Apparatus and method for training liver function evaluation model

Abstract

The present invention relates to a method for training a liver function evaluation model. The method for training a liver function evaluation model may comprise the steps of: refining a segmentation algorithm on the basis of feedback information; collecting image data for a plurality of tissues of the liver on the basis of the segmentation algorithm; collecting clinical data for the plurality of tissues; and constructing an evaluation model that receives the image data and the clinical data as input and outputs liver function prediction data.

Description

Liver function assessment model learning device and method

[0001] The present invention relates to a liver function evaluation model learning device and method.

[0002] The pathophysiology of progressive chronic liver disease consists of hepatocyte inflammation, fibrosis, and the liver's regenerative capacity, and these factors need to be considered comprehensively to evaluate the degree of progression of chronic liver disease.

[0003] Generally, histopathological examination via liver biopsy is considered the standard method for diagnosing liver disease; however, this is an invasive procedure that can cause bleeding in patients with liver disease who have impaired blood coagulation ability. It also has the disadvantage of being difficult to evaluate function while enabling structural diagnosis, and faces issues regarding the concordance of tissue interpretations and the suitability of samples. Additionally, while liver elastography using ultrasound and MRI is a non-invasive test commonly used in routine clinical practice, it is currently difficult to use for functional evaluation, similar to liver biopsy, although it allows for structural diagnosis.

[0004] Meanwhile, gadoxetic acid is one of the hepatocellular-specific contrast agents that is injected intravenously. It is distributed within the blood vessels and extravascular space, gradually absorbed by hepatocellular cells, and then excreted into the bile ducts. During this pharmacodynamic process involving gadoxetic acid, specific images representing the hepatobiliary phase can be obtained.

[0005] In these images, signal intensity varies depending on the presence or absence of liver lesions; specifically, signal intensity is high in the case of a normal liver and low in the case of focal lesions, resulting in high contrast for the focal lesions. Accordingly, images representing hepatobiliary stages demonstrate high sensitivity for detecting focal liver lesions, and thus can be critically utilized in the evaluation of focal lesions in patients with chronic liver disease.

[0006] In particular, given the characteristics of Asian countries where local treatment is preferred over liver transplantation for liver cancer, highly sensitive gastrocetate MRI occupies an important position in various guidelines. Furthermore, since the uptake of hepatocyte-specific contrast agents by hepatocytes is influenced by liver function—specifically liver function—increases significantly when liver function declines, active research and development are underway to utilize this phenomenon as a biomarker for liver function deterioration.

[0007] The technology forming the background of the present invention is disclosed in Korean Registered Patent Publication No. 10-2455051.

[0008] The present invention aims to solve the problems of the aforementioned conventional technology by providing a liver function evaluation model training device and method capable of constructing an evaluation model that enables more accurate prediction of liver function, by producing liver function prediction data that sufficiently reflects liver set acid dynamics through the segmentation of regions for the liver parenchyma, hepatobiliary tract, hepatic portal vein, kidney, and spleen in image data and training a model using data extracted from each segmented tissue, such as signal intensity and volume information.

[0009] However, the technical problems that the embodiments of the present invention aim to solve are not limited to the technical problems described above, and other technical problems may exist.

[0010] As a technical means for achieving the above-mentioned technical task, a method for learning a liver function evaluation model according to one embodiment of the present invention may include the steps of: refining a segmentation algorithm based on feedback information; collecting image data of a plurality of tissues of the liver based on the segmentation algorithm; collecting clinical data of the plurality of tissues; and constructing an evaluation model that outputs liver function prediction data using the image data and the clinical data as inputs.

[0011] According to one embodiment of the present invention, the image data may be generated based on magnetic resonance imaging data based on a reading set acid.

[0012] According to one embodiment of the present invention, the plurality of tissues may include at least one of liver parenchyma, hepatobiliary tract, hepatic portal vein, kidney, and spleen.

[0013] According to one embodiment of the present invention, the step of collecting image data for the plurality of tissues may involve generating analysis data by performing an analysis on at least one of signal intensity, volume, and texture for each of the plurality of tissues from segmented data obtained through the segmentation algorithm, and collecting the image data that reflects the analysis data.

