Method and system for analyzing medical image by using artificial intelligence

Abstract

The present invention relates to a method and a system for analyzing a medical image by using artificial intelligence. More specifically, a method for analyzing a medical image according to the present invention may comprise the steps of: receiving, from a user terminal, a selection of at least one object to be measured; on the basis of the selection of the at least one object to be measured, specifying at least one human body component corresponding to the object to be measured; selecting at least one analysis model trained to perform analysis on the specified human body component from among a plurality of pre-trained analysis models; processing the medical image as an input of the analysis model; and generating, using output data of the analysis model, a measurement value for the object to be measured.

Description

Medical image analysis method and system using artificial intelligence

[0001] The present invention relates to a method and system for medical image analysis using artificial intelligence.

[0002] With the recent development of artificial intelligence, there has been a surge in cases of achieving outstanding results across various industrial sectors. In particular, there has been a rapid increase in cases where patterns are learned from vast amounts of data and results are achieved in the analysis of vast amounts of unstructured data through machine learning (ML) and / or deep learning technologies.

[0003] In this regard, attempts to apply artificial intelligence technology in the medical field have recently become active. For instance, the medical sector is utilizing AI technology to diagnose patients' diseases by analyzing various medical data (e.g., X-ray, CT, and MRI images).

[0004] However, artificial intelligence technology currently used in the medical field is primarily focused on determining the presence of lesions or diagnosing specific diseases. Furthermore, due to the nature of X-ray images, multiple structures (e.g., bones) often appear overlapping, which limits the ability to extract objective indicators such as the angle or length of structures from images using conventional methods. Accordingly,

[0005] There is still a need for methods to be introduced in the medical field to significantly improve the accuracy and efficiency of patient care.

[0006] (Patent Document 1) Korean Registered Patent Publication No. 10-2566183 (Aug. 10, 2023)

[0007] The present invention aims to provide a medical image analysis method and system using artificial intelligence that can be flexibly utilized for various medical image analyses.

[0008] More specifically, the present invention aims to provide a medical image analysis method and system using artificial intelligence capable of recognizing human body components in medical images and generating analysis results for the recognized human body components.

[0009] Furthermore, the present invention aims to provide a medical image analysis method and system using artificial intelligence that can increase the accuracy of medical image analysis and improve the accuracy and efficiency of patient care.

[0010] Furthermore, the present invention aims to provide a medical image analysis method and system using artificial intelligence that can improve the efficiency of a doctor's consultation time and medical research time by flexibly responding to user requirements and processing multiple images efficiently and quickly.

[0011] To solve the problem described above, the medical image analysis method according to the present invention may include the steps of: specifying a medical image; receiving a selection of at least one measurement target from a user terminal; specifying at least one human body component corresponding to the measurement target based on the selection of the at least one measurement target; selecting at least one analysis model trained to perform analysis on the specified human body component among a plurality of previously trained analysis models; processing the medical image as input to the analysis model; and generating a measurement value for the measurement target using the output data of the analysis model.

[0012] Furthermore, each of the above multiple analysis models may be trained for each of the different human body components to perform analysis on each of the multiple different human body components.

[0013] Furthermore, the step of generating the measurement value may include obtaining at least one of a mask, a specific point, and a specific location corresponding to the specified human body component from the analysis model, performing contouring on the obtained mask, and generating the measurement value for the measurement target based on the result of the contouring and the specific point.

[0014] Furthermore, if the specified human body component includes a first human body component and a second human body component different from the first human body component, the specific analysis model may include a first analysis model trained to perform an analysis on the first human body component and a second analysis model trained to perform an analysis on the second human body component.

[0015] Furthermore, the first analysis model can recognize a first region corresponding to the set of the first human body components in the medical image, and based on the recognized first region, obtain at least one of a first mask, a first specific point, and a first specific location corresponding to the set of the first human body components.

[0016] Furthermore, the second analysis model, which is different from the first analysis model, can recognize a second region corresponding to the set of the second human body components in the medical image, and based on the recognized second region, can obtain at least one of a second mask, a second specific point, and a second specific location corresponding to the set of the second human body components.

[0017] In one embodiment, a third analysis model different from the first analysis model and the second analysis model can recognize the specific point in the medical image and generate information about the recognized specific point.

[0018] Furthermore, in the step of generating the measurement value, the boundaries of each of the first human body component and the second human body component are recognized based on the first mask, the second mask, and the specific point, and the contours of each of the first human body component and the second human body component are extracted according to the recognized boundaries of each of the first human body component and the second human body component, and the measurement value for the measurement target can be generated based on the extracted contours of each of the first human body component and the second human body component and the specific point.

[0019] Furthermore, the above measurement value may include at least one of the angle, length, and area calculated for the specified human body component.

[0020] Meanwhile, the medical image analysis system according to the present invention includes a control unit for specifying a medical image, and the control unit receives at least one measurement target from a user terminal, and based on the selection of the at least one measurement target, specifies at least one human body component corresponding to the measurement target, selects at least one analysis model trained to perform analysis on the specified human body component among a plurality of previously trained analysis models, processes the medical image as an input to the analysis model, and generates a measurement value for the measurement target using the output data of the analysis model.

[0021] Meanwhile, the program according to the present invention is a program that is executed by one or more processes in an electronic device and can be stored on a computer-readable recording medium, and the program may include instructions for performing the steps of: specifying a medical image; receiving a selection of at least one measurement target from a user terminal; specifying at least one human body component corresponding to the measurement target based on the selection of the at least one measurement target; selecting at least one analysis model trained to perform an analysis of the specified human body component among a plurality of previously trained analysis models; processing the medical image as an input to the analysis model; and generating a measurement value for the measurement target using the output data of the analysis model.

[0022] As described above, according to the medical image analysis method and system of the present invention, specific human body components can be accurately recognized and measured in medical images by analyzing medical images using analysis models specialized for each human body component. Through this, the present invention enables accurate analysis even for overlapping elements, which reduces errors that may occur when analyzing with a single model and allows for the effective analysis of complex skeletal structures.

[0023] Furthermore, according to the medical image analysis method and system of the present invention, the accuracy of diagnosis can be significantly improved by analyzing human body components in a medical image using multiple analysis models and generating measurement values ​​based on the analysis results. In particular, the present invention enables the derivation of more precise diagnostic results by selecting and applying an optimized analysis model for each of the various human body components.

