KR20230068060A - Method and apparatus for processing a medical image based on AI Inference - Google Patents

Abstract

The present invention selects at least one or more of a plurality of AI inference models as a medical image is input through linkage with a medical device, and analyzes the input medical image using the selected AI inference model to apply to a body part. An AI inference unit that generates AI inference result data on whether or not there is a lesion on the disease, matches the AI inference result data to the input medical image, stores the result data in a database, and when an image request is received from the viewer terminal, the AI inference result data It is possible to provide an AI inference-based medical image processing device including a medical image management unit that transmits a matched medical image to the viewer terminal and displays it on a display of the viewer terminal.

Description

Medical image processing apparatus and method based on AI inference {Method and apparatus for processing a medical image based on AI Inference}

The present invention relates to artificial intelligence analysis of medical images.

PACS (Picture Archiving Communication System) is a medical image storage and transmission system that integrates functions that acquire and store image information obtained through an image diagnosis device in a digital state, and transmit and retrieve the reading and medical records together. refers to the system provided by

In particular, PACS converts all radiological examination results taken by X-Ray, CT, MRI, PET, and SPECT into digital images, and stores them in a mass storage device at the same time as they are taken so that radiologists can read them through the monitor. A film taken of a patient's condition can be uploaded to a computer image by a doctor in a matter of seconds and used for treatment.

Accordingly, PACS increases the accuracy and efficiency of diagnosis by integrating and managing all medical images in a medical institution, reduces waiting time for patients and simplifies transfer between medical institutions, and provides digital files instead of film used in the existing system. It has the advantage of being easy to store and manage. In addition, the patient's current and past images can be freely inquired and used for reading, and there is an effect of strengthening the quality of reading through image quality correction.

On the other hand, with the recent development of technologies such as deep learning and artificial intelligence, various forms of AI inference models are being developed to analyze medical images with real-time artificial intelligence models for accurate prediction and diagnosis of lesions and to process them quickly and accurately. .

Patent Publication No. 10-2016-0103578 (published date: 2016.09.02.)

According to the present invention, as a medical image is input, analysis is performed using any one of a plurality of AI inference models to generate AI inference result data, which is then matched to a medical image and stored, and AI inference result based on a user's request. An AI inference-based medical image processing device and method capable of providing data-matched medical images are provided.

In order to achieve the above object, an AI inference-based medical image processing apparatus according to an embodiment of the present invention selects at least one or more of a plurality of AI inference models as a medical image is input through interworking with a medical device. and an AI inference unit that analyzes the input medical image using the selected AI inference model and generates AI inference result data on whether a body part has a lesion, and the AI inference on the input medical image. A medical image management unit that matches result data and stores them in a database, and transmits a medical image matched with the AI inference result data to the viewer terminal when an image request is received from the viewer terminal and displays it on a display of the viewer terminal. can

According to an embodiment of the present invention, the medical image processing device interworks with a plurality of AI service provider devices that provide AI inference services for medical images based on an AI inference model, and based on the selected AI inference model. An external linkage unit requesting and receiving a setting file and weight parameters for the selected AI inference model from an AI service provider device providing AI inference service, wherein the AI inference unit transmits the received setting file and weight parameters It is possible to generate AI inference result data on whether there is a lesion on a body part by analyzing the input medical image.

According to an embodiment of the present invention, the device of the AI service provider providing the AI inference service based on the selected AI inference model encrypts the setting file and weight parameters using the private key of the AI service provider. After providing this to the medical image processing device, the external interlocking unit may decrypt the encrypted setting file and weight file using the public key of the AI service provider.

According to an embodiment of the present invention, the medical image processing device sets data weights for an acceleration model for accelerating artificial intelligence execution and weight parameters based on a setting file and weight parameters for an AI inference model received through the external interlocking unit. The AI inference unit further includes an AI inference result for the medical image based on the execution model data can be generated.

