KR20220073582A - Method for training and testing obfuscation network capable of performing distinct concealing processes for distinct parts of original medical image, and training device and testing device using the same - Google Patents

Abstract

The present invention mixes characteristic concealing processes corresponding to respective regions of the original medical image to generate a tampered medical image by concealing the original medical image, so that the human being recognized that the tampered medical image is different from the original medical image It is to enable the tampered medical image to be recognized as the original medical image in the network.

Description

A method for learning and testing a modulation network that conceals an original medical image by performing different concealing processes for each area of the original medical image, and a learning and testing apparatus using the same. CAPABLE OF PERFORMING DISTINCT CONCEALING PROCESSES FOR DISTINCT PARTS OF ORIGINAL MEDICAL IMAGE, AND TRAINING DEVICE AND TESTING DEVICE USING THE SAME}

The present invention relates to an obfuscation network for concealing an original medical image, and more particularly, to an area including medical information used for medical condition determination on an original medical image and a different control to other areas. A method of learning and testing a modulation network that conceals an original medical image by applying a sealing process, and a learning apparatus and a testing apparatus using the same.

Big data is not only standardized data used in the existing corporate environment or public institutions, but also e-commerce data, metadata, web log data, Radio Frequency Identification (RFID) data, sensor network data, Data that includes unstructured or semi-structured data that has not been utilized before, such as social network data, Internet text and document data, and Internet search indexing data. It is called big data in the sense that it will have a level of data that is difficult to handle.

Such big data may have no meaning by itself, but it can be usefully used to create new data in various fields through machine learning on patterns, etc., or to make judgments or predictions.

As a representative example, medical big data, which is generated by each medical institution into big data of patient information, is analyzed and refined. It is expected to greatly contribute to the creation and analysis of

However, although many fields are making rapid progress through open innovation based on big data analysis, data in the medical industry contains many sensitive data, so there have been many difficulties in external cooperation using the data. In recent years, expectations for the use of big data have increased with the passage of the Data 3 Act, but it can be seen that the demand for anonymization has grown proportionally.

A high-level anonymization method that goes beyond simply deleting the patient's personal information is expected to further increase the utilization value of medical big data. Medical images that can be treated as personal information: Patient observation videos, pictures of affected areas, pictures before and after plastic surgery, X-rays, ultrasound, CT scans, PET (Positron Emission Tomography), f-MRI (Functional Magnetics) Resonance imaging (Functional Magnetic Resonance Imaging), etc., can be personal image information that can identify an individual in a specific situation because many complex situations occur. In fact, there have been cases where individuals were identified using leaked X-rays overseas. Therefore, there is a need for a method that can safely use medical images in a form that protects personal information and copyright.

In particular, in recent years, various types of concealing processing methods for preventing leakage of personally identifiable information by irreversibly converting information included in original medical data have been proposed.

For example, a method of erasing or hiding all objects included in the original medical data by masking or blurring has the advantage of relatively simple concealing processing, but all objects included in the original medical data In the case of concealment, all meaningful data necessary to determine the disease name is also concealed, so there is a disadvantage in that the utility value as medical big data is lowered.

As another example, the method of determining the area including identification information among the information included in the original medical data and concealing the area is advantageous for data utilization by leaving the other parts except for the identification information included area, but If an omission occurs in the process of determining a domain, there is a limit to being vulnerable to personal information leakage.

As another method, as disclosed in Korean Patent Application Laid-Open No. 10-2020-0014396, there is a method of irreversibly processing other areas while leaving only important areas necessary for data analysis such as disease name determination on original medical data. Specifically, the medical image concealing system described in the relevant patent application separates a facial region from a medical image including the patient's facial region and conceals the rest of the facial region except for a region necessary for analysis.

Such concealing technology may be more suitable for preventing personal information leakage, but since the area that is not concealed is recognizable by the human eye, if the area contains important identification information for personal identification, There is still a limitation that it cannot be completely secure against leakage of personal information.

Accordingly, there is a need for an improvement method to solve the above problems.

An object of the present invention is to solve all of the above problems.

In addition, an object of the present invention is to minimize the risk of personal information exposure while being relatively simple by applying different concealing processes for each area of the original medical image instead of a separate task of finding personal identification information included in the original medical image.

Another object of the present invention is to prevent security or privacy-related problems that may occur by restoring the original medical image by generating data obtained by irreversibly concealing the original medical image.

In addition, the present invention applies a sufficient concealing processor to the original medical image, but preserves important features that the machine learning network can interpret, so that it is impossible for humans to infer the original medical image from the tampered medical image, but the machine learning network Another object is to make it possible to generate a recognition result identical to or similar to the original medical image from a tampered medical image.

In addition, another object of the present invention is to further activate the big data trading market.

In order to achieve the object of the present invention as described above and to realize the characteristic effects of the present invention to be described later, the characteristic configuration of the present invention is as follows.

According to an aspect of the present invention, in a method for learning a modulation network for concealing an original medical image by performing different concealing processes for each region of an original medical image, (a) a learning apparatus includes: a medical image for training is input to the modulation network to cause the modulation network to perform the first and second concealing processes respectively corresponding to the first learning area and the second learning area of the training medical image, respectively, the first area for training and the second area for training. generating a modulated medical image for learning corresponding to the medical image for learning by performing in the second region; (b) the learning device, (i) a first discriminator for realizing a first reference medical image that has undergone the first concealing process on the entire region of the medical image for training and the entire region of the medical image for training By inputting the training modulated medical image to each of the second discriminators for judging the second reference medical image that has undergone the second concealing process to be genuine, the first discriminator and the second discriminator cause the training modulated medical image outputting a first modulated medical image score to a second modulated medical image score for determining whether an image is real or fake, and (ii) inputting the modulated medical image for training into an image identification network, causing the image identification network to , performing a process of performing a learning operation on the modulated medical image for learning and outputting medical feature information for learning corresponding to the modulated medical image for learning; and (c) maximizing a summation loss obtained by weighted summing, by the learning apparatus, a first_1 discriminant loss and a 2_1 discriminant loss generated by referring to the first modulated medical image score and the second modulated medical image score, and the image Training the modulation network to minimize an accuracy loss generated by referring to a task specific output generated using the medical feature information for training or the medical feature information for training of an identification network and a ground truth corresponding to the training medical image. ; A method comprising

As an example, (d) the learning device generates the first_1 discriminant loss and the first_1 discriminant loss generated by the first discriminator and the second discriminator with reference to the first modulated medical image score and the second modulated medical image score. The second_1 discrimination loss is minimized, respectively, and the first reference medical image score and the second reference medical image corresponding to the first and second reference medical images respectively input to the first discriminator and the second discriminator, respectively. training each of the first discriminator and the second discriminator to maximize a 1_2 discriminant loss and a 2_2 discriminant loss generated with reference to the image score, respectively; Disclosed is a method characterized in that it further comprises.