[0014] According to one embodiment of the present invention, the signal intensity may be analyzed based on at least one of the degree of absorption, degree of release, and degree of contrast enhancement of the acetic acid.

[0015] According to one embodiment of the present invention, the step of enhancing the segmentation algorithm may include: collecting standard medical data; inputting the standard medical data into an initial segmentation algorithm to generate initial mask data; modifying the initial mask data based on feedback information based on expert evaluation of the initial mask data to generate correct mask data; and tuning the initial segmentation algorithm using the correct mask data to construct the enhanced segmentation algorithm.

[0016] As a technical means for achieving the above-mentioned technical task, a liver function evaluation model learning device according to one embodiment of the present invention may include an enhancement unit that enhances a segmentation algorithm based on feedback information, an image collection unit that collects image data of a plurality of tissues of the liver based on the segmentation algorithm, a clinical collection unit that collects clinical data of the plurality of tissues, and a model construction unit that constructs an evaluation model that outputs liver function prediction data using the image data and the clinical data as inputs.

[0017] According to one embodiment of the present invention, the image data may be generated based on magnetic resonance imaging data based on a reading set acid.

[0018] According to one embodiment of the present invention, the plurality of tissues may include at least one of liver parenchyma, hepatobiliary tract, hepatic portal vein, kidney, and spleen.

[0019] According to one embodiment of the present invention, the image acquisition unit may generate analysis data by performing an analysis of at least one of the signal strength, volume, and texture of each of the plurality of tissues from segmented data obtained through the segmentation algorithm, and collect the image data that reflects the analysis data.

[0020] According to one embodiment of the present invention, the signal intensity may be analyzed based on at least one of the degree of absorption, degree of release, and degree of contrast enhancement of the acetic acid.

[0021] According to one embodiment of the present invention, the enhancement unit may include a standard collection unit that collects standard medical data, an initial generation unit that inputs the standard medical data into an initial segmentation algorithm to generate initial mask data, a correct answer generation unit that modifies the initial mask data based on feedback information based on expert evaluation of the initial mask data to generate correct answer mask data, and an algorithm construction unit that constructs an enhanced segmentation algorithm by tuning the initial segmentation algorithm using the correct answer mask data.

[0022] As a technical means for achieving the above-mentioned technical task, a liver function evaluation method according to one embodiment of the present invention may include the steps of acquiring medical image data of a subject, inputting the medical image data into an advanced segmentation algorithm based on feedback information to extract image data of a plurality of tissues of the liver, and inputting the image data into a pre-trained evaluation model to output liver function prediction data.

[0023] As a technical means for achieving the above-mentioned technical task, a liver function evaluation device according to one embodiment of the present invention may include an acquisition unit for acquiring medical image data of a subject, an extraction unit for extracting image data of a plurality of tissues of the liver by inputting the medical image data into an advanced segmentation algorithm based on feedback information, and an output unit for outputting liver function prediction data by inputting the image data into a pre-trained evaluation model.

[0024] The means for solving the problem described above are merely exemplary and should not be interpreted as intended to limit the present invention. In addition to the exemplary embodiments described above, additional embodiments may exist in the drawings and the detailed description of the invention.

[0025] According to the solution to the problem of the present invention described above, regions for the liver parenchyma, hepatobiliary tract, hepatobiliary portal vein, kidney, and spleen are segmented in image data, and by training a model using data extracted for each segmented tissue, liver function prediction data that sufficiently reflects the biliary set acid dynamics is produced, thereby enabling the construction of an evaluation model capable of more accurate prediction of liver function.

[0026] However, the effects obtainable from this invention are not limited to those described above, and other effects may exist.

[0027] FIG. 1 is a schematic diagram of a liver function evaluation system according to one embodiment of the present invention.

[0028] FIG. 2 is a diagram illustrating the construction and application process of a liver function evaluation model according to one embodiment of the present invention.

[0029] FIG. 3 is a schematic block diagram of a liver function evaluation model learning device according to one embodiment of the present invention.