[0024] Furthermore, according to the medical image analysis method and system of the present invention, when analyzing multiple (or large volume) images, the time and volume of analysis can be efficiently reduced by selecting and analyzing models suitable for the information required by the user, and the reduction of time can be further maximized in terms of the algorithm capable of parallel processing.

[0025] Furthermore, according to the medical image analysis method and system of the present invention, by automatically recognizing and analyzing multiple human body components in a medical image and providing the necessary data to medical staff, clinical and research time can be reduced. In other words, the present invention can contribute to reducing medical research time along with reducing the time spent on clinical consultations and improving accuracy for medical staff.

[0026] Furthermore, the medical image analysis method and system according to the present invention can be usefully applied in various diagnostic environments. That is, the present invention can be usefully utilized in various medical settings or clinical fields.

[0027] FIG. 1 is a conceptual diagram illustrating a medical image analysis system according to the present invention.

[0028] FIGS. 2a and FIGS. 2b are conceptual diagrams for explaining a learning method of a medical image analysis system according to the present invention.

[0029] FIG. 3 is a flowchart illustrating a medical image analysis method according to the present invention.

[0030] FIGS. 4a, FIGS. 4b, FIGS. 5a, FIGS. 5b, FIGS. 5c and FIGS. 6 are conceptual diagrams for explaining a medical image analysis method according to the present invention.

[0031] FIGS. 7a and 7b are conceptual diagrams for explaining an embodiment of a user interface screen according to the present invention.

[0032] Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Identical or similar components are assigned the same reference number regardless of the drawing symbols, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" used for components in the following description are assigned or used interchangeably solely for the ease of drafting the specification and do not inherently possess distinct meanings or roles. Furthermore, in describing embodiments disclosed in this specification, if it is determined that a detailed description of related prior art could obscure the essence of the embodiments disclosed in this specification, such detailed description will be omitted. Additionally, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification; the technical concept disclosed in this specification is not limited by the attached drawings, and it should be understood that they include all modifications, equivalents, and substitutions that fall within the spirit and technical scope of the present invention.

[0033] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. These terms are used solely for the purpose of distinguishing one component from another.

[0034] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0035] Singular expressions include plural expressions unless the context clearly indicates otherwise.

[0036] In this application, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0037] The present invention relates to a method and system for medical image analysis using artificial intelligence. The medical image analysis system according to the present invention may be a system that recognizes human body components (or human body structures) according to a measurement target requested by a user in a medical image and generates an analysis result for the recognized elements.

[0038] Here, human body components may refer to various bones constituting the skeletal structure of the human body (or body) and various anatomical sites (or anatomical structures, anatomical components, etc.) that can be identified in medical imaging. For example, human body components may include individual bones such as the frontal bone, temporal bone, clavicle, scaphoid, femur, talus, calcaneus, and metatarsal bones, and may also include detailed parts of each bone (e.g., specific parts of the bone such as articular surfaces, articular end faces, diaphyses, epiphyses, etc.). Additionally, human body components may include individual human body structures such as the heart, lung, vessel, liver, and tumor mass. In the present invention, human body components may also be referred to as “human body structures” or “human body structure elements.”

[0039] Hereinafter, the present invention will be examined in more detail with reference to the attached drawings. FIG. 1 is a conceptual diagram illustrating a medical image analysis system according to the present invention. FIG. 2a and FIG. 2b are conceptual diagrams illustrating a learning method of a medical image analysis system according to the present invention, FIG. 3 is a flowchart illustrating a medical image analysis method according to the present invention, and FIG. 4a, FIG. 4b, FIG. 5a, FIG. 5b, FIG. 5c, and FIG. 6 are conceptual diagrams illustrating a medical image analysis method according to the present invention. Furthermore, FIG. 7a and FIG. 7b are conceptual diagrams illustrating an embodiment of a user interface screen according to the present invention.

[0040] Meanwhile, as illustrated in FIG. 1, the medical image analysis system (100) according to the present invention may include at least one of an input unit (110), an output unit (120), a communication unit (130), a storage unit (140), and a control unit (150).

[0041] Although not described, the medical image analysis system (100) according to the present invention may include one or more processors, and such processors may include one or more general-purpose processors and / or one or more special-purpose processors (e.g., digital signal processors, tensor processing units (TPUs), graphics processing units (GPUs), neural network processing units (NPUs), application integrated circuits, application semiconductors (ASICs), etc.). One or more processors may be configured to execute instructions, computer-readable instructions, and / or other instructions described herein that are stored (or included) in the storage unit (140). The medical image analysis system and method according to the present invention may perform data processing described below in cooperation with memory and at least one processor. The processor may perform a series of operations and data processing using data and information stored in memory. In this case, memory may be a component of the storage unit (140).

[0042] Meanwhile, the input unit (110) can be configured in various ways as a means of data input. For example, the input unit (110) can be configured to receive user input. The input unit (110) can be configured to receive user input from a user terminal (10). Here, “receiving input” may mean receiving an input signal (or selection signal) corresponding to the user’s input based on input made by the user through the input unit configuration provided in the user terminal (10).

[0043] In addition, the input unit (110) in the present invention does not necessarily mean a hardware means, but can be understood as a channel for receiving input from a user.

[0044] The input unit (110) may also be referred to as a user interface module. The input unit (110) may include a touch screen, a computer mouse, a keyboard, a keypad, a touchpad, a trackball, a joystick, a voice recognition module, or other similar devices. However, the present invention does not limit the type of input unit (110).

[0045] Here, user input may include documents, text, images (or videos), voice, etc. In this case, the medical image analysis system (100) may further include a module that converts voice into text.

[0046] Next, the output unit (120) can output information through an output unit configuration (e.g., a display unit, a touch screen, a speaker, etc.) provided in a user terminal (10) linked to the medical image analysis system (100) according to the present invention. For example, the output unit (120) can output a page (or service page, 1000) linked to the medical image analysis system (100) according to the present invention to the display unit of the user terminal (10). In addition, the output unit (120) does not necessarily mean a hardware means, but can be understood as a channel for outputting results to a user.

[0047] Next, the communication unit (130) may be connected via a wireless or wired network to a user terminal (10), a server (e.g., a central server, an external server, etc.), a device, and at least one network, etc., to receive or transmit overall data and information necessary for the operation of the medical image analysis system (100) according to the present invention.

[0048] Here, the user terminal (10) may include at least one of a mobile phone, a smartphone, a notebook computer, a laptop computer, a slate PC, a tablet PC, an ultrabook, a desktop computer, a digital broadcasting terminal, a PDA (personal digital assistants), a PMP (portable multimedia player), a navigation device, and a wearable device (e.g., a smartwatch, a smart glass, a head-mounted display).