As an example, (e) the training device training the image identification network to minimize the accuracy loss; Disclosed is a method characterized in that it further comprises.

As an example, a method is disclosed, wherein the learning device is in a state in which the image identification network has been trained to recognize the medical image for training.

As an example, when the learning apparatus performs learning of the modulation network using the summation loss, (i) for the first region for learning, the first_1 discrimination loss corresponding to the first region for learning is Learning is carried out by applying a higher weight than the second_1 discrimination loss corresponding to the second area, and (ii) for the second area for learning, the second_1 determination loss corresponding to the second area for learning is applied to the second_1 determination loss corresponding to the second area for learning. Disclosed is a method characterized in that the learning is performed by applying a weight higher than the first_1 discrimination loss corresponding to the first area.

As an example, the learning device causes the modulation network to conceal the first region for learning so as to have a relatively low noise density according to the degree of concealing through the first concealing process, and Disclosed is a method characterized in that the second area for learning is subjected to a concealing process to have a relatively high noise density according to a concealing degree through a concealing process.

As an example, the learning apparatus sets, among the areas of the medical image for learning, a region including medical information used for medical condition determination as the first region for learning, and selects the first region for learning from the medical image for learning. Disclosed is a method characterized in that the excluded area is set as the second area for learning.

As an example, the learning apparatus obtains the first area information for learning and the second area information for learning of the medical image for learning from labeling information for learning that sets a first area for learning and a second area for learning in the medical image for training A method characterized in that the input to the modulation network is disclosed.

As an example, the learning device inputs the medical image for training into the modulation network so that the modulation network performs a convolution operation on the medical image for training at least once through an encoder to generate a feature map for learning, through a decoder Disclosed is a method characterized in that the learning feature map is deconvolved at least once to output the modulated medical image for learning.

According to another aspect of the present invention, there is provided a method of testing a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image, (a) by a learning device, (i) ) input a medical image for training into a modulation network to cause the modulation network to perform each of the first and second learning concealing processes corresponding to each of the first learning region and the second learning region of the training medical image for the learning The first area and the second area for learning are performed to generate a modulated medical image for training corresponding to the medical image for training, and (ii) (ii-1) the entire area of the medical image for training is covered with the first control for training. A first discriminator for judging the first reference medical image that has undergone the sealing process as genuine, and a second discriminant that allows the entire area of the training medical image to be judged as genuine and a second reference medical image that has undergone the second concealing process for learning By inputting the modulated medical image for learning to each device, the first discriminator and the second discriminator generate a first modulated medical image score to a second modulated medical image score for judging whether the modulated medical image for training is real or fake and (ii-2) inputting the modulated medical image for training into an image identification network, causing the image identification network to perform a learning operation on the modulated medical image for training to correspond to the modulated medical image for training. performing a process to output feature information, and (iii) a summation loss obtained by weighted summing a first_1 discriminant loss and a 2_1 discriminant loss generated by referring to the first modulated medical image score and the second modulated medical image score. maximizing, and referring to the task specific output generated using the medical feature information for learning or the medical feature information for learning of the image identification network and the ground truth corresponding to the medical image for learning acquiring, by a test device, a medical image for testing in a state in which the modulation network is trained to minimize the generated accuracy loss; and (b) the test device inputs a medical image for testing into the modulation network to cause the modulation network to cause a first test area corresponding to each of the first area for testing and the second area for testing of the medical image for testing. generating a modulated medical image for testing corresponding to the medical image for testing by performing a concealing process and a second concealing process for testing on the first region for testing and the second region for testing, respectively; A method comprising

As an example, in step (a),

The first discriminator and the second discriminator respectively minimize the first_1 discrimination loss and the second_1 discrimination loss by the learning device, and the first discriminator and the second discriminator respectively input to the first discriminator and the second discriminator that the first_2 discriminant loss and the second_2 discriminant loss generated by referring to the first reference medical image score and the second reference medical image score corresponding to the reference medical image and the second reference medical image, respectively, are respectively maximized A method of characterization is disclosed.

As an example, a method is disclosed, wherein the image identification network is trained by the learning device to minimize the accuracy loss or is trained to recognize the medical image for training.

As an example, the test device causes the modulation network to conceal the first region for the test to have a relatively low noise density according to the degree of concealing through the first concealing process for the test, Disclosed is a method characterized in that the second area for the test is subjected to a concealing process to have a relatively high noise density according to the degree of concealing through the second concealing process for the test.

As an example, the test apparatus sets, among the areas of the test medical image, an area including medical information used for medical condition determination as the first test area, and in the test medical image, the test area Disclosed is a method characterized in that an area other than the first area is set as the second area for the test.

As an example, the test device may include, from labeling information for testing in which a first region for testing and a second region for testing in the medical image for testing are set, the first region information for testing and the test in the medical image for testing. A method is disclosed, comprising obtaining second domain information for use and inputting it into the modulation network.

According to another aspect of the present invention, there is provided a learning apparatus for learning a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image, comprising: at least one memory for storing instructions; and at least one processor configured to execute the instructions, wherein the processor is configured to: (I) input a medical image for training into a modulation network to cause the modulation network to form a first region for training and a first region for training of the medical image for training. A process of generating a modulated medical image for training corresponding to the medical image for training by performing a first concealing process and a second concealing process corresponding to each of the two regions in the first region for learning and the second region for learning, respectively , (II) (i) a first discriminator for realizing a first reference medical image that has undergone the first concealing process on the entire region of the medical image for training, and a second control unit on the entire region of the medical image for training By inputting the training modulated medical image to each of the second discriminators for judging the second reference medical image that has undergone the sealing process to be genuine, the first discriminator and the second discriminator determine whether the training modulated medical image is genuine. a sub-process for outputting a first modulated medical image score to a second modulated medical image score for determining whether it is fake, and (ii) inputting the modulated medical image for training into an image identification network, causing the image identification network to: a process of performing a sub-process of performing a learning operation on a modulated medical image and outputting medical feature information for learning corresponding to the modulated medical image for learning; and (III) maximizing a summation loss obtained by weighted summing a first_1 discriminant loss and a 2_1 discriminant loss generated with reference to the first modulated medical image score and the second modulated medical image score, for the training of the image identification network a process of training the modulation network to minimize an accuracy loss generated by referring to a task specific output generated using the medical characteristic information or the medical characteristic information for training and a ground truth corresponding to the training medical image; A learning apparatus for performing is disclosed.

As an example, (IV) the processor generates the first_1 discriminant loss and the second discriminator generated by the first discriminator and the second discriminator with reference to the first modulated medical image score and the second modulated medical image score, respectively. 2_1 discrimination loss is minimized, respectively, and a first reference medical image score and a second reference medical image corresponding to the first and second reference medical images respectively input to the first discriminator and the second discriminator a process of training each of the first discriminator and the second discriminator to maximize the first_2 discriminant loss and the second_2 discriminant loss generated with reference to the score, respectively; Disclosed is a learning apparatus further comprising a.