[0030] FIG. 4 is a schematic block diagram of a liver function evaluation device according to one embodiment of the present invention.

[0031] FIG. 5 is a flowchart of the operation of a liver function evaluation model learning method according to one embodiment of the present invention.

[0032] FIG. 6 is an operation flowchart of the advanced process of a partitioning algorithm according to one embodiment of the present invention.

[0033] FIG. 7 is an operation flowchart of a liver function evaluation method according to one embodiment of the present invention.

[0034] Embodiments of the present invention are described below with reference to the attached drawings to enable those skilled in the art to easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0035] Throughout this specification, when a part is described as being "connected" to another part, this includes not only cases where they are "directly connected," but also cases where they are "electrically connected" or "indirectly connected" with other elements interposed between them.

[0036] Throughout the entire specification, when a component is described as being located "on," "on top," "on top," "under," "on bottom," or "on bottom" of another component, this includes not only cases where the component is in contact with the other component but also cases where another component exists between the two components.

[0037] Throughout this specification, when a part is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0038] The present invention relates to a liver function evaluation model learning device and method.

[0039] FIG. 1 is a schematic diagram of a liver function evaluation system according to one embodiment of the present invention.

[0040] Referring to FIG. 1, a liver function evaluation system (10) (hereinafter referred to as the evaluation system (10)) may include a liver function evaluation model learning device (100) (hereinafter referred to as the learning device (100)), a liver function evaluation device (200) (hereinafter referred to as the evaluation device (200)), and a database (300).

[0041] According to one embodiment of the present invention, the evaluation system (10) divides regions for multiple tissues such as liver parenchyma, hepatic portal vein, bile duct, spleen, and kidney from a diuretic acid contrast MRI and trains an artificial intelligence-based model thereon to comprehensively perform non-invasive functional and structural evaluations of the liver, thereby serving as a biomarker for liver function and liver disease that can be utilized for the diagnosis of liver disease, assessment of severity, and prediction of disease progression / death.

[0042] Specifically, the learning device (100) may construct an evaluation model for a segmentation algorithm and liver function used in the evaluation system (10). The learning device (100) may collect multiple source data from the database (300) to prepare learning data to be used for constructing the evaluation model and train the evaluation model. At this time, the database (300) may receive and store data from international standard formats such as DICOM (Digital Imaging and Communications in Medicine), but is not limited thereto, and the learning device (100) may receive various medical data to be used as learning data from various external servers including DICOM. Additionally, the learning device (100) may store various data collected and generated to be used as learning data when learning the segmentation algorithm and constructing the evaluation model in the database (300).

[0043] Additionally, the evaluation device (200) may provide prediction and evaluation results regarding liver function for a subject by utilizing the segmentation algorithm and evaluation model established in the learning device (100). Here, the segmentation algorithm may be a segmentation model learned based on artificial intelligence. Furthermore, the segmentation algorithm and the evaluation model may each be independent AI-based learning models, but are not limited thereto; the segmentation algorithm and the evaluation model may be substantially included in a single AI-based learning model and implemented as a single model. Therefore, the segmentation algorithm and the evaluation model in this invention should be interpreted as being distinguished and described according to their roles for the convenience of explanation.

[0044] In this regard, the learning device (100), the evaluation device (200), and the database (300) may be interconnected via a network for mutual data sharing, and examples of such networks may include, but are not limited to, a 3GPP (3rd Generation Partnership Project) network, an LTE (Long Term Evolution) network, a 5G network, a WIMAX (World Interoperability for Microwave Access) network, a wired / wireless Internet (Internet), a LAN (Local Area Network), a Wireless LAN (Wireless Local Area Network), a WAN (Wide Area Network), a PAN (Personal Area Network), a Bluetooth network, a Wi-Fi network, a NFC (Near Field Communication) network, a satellite broadcasting network, an analog broadcasting network, a DMB (Digital Multimedia Broadcasting) network.

[0045] Hereinafter, the construction and application process of a liver function evaluation model according to one embodiment of the present invention will be described in detail.

[0046] FIG. 2 is a diagram illustrating the construction and application process of a liver function evaluation model according to one embodiment of the present invention.