[0049] Furthermore, the communication unit (130) can support various communication methods according to the communication standards of the communicating device.

[0050] For example, the communication unit (130) may be configured to communicate with a communication target using at least one of the following technologies: WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Wi-Fi (Wireless Fidelity) Direct, DLNA (Digital Living Network Alliance), WiBro (Wireless Broadband), WiMAX (World Interoperability for Microwave Access), HSDPA (High Speed ​​Downlink Packet Access), HSUPA (High Speed ​​Uplink Packet Access), LTE (Long Term Evolution), LTE-A (Long Term Evolution-Advanced), 5G (5th Generation Mobile Telecommunication), Bluetooth (Bluetooth™ RFID (Radio Frequency Identification), Infrared Communication (Infrared Data Association; IrDA), UWB (Ultra-Wideband), ZigBee, NFC (Near Field Communication), Wi-Fi Direct, and Wireless USB (Wireless Universal Serial Bus).

[0051] Meanwhile, the storage unit (140) serves to store various data related to the present invention and may include one or more non-transient computer-readable storage media that can be read and / or accessed by at least one of one or more processors (140).

[0052] One or more computer-readable storage media may include volatile and / or non-volatile storage components, such as optical, magnetic, organic, or other memory or disk storage devices. In some examples, the storage unit (140) may be implemented using a single physical device (e.g., one optical, magnetic, organic, or other memory or disk storage device), whereas in other examples, the storage unit (140) may be implemented using two or more physical devices.

[0053] The storage unit (140) may include computer-readable instructions and additional data. The storage unit (140) may include a storage necessary to perform at least some of the methods, scenarios, and techniques described herein and / or at least some of the functions of the device and network.

[0054] Furthermore, at least a portion of the storage unit (140) may be a cloud storage or a cloud server. At least a portion of the data corresponding to user input received from the input unit (110) and the training data may be stored in the storage unit (140).

[0055] That is, the storage unit (140) is sufficient as a space where information necessary for the operation of the medical image analysis system (100) according to the present invention is stored, and it can be understood that there are no restrictions on the physical space.

[0056] Meanwhile, the control unit (150) can perform the role of controlling the overall operation of the medical image analysis system (100) related to the present invention. The control unit (150) can process signals, data, information, etc. that are input or output through the components described above, or perform a series of data processing to provide or process appropriate information and functions to the user.

[0057] The control unit (150) may include at least one analysis model (or module) used in the process of performing an analysis of a plurality of different human body components included in a medical image (or image). In the present invention, the analysis model may be configured to perform an analysis of a human body component corresponding to a measurement target and output a result value (or output data) according to the result of the analysis.

[0058] Here, there may be various types of medical images. For example, a medical image may include at least one of an X-ray image, a CT (Computed Tomography) image, or an MRI (Magnetic Resonance Imaging) image. However, the types of medical images in the present invention are not necessarily limited to the examples mentioned, and may include various other types of medical images. In the present invention, "medical image" may also be referred to as "medical image," "medical data," "medical image data," or "medical image data."

[0059] The medical images described above may include anatomical (or pathological) elements that can be classified (or distinguished) into specific parts or functional units within the human body (or body), such as bones, muscles, nerves, and blood vessels. These elements may be formed to have specific locations and specific shapes. For convenience of explanation, this specification describes the medical images used in the learning and / or inference processes on the premise that they are medical images of a specific part (e.g., the foot) (e.g., X-ray images of the bones constituting the foot). However, the medical image analysis system and method according to the present invention may be configured to enable analysis of various structures included in the human body (e.g., muscle structures, nerve structures, circulatory system structures, blood vessel structures, internal organs, cartilage structures, etc.) in addition to skeletal structures.

[0060] Hereinafter, the learning method of the analysis model according to the present invention will be examined in more detail. At this time, the analysis model utilized (or used) in the present invention is not necessarily limited to the models mentioned, and may include other models in addition to the first analysis model (151), the second analysis model (152), and the third analysis model (153). It is obvious that the analysis model utilized in the present invention may be one or multiple (N) and may be varied depending on the case. Furthermore, one specific analysis model is a method of analyzing at least one human body component, and depending on the case, it may be a method of analyzing including one or multiple (N) human body components.

[0061] The control unit (150) can specify a training data set (or training data) to be used for training at least one analysis model (or a learning target analysis model).

[0062] In this case, the criteria for specifying the training data set may vary. The control unit (150) may specify the training data set to be used for training each of the multiple analysis models (151, 152, 153) based on various criteria.

[0063] In one embodiment, the control unit (150) collects medical images (or images) from at least one of various sources (e.g., a database (DB), web crawling, API, a server linked to the medical image analysis system (100), an external server, etc.) and can specify the collected medical images as a training data set to be used for training each of a plurality of analysis models (151, 152, 153).

[0064] In another embodiment, the control unit (150) may specify a medical image stored in at least one of various storage units (e.g., storage unit (140), memory, database (DB), etc.) as a training data set to be used for training each of a plurality of analysis models (151, 152, 153).

[0065] As illustrated in FIGS. 2a and 2b, the control unit (150) can train each of the plurality of analysis models (151, 152, 153) by processing at least one medical image (e.g., “Foot x-ray”, 210) specified as a learning data set as an input to each of the plurality of analysis models (151, 152, 153).

[0066] Here, the learning target analysis models (151, 152, 153) may be of various types of models. For example, they may include at least one of various architectures such as Unet, RetinaNet, and yolo. However, the types of models used in the present invention are not necessarily limited to the examples mentioned, and may include various other types of models. For convenience of explanation, the following description will be based on the premise that the learning target analysis model is a U-Net-based model.

[0067] Each of the learning target analysis models (151, 152, 153) consists of an encoder (or contracting path) and a decoder (or expanding path), and the structures of the encoder and decoder may be symmetrical to each other. The part connecting these encoder and decoder may be named a bridge.

[0068] In one embodiment, the encoder receives image data as input and can convert it into a feature map that becomes progressively smaller through a plurality of convolution layers and pooling layers. Each convolution block consists of two 3x3 convolution layers and a ReLU activation function, and the max pooling layers can reduce the size of the feature map by half and extract important features (or characteristics).