As an example, (V) the processor trains the image identification network to minimize the accuracy loss; Disclosed is a learning apparatus further comprising a.

As an example, a method is disclosed, wherein the processor is in a state in which the image identification network has been trained to recognize the medical image for training.

As an example, when the processor performs learning of the modulation network using the summation loss, (i) for the first area for learning, the first_1 discrimination loss corresponding to the first area for learning is Learning is carried out by applying a higher weight than the second_1 discrimination loss corresponding to the second area, and (ii) for the second learning area, the second learning loss is applied to the second_1 determination loss corresponding to the second learning area. Disclosed is a learning apparatus characterized in that the learning is performed by applying a weight higher than the first_1 discrimination loss corresponding to the region.

As an example, the processor causes the modulation network to conceal the first region for learning so as to have a relatively low noise density according to the degree of concealing through the first concealing process, and Disclosed is a learning apparatus characterized in that the second area for learning is subjected to a concealing process to have a relatively high noise density according to a degree of concealing through a sealing process.

As an example, the processor sets, among the areas of the medical image for learning, an area including medical information used for medical condition determination as the first area for learning, except for the first area for learning in the medical image for learning. Disclosed is a learning apparatus characterized in that an area is set as the second area for learning.

As an example, the processor obtains the first area information for learning and the second area information for learning of the medical image for learning from the labeling information for learning that sets the first area for learning and the second area for learning in the medical image for training. Disclosed is a learning apparatus, characterized in that the input to the modulation network.

As an example, the processor inputs the medical image for training into the modulation network to cause the modulation network to perform a convolution operation on the medical image for training at least once through an encoder to generate a feature map for training, and via a decoder Disclosed is a learning apparatus characterized in that the learning feature map is deconvolved at least once to output the modulated medical image for learning.

According to another aspect of the present invention, there is provided a test apparatus for testing a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image, comprising: at least one memory for storing instructions; and at least one processor configured to execute the instructions, wherein the processor (I) by means of a learning device (i) inputs a medical image for training into a modulation network to cause the modulation network to cause the medical image for training. Corresponding to the medical image for learning by performing a first concealing process for learning and a second concealing process for learning corresponding to each of the first region for learning and the second region for learning, respectively, in the first region for learning and the second region for learning a first discriminator configured to generate a modulated medical image for learning, and (ii) (ii-1) to truly determine a first reference medical image that has undergone the first concealing process for learning on the entire area of the medical image for learning; By inputting the modulated medical image for training into each of the second discriminators for truly judging the second reference medical image that has undergone the second concealing process for learning in the entire area of the medical image for training, the first discriminator and the first discriminator 2 A process for causing a discriminator to output a first modulated medical image score to a second modulated medical image score for determining whether the modulated medical image for training is real or fake, and (ii-2) image identification of the modulated medical image for training input to a network, causing the image identification network to perform a learning operation on the modulated medical image for training and output medical feature information for training corresponding to the modulated medical image for training; (iii) the first modulated medical image; Maximizes the summation loss obtained by weighted summing the first_1 discriminant loss and the 2_1 discriminant loss generated by referring to the image score and the second modulated medical image score, and the medical feature information for learning or the medical feature information for learning of the image identification network In a state in which the modulation network is trained to minimize the accuracy loss generated by referring to the task specific output generated using a process of acquiring a medical image for and (II) inputting a medical image for testing into the modulation network to cause the modulation network to perform a first concealing process for testing and a test corresponding to each of the first region for testing and the second region for testing of the medical image for testing. a process of generating a modulated medical image for testing corresponding to the medical image for testing by performing a second concealing process on the first region for testing and the second region for testing, respectively; A test device for performing is disclosed.

As an example, in the process (I), the first discriminator and the second discriminator respectively minimize the first_1 discriminant loss and the second_1 discriminant loss by the learning apparatus, and the first discriminator and the second discriminator respectively The first_2 determination loss and the second_2 determination generated by referring to the first reference medical image score and the second reference medical image score corresponding to the first and second reference medical images respectively input to the second discriminator Disclosed is a test apparatus characterized in that the state is learned to maximize each loss.

As an example, the image identification network is trained by the learning device to minimize the accuracy loss, or a test device is disclosed, characterized in that it is trained to recognize the training medical image.

As an example, the processor causes the modulation network to conceal the first region for the test to have a relatively low noise density according to the degree of concealing through the first concealing process for the test, Disclosed is a test apparatus characterized in that the second area for testing is subjected to a concealing process to have a relatively high noise density according to a degree of concealing through a second concealing process for testing.

As an example, the processor sets, among the regions of the test medical image, a region including medical information used for determining a medical condition as the first region for the test, and in the test medical image, Disclosed is a test apparatus characterized in that an area other than the first area is set as the second area for the test.

As an example, the processor may include, from labeling information for testing in which a first area for testing and a second area for testing in the medical image for testing are set, the first area information for testing and the test area of the medical image for testing. A test apparatus is disclosed, characterized in that the second area information is obtained and input to the modulation network.

In addition to this, a computer-readable recording medium for recording a computer program for executing the method of the present invention is further provided.

The present invention is relatively simple and effective in reducing the risk of personal information exposure by applying different concealing processes for each area of the original medical image instead of a separate task of finding personal identification information included in the original medical image.

In addition, the present invention has an effect of preventing security or privacy-related problems that may occur when the original medical image is restored by generating data obtained by irreversibly concealing the original medical image.

In addition, the present invention applies a sufficient concealing processor to the original medical image, but preserves important features that the machine learning network can interpret, so that it is impossible for humans to infer the original medical image from the tampered medical image, but the machine learning network It is possible to generate a recognition result identical to or similar to the original medical image from the tampered medical image.

In addition, the present invention has the effect of activating the big data trading market more.

The accompanying drawings for use in the description of the embodiments of the present invention are only a part of the embodiments of the present invention, and for those of ordinary skill in the art to which the present invention pertains (hereinafter, "those of ordinary skill in the art"), the invention Other drawings may be obtained based on these drawings without any work being done.
1 schematically shows a learning apparatus for learning an obfuscation network for concealing an original medical image by performing different concealing processes for each region of an original medical image according to an embodiment of the present invention; did it,
2 schematically illustrates a method for learning a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention;
3 is a medical image for training and labeling information for learning used in a method for learning a modulation network that conceals an original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention; , which shows the concealing process reference value by way of example,
4 is a modulated medical image for learning to which concealing is applied in a method for learning a modulation network that conceals the original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention; It is shown as an example,
5 is a method for learning a discriminator in a method for learning a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention; is schematically shown,
6 is an image recognition network in a method for learning a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention; It schematically shows how to learn
7 schematically shows a test apparatus for testing a modulation network that conceals an original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention;
8 schematically illustrates a method of testing a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following detailed description of the present invention refers to the accompanying drawings, which show by way of illustration a specific embodiment in which the present invention may be practiced, in order to clarify the objects, technical solutions and advantages of the present invention. These embodiments are described in sufficient detail to enable those skilled in the art to practice the present invention.