[0047] Referring to FIG. 2, FIG. 2(a) is a diagram schematically illustrating a process for enhancing an existing segmentation algorithm to acquire image data to be used as training data for building a liver function evaluation model.

[0048] Referring to FIG. 2(a), the learning device (100) can improve the segmentation algorithm based on feedback information.

[0049] According to one embodiment of the present invention, a learning device (100) can construct a learning data set for building an evaluation model that can be used to evaluate liver function of a subject from medical imaging data of the subject and predict symptoms and severity of liver function.

[0050] Specifically, the learning device (100) can collect standard medical data. At this time, the standard medical data may include various medical imaging data such as DICOM (Digital Imaging and Communications in Medicine) files and patient data for various patients, but is not limited thereto. The standard medical data may further include imaging data and patient data obtained in various environments, and virtual data generated by augmentation or partial transformation for the training of the model. Additionally, the standard medical data may include magnetic resonance imaging (MRI) data based on a diaphragmatic contrast agent.

[0051] Additionally, the learning device (100) can input standard medical data into an initial segmentation algorithm to extract initial mask data. Here, the initial segmentation algorithm refers to an artificial intelligence-based algorithm that segments regions and features within an image, which has been previously researched and developed. For example, a deep learning algorithm (deep learning model) of a previously researched prototype may be used, but is not limited thereto. In other words, the learning device (100) may extract initial mask data for standard medical data using a predetermined segmentation algorithm that has been previously established. Here, mask data may include data containing binary representing a specific region of interest (ROI) of the image data.

[0052] Additionally, the learning device (100) can generate correct answer mask data by modifying the initial mask data based on feedback information derived from expert evaluation of the initial mask data. Specifically, the learning device (100) can transmit the initial mask data to a terminal possessed by an expert, including a radiologist, and request input of feedback information regarding it. Furthermore, when the requested feedback information is input through the terminal, the learning device (100) may modify the initial mask data into correct answer mask data to be used as correct answer data during the training of the segmentation algorithm by reflecting the feedback information.

[0053] However, it is not limited thereto, and the learning device (100) may generate correct mask data by independently performing analysis on the initial mask data and automatically modifying a predetermined area of ​​the initial mask data. In such a case, the practical applicability of the learning device (100) may not be limited, considering the possibility that additional equipment for separate analysis and data processing may be required.

[0054] In addition, the learning device (100) can build an advanced partitioning algorithm by tuning the initial partitioning algorithm using the correct answer mask data.

[0055] Specifically, the learning device (100) may retrain an initial segmentation algorithm using collected standard medical data, initial mask data, and correct mask data. The learning device (100) may advance the segmentation algorithm in a direction such that the segmentation algorithm produces mask data close to the correct mask data from the standard medical data and initial mask data.

[0056] In this regard, the learning device (100) may use an advanced segmentation algorithm to receive magnetic resonance imaging data as standard medical data, calculate mask data for multiple tissues, and thereby generate image data for each of the multiple tissues.

[0057] Referring to FIG. 2, FIG. 2(b) is a diagram schematically illustrating the process of constructing an evaluation model that calculates liver function prediction data from magnetic resonance imaging data using image data obtained through an advanced segmentation algorithm.

[0058] Referring to FIG. 2(b), the learning device (100) can collect image data of multiple tissues of the liver based on an advanced segmentation algorithm. Here, the image data may be generated based on magnetic resonance imaging data based on gadoxetic acid, as described above. Gadoxetic acid is a liver cell-specific contrast agent that is selectively absorbed by liver cells and subsequently excreted through the bile duct. The learning device (100) may acquire segmented image data of multiple tissues using magnetic resonance imaging data taken after the injection of gadoxetic acid.

[0059] Additionally, the plurality of tissues may include at least one of the liver parenchyma, hepatobiliary tract, hepatic portal vein, kidney, and spleen. In this regard, the learning device (100) may use a segmentation algorithm to identify and segment each tissue based on the signal intensity of the biliary settl appearing in the magnetic resonance imaging data, and generate image data for each tissue by reflecting analysis elements such as volume and texture for each segmented tissue.