According to an embodiment of the present invention, the medical image processing device receives an AI inference model that it serves from each of the plurality of AI service provider-side devices, a setting file for the AI inference model, and weight parameters, and the provided An AI model register that generates execution models for artificial intelligence execution acceleration using the received setting file and weight parameters, matches each of the generated execution models with identifiers of multiple AI service providers, and stores them in the AI model database. The AI inference unit may select an execution model stored in the AI model database as the AI inference model.

According to an embodiment of the present invention, the AI inference unit receives AI inference model selection information from an external device and selects one of the plurality of AI inference models, or uses attribute values in the medical image to infer the plurality of AIs. You can choose any one of the models.

According to an embodiment of the present invention, the medical image processing device receives the medical image having a DICOM format through a DIOCM protocol, and if the medical image is a non-DICOM image, the non-DICOM image is processed using a predetermined plug-in. A DICOM image may be received from the pre-processing unit in conjunction with a pre-processing unit that converts a non-DICOM image into a DICOM image.

In order to achieve the above object, an AI inference-based medical image processing method according to an embodiment of the present invention includes the steps of receiving a medical image through interworking with a medical device in a medical image processing device, and processing the medical image. Selecting at least one of a plurality of AI inference models in a device, and analyzing the input medical image using the selected AI inference model in the medical image processing device to determine whether or not a body part has a lesion. Generating AI inference result data, matching the AI inference result data to the input medical image in the medical image processing device and storing the result data in a database, and receiving an image request from a viewer terminal in the medical image processing device. Then, a medical image management unit may include transmitting a medical image matched with the AI inference result data to the viewer terminal and displaying the medical image on a display of the viewer terminal.

According to an embodiment of the present invention, the medical image processing device interworks with a plurality of AI service provider devices that provide AI inference services for medical images based on an AI inference model, and generates the AI inference result data. The step includes requesting and receiving a configuration file and weight parameters for the selected AI inference model from an AI service provider device providing an AI inference service based on the selected AI inference model, and receiving the received configuration file and weight parameters. The method may include performing an analysis on the input medical image using parameters to generate AI inference result data on whether a body part has a lesion or not.

According to an embodiment of the present invention, the receiving step may include the setting encrypted using a private key of the AI service provider in a device of an AI service provider providing an AI inference service based on the selected AI inference model. In the step of receiving a file and weight parameters and generating AI inference result data, the medical image processing device decrypts the encrypted setting file and weight file using the public key of the AI service provider, and then based on this The AI inference result data may be generated.

According to an embodiment of the present invention, the medical image processing method includes calculating data weights for an acceleration model for accelerating artificial intelligence execution and weight parameters based on a setting file and weight parameters for the received AI inference model; The step of converting the medical image into a form usable for the acceleration model and then generating an execution model based thereon, wherein the step of generating the AI inference result data sets the execution model as an AI inference model, AI inference result data for medical images can be created.

According to an embodiment of the present invention, the medical image processing method includes the steps of receiving an AI inference model serviced by the device from each of a plurality of AI service provider side devices, a setting file for the AI inference model, and weight parameters; Creating execution models for artificial intelligence execution acceleration using provided setting files and weight parameters, and matching each of the generated execution models with identifiers of a plurality of AI service providers and storing them in an AI model database. Further, the selecting may select an execution model stored in the AI model database as the AI inference model.

According to an embodiment of the present invention, the selecting may include selecting one of the plurality of AI inference models by receiving AI inference model selection information from an external device, or using attribute values in the medical image to select the plurality of AI inference models. Any one of the inference models can be selected.

According to an embodiment of the present invention, the receiving of the medical image may include converting the non-DICOM image into a DICOM image using a predetermined plug-in when the medical image is a non-DICOM image, and then receiving the non-DICOM image. can

According to the above-described embodiment of the present invention, as a medical image is input, analysis is performed using any one of a plurality of AI inference models to generate AI inference result data, which is matched to the medical image and stored, and user By providing medical images matched with AI inference result data upon request, various types of medical-related AI services are possible.