[0069] In addition, the decoder restores the size of the feature map back to its original size through an upsampling layer and a convolution layer, increases the size of the feature map using upsampling or transpose convolution, and can fine-tune the upsampled feature map through a layer consisting of two 3x3 convolution layers and a ReLU activation function. In this process, a skip connection is used to pass the feature map from the encoding process to the decoder, allowing it to be combined with the upsampled feature map.

[0070] Furthermore, in the final output layer, 1x1 convolution can be used to calculate the probability that each pixel belongs to a specific class. The class probability of each pixel is output using a Softmax or Sigmoid activation function, thereby generating the final segmented image.

[0071] However, in addition to the models examined above, additional learning target analysis models such as semantic segmentation, instance segmentation, object detection, and key point detection may be further included (or utilized). For example, the present invention may provide a user with a function to select an appropriate learning target analysis model for each structure.

[0072] In this regard, each of the plurality of analysis models (151, 152, 153) according to the present invention may exist by being learned for each of the different elements so as to perform analysis on each of the plurality of different human body components. That is, each of the plurality of analysis models (151, 152, 153) may be learned independently (or individually) for each of the plurality of different human body structural elements, and this can be understood as each existing as having models that are individually optimized (or learned) to fit a specific human body component.

[0073] In one embodiment, the control unit (150) can train the first analysis model (151) by processing a medical image (210) as an input to the first analysis model (151) so that the first analysis model (151) performs an analysis of a first human body component (e.g., calcaneus) among a plurality of different human body components. In this case, the first analysis model (151) can be trained to recognize a first region (or object) corresponding to a set of first human body components in the medical image (210), and based on the recognized first region, to acquire at least one of a first mask, a first specific point, and a first specific location corresponding to a set of first human body components.

[0074] In another embodiment, the control unit (150) can train the second analysis model (152) by processing a medical image (210) as input to the second analysis model (152) so that the second analysis model (152) performs an analysis of a second human body component (e.g., talus) among a plurality of different human body components. In this case, the second analysis model (152) [trains] the second human body in the medical image (210).

[0075] It can be learned to recognize a second region (or object) corresponding to a set of components, and based on the recognized second region, to acquire at least one of a second mask, a second specific point, and a second specific location corresponding to a set of second human body components.

[0076] In another embodiment, the control unit (150) can train the third analysis model (153) by processing a medical image (210) as an input to the third analysis model (153) so that the third analysis model (or the Nth analysis model, 153) performs an analysis of a third human body component (e.g., metatarsal bone) among a plurality of different human body components. In this case, the third analysis model (153) can be trained to recognize a third region (or object) corresponding to the third human body component in the medical image (210), and based on the recognized third region, to acquire at least one of a third mask, a third specific point, and a third specific location corresponding to a set of the third human body components.

[0077] Additionally, although not shown, the fourth analysis model (154) may be configured to recognize specific points rather than specific regions in the medical image (210) and generate information about the recognized points.

[0078] Furthermore, the control unit (150) can generate an analysis result (or measurement value) for at least one human body component corresponding to the measurement target by using the output data (211, 212, 213) of each of the plurality of analysis models (151, 152, 153, 154).

[0079] More specifically, the control unit (150) can generate a synthetic image (220) based on a first mask (211), a second mask (212), and a third mask (213) obtained from each of a plurality of analysis models (151, 152, 153, 154), and generate a measurement value (230, or indicator result value) for a measurement target through post-processing of the synthetic image (220). For example, the measurement value (230) may include numerical (or quantitative) data for at least one of an angle, length, or area calculated (or measured) for at least one human body component corresponding to the measurement target. In this case, the control unit (150) can visualize the measurement value (230) so that medical personnel (e.g., a doctor) can utilize it during the treatment process of a patient, and the measurement value (230) can be made editable (or modified) by a user (e.g., medical personnel). More specific details regarding this will be described later.

[0080] As seen above, various analysis models included in the medical image analysis system (100) according to the present invention may correspond to at least one of, for example, i) a model that extracts a contour and extracts a specific point or axis through the extracted contour (or a model that forms a mask), ii) a model that extracts a bounding box and specifies (or estimates) the location and contour of a structure based on the extracted bounding box (or a model that forms a bounding box), and iii) a model that extracts a characteristic point in a medical image (or a model that extracts a specific point).

[0081] In other words, the present invention enables rapid analysis of medical images through the parallel processing of various analysis models, which can be useful for processing a large volume of X-rays. Furthermore, the present invention allows for flexible resource utilization by employing analysis models that correspond to the targets required by the user. Moreover, rather than analyzing medical images using only a single model, the present invention enables precise analysis of analysis elements by performing analysis using an analysis model trained to suit specific human body components.

[0082] Meanwhile, the medical image analysis system (100) according to the present invention may be implemented in various platform forms such as an application, software, or website. For convenience of explanation, the form in which the medical image analysis system (100) is implemented is not limited to any one form.

[0083] Furthermore, the user described above may possess a user account already registered in the medical image analysis system (100) according to the present invention. In this case, the account may be created through a page (or screen) linked to the medical image analysis system (100). Alternatively, the account may also be created in at least one other system linked to the medical image analysis system (100). However, in this specification, the system in which the user account was issued is not distinguished separately, and all accounts capable of using various services (or functions) provided by the medical image analysis system (100) according to the present invention are referred to as “accounts already registered in the medical image analysis system (100).”

[0084] Meanwhile, the present invention aims to provide a medical image analysis method and system that can be flexibly utilized for various medical image analyses. More specifically, the present invention aims to recognize human body structures, including human body components, in medical images and generate analysis results for the recognized structures. Below, the medical image analysis method according to the present invention will be examined in more detail.

[0085] First, in the present invention, a process of specifying a medical image may be carried out (S310, see FIG. 3).

[0086] In the present invention, there may be various methods (or methods or criteria) for specifying a medical image. For example, a medical image may be specified based on i) user input received through a user terminal (10), ii) user input received through a page linked to a medical image analysis system (100), or iii) information received through a user terminal (10).

[0087] In one embodiment, as illustrated in FIG. 1, a control unit (150) may provide a service page (1000) associated with a medical image analysis system (100) to a user terminal (10). In order to receive user input regarding a medical image, the control unit (150) may provide (or display) a graphic object (1001) associated with an image input (or upload) function in an area of ​​the service page (1000). When the graphic object (1001) is selected by the user terminal (10), the control unit (150) may activate (or output) an image upload page (or window) to the user terminal (10). Through the image upload page, the user may select at least one of the medical images stored (or embedded) in the storage (or memory or storage space or database) of the user terminal (10) or upload a medical image using a drag-and-drop method. The control unit (150) can identify the medical image as the medical image to be analyzed based on the input of a medical image corresponding to the user selection (or input).