Also, throughout this description and claims, the word "comprises" and variations thereof are not intended to exclude other technical features, additions, components or steps. Other objects, advantages and characteristics of the present invention will become apparent to a person skilled in the art in part from this description and in part from practice of the present invention. The following illustrations and drawings are provided by way of illustration and are not intended to limit the invention.

Moreover, the invention encompasses all possible combinations of the embodiments indicated herein. It should be understood that the various embodiments of the present invention are different but need not be mutually exclusive. For example, certain shapes, structures, and characteristics described herein with respect to one embodiment may be implemented in other embodiments without departing from the spirit and scope of the invention. In addition, it should be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the present invention. Accordingly, the detailed description set forth below is not intended to be taken in a limiting sense, and the scope of the present invention, if properly described, is limited only by the appended claims, along with all scope equivalents as those claimed. Like reference numerals in the drawings refer to the same or similar functions throughout the various aspects.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings in order to enable those of ordinary skill in the art to easily practice the present invention.

1 is a learning apparatus 1000 for learning an obfuscation network for concealing an original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention. schematically shown.

Referring to FIG. 1 , the learning apparatus 1000 according to an embodiment of the present invention provides instructions for learning a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image. The operation of learning the modulation network for concealing the original medical image by performing different concealing processes for each region of the original medical image according to the memory 1001 for storing them and the instructions stored in the memory 1001 is performed. It may include a processor 1002 .

Specifically, the learning device 1000 is typically a computing device (eg, a device that may include a computer processor, memory, storage, input and output devices, other components of a conventional computing device; electronic devices such as routers, switches, etc.) A desired system using a combination of communication devices; electronic information storage systems such as network attached storage (NAS) and storage area networks (SANs)) and computer software (ie, instructions that cause the computing device to function in a particular way). performance may be achieved.

In addition, the processor of the computing device may include a hardware configuration such as a micro processing unit (MPU) or a central processing unit (CPU), a cache memory, and a data bus. In addition, the computing device may further include an operating system and a software configuration of an application for performing a specific purpose.

However, a case in which the computing device includes an integrated processor in which a medium, a processor, and a memory are integrated for implementing the present invention is not excluded.

On the other hand, the learning device 1000 may be a device or a server for performing a concealing process for an original medical image, and is mounted on a medical device for producing an original medical image and a computer device interlocked therewith. For a concealing process may be performed, but the present invention is not limited thereto.

A method of learning a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image using the learning apparatus 1000 according to an embodiment of the present invention configured as described above. It will be described with reference to 2 to 6 as follows.

First, FIG. 2 schematically illustrates a method for learning a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention.

Referring to FIG. 2 , when a training medical image (x) is acquired, the learning apparatus 1000 inputs the acquired medical training image x into the modulation network 100 so that the modulation network 100 causes the training medical image (x) to be obtained. Medical image (x) for training by performing a first concealing process and a second concealing process respectively corresponding to the first learning area and the second learning area of the image (x) in the first learning area and the second learning area, respectively It is possible to generate a modulated medical image (x') for learning corresponding to .

At this time, the modulation network 100 may conceal the training medical image (x) into the training modulation medical image (x') by performing different concealing processes for each area of the training medical image, for example, The network 100 may generate a modulated medical image for learning x' in which the degree of concealing is different for each of the first region for learning and the second region for learning of the medical image for learning (x). And, the modulation network 100 simultaneously performs the first concealing process and the second concealing process, so that the first area for learning is concealed by the first concealing process, and the second area for learning is the second concealing process It can be made to be concealed by That is, the modulation network 100 mixes the first concealing process and the second concealing process to conceal the medical image for training (x), so that the first region for learning in the medical image for training (x) is the first concealing process. Concealed by a process and the second area for learning may be concealed by a second concealing process.

To this end, the learning apparatus 1000 sets an area including medical information used for medical condition determination among areas of the medical image for learning x as a first area for learning, and the first area for learning in the medical image for learning. A region except for is set as a second region for learning, and the modulation network 100 conceals the first region for learning to have a relatively low noise density according to the degree of concealing through the first concealing process, , through the second concealing process, the second area for learning may be concealed to have a relatively high noise density according to the degree of concealing.

Here, the learning apparatus 1000 is configured to set the first area for learning and the second area for learning of the medical image for learning (x) from the labeling information for learning in which the first area for learning and the second area for learning are set in the medical image for learning (x). information can be obtained.

In this case, the labeling information for learning on the first area for learning and the second area for learning in the medical image for learning (x) is generated by the human labeler setting the first area for learning and the second area for learning in the medical image for learning (x) or , may be generated by segmenting the medical image (x) for training into a first region for learning and a second region for learning through a machine learning-based segmentation network.

Accordingly, when the learning apparatus 1000 obtains the first area information for learning and the second area information for learning of the medical image for training x, the first area information for learning and the second area information for learning are transferred to the modulation network 100 . input to cause the modulation network 100 to perform a first concealing process on the first region for learning with reference to information on the first region for learning and information on the second region for learning, and to perform a second concealing process on the second region for learning can do.

The training modulated medical image (x') generated through the above process is recognized differently from the training medical image (x) by humans, but in a learning network, it may be recognized as the same or similar to the training medical image (x).

As an example, FIG. 3 shows a method for learning a modulation network 100 that conceals an original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention. Assuming that the medical image, the labeling information for learning, and the reference value of the concealing process are exemplarily shown, the modulated medical image (x') for learning to which the concealing is applied may be generated as shown in FIG. 4 .

Here, the area indicated by A1 of the modulated medical image (x') for training in FIG. 4 is determined to include medical information used for medical condition determination and is concealed through a first concealing process to have a relatively low noise density. It may be a processed first area for learning, and the area marked A2 may be a second area for learning that has been concealed through a second concealing process to have a relatively high noise density with respect to an area other than the first area for learning .

On the other hand, the modulation network 100 is, for example, an encoder including a plurality of convolutional layers for convolutional operation of an input medical image (x) for training, and deconvolves a feature map output from the encoder for learning It may include, but is not limited to, a decoder including a plurality of deconvolutional layers for outputting a modulated medical image (x'), and is not limited thereto. ) may include a learning network of various structures modulated by

Next, referring back to FIG. 2 , the learning apparatus 1000 is configured to truly determine the first reference medical image (x + noise_1) that has undergone the first concealing process on the entire area of the training medical image (x). In each of the first discriminator 200_1 and the second discriminator 200_2 to truly judge the second reference medical image (x + noise_2) that has undergone the second concealing process on the entire area of the medical image for training (x) The first modulated medical image in which the modulated medical image for learning (x') is input, and each of the first discriminator 200_1 and the second discriminator 200_2 determines whether the modulated medical image for learning (x') is real or fake. A process may be performed to output an image score and a second modulated medical image score.