1 is a diagram showing the overall configuration of a medical image processing system based on artificial intelligence inference according to an embodiment of the present invention.
2 is a block diagram showing the detailed configuration of a medical image processing apparatus according to an embodiment of the present invention.
3 is a flowchart illustrating a process of providing AI inference result data for a medical image according to an embodiment of the present invention.
4 is a flowchart illustrating a process of servicing AI inference result data for a medical image according to another embodiment of the present invention.
5 is a flowchart illustrating a process of servicing AI inference result data for medical images according to another embodiment of the present invention.

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings. The detailed descriptions that follow are provided to provide a comprehensive understanding of the methods, devices and/or systems described herein. However, this is only an example and the present invention is not limited thereto.

In describing the embodiments of the present invention, if it is determined that the detailed description of the known technology related to the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted. In addition, terms to be described later are terms defined in consideration of functions in the present invention, which may vary according to the intention or custom of a user or operator. Therefore, the definition should be made based on the contents throughout this specification. Terminology used in the detailed description is merely for describing the embodiments of the present invention and should in no way be limiting.

Hereinafter, a medical image processing system and method based on AI (Artificial Intelligence) inference will be described with reference to the accompanying drawings.

Prior to description, the AI inference model used in the embodiments of the present invention is generated through artificial intelligence learning, collects learning data necessary for performing artificial intelligence such as machine learning and deep learning, and extracts features from the collected learning data. In order to do this, weight values are calculated for each learning data dimension, and after learning data patterns by accumulating the calculated weight values, it can be generated based on this. That is, the AI inference model may be composed of weight parameters and an artificial intelligence model as a learning result.

1 is a diagram showing the overall configuration of a medical image processing system based on artificial intelligence inference according to an embodiment of the present invention, and FIG. 2 is a block diagram showing the detailed configuration of a medical image processing apparatus according to an embodiment of the present invention. am.

As shown in FIG. 1, the artificial intelligence inference-based medical image processing system receives a medical device 10 that generates a medical image and the medical image generated by the medical device 10, and then converts the medical image into an arbitrary AI inference model. It may include a medical image processing device 20 and a viewer terminal 30 that determine through AI inference and generate AI inference result data and store them in the database 22 .

The medical device 10 , the medical image processing device 20 , and the viewer terminal 40 may be interconnected through a wired or wireless network (not shown) to transmit and receive data.

In an embodiment of the present invention, the medical device 10 is a DICOM form (such as X-ray, ultrasound, CT (Computed Tomography), MRI (Magnetic Resonance Imaging), PET (Positron Emission Tomography), MRA (Magnetic Resonance Angiography), etc. Modality) or non-DICOM format medical images.

Examples of the medical device 10 that generates a medical image having a non-DICOM format include equipment such as OT, ES, and SC, or dental equipment.

In the case of a medical image having a non-DICOM format, it may be converted into a DICOM image by the pre-processing unit 15 interworking with the medical image processing device 20 and then provided to the medical image processing device 20 . Specifically, as the non-DIOCM image is received, the pre-processing unit 15 converts the non-DICOM image into a DIOCM image using a predetermined plug-in for conversion, and then transmits the non-DICOM image to the medical image processing device 20 using a DICOM protocol. there is.

Meanwhile, as the DIOCM image is received, the pre-processing unit 15 may transmit the DICOM image to the medical image processing device 20 based on the DICOM protocol.

As shown in FIG. 2 , the medical image processing apparatus 20 selects one of a plurality of AI inference models, and analyzes the input medical image using the selected AI inference model to analyze the body part. It may include an AI inference unit 200 that generates AI inference result data on whether or not a lesion is present. In this case, the AI inference unit 200 may store and manage medical images and AI inference result data in the database 22 .