[0060] In other words, the learning device (100) may generate analysis data by performing an analysis of at least one of the signal strength, volume, and texture of each of a plurality of tissues from segmented data obtained through a segmentation algorithm, and collect image data reflecting the analysis data. Here, the signal strength may be analyzed based on at least one of the degree of absorption, degree of emission, and degree of contrast enhancement of dextrin.

[0061] Specifically, the learning device (100) may reflect analysis elements regarding signal intensity, liver volume, and texture according to the degree of uptake of dexacetate among the plurality of tissues for the liver parenchyma. Additionally, the learning device (100) may reflect analysis elements regarding signal intensity and spleen volume considering a reference value for signal intensity correction for the spleen among the plurality of tissues. Additionally, the learning device (100) may reflect analysis elements regarding signal intensity according to the degree of drainage of dexacetate for the bile duct among the plurality of tissues. Additionally, the learning device (100) may reflect analysis elements regarding signal intensity for comparing relative signal intensity with the liver and bile duct based on the degree of contrast enhancement of dexacetate for the kidney parenchyma and hepatic portal vein among the plurality of tissues. However, it is not limited thereto.

[0062] Additionally, the learning device (100) may collect clinical data for multiple tissues. At this time, the clinical data may include data regarding demographic factors, blood test results, and clinical findings, and more specifically, the demographic factors may include statistical information regarding the causes of chronic liver disease regarding the patient's gender and age, the blood test results may include test results for albumin and bilirubin, and the clinical findings may include various clinical data regarding patients with liver disease, such as the presence of ascites, hepatic encephalopathy, variceal procedure, liver transplant, and death. At this time, the clinical data may be coded information for the training of an artificial intelligence-based evaluation model. However, it is not limited thereto.

[0063] The learning device (100) can construct an evaluation model that outputs liver function prediction data using collected image data and clinical data as inputs. In this regard, the learning device (100) may determine an optimal combination of image data and clinical data inputs and an artificial intelligence-based learning model by utilizing evaluation results from an existing liver function evaluation system.

[0064] Specifically, the learning device (100) may consider various learning models for building an evaluation model. Such learning models may, for example, refer to deep learning models, machine learning models, neural network models (artificial neural network models), neuro-fuzzy models, etc. As for the artificial intelligence models considered herein, various neural network models for machine learning that are already known in the past or will be developed in the future, such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Deep Neural Networks, may be applied.

[0065] Additionally, the learning device (100) may form various combinations of image data and clinical data as input data to build an evaluation model, perform learning using various learning models with the formed input combinations, compare the learning results with the evaluation results through an existing liver function evaluation system, select an input combination and learning model with high performance, accuracy, and reliability, and finally learn and build an evaluation model.

[0066] Referring to FIG. 2, FIG. 2(c) is a diagram illustrating additional advancements and application examples of the constructed evaluation model.

[0067] Referring to FIG. 2(c), the learning device (100) may further perform external validation on the constructed evaluation model using external data. For example, the learning device (100) may perform validation on the model's precision, recall, accuracy, and harmonic mean (F1 Score), etc. The learning device (100) may validate the liver function prediction data output from the evaluation model using a number of external data, and further enhance the evaluation model by retraining the evaluation model and adjusting the model structure and the combination of training data based on the validation results.

[0068] Additionally, referring to (c) of FIG. 2, the evaluation device (200) can take medical imaging data of the subject as input, calculate liver function prediction data of the subject, and provide information on the evaluation of the subject's liver function.

[0069] In this regard, the evaluation device (200) can acquire medical imaging data of a subject. Here, the subject refers to a subject undergoing evaluation and examination of liver function, and may refer to a patient requiring evaluation and examination of liver function. Additionally, the medical imaging data may be magnetic resonance imaging data based on a liver function test of the subject.

[0070] Additionally, the evaluation device (200) can input medical image data into an advanced segmentation algorithm based on feedback information from the learning device (100) to extract image data for multiple tissues of the liver. At this time, the image data for multiple tissues may be, as described above, respective segmented image data for the liver parenchyma, hepatic portal vein, spleen, bile duct, and kidney of the subject.