[0088] In another embodiment, the control unit (150) receives from the user terminal (10) a medical image

[0089] Link information (e.g., URL) or link information of an external storage service (e.g., Google Drive, Dropbox, etc.) storing medical images can be received. Then, the control unit (150) can directly access the medical images or download the medical images through the link information to identify the medical images to be analyzed.

[0090] However, the method of specifying a medical image in the present invention is not necessarily limited to the embodiments described above. For convenience of explanation, the following description assumes that a medical image has been specified based on user input received through a service page (1000). Furthermore, the control unit (150) may receive one or more medical images (i.e., multiple images are possible) from a user terminal (10), and for convenience of explanation, the present specification assumes that one medical image has been received.

[0091] Meanwhile, in the present invention, at least one measurement target is selected from a user terminal (S320, see FIG. 3), and based on the selection of at least one measurement target, a process of specifying at least one human body component corresponding to the measurement target may be performed (S330, see FIG. 3).

[0092] Furthermore, in the present invention, a process of selecting at least one analysis model trained to perform an analysis of a specific human body component among a plurality of analysis models may be carried out (S340, see FIG. 3).

[0093] In the present invention, there may be various methods (or ways) for selecting a measurement target. For example, the control unit (150) may select a measurement target based on at least one of i) a user's selection of a measurement target item (or measurement target, measurement item, etc.) provided on a user terminal (10) and / or a service page (1000), ii) a user's input regarding a region (or object) corresponding to at least one of a plurality of different human body components included in a medical image, and iii) a user input received through the user terminal (10).

[0094] However, the method of selecting a measurement target in the present invention is not necessarily limited to the embodiments described above. Hereinafter, the description will be based on the premise that a measurement target is selected based on the measurement target items provided on the service page (1000).

[0095] In one embodiment, as illustrated in FIG. 4a, the control unit (150) may provide a plurality of measurement target items (401, 402, 403) related to the analysis of human body components in a part of the service page (1000).

[0096] As an example related thereto, the measurement target item may be related to an indicator presented to the medical community in relation to the measurement of objective indicators through a patient's medical image. In this case, for each of the multiple measurement target items (401, 402, 403), information regarding a specific human body component corresponding to the measurement target (or to be measured) and information regarding a specific analysis model that performs analysis on (or is trained to perform analysis on) said specific human body component may be matched and exist.

[0097] For example, as illustrated in FIG. 4b, the first measurement target item (ex: “TC”, 401) may have information about the first human body component (411) and the second human body component (412) corresponding to the measurement target, and information about the first analysis model (151) and the second analysis model (152) that perform analysis on each of the first human body component (411) and the second human body component (412), matched therein.

[0098] For another example, in the second measurement target item (ex: “Calcaneal pitch angle”, 402), information regarding the first human body component (411) and the second human body component (412) corresponding to the measurement target, and information regarding the first analysis model (151) and the second analysis model (152) that perform analysis on each of the first human body component (411) and the second human body component (412) may be matched and present.

[0099] As another example, in the third measurement target item (ex: “Meary’s angle”, 403), information regarding the first human body component (411), the second human body component (412), and the third human body component (413) corresponding to the measurement target, and information regarding the first analysis model (151), the second analysis model (152), and the third analysis model (153) that perform analysis on each of the first human body component (411), the second human body component (412), and the third human body component (413) may be matched and present.

[0100] However, the items subject to measurement are not limited to this and may include various additional items subject to measurement. Accordingly, it is obvious that multiple different bones may also include various additional bones in addition to the first human body component (411), the second human body component (412), and the third human body component (413).

[0101] The control unit (150) may receive a selection from the user terminal (10) of at least one of a plurality of measurement target items (401, 402, 403). For example, the control unit (150) may receive a user selection (or input) for the first measurement target item (401) and the third measurement target item (403) based on the selection of the first measurement target item (401) and the third measurement target item (403) from the user terminal (10).

[0102] Additionally, the control unit (150) can identify at least one human body component corresponding to the measurement target among a plurality of different human body components based on a selection for at least one measurement target item. For example, the control unit (150) can identify the first human body component (411) and the second human body component (412) matched to the first measurement target item (401), and the first human body component (411), the second human body component (412), and the third human body component (413) matched to the third measurement target item (403) as human body components corresponding to the measurement target, based on the reception of a user selection for the first measurement target item (401) and the third measurement target item (403).

[0103] Furthermore, the control unit (150) may select at least one analysis model trained to perform an analysis on a specific human body component (or measurement target) among a plurality of analysis models. Based on the fact that at least one human body component corresponding to the measurement target is specified according to user input, the control unit (150) may select a specific analysis model to perform an analysis on the specified human body component.

[0104] In this case, a specific analysis model (or at least one analysis model) may be understood to include one or more models depending on the specific human body component.

[0105] Based on the fact that the specified human body component includes a first human body component (411), a second human body component (412), and a third human body component (413), the control unit (150) may select (or specify) a first analysis model (151) trained to perform an analysis on the first human body component (411), a second analysis model (152) trained to perform an analysis on the second human body component (412), and a third analysis model (153) trained to perform an analysis on the third human body component (413) as specific analysis models to perform an analysis on each of the specified human body components (411, 412, 413).

[0106] However, the method (or method or criterion) for specifying a human body component corresponding to the measurement target is not necessarily limited to the embodiments described above. For example, a specific human body component corresponding to the measurement target may be specified based on i) receiving user input (or selection) for a region (or object) corresponding to at least one human body component among a plurality of different human body components included in a medical image, ii) being specified based on a request (or information) received from a user terminal (10), or iii) being measured through self-analysis of the medical image (400) of the medical image analysis system (100).

[0107] Meanwhile, in the present invention, a process of processing a medical image as input to an analysis model may be performed (S350, see FIG. 3).

[0108] As seen above, each of the multiple analysis models (151, 152, 153) may be trained for each of the different human body components to perform analysis on each of the multiple different human body components. A specific analysis model may be understood to include one or more structures depending on the specific human body component.

[0109] There may be one or more types of models that can be specified, as shown in the example below.

[0110] When a medical image is input, the specified analysis model can detect (or identify) specific anatomical landmarks (or keypoints) within the image and predict their precise locations. This can also be understood as each analysis model, trained to perform analysis on specific anatomical landmarks, detecting the specific landmark of interest within the input medical image and accurately estimating its coordinates.