In this case, the maximum value of the image score generated by the discriminator for the input image may be 1, which determines the input image as real, and the minimum value of the image score may be 0, which determines the input image as fake.

In addition, the learning apparatus 1000 inputs the modulated medical image (x') for learning into the image identification network 300, and causes the image identification network 300 to perform a learning operation on the modulated medical image for learning (x') for learning. A process of outputting medical feature information for learning corresponding to the modulated medical image (x') may be performed.

In this case, the image identification network 300 may include a machine learning network, and the machine learning network is K-Nearest Neighbors, Linear Regression, Logistic Regression, and a support vector. Machine (SVM, Support Vector Machine), Decision Tree and Random Forest, Neural Network, Clustering, Visualization and Dimensionality Reduction, Association Rule Learning ( Association Rule Learning), a deep trust neural network (Deep Belief Network), reinforcement learning (Reinforcement Learning), and may include at least one of a deep learning algorithm, but the machine learning network is not limited thereto, and may include various learning algorithms. .

In addition, the medical feature information for learning may be a feature or logits of the modulated medical image (x') for learning. In addition, the medical feature information for training includes a feature value related to a predetermined feature in the modulated medical image for training x', or a value related to at least one of a vector, a matrix, and a coordinate related to the predetermined feature. It can be a logit that For example, it may be a class for determining cancer, a characteristic related to cancer, for example, a metastasis state of cancer, and coordinates regarding a location of the cancer.

Next, the learning apparatus 1000 uses the respective losses generated from the first discriminator 200_1 , the second discriminator 200_2 , and the image identification network 300 , the modulation network 100 , the first discriminator Each of (200_1), the second discriminator 200_2, and the image identification network 300 may be trained.

First, the learning apparatus 1000 may generate a first_1 discriminant loss by referring to the first modulated medical image score from the first discriminator 200_1 , and a second modulated medical image score from the second discriminator 200_2 . The 2_1 discrimination loss may be generated by referring to , and the summed loss may be generated by weighted summing the 1_1 discrimination loss and the 2_1 discrimination loss.

Here, the learning apparatus 1000 may obtain the summation loss by weighted summing the first_1 discrimination loss and the 2_1 discrimination loss in correspondence to the weights with reference to the first area information for learning and the second area information for learning. Here, the weights referring to the first region information for learning and the second region information for learning may be further adjusted according to the learned performance of the modulation network 100 .

In addition, the learning apparatus 1000 includes a task specific output generated by using the medical characteristic information for training or the medical characteristic information for learning of the image identification network 300 from the image identification network 300 and a ground truth corresponding to the medical image for training. You can refer to the (ground truth) to generate the accuracy loss you created. Here, the task specific output may be a probability distribution for a modulated medical image (x') for training in which medical feature information for training is mapped for each class, and classification information corresponding to the first area for learning on the training medical image using the task specific output and location information.

Accordingly, the learning apparatus 1000 may train the modulation network 100 to maximize the summation loss and minimize the accuracy loss. Here, when learning for the modulation network 100 is performed, by learning the modulation network 100 by further including a regularization loss for the parameters of the modulation network 100, overfitting is prevented and It can improve the generalization performance.

At this time, when the learning apparatus 1000 performs learning of the modulation network 100 using the summation loss, the first area for learning corresponds to the first_1 discrimination loss corresponding to the first area for learning and the second area for learning. Learning is carried out by applying a weight higher than the second_1 discrimination loss, and for the second area for learning, a weight higher than the first_1 determination loss corresponding to the first area for learning is applied to the second_1 determination loss corresponding to the second area for learning. You can apply it to learn. That is, the parameter of the modulation network 100 corresponding to the first concealing process for concealing the first region for learning may be updated by backpropagation using the summation loss with a high weight applied to the first_1 discrimination loss. Also, the parameter of the modulation network 100 corresponding to the second concealing process for concealing the second region for learning may be updated by backpropagation using the summation loss in which a weight is applied to the second_1 discrimination loss with a high weight. Accordingly, the modulated medical image x' for training may be recognized as genuine by the first discriminator 200_1 and the second discriminator 200_2.

While learning the modulation network 100 as described above, referring to FIG. 5 , the learning apparatus 1000 provides a first modulated medical image score to a first discriminator 200_1 and a second discriminator 200_2, respectively. A first reference medical image (x) that is input to the first discriminator 200_1 and the second discriminator 200_2, respectively, and minimizes the first_1 discriminant loss and the 2_1 discriminant loss generated by referring to the second modulated medical image score + noise_1) and the second reference medical image (x + noise_2) to maximize the first_2 discriminant loss and the second_2 discriminant loss generated by referring to the first reference medical image score and the second reference medical image score, respectively. Each of the discriminator 200_1 and the second discriminator 200_2 may be learned. That is, the learning apparatus 100 trains the first discriminator 200_1 to minimize the 1_1 discrimination loss and maximize the 1_2 discrimination loss, and to minimize the 2_1 discrimination loss and maximize the 2_2 discrimination loss to maximize the second discrimination loss. It is possible to learn the machine 200_2. Accordingly, the first reference medical image (x + noise_1) and the second reference medical image (x + noise_2) can be determined to be genuine, and the modulated medical image for training (x') can be determined to be fake.

Also, referring to FIG. 6 , the learning apparatus 1000 may train the image identification network 300 to minimize an accuracy loss. On the other hand, the image identification network 300 may be trained in advance using the medical image (x) for training, unlike learning using the accuracy loss.

By this learning method, for each batch in each epoch, the modulation network 100, the first discriminator 200_1, the second discriminator 200_2, and the image identification network 300 can be updated individually.

And, in the state in which the modulation network 100 is learned as described above, the test device inputs a medical image for testing into the modulation network 100 that has been trained, and causes the modulation network 100 to modulate the test medical image, and It is prevented from being recognized as a medical image for testing, and the image identification network 300 can generate a modulated medical image for testing that can output the same or similar recognition result as the medical image for testing.

Accordingly, FIG. 7 shows a test apparatus 2000 for testing the modulation network 100 for concealing the original medical image by performing different concealing processes for each region of the original medical image according to an embodiment of the present invention. ) is schematically shown.

Referring to FIG. 7 , the test apparatus 2000 according to an embodiment of the present invention tests the modulation network 100 for concealing the original medical image by performing different concealing processes for each region of the original medical image. A modulation network 100 for concealing the original medical image by performing different concealing processes for each region of the original medical image according to the memory 1001 for storing instructions for It may include a processor 1002 that performs the operation under test.