In addition, upon receiving an external request, for example, a user search request, the medical image processing apparatus 20 transmits the medical image and AI inference result data to the viewer terminal 30 that requested the search, and displays the viewer terminal 30 (not shown). ) may further include a medical image management unit 210 capable of being displayed. Here, the AI inference unit 200 and the medical image management unit 210 may be implemented by a Picture Archiving Communication System (PACS) server.

As described above, in order to select one AI inference model based on a plurality of AI inference models, the medical image processing apparatus 20 according to an embodiment of the present invention provides an AI inference service for medical images based on the AI inference model. Integrates with a plurality of AI service provider side devices 40 that provide AI inference model setting files and AI inference from the AI service provider side device 40 that provides AI inference service based on the selected AI inference model An external interlocking unit 220 requesting and receiving weight parameters to be applied to the model may be further included. In this case, the external interlocking unit 220 may provide the received setting file and weight parameters to the AI inference unit 200.

Accordingly, the AI inference unit 200 of the medical image processing device 20 converts a medical image having a DICOM format into an image that can be applied to an AI inference model based on the setting file, and then converts the converted image and weight parameters Analysis of medical images is performed using , and AI inference result data can be generated through the analysis.

Meanwhile, the external interlocking unit 220 of the medical image processing device 20 according to an embodiment of the present invention may manage the public key of the AI service provider. In this case, the device 40 of the AI service provider may encrypt the setting file and weight parameters using the private key of the AI service provider, and then provide them to the medical image processing device 20 . Accordingly, the external interlocking unit 220 of the medical image processing device 20 may obtain the setting file and weight parameters by decrypting the encrypted setting file and weight parameters using the public key.

Meanwhile, the medical image processing apparatus 20 according to an embodiment of the present invention may generate an execution model for an AI inference model through the execution model generator 230 . In this case, the AI inference unit 200 may set the execution model generated by the execution model generation unit 230 as an AI inference model to generate AI inference result data for the medical image.

A process of generating an execution model by the execution model generation unit 230 will be described below.

The execution model generation unit 230 of the medical image processing device 20 according to an embodiment of the present invention is an AI inference model for artificial intelligence execution acceleration (hereinafter referred to as 'acceleration After converting the medical image into a form usable in the acceleration model, that is, a custom layer, the process of adjusting the position of the operation function and module and performing the operation method is performed based on the custom layer. Optimization can be performed to generate an execution model for an accelerated model.

In addition, the execution model generation unit 230 calculates the data weight for the weight parameter based on the setting file and the weight parameter, separates the learning function from the execution function in the AI inference model, and then removes the learning function. You can create an action model.

The generated execution model may be provided to the AI reasoning unit 200 .

Accordingly, the AI inference unit 200 may generate AI inference result data by performing analysis on the medical image after implementing an execution environment corresponding to the execution model.

In addition, the AI reasoning unit 200 calculates the optimal value of the memory allocation amount required for each execution stage of the execution model in order to accelerate the speed of generating AI inference result data, determines whether the execution stage including parallel processing is completed, and completes Memory optimization can be implemented by reusing a memory area from which data that is not used again in a step is deleted.

That is, the AI reasoning unit 200 determines whether each of the many calculation processing processes is completed, and deletes all unnecessary data at the stage in which data processing is completed. At this time, only the data calculated as the result value in each step is left, and the data used for calculating the result value may be regarded as unnecessary data that is not used again and may be deleted.

In addition, the AI reasoning unit 200 may optimize the memory of the execution model by reusing the generated spare memory area after deleting unnecessary data in the process completion stage.

Meanwhile, the medical image processing device 20 according to an embodiment of the present invention may further include an AI model registration unit 240 for processing an AI inference model registration request from the device 40 of an external AI service provider. .

The AI model registration unit 240 receives an AI inference model registration request using the AI inference model it serves, a setting file for the AI inference model, weight parameters, etc. from the AI service provider side device 40, and receives the AI inference model registration request. After providing an inference model, a setting file for the AI inference model, weight parameters, etc. to the execution model generator 230, an execution model for the AI inference model may be provided in response thereto.

