including diagnosis, correction, or surgery, of the target body part of the medical image, and the medical person during the treatment among the target body parts. By identifying the non-identification part, which is a part that does not need to be provided to the subject, and mapping the non-identification part using a processed image, which is an image of a target body part of another subject, using the region of the identified part and the processed image generating pseudonymized images each including the mapped pseudonymized regions;
The pseudonymized image is a fake image that has a body structure for a corresponding target body part and is recognized as a corresponding target body part of another target, and is an image of a corresponding target body part of another target, but is a corresponding target of the target An image including landmark characteristics of the identification part included in the body part,
The control unit divides the target body part of the subject into a plurality of subregions according to weight ranges of importance, and adjusts the type of the subregion included in the identification part according to an applied weight selected by a user from among the plurality of subregions. pseudonymization device.
A server for pseudonymizing a medical image including a target body part received from a terminal, which is a body part of the subject's head, comprising:
a communication unit that receives the medical image from the terminal and transmits the pseudonymized image to the terminal; and
Including; a control unit for controlling the pseudonymization process,
The control unit determines an identification part, which is a feature part to be provided to a medical person for treatment including diagnosis, correction, or surgery, of the target body part of the medical image, and the medical person during the treatment among the target body parts. By identifying the non-identification part, which is a part that does not need to be provided to the subject, and mapping the non-identification part using a processed image, which is an image of a target body part of another subject, using the region of the identified part and the processed image generating pseudonymized images each including the mapped pseudonymized regions;
The pseudonymized image is a fake image that has a body structure for a corresponding target body part and is recognized as a corresponding target body part of another target, and is an image of a corresponding target body part of another target, but is a corresponding target of the target An image including landmark characteristics of the identification part included in the body part,
The control unit divides the target body part of the subject into a plurality of subregions according to weight ranges of importance, and adjusts the type of the subregion included in the identification part according to an applied weight selected by a user from among the plurality of subregions. pseudonymization server.

Images