[0071] Additionally, the evaluation device (200) can input image data into a pre-trained evaluation model to output liver function prediction data. Here, the liver function prediction data may, for example, be data expressed in numerical and categorical formats regarding liver function, and may be data that can be used for the diagnosis of liver disease. Specifically, the liver function prediction data may be used for the early diagnosis of chronic liver disease, assessment of the severity of liver cirrhosis, prediction of the risk of progression and death from decompensated liver cirrhosis, and individualization of treatment based on stratification according to severity.

[0072] As described above, the evaluation device (200) calculates liver function prediction data using an evaluation model learned from data on multiple tissues segmented based on dexacetate-based magnetic resonance imaging data, thereby obtaining data that more sufficiently and accurately reflects the pharmacokinetics of dexacetate compared to existing technology, more specifically, the pharmacokinetics of dexacetate including its distribution in the hepatic portal vein, bile duct, and kidney, and thereby can achieve the effect of comprehensively performing non-invasive functional and structural evaluations of the liver.

[0073] That is, the learning device (100) divides the regions for the liver parenchyma, hepatobiliary ducts, hepatobiliary portal vein, kidney, and spleen in the image data, and by training a model using data extracted for each divided tissue, such as signal intensity and volume information, it can produce liver function prediction data that sufficiently reflects the biliary set dynamics, thereby enabling the construction of an evaluation model capable of more accurate prediction of liver function.

[0074] FIG. 3 is a schematic block diagram of a liver function evaluation model learning device according to one embodiment of the present invention.

[0075] Referring to FIG. 3, the learning device (100) may include an enhancement unit (110), an image acquisition unit (120), a clinical acquisition unit (130), and a model building unit (140).

[0076] According to one embodiment of the present invention, the enhancement unit (110) can enhance the division algorithm based on feedback information. Specifically, with reference to FIG. 3, the enhancement unit (110) may include a standard collection unit (111), an initial generation unit (112), a correct answer generation unit (113), and an algorithm construction unit (114).

[0077] Here, the standard collection unit (111) can collect standard medical data. Additionally, the initial generation unit (112) can input the standard medical data into an initial segmentation algorithm to generate initial mask data. Additionally, the correct answer generation unit (113) can generate correct answer mask data by modifying the initial mask data based on feedback information derived from expert evaluation of the initial mask data. Additionally, the algorithm construction unit (114) can construct an advanced segmentation algorithm by tuning the initial segmentation algorithm using the correct answer mask data.

[0078] According to one embodiment of the present invention, the image acquisition unit (120) can collect image data for a plurality of tissues of the liver based on a segmentation algorithm. Here, the plurality of tissues may include at least one of liver parenchyma, hepatobiliary tract, hepatic portal vein, kidney, and spleen. Additionally, the image data may be generated based on magnetic resonance imaging data based on a cadaveric acid-based cadaveric acid.

[0079] Specifically, the image acquisition unit (120) may generate analysis data by performing an analysis on at least one of the signal strength, volume, and texture of each of a plurality of tissues from segmented data obtained through a segmentation algorithm, and collect image data reflecting the analysis data. At this time, the signal strength may be analyzed based on at least one of the degree of absorption, degree of emission, and degree of contrast enhancement of the dextrin.

[0080] According to one embodiment of the present invention, the clinical collection unit (130) can collect clinical data for a plurality of tissues.

[0081] According to one embodiment of the present invention, the model building unit (140) can build an evaluation model that outputs liver function prediction data with image data and clinical data as inputs.

[0082] FIG. 4 is a schematic block diagram of a liver function evaluation device according to one embodiment of the present invention.

[0083] Referring to FIG. 4, the evaluation device (200) may include an acquisition unit (210), an extraction unit (220), and an output unit (230).

[0084] According to one embodiment of the present invention, the acquisition unit (210) can acquire medical image data of a subject.

[0085] According to one embodiment of the present invention, the extraction unit (220) can extract image data of a plurality of tissues of the liver by inputting medical image data into an advanced segmentation algorithm based on feedback information.