In addition, the AI model registration unit 240 may match the execution model provided from the execution model generation unit 230 to an identifier of an AI service provider, store it in the AI model database 24, and manage the execution model.

In addition, the AI model registration unit 240 provides setting information for implementing an execution environment for an execution model generated in the course of the AI inference unit 200 generating AI inference data through an arbitrary execution model to an AI service provider. It can be matched with the identifier of and stored in the AI model database 24.

In this case, the AI inference unit 200 searches the AI model database 24 for an execution model and setting information for the identifier of an AI service provider servicing the selected AI inference model, and then generates AI inference result data based thereon. can do.

On the other hand, in an embodiment of the present invention, selection of an AI inference model is automatically selected based on attribute values in a medical image, for example, a photographed region, examination information, etc., or based on a request from the viewer terminal 30, a request from a doctor, or a request from a hospital side. can

A process of generating AI inference result data by the medical image processing apparatus 20 having the configuration described above will be described with reference to FIGS. 3 to 5 .

3 is a flowchart illustrating a process of providing AI inference result data for a medical image according to an embodiment of the present invention.

As shown in FIG. 3 , as a medical image is received from the medical device 10 (S300), the medical image processing apparatus 20 determines the attribute value of the medical image, a request from the viewer terminal 30, a doctor's request, or a hospital side. Based on the request, an AI inference model to be applied to the medical image is selected (S302).

Then, the medical image processing device 20 generates AI inference result data by applying the selected AI inference model to the medical image, matches the AI inference result data to the medical image, and stores the data in the database 22 (S304).

Subsequently, when a user's request, for example, a medical image request is received from the viewer terminal 30 (S306), the medical image processing device 20 converts the medical image stored in the database 22 and AI inference result data matched thereto to the viewer terminal. (30) to be displayed on the display of the viewer terminal 30 (S308).

4 is a flowchart illustrating a process of servicing AI inference result data for a medical image according to another embodiment of the present invention.

As shown in FIG. 4 , as a medical image is received from the medical device 10 (S400), the medical image processing apparatus 20 determines the attribute value of the medical image, a request from the viewer terminal 30, a doctor's request, or a hospital side. Based on the request, an AI inference model to be applied to the medical image is selected (S402).

Then, the medical image processing device 20 requests and receives a setting file and weight parameters from the AI service provider-side device 40 serving the selected AI inference model (S404).

Then, the medical image processing device 20 generates an execution model for the AI inference model based on the setting file and weight parameters (S406). Specifically, the medical image processing apparatus 20 calculates an acceleration model for artificial intelligence execution acceleration and data weights for the weight parameters using a setting file and weight parameters, and converts the medical image into a form usable in the acceleration model, that is, custom After converting to a layer, an execution model for the acceleration model can be created by optimizing through the process of adjusting the location of calculation functions and modules and performing calculation methods based on the custom layer.

Then, the medical image processing device 20 analyzes the medical image as an execution model to generate AI inference result data for the medical image, matches the AI inference result data to the medical image, and stores the data in the database 22. Do (S408).

Subsequently, when a user's request, for example, a request for a medical image is received from the viewer terminal 30 (S410), the medical image processing apparatus 20 converts the medical image stored in the database 22 and AI inference result data matched thereto to the viewer terminal. (30) to be displayed on the display of the viewer terminal 30 (S412).

5 is a flowchart illustrating a process of servicing AI inference result data for medical images according to another embodiment of the present invention.

As shown in FIG. 5 , the medical image processing device 20 infers AI by using the AI inference model it serves from the device 40 of the AI service provider, the setting file for the AI inference model, and weight parameters. A model registration request is received (S500).

Thereafter, the medical image processing device 20 generates an execution model using the AI inference model, setting files and weight parameters for the AI inference model, etc. (S502), and then matches it to an identifier of an AI service provider to generate an AI model database. It is stored in (24) (S504).