[0111] Another specific analysis model, when a medical image is input, can recognize (or identify) a region (or object) corresponding to a specific human body component in the medical image and segment (or classify, separate, etc.) said region from other regions. This can also be understood as each analysis model trained to perform analysis on a specific human body component recognizing a specific region corresponding to that specific human body component, defining the boundaries of said specific region, and dividing said specific region into segments.

[0112] A specific analysis model can mask a recognized (or segmented) specific region to generate a mask corresponding to one or more human body components. Here, "masking a recognized specific region" can be understood as a process of representing a recognized specific region as a binarized value, a multi-class value, or an instance-specific unique value.

[0113] In the case of binarization, it may mean that a specific analysis model assigns (or assigns, sets, etc.) a first value (e.g., "1") to pixels belonging to a specific region recognized by the model, and assigns a second value (e.g., "0") to pixels in regions that do not belong to said specific region. Here, the first value indicates that the pixel belongs to a specific region corresponding to a specific human body component, and the second value may indicate that the pixel belongs to the remaining region excluding the specific region corresponding to the specific human body component.

[0114] In the case of multiclass, a specific analysis model may assign a unique value representing the class to the pixel to which each specific recognized region belongs. For example, if a model is trained to recognize N different human body components, each pixel may have a value from 0 to N, where 0 represents the background and values ​​from 1 to N correspond to specific human body components.

[0115] In the case of instance segmentation, specific analysis models can assign a unique identifier to each pixel that distinguishes individual instances, along with class information. This enables the differentiation of individual objects even within the same class. For example, the Mth instance of class C can be represented in the form CM, where C represents the class and M represents the instance number within that class.

[0116] The generated mask can be represented in a binary, multi-class, or instance-specific form. In the binary form, parts having a first value may be displayed with a first visual appearance (e.g., light color), and parts having a second value may be displayed with a second visual appearance (e.g., dark color). In the multi-class form, parts having values ​​representing each class may be displayed with different visual appearances (e.g., different colors or shades).

[0117] In the case of instance-specific forms, each instance can be distinguished and displayed by a unique color or pattern, which allows individual objects within the same class to be visually distinguished.

[0118] In this regard, as illustrated in FIG. 5, the control unit (150) can process the medical image (500) as input to each of the specified analysis models (151, 152, 153). In this case, the specified analysis models (151, 152, 153) can be understood as models selected based on the user's selection of the first measurement target item (501) and the third measurement target item (503).

[0119] First, when a medical image (500) is input to the first analysis model (151), the first analysis model (151) can recognize a first region (511a) corresponding to a first human body component among a plurality of regions corresponding to a plurality of different human body components included in the medical image (500), and mask the recognized first region (511a) to generate a first mask (511b) corresponding to the first human body component. Additionally, the first analysis model (151) can obtain at least one of a first specific point and a first specific location corresponding to a set of the first human body components based on the recognized first region (511a).

[0120] Next, when a medical image (500) is input to a second analysis model (152) different from the first analysis model (151), the second analysis model (152) can recognize a second region (512a) corresponding to a second human body component among a plurality of regions corresponding to a plurality of different human body components included in the medical image (500), and mask the recognized second region (512a) to generate a second mask (512b) corresponding to the second human body component. Additionally, the second analysis model (152) can obtain at least one of a second specific point and a second specific location corresponding to a set of second human body components based on the recognized first region (512a).

[0121] Furthermore, when a medical image (500) is input, the third analysis model (153), which is different from the first analysis model (151) and the second analysis model (152), recognizes a third region (513a) corresponding to a third human body component among a plurality of regions corresponding to a plurality of different human body components included in the medical image (500), and can mask the recognized third region (513a) to generate a third mask (513b) corresponding to the third human body component. Additionally, the third analysis model (153) can obtain at least one of a third specific point and a third specific location corresponding to a set of the third human body components based on the recognized third region (513a).

[0122] Meanwhile, in the present invention, a process of generating a measurement value for a measurement target can be carried out using the output data of an analysis model (S360, see FIG. 3).

[0123] In the present invention, there may be various methods for extracting (or obtaining) measurement values ​​(or information) regarding a measurement target from a medical image.

[0124] In one embodiment, the present invention obtains a contour using a mask obtained through an analysis model, and based thereon, extracts information such as characteristic points (e.g., points where tangents meet, lowest points, etc.), the elliptical axis most suitable for the contour, and the axis that best represents the contour.

[0125] In another embodiment, the present invention utilizes a model most suitable for the image situation among a plurality of analysis models to detect key points (or specific points) in a medical image and extract characteristic points based on the detection results.

[0126] In another embodiment, the present invention recognizes a bounding box corresponding to a specific human body component through object detection (or an analysis model that performs object detection) and can extract the specific human body component based thereon.

[0127] The control unit (150) can obtain at least one of a mask corresponding to a specific human body component, a specific point, and a specific location as output data of a specific analysis model.

[0128] Specifically, the control unit (150) can obtain a first mask (511b), a second mask (512b), and a third mask (513b) generated from each of the first analysis model (151), the second analysis model (152), and the third analysis model (153) described above.

[0129] And, the control unit (150) can perform contouring on the acquisition of specific points required or on the acquired mask.

[0130] More specifically, the control unit (150) can recognize the boundaries of a specific human body component based on the acquired mask and generate (or extract) a contour of the specific human body component according to the recognized boundaries.

[0131] In this case, the control unit (150) can combine the first mask (511b), the second mask (512b), and the third mask (513b) with the medical image (500) to generate a composite image and perform contouring on the generated composite image. In the composite image, specific human body components corresponding to each of the acquired masks (511b, 512b, 513c) may be visually highlighted and displayed. For example, in the composite image, a first human body component (521) corresponding to the first mask (511b) may be displayed as a first visual appearance, a second human body component (522) corresponding to the second mask (512b) may be displayed as a second visual appearance, and a third human body component (523) corresponding to the third mask (513b) may be displayed as a third visual appearance.

[0132] In one embodiment, the control unit (150) can recognize the boundary of the first human body component (521) in the composite image and extract the contour of the first human body component (521) according to the recognized boundary.

[0133] In another embodiment, the control unit (150) can recognize the boundary of the second human body component (522) in the composite image and extract the contour of the second human body component (522) according to the recognized boundary.