Specifically, test device 2000 is typically a computing device (eg, a device that may include a computer processor, memory, storage, input and output devices, other components of a conventional computing device; electronic devices such as routers, switches, etc.) A desired system using a combination of communication devices; electronic information storage systems such as network attached storage (NAS) and storage area networks (SANs)) and computer software (ie, instructions that cause the computing device to function in a particular way). performance may be achieved.

In addition, the processor of the computing device may include a hardware configuration such as a micro processing unit (MPU) or a central processing unit (CPU), a cache memory, and a data bus. In addition, the computing device may further include an operating system and a software configuration of an application for performing a specific purpose.

However, a case in which the computing device includes an integrated processor in which a medium, a processor, and a memory are integrated for implementing the present invention is not excluded.

On the other hand, the test device 2000 may be a device or a server for performing a concealing process for an original medical image, and is mounted on a medical device that produces an original medical image and a computer device interlocked therewith. For a concealing process may be performed, but the present invention is not limited thereto.

A method of testing a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image using the test device 2000 according to an embodiment of the present invention configured as described above. 8 as follows. Hereinafter, detailed description of parts that can be easily understood from the description with reference to FIGS. 2 to 6 will be omitted.

In a state in which the modulation network 100 is learned by the learning method with reference to FIGS. 2 to 6 , the test device 2000 obtains the test medical image x, By inputting into the modulation network 100, the modulation network 100 causes the first concealing process and the second concealing process corresponding to each of the first area for testing and the second area for testing of the medical image for testing x. Each process may be performed in the first area for testing and the second area for testing to generate a modulated medical image for testing (x′) corresponding to the medical image for testing (x).

At this time, the modulation network 100 may conceal the medical image for testing (x) into the modulation medical image (x') for testing by performing different concealing processes for each area of the medical image for testing, For example, the modulation network 100 may generate a modulated medical image for test x' in which the degree of concealing is different for each of the first test area and the second test area of the test medical image x. And, the modulation network 100 performs the first concealing process and the second concealing process at the same time, so that the first area for the test is concealed by the first concealing process, and the second area for the test is the second concealing process. It can be made to be concealed by a sealing process. That is, the modulation network 100 mixes the first concealing process and the second concealing process to conceal the medical image for testing (x), so that the first area for testing in the medical image for testing (x) is the first area for testing. 1 to be concealed by a concealing process, and the second area for testing may be to be concealed by a second concealing process.

To this end, the test apparatus 2000 sets, among the areas of the test medical image x, an area including medical information used to determine a medical state as a first area for testing, and performs the test in the test medical image. An area other than the first area for testing is set as a second area for testing, and the first area for testing is set so that the modulation network 100 has a relatively low noise density according to the degree of concealing through the first concealing process. may be subjected to a concealing process, and the second region for testing may be subjected to a concealing process to have a relatively high noise density according to the degree of concealing through the second concealing process.

Here, the test device 2000 sets the first area for testing of the medical image for testing x based on the labeling information for testing in which the first area for testing and the second area for testing are set in the medical image for testing (x). Information and second area information for testing may be obtained and input to the modulation network 100 .

At this time, the test labeling information for the first area for testing and the second area for testing in the medical image for testing (x) is obtained by the human labeler from the first area for testing and the first area for testing in the medical image for testing (x). It may be generated by setting two regions, or it may be generated by segmenting a medical image (x) for testing into a first region for testing and a second region for testing through a machine learning-based segmentation network.

Accordingly, the test apparatus 2000 modulates the first area information for testing and the second area information for testing when acquiring the first area information for testing and the second area information for testing of the medical image x for testing Input to the network 100 to cause the modulation network 100 to perform a first concealing process on the first area for testing with reference to the first area information for testing and the second area information for testing, and the second area for testing may be configured to perform a second concealing process.

The modulated medical image for testing (x') generated through the above process is recognized differently from the medical image for testing (x) by humans, but in a learning network, it can be recognized as the same as or similar to the medical image for testing (x). .

On the other hand, the modulation medical data concealed in the learned modulation network according to the embodiment of the present invention may be provided (or sold) to a purchaser who wants to purchase the medical image big data. In addition, the falsified medical data is transmitted to a third party or other institution, so that the third party or other institution can use the tampered medical data to diagnose the user's health condition without exposing the user's personal information. can

In addition, according to an embodiment of the present invention, while the concealed image data is provided (or sold) to the above purchaser, the method of testing the learned modulation network of program instructions that can be executed through various computer components It may be implemented in a form and recorded on a computer-readable recording medium and provided. According to an embodiment of the present invention, the purchaser executes the program instructions recorded on the recording medium using the computing device to generate concealed-processed data from the original data that the purchaser owns separately or obtained from other sources or paths. and the concealed data can be used in its own learning network. In addition, the purchaser may transfer at least two or more of the concealed data, the original image data possessed by him or obtained from other sources or channels, and the concealed image data provided (or sold) above to his/her own learning network. can also be used together.

On the other hand, according to an embodiment of the present invention, when the learned modulation network testing method is implemented in the form of program instructions that can be executed through various computer components, the accuracy is set high and the computing device of the purchaser overcomputing It can incur computational overhead, allowing buyers to reduce accuracy through settings, avoiding the computational overhead.

The embodiments according to the present invention described above may be implemented in the form of program instructions that can be executed through various computer components and recorded in a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the computer-readable recording medium may be specially designed and configured for the present invention, or may be known and available to those skilled in the computer software field. Examples of the computer-readable recording medium include hard disks, magnetic media such as floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floppy disks. media), and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules for carrying out the processing according to the present invention, and vice versa.

In the above, the present invention has been described with specific matters such as specific components and limited embodiments and drawings, but these are provided to help a more general understanding of the present invention, and the present invention is not limited to the above embodiments. , various modifications and variations can be devised from these descriptions by those of ordinary skill in the art to which the present invention pertains.