The above steps may be performed with a plurality of AI service provider side devices 40 to build an AI model database 24 composed of different execution models matched to identifiers of a plurality of AI service providers.

Then, as the medical image is received from the medical device 10 as the medical image is received (S506), the medical image processing device 20 determines the attribute value of the medical image, the request of the viewer terminal 30, the doctor's request, or the hospital side. Based on the request, the type of AI inference model to be applied to the medical image is selected, and an execution model corresponding to the type of the selected AI inference model is searched in the AI model database 24 (S508).

Then, the medical image processing device 20 sets setting information for implementing an execution environment for the searched execution model (S510), and executes the execution model in the implemented execution environment based on the set setting information to obtain medical data. After generating the AI inference result data for the image, the AI inference result data is matched with the medical image and stored in the database 22 (S512).

In addition, the medical image processing device 20 stores setting information set for implementing an execution environment in the AI model database 24 (S514). Specifically, the medical image processing device 20 may match setting information to an identifier of an AI service provider of a searched execution model and store the searched execution model in the AI model database 24 .

Subsequently, when a user's request, for example, a request for a medical image is received from the viewer terminal 30 (S516), the medical image processing device 20 converts the medical image stored in the database 22 and AI inference result data matched thereto to the viewer terminal. (30) to be displayed on the display of the viewer terminal 30 (S518).

Through the steps described above, an AI model database 24 composed of execution models, setting information, etc. for AI inference models serviced by a plurality of AI service provider side devices 40 can be built. Accordingly, the medical image processing apparatus 20 may determine whether setting information exists in the searched execution model, and if the setting information exists, implement an execution environment using the setting information.

The above description is merely an example of the technical idea of the present invention, and various modifications and variations can be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments expressed in the present invention are not intended to limit the technical idea of the present invention, but to explain, and the scope of the present invention is not limited by these embodiments. The protection scope of the present invention should be construed according to the claims below, and all technical ideas that are equivalent or within the scope of equivalents should be construed as being included in the scope of the present invention.

10: Medical device
15: pre-processing unit
20: medical image processing device
30: viewer terminal
40: AI service provider side device
200: AI reasoning unit
210: medical image management unit
220: external interlocking unit
230: execution model generation unit
240: AI model register

Claims (14)