[0134] In another embodiment, the control unit (150) can recognize the boundary of the third human body component (523) in the composite image and extract the contour of the third human body component (523) according to the recognized boundary.

[0135] The control unit (150) can generate a measurement value for a measurement target based on the extracted contour.

[0136] In this case, the control unit (150) can specify (or select, set, etc.) a specific point (or specific point, point of interest, reference point, feature point, landmark, etc.) in the extracted contour based on a preset standard.

[0137] There may be various criteria for determining a specific point. For example, the criteria for determining a specific point may include at least one of i) anatomically significant points (e.g., joint sites, bony protrusions, bone ends), ii) geometric centers (e.g., bone centers or axes of symmetry), iii) joints between bones (e.g., joints between bones), iv) bone boundary points (e.g., bone ends or edges), v) specific reference lines (e.g., a line parallel to the sole of the foot), and vi) mathematical calculations (e.g., points calculated based on specific patterns or angles). However, the criteria for determining a specific point are not limited to the examples mentioned and may include additional criteria.

[0138] The control unit (150) can generate a measurement value for a measurement target based on a specific point specified on the extracted contour.

[0139] In one embodiment, as illustrated in FIG. 5b, the control unit (150) may calculate (or measure) an angle formed by at least one of the first human body component (521) to the third human body component (523) in order to generate a measurement value for a first measurement target item selected from the user terminal (10). In this case, the control unit (150) may calculate (or generate) a measurement value (e.g., an angle (e.g., “59.5˚”)) for the first measurement target item by connecting specific points (530a, 530b, 530c) specified on the contour of at least one of the first human body component (521) to the third human body component (523).

[0140] In another embodiment, the control unit (150) can calculate (or measure) the angle formed by at least one of the first human body component (521) to the third human body component (523) in order to generate a measurement value for a third measurement target item selected from the user terminal (10).

[0141] In this case, the control unit (150) can calculate (or generate) a measurement value (e.g., an angle (e.g., “21.7°”)) for a third measurement target item by connecting specific points (530a, 530d, 530e) specified on the contour of at least one of the first human body components (521) to the third human body components (523).

[0142] Furthermore, as illustrated in FIG. 5c, the control unit (150) can generate a calculated measurement value (540) for human body components (521, 522, 523) corresponding to the measurement target. In this case, the control unit (150) can generate a measurement value for the measurement target using images of human body components extracted from at least one analysis model and data extracted to generate a measurement value for said human body components. As an example, the control unit (150) can display the measurement value by overlaying it on a medical image (500) to provide the measurement value to the user for visualization.

[0143] Meanwhile, the control unit (150) can provide the generated measurement value to the user terminal (10).

[0144] For example, as illustrated in FIG. 6, the control unit (150) can provide a measurement value for a measurement target overlaid on a medical image (600) to a service page (1000) in one area (e.g., a first area).

[0145] Additionally, in an area of ​​the service page (1000) that is different from the area containing the medical image (600) (e.g., a second area), measurement values ​​for a measurement target may be provided in text form. For example, the control unit (150) may provide a measurement value (e.g., “40”, 611) for a first measurement target (e.g., “TC angle”) and a measurement value (e.g., “20”, 612) for a third measurement target (e.g., “Meary’s angle”) as text (or number) based information in an area of ​​the service page (1000).

[0146] Furthermore, the measurement value for the measurement target can be made editable (or modified) by a user (e.g., medical staff). For example, the user can adjust an object (e.g., a line) overlaid on the medical image (600). In this case, when the object is adjusted by the user, the measurement value corresponding to the adjusted object among the numerical measurement values ​​(611, 612) corresponding to the adjusted object can be updated and displayed on the service page (1000).

[0147] Meanwhile, the service page according to the present invention may be formed in various forms in addition to the form described above.

[0148] In one embodiment, as illustrated in FIG. 7a, the control unit (150) may provide a service page (700) to the user terminal (10). In this case, the service page (700) illustrated in FIG. 7a may include at least one of the following: i) a function to select a measurement target to be analyzed; ii) a function to input (or select) a medical image; iii) a function to select a type of angle to be generated; iv) a function to receive the input medical image; v) a function to generate a file corresponding to the input data; vi) a function to delete the file; vii) a function to generate an artificial intelligence angle of the input medical image; and viii) a function to display the progress of generating the artificial intelligence angle.

[0149] In another embodiment, as illustrated in FIG. 7b, the service page (700) may include at least one of the following: i) a function to select an angle to check among the generated angles; ii) a function to modify the generated angle; iii) a function to enlarge and check and modify the generated angle and image; iv) a function to input at least one image; v) a function to move to an image corresponding to the generated data; and iv) a function to convert the measured result into an Excel file.

[0150] In this way, the present invention can provide various functions to the user through a service page.

[0151] As described above, according to the medical image analysis method and system of the present invention, by analyzing medical images using an analysis model specialized for each human body component, specific human body components in medical images can be accurately recognized and measured.

[0152] Through this, the present invention enables accurate analysis of human body components that overlap each other, which reduces errors that may occur when analyzing with a single model and allows for the effective analysis of complex skeletal structures.

[0153] Furthermore, according to the medical image analysis method and system of the present invention, the accuracy of diagnosis can be significantly improved by analyzing human body components in a medical image using multiple analysis models and generating measurement values ​​based on the analysis results. In particular, the present invention enables the derivation of more precise diagnostic results by selecting and applying an optimized analysis model for each of the various human body components.

[0154] Furthermore, according to the medical image analysis method and system of the present invention, by automatically recognizing and analyzing multiple human body components in a medical image and providing the necessary data to medical staff, clinical and research time can be reduced. In other words, the present invention can contribute to reducing medical research time along with reducing the time spent on clinical consultations and improving accuracy for medical staff.

[0155] Furthermore, the medical image analysis method and system according to the present invention can be usefully applied in various diagnostic environments. That is, the present invention can be usefully utilized in various medical settings or clinical fields.

[0156] Meanwhile, the present invention described above can be implemented as a program that is executed by one or more processes on a computer and can be stored on a computer-readable medium (or recording medium).

[0157] Furthermore, the present invention described above can be implemented as computer-readable code or instructions on a medium on which a program is recorded. That is, the present invention can be provided in the form of a program.

[0158] Meanwhile, computer-readable media include all types of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable media include HDD (Hard Disk Drive), SSD (Solid State Disk), SSD (Silicon Disk Drive), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

[0159] Furthermore, the computer-readable medium may be a server or cloud storage that includes a storage and is accessible to an electronic device via communication. In this case, the computer may download the program according to the present invention from the server or cloud storage via wired or wireless communication.