Accordingly, the spirit of the present invention should not be limited to the above-described embodiments, and not only the claims described below, but also all modifications equivalently or equivalently to the claims described below belong to the scope of the spirit of the present invention. will do it

1000: learning device
1001: the memory of the learning device
1002: processor of the learning device
2000: test device
2001: Memory of the test device
2002: Processor in the test unit
100: modulation network
200_1: first discriminator
200_2: second discriminator
300: image identification network
x: medical image for training or medical image for testing
x': Modulated medical image for training or modulated medical image for testing
x + noise_1: first reference medical image
x + noise_2: second reference medical image

Claims (30)

A method of learning a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image, the method comprising:
(a) a learning device inputs a medical image for training into a modulation network so that the modulation network causes a first concealing process and a second concealing process corresponding to each of the first region for learning and the second region for learning of the medical image for training generating a modulated medical image for learning corresponding to the medical image for learning by performing each in the first region for learning and the second region for learning;
(b) the learning device, (i) a first discriminator for realizing a first reference medical image that has undergone the first concealing process on the entire region of the medical image for training and the entire region of the medical image for training By inputting the training modulated medical image to each of the second discriminators for judging the second reference medical image that has undergone the second concealing process to be genuine, the first discriminator and the second discriminator cause the training modulated medical image outputting a first modulated medical image score to a second modulated medical image score for determining whether an image is real or fake, and (ii) inputting the modulated medical image for training into an image identification network, causing the image identification network to , performing a process of performing a learning operation on the modulated medical image for learning and outputting medical feature information for learning corresponding to the modulated medical image for learning; and
(c) the learning apparatus maximizes a summation loss obtained by weighted summing a first_1 discriminant loss and a 2_1 discriminant loss generated by referring to the first modulated medical image score and the second modulated medical image score, and identifies the image training the modulation network to minimize an accuracy loss generated by referring to a task specific output generated using the medical characteristic information for training or the medical characteristic information for training of a network and a ground truth corresponding to the training medical image;
How to include. According to claim 1,
(d) the first discriminator and the second discriminator generated by the learning device with reference to the first modulated medical image score and the second modulated medical image score, respectively, and the first_1 discrimination loss and the second_1 discriminant The loss is minimized, respectively, and the first reference medical image score and the second reference medical image score corresponding to the first and second reference medical images respectively input to the first discriminator and the second discriminator are obtained. training each of the first discriminator and the second discriminator to maximize the first_2 discriminant loss and the second_2 discriminant loss generated with reference to each other;
Method, characterized in that it further comprises. 3. The method of claim 2,
(e) training, by the training device, the image identification network to minimize the accuracy loss;
Method, characterized in that it further comprises. According to claim 1,
Method, characterized in that the learning device is in a state in which the image identification network has been trained to recognize the medical image for training. According to claim 1,
When the learning apparatus performs learning of the modulation network using the summation loss, (i) with respect to the first region for learning, the second region for learning corresponds to the first_1 discrimination loss corresponding to the first region for learning. Learning is carried out by applying a weight higher than the second_1 discrimination loss corresponding to (ii) in the second area for learning, the second_1 discrimination loss corresponding to the second area for learning is applied to the first area for learning Method characterized in that the learning is carried out by applying a weight higher than the corresponding first_1 discrimination loss. According to claim 1,
The learning apparatus causes the modulation network to conceal the first region for learning to have a relatively low noise density according to a concealing degree through the first concealing process, and the second concealing process Method according to claim 1, characterized in that the second area for learning is subjected to a concealing process to have a relatively high noise density according to the degree of concealing through . 7. The method of claim 6,
The learning apparatus sets, among the areas of the medical image for learning, an area including medical information used for medical condition determination as the first area for learning, and selects an area other than the first area for learning from the medical image for learning. Method characterized in that it is set as the second area for learning. According to claim 1,
The learning device is configured to obtain the first area information for learning and the second area information for learning of the medical image for learning from the labeling information for learning in which the first area for learning and the second area for learning are set in the medical image for training, and the modulation A method characterized in that input into the network. According to claim 1,
The learning device inputs the medical image for training into the modulation network so that the modulation network performs a convolution operation on the medical image for training at least once through an encoder to generate a feature map for learning, and the learning feature through a decoder A method characterized in that the map is deconvolved at least once to output the modulated medical image for training. A method of testing a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image, the method comprising:
(a) by a learning device, (i) inputting a medical image for training into a modulation network, causing the modulation network to cause a first concealing process for learning corresponding to each of the first region for learning and the second region for learning of the medical image for training; A second concealing process for learning is performed in the first region for learning and the second region for learning to generate a modulated medical image for learning corresponding to the medical image for learning, (ii) (ii-1) the medical treatment for learning A first discriminator for realizing a first reference medical image that has undergone the first concealing process for learning on the entire area of the image, and a second reference for performing the second concealing process for learning on the entire region of the medical image for training The first modulated medical treatment image is input to each of the second discriminators for judging the medical image as genuine, and the first discriminator and the second discriminator determine whether the modulated medical image for learning is real or fake. a process of outputting an image score or a second modulated medical image score, and (ii-2) inputting the modulated medical image for training into an image identification network, causing the image identification network to perform a learning operation on the modulated medical image for training; performing a process of outputting medical feature information for training corresponding to the modulated medical image for training, (iii) a first_1 discriminative loss and a first 2_1 Maximize the summation loss by weighted summing the discrimination loss, and refer to the task specific output generated using the medical feature information for learning or the medical feature information for learning of the image identification network and the ground truth corresponding to the medical image for learning acquiring, by a test device, a medical image for testing in a state in which the modulation network is trained to minimize the generated accuracy loss; and
(b) the test device inputs a medical image for testing into the modulation network so that the modulation network causes the first control for testing corresponding to each of the first region for testing and the second region for testing of the medical image for testing. performing a sealing process and a second concealing process for testing respectively on the first area for testing and the second area for testing to generate a modulated medical image for testing corresponding to the medical image for testing;
How to include. 11. The method of claim 10,
In step (a),
The first discriminator and the second discriminator respectively minimize the first_1 discrimination loss and the second_1 discrimination loss by the learning device, and the first discriminator and the second discriminator respectively input to the first discriminator and the second discriminator that the first_2 discriminant loss and the second_2 discriminant loss generated by referring to the first reference medical image score and the second reference medical image score corresponding to the reference medical image and the second reference medical image, respectively, are respectively maximized How to characterize. 12. The method of claim 11,
The image identification network is trained by the learning device to minimize the accuracy loss, or is in a state in which it has been trained to recognize the medical image for training. 11. The method of claim 10,
The test device causes the modulation network to conceal the first region for the test to have a relatively low noise density according to the degree of concealment through the first concealing process for the test, The method of claim 1, wherein the second area for the test is subjected to a concealing process to have a relatively high noise density according to the degree of concealing through a second concealing process. 14. The method of claim 13,
The test apparatus sets, among the areas of the medical image for testing, an area including medical information used to determine a medical condition as the first area for testing, and in the medical image for testing, the first area for testing A method, characterized in that the area except for is set as the second area for the test. 11. The method of claim 10,