As a medical image is input through interworking with a medical device, at least one of a plurality of AI inference models is selected, and the input medical image is analyzed using the selected AI inference model to determine whether a body part has a lesion or not. An AI reasoning unit that generates AI reasoning result data for;
The AI inference result data is matched with the input medical image and stored in a database, and when an image request is received from the viewer terminal, the medical image matched with the AI inference result data is transmitted to the viewer terminal to display the viewer terminal. An AI inference-based medical image processing device including a medical image management unit for displaying on.
According to claim 1,
The medical image processing device,
It works with multiple AI service provider devices that provide AI inference services for medical images based on the AI inference model, and reminds the AI service provider device that provides AI inference services based on the selected AI inference model. Further comprising an external interlocking unit that requests and receives a setting file and weight parameters for the selected AI inference model,
The AI reasoning unit,
An AI inference-based medical image processing device that analyzes the input medical image using the received setting file and weight parameters to generate AI inference result data on whether a body part has a lesion.
According to claim 2,
The device of the AI service provider that provides the AI inference service based on the selected AI inference model encrypts the setting file and weight parameters using the private key of the AI service provider and sends them to the medical image processing device. provide,
The external interlocking unit,
An AI inference-based medical image processing device for decrypting the encrypted setting file and weight file using a public key of the AI service provider.
According to claim 2,
The medical image processing device,
An acceleration model for accelerating artificial intelligence execution and data weights for weight parameters are calculated based on the setting file and weight parameters for the AI inference model received through the external interlocking unit, and the medical image can be used for the acceleration model. Further comprising an execution model generation unit for generating an execution model based on the conversion to
The AI reasoning unit,
An AI inference-based medical image processing device that generates AI inference result data for the medical image based on the execution model.
According to claim 1,
The medical image processing device,
Receives an AI inference model that it serves, a setting file for the AI inference model, and weight parameters from each of the devices of a plurality of AI service providers, and uses the provided setting file and weight parameters to accelerate artificial intelligence execution. Further comprising an AI model register for generating each execution model, matching each of the generated execution models with an identifier of each of a plurality of AI service providers, and storing them in an AI model database,
The AI reasoning unit,
An AI inference-based medical image processing device that selects an execution model stored in the AI model database as the AI inference model.
According to claim 1,
The AI reasoning unit,
AI inference-based medical care that selects one of the plurality of AI inference models by receiving AI inference model selection information from an external device or selects one of the plurality of AI inference models using attribute values in the medical image image processing device.
According to claim 1,
The medical image processing device,
Receiving the medical image having a DICOM format through a DIOCM protocol,
If the medical image is a non-DICOM image, the non-DICOM image is interlocked with a pre-processing unit that converts the non-DICOM image into a DICOM image using a predetermined plug-in and receives the DICOM image from the pre-processing unit based on AI reasoning. processing unit.
Receiving a medical image through interworking with a medical device in a medical image processing device;
selecting at least one of a plurality of AI inference models in the medical image processing device;
generating AI inference result data on whether a body part has a lesion by analyzing the input medical image using the selected AI inference model in the medical image processing device;
matching the AI inference result data to the input medical image in the medical image processing device and storing the result data in a database;
When the medical image processing device receives an image request from a viewer terminal, a medical image management unit including a step of transmitting a medical image matched with the AI inference result data to the viewer terminal and displaying it on a display of the viewer terminal. A medical image processing method based on AI inference.
According to claim 8,
The medical image processing device,
Interlocks with multiple AI service provider devices that provide AI inference services for medical images based on AI inference models,
The step of generating the AI inference result data,
Requesting and receiving a setting file and weight parameters for the selected AI inference model from an AI service provider device providing an AI inference service based on the selected AI inference model;
An AI inference-based medical image processing method comprising performing an analysis on the input medical image using the received setting file and weight parameters to generate AI inference result data on whether a body part has a lesion.
According to claim 9,
In the receiving step,
Receiving the setting file and weight parameters encrypted using the private key of the AI service provider in a device of an AI service provider providing an AI inference service based on the selected AI inference model;
The step of generating the AI inference result data,
The AI inference-based medical image processing method of decrypting the encrypted setting file and weight file using the public key of the AI service provider in the medical image processing device and generating the AI inference result data based thereon.
According to claim 9,
The medical image processing method,
Calculating data weights for an acceleration model for accelerating artificial intelligence execution and weight parameters based on the setting file and weight parameters for the received AI inference model;
Converting the medical image into a form usable for the acceleration model and then generating an execution model based thereon;
The step of generating the AI inference result data,
An AI inference-based medical image processing method of generating AI inference result data for the medical image by setting the execution model as an AI inference model.
According to claim 8,
The medical image processing method,
Receiving from each of a plurality of AI service provider-side devices an AI inference model that they serve, a setting file for the AI inference model, and weight parameters;
generating an execution model for artificial intelligence execution acceleration using the provided setting file and weight parameters;
Further comprising matching each of the generated execution models with an identifier of each of a plurality of AI service providers and storing them in an AI model database,
The selection step is
An AI inference-based medical image processing method for selecting an execution model stored in the AI model database as the AI inference model.
According to claim 8,
The selection step is
AI inference-based medical care that selects one of the plurality of AI inference models by receiving AI inference model selection information from an external device or selects one of the plurality of AI inference models using attribute values in the medical image Image processing method.
According to claim 8,
Receiving the medical image,
If the medical image is a non-DICOM image, the AI inference-based medical image processing method converts the non-DICOM image into a DICOM image using a predetermined plug-in and receives the non-DICOM image.

Images