[0160] Furthermore, in the present invention, the computer described above is an electronic device equipped with a processor, namely a CPU (Central Processing Unit), and no specific limitations are placed on its type. Furthermore, in the present invention, the computer described above is an electronic device equipped with one or more processing units, and no specific limitations are placed on its type. Such processing units include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), TPUs (Tensor Processing Units), FPGAs (Field-Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), etc. The implementation of the present invention may be performed using one or more of these various processing units and is not limited to a specific hardware configuration.

[0161] Hereinafter, a medical image analysis method based on FIG. 2b of the present specification is described as follows.

[0162] A medical image analysis method according to one embodiment of the present invention may include the steps of receiving the medical image from a user terminal (10), selecting a plurality of analysis models (151, 152, 153) that are pre-trained for each of a plurality of human body components, processing the medical image (210) as an input to the selected analysis models (151, 152, 153), obtaining output data (211, 212, 213) for the plurality of human body components from each of the plurality of analysis models (151, 152, 153), and generating a measurement value (230) for the measurement target using the output data (211, 212, 213). At this time, each of the plurality of analysis models (151, 152, 153) may be trained on different human body components.

[0163] As illustrated in FIG. 2b, the input data input to the medical image analysis system (100) is a medical image (210), which may be a single medical image. A single medical image (210) may contain multiple structures (human body components), and it is desirable to select the most appropriate training architecture for each structure.

[0164] Multiple analysis models (151, 152, 153) are repeatedly trained for each structure, and one analysis model may have trained on one structure. The same input data may be input to the multiple analysis models (151, 152, 153) included in the medical image analysis system (100). For example, a single medical image (210) containing multiple structures may be input to the multiple analysis models (151, 152, 153) in common. The multiple analysis models (151, 152, 153) may output output data (211, 212, 213) for each structure included in the single medical image (210).

[0165] For example, a foot X-ray image may be input as a single medical image (210). This foot X-ray image may contain multiple bone structures, such as the calcaneus, talus, and first metatarsal. In the present invention, a first analysis model (151) specialized for calcaneus analysis, a second analysis model (152) specialized for talus analysis, and a third analysis model (153) specialized for first metatarsal analysis can be independently trained and utilized.

[0166] Each analysis model may have a structure optimized for the characteristics of the corresponding human body component. For example, since the calcaneus is a relatively large bone with clear boundaries, the first analysis model (151) may be specialized in accurately segmenting the entire area of ​​the calcaneus using semantic segmentation techniques.

[0167] On the other hand, for structures that are complexly connected to other bones and overlap with each other, such as the talus, the second analysis model (152) may be specialized in distinguishing individual instances or segmented planes by utilizing a model of another segmentation (instance segmentation or sementic segmentation) technique. Additionally, if important feature points are required for specific angle measurements, the third analysis model (153) may utilize a keypoint detection technique.

[0168] A characteristic aspect of the present invention may be that a single identical medical image is input to multiple analysis models. Although a foot X-ray image (210) is input identically to the first analysis model (151), the second analysis model (152), and the third analysis model (153), each model can generate output data only for specific human body components it has learned. This can produce an effect similar to multiple specialists analyzing the same image from the perspective of their respective fields of expertise.

[0169] The output data of each analysis model may have different forms. The first analysis model (151) may generate a first mask (211) corresponding to the calcaneus, the second analysis model (152) may generate a second mask (212) corresponding to the talus, and the third analysis model (153) may generate a third mask (213) corresponding to the first metatarsal or a specific key point. The output data generated in this way may be combined into a composite image (220) so that each human body component can be visualized in different colors or patterns.

[0170] Meanwhile, the above detailed description should not be interpreted restrictively in all respects but should be considered exemplary. The scope of the invention shall be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the invention are included within the scope of the invention.

Claims

1. A method for analyzing medical images performed by a control unit, A step of receiving the medical image from a user terminal; - the medical image includes a plurality of human body components, and - A step of selecting a plurality of pre-trained analysis models for each of the plurality of human body components; A step of processing the medical image as input to the selected analysis models; A step of obtaining output data for the plurality of human body components from each of the plurality of analysis models; and It includes the step of generating a measurement value for the measurement target using the output data above, A medical image analysis method in which each of the above multiple analysis models is trained with different human body components.

2. In Paragraph 1, The step of obtaining the above output data is, A medical image analysis method comprising obtaining different types of output data including at least one of a mask, a feature point, and a specific location from the above analysis models.

3. In Paragraph 2, The step of generating the above measurement value is, A step of performing post-processing on the above-mentioned acquired output data, including contouring, boundary recognition, contour extraction, and specific point identification; and A medical image analysis method comprising the step of generating at least one of an angle, length, and area for a plurality of human body components based on the above post-processed output data.

4. In Paragraph 1, The step of obtaining the above output data is, A medical image analysis method comprising obtaining a first mask, a first feature point, or a first feature location corresponding to a set of first human body components learned in the first analysis model among the plurality of analysis models from a first analysis model among the plurality of analysis models.

5. In Paragraph 4, The step of obtaining the above output data is, A medical image analysis method comprising obtaining a second mask, a second feature point, or a second feature location corresponding to a set of second human body components learned in the second analysis model among the plurality of analysis models above from a second analysis model.

6. In Paragraph 1, A medical image analysis method comprising at least one of the plurality of analysis models, a semantic segmentation model, an instance segmentation model, an object detection model, and a key point detection model.

7. In Paragraph 1, A medical image analysis method in which the input values ​​input to each of the above plurality of analysis models are all the same medical image as the medical image.

8. In Paragraph 1, The step of obtaining the above output data is, A medical image analysis method comprising obtaining masks corresponding to different human body components for each of the plurality of analysis models and generating a synthetic image based on the obtained masks.

9. In Paragraph 8, A medical image analysis method in which the above composite image is obtained by superimposing the above acquired masks onto the above medical image.

10. A medical image analysis system comprising a communication unit, a storage unit, and a control unit, The above system is, The medical image is received from a user terminal, and the medical image includes a plurality of human body components, and Select multiple analysis models trained for each of the above multiple human body components, and The above medical image is processed as input to the above-selected analysis models, and Output data for the plurality of human body components is obtained from each of the plurality of analysis models, and A measurement value for the measurement target is generated using the above output data, and A medical image analysis system in which each of the above multiple analysis models is trained with different human body components.

Images