The test device may include, from the test labeling information in which the test first area and the test second area are set in the test medical image, the test first area information and the test second area information of the test medical image A method, characterized in that obtaining area information and inputting it into the modulation network. A learning apparatus for learning a modulation network for concealing an original medical image by performing different concealing processes for each region of the original medical image, the learning apparatus comprising:
at least one memory storing instructions; and
at least one processor configured to execute the instructions, wherein the processor is configured to: (I) input a medical image for training into a modulation network to cause the modulation network to form a first region for training and a second region for training of the medical image for training. A process of generating a modulated medical image for learning corresponding to the medical image for training by performing a first concealing process and a second concealing process corresponding to each region in the first region for learning and the second region for learning, respectively; (II) (i) a first discriminator for realizing a first reference medical image that has undergone the first concealing process on the entire region of the medical image for training and the second concealing on the entire region of the medical image for training The modulated medical image for training is input to each of the second discriminators for judging the processed second reference medical image as real, so that the first discriminator and the second discriminator determine whether the modulated medical image for training is real or fake. a sub-process for outputting a first modulated medical image score to a second modulated medical image score for determining recognition, and (ii) inputting the modulated medical image for training into an image identification network, causing the image identification network to: a process of performing a sub-process of performing a learning operation on a medical image and outputting medical feature information for learning corresponding to the modulated medical image for learning; and (III) maximizing a summation loss obtained by weighted summing a first_1 discriminant loss and a 2_1 discriminant loss generated with reference to the first modulated medical image score and the second modulated medical image score, for the training of the image identification network a process of training the modulation network to minimize an accuracy loss generated by referring to a task specific output generated using the medical characteristic information or the medical characteristic information for training and a ground truth corresponding to the training medical image; A learning device to do. 17. The method of claim 16,
(IV) the first_1 discriminant loss and the second_1 discriminative loss generated by the processor with reference to the first modulated medical image score and the second modulated medical image score, respectively, by the first discriminator and the second discriminator , and referring to the first reference medical image score and the second reference medical image score respectively corresponding to the first and second reference medical images input to the first discriminator and the second discriminator, respectively. a process of training each of the first discriminator and the second discriminator to maximize the first_2 discriminant loss and the second_2 discriminant loss, respectively;
Learning apparatus further comprising a. 18. The method of claim 17,
(V) the processor training the image identification network to minimize the accuracy loss;
Learning apparatus further comprising a. 17. The method of claim 16,
Learning apparatus, characterized in that the processor is in a state in which the image identification network has been trained to recognize the medical image for training. 17. The method of claim 16,
The processor, when learning of the modulation network using the summation loss, (i) for the first area for learning, to the first_1 discrimination loss corresponding to the first area for learning, to the second area for learning Learning is carried out by applying a weight higher than the corresponding second_1 discrimination loss, and (ii) for the second area for learning, the second_1 determination loss corresponding to the second area for learning corresponds to the first area for learning Learning apparatus, characterized in that the learning is carried out by applying a weight higher than that of the first_1 discrimination loss. 17. The method of claim 16,
The processor causes the modulation network to conceal the first region for learning to have a relatively low noise density according to the degree of concealing through the first concealing process, and perform the second concealing process The learning apparatus according to claim 1, wherein the second area for learning is subjected to a concealing process to have a relatively high noise density according to the degree of concealing. 22. The method of claim 21,
The processor sets, among the areas of the medical image for training, an area including medical information used for medical condition determination as the first area for learning, and sets an area other than the first area for learning in the medical image for training A learning apparatus, characterized in that it is set as a second area for learning. 17. The method of claim 16,
The processor is configured to obtain the first area information for learning and the second area information for learning of the medical image for training from the labeling information for learning in which a first area for learning and a second area for learning are set in the medical image for training, and the modulation network Learning device, characterized in that input as. 17. The method of claim 16,
The processor inputs the medical image for training into the modulation network to cause the modulation network to perform a convolution operation on the medical image for training at least once through an encoder to generate a feature map for learning, and the feature map for learning through a decoder A learning apparatus, characterized in that by performing a deconvolution operation at least once to output the modulated medical image for learning. A test apparatus for testing a modulation network that conceals an original medical image by performing different concealing processes for each area of the original medical image,
at least one memory storing instructions; and
at least one processor configured to execute the instructions, wherein the processor (I) by means of a learning device (i) inputs a medical image for training into a modulation network to cause the modulation network to modify the medical image for training. A first concealing process for learning and a second concealing process for learning, respectively, corresponding to the first region for learning and the second region for learning, respectively, are performed in the first region for learning and the second region for learning to correspond to the medical image for learning a first discriminator configured to generate a modulated medical image for training, and (ii) (ii-1) to truly determine a first reference medical image that has undergone the first concealing process for training on the entire area of the training medical image; By inputting the modulated medical image for training into each of the second discriminators for truly judging the second reference medical image that has undergone the second concealing process for learning in the entire area of the training medical image, the first discriminator and the second a process for causing a discriminator to output a first modulated medical image score to a second modulated medical image score for determining whether the modulated medical image for training is real or fake, and (ii-2) converting the modulated medical image for training into an image identification network to perform a process for causing the image identification network to perform a learning operation on the modulated medical image for training and output medical feature information for training corresponding to the modulated medical image for training, (iii) the first modulated medical image Maximizes the summation loss obtained by weighted summing the first_1 discriminant loss and the 2_1 discriminant loss generated by referring to the score and the second modulated medical image score, and obtains the medical feature information for learning or the medical feature information for learning of the image identification network In a state in which the modulation network is trained to minimize the accuracy loss generated by referring to the task specific output generated using the training image and the ground truth corresponding to the training medical image, the process of acquiring medical images; and (II) inputting a medical image for testing into the modulation network to cause the modulation network to perform a first concealing process for testing and a test corresponding to each of the first region for testing and the second region for testing of the medical image for testing. a process of generating a modulated medical image for testing corresponding to the medical image for testing by performing a second concealing process on the first region for testing and the second region for testing, respectively; A test device that performs 26. The method of claim 25,
In the process (I) above,
The first discriminator and the second discriminator respectively minimize the first_1 discrimination loss and the second_1 discrimination loss by the learning device, and the first discriminator and the second discriminator respectively input to the first discriminator and the second discriminator that the first_2 discriminant loss and the second_2 discriminant loss generated by referring to the first reference medical image score and the second reference medical image score corresponding to the reference medical image and the second reference medical image, respectively, are respectively maximized Characterized test device. 27. The method of claim 26,
The image identification network is trained by the learning device to minimize the accuracy loss or to be trained to recognize the training medical image. 26. The method of claim 25,
The processor causes the modulation network to conceal the first region for the test to have a relatively low noise density according to the degree of concealing through the first concealing process for the test, 2 . A test apparatus, characterized in that the second area for the test is subjected to a concealing process to have a relatively high noise density according to the degree of concealing through a concealing process. 29. The method of claim 28,
The processor sets, among the regions of the test medical image, a region including medical information used for medical condition determination as the first test region, and sets the test first region in the test medical image. A test apparatus, characterized in that the excluded area is set as the second area for the test. 26. The method of claim 25,
The processor is configured to: the test first area information and the test second area of the test medical image from the test labeling information in which the test first area and the test second area are set in the test medical image A test device, characterized in that obtaining information and inputting it into the modulation network.

Images