[0086] According to one embodiment of the present invention, the output unit (230) can output liver function prediction data by inputting image data into a pre-trained evaluation model.

[0087] Below, based on the details described above, we will briefly examine the operation flow of the present invention.

[0088] FIG. 5 is a flowchart of the operation of a liver function evaluation model learning method according to one embodiment of the present invention.

[0089] The liver function evaluation model learning method illustrated in FIG. 5 can be performed by the liver function evaluation model learning device (100) described above. Therefore, even if the content described below is omitted, the description of the liver function evaluation model learning device (100) can be equally applied to the description of the liver function evaluation model learning method.

[0090] Referring to FIG. 5, in step S11, the enhancement unit (110) can enhance the division algorithm based on feedback information.

[0091] Next, in step S12, the image acquisition unit (120) can collect image data for multiple tissues of the liver based on a segmentation algorithm. Here, the multiple tissues may include at least one of the liver parenchyma, hepatobiliary ducts, hepatic portal vein, kidney, and spleen. Additionally, the image data may be generated based on magnetic resonance imaging data based on a cadaveric acid-based system.

[0092] Specifically, in step S12, the image acquisition unit (120) may generate analysis data by performing an analysis on at least one of the signal strength, volume, and texture of each of a plurality of tissues from segmented data obtained through a segmentation algorithm, and collect image data reflecting the analysis data. At this time, the signal strength may be analyzed based on at least one of the degree of absorption, degree of emission, and degree of contrast enhancement of acetylcholine.

[0093] Next, in step S13, the clinical collection unit (130) can collect clinical data for multiple tissues.

[0094] Next, in step S14, the model building unit (140) can build an evaluation model that outputs liver function prediction data with image data and clinical data as inputs.

[0095] FIG. 6 is an operation flowchart of the advancement process of a partitioning algorithm according to one embodiment of the present invention. That is, FIG. 6 may be a diagram showing more specifically the operation performed in the advancement unit (110) in the liver function evaluation model learning method.

[0096] Referring to FIG. 6, in step S101, the standard collection unit (111) can collect standard medical data.

[0097] Next, in step S102, the initial generation unit (112) can generate initial mask data by inputting standard medical data into an initial segmentation algorithm.

[0098] Next, in step S103, the answer generation unit (113) can generate answer mask data by modifying the initial mask data based on feedback information based on expert evaluation of the initial mask data.

[0099] Next, in step S104, the algorithm building unit (114) can build an advanced partitioning algorithm by tuning the initial partitioning algorithm using the correct answer mask data.

[0100] FIG. 7 is an operation flowchart of a liver function evaluation method according to one embodiment of the present invention.

[0101] The liver function evaluation method illustrated in FIG. 7 can be performed by the liver function evaluation device (200) described above. Therefore, even if the content described below is omitted, the description of the liver function evaluation device (200) can be equally applied to the description of the liver function evaluation method.

[0102] Referring to FIG. 7, in step S21, the acquisition unit (210) can acquire medical image data of the subject.

[0103] Next, in step S22, the extraction unit (220) can input medical image data into an advanced segmentation algorithm based on feedback information to extract image data for multiple tissues of the liver.

[0104] Next, in step S23, the output unit (230) can output liver function prediction data by inputting image data into a pre-trained evaluation model.

[0105] In the description above, steps S11 to S14, S101 to S104, and S21 to S24 may be further divided into additional steps or combined into fewer steps according to an embodiment of the present invention. Additionally, some steps may be omitted as necessary, and the order between steps may be changed.

[0106] A liver function evaluation model learning method and a liver function evaluation method according to one embodiment of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either individually or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the present invention, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The above-described hardware device may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

[0107] In addition, the aforementioned liver function evaluation model learning method and liver function evaluation method may also be implemented in the form of a computer program or application executed by a computer stored on a recording medium.

[0108] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical concept or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0109] The scope of the present invention is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and the concept of equivalents thereof should be interpreted as being included within the scope of the present invention.

[0110] [Explanation of the symbol]

[0111] 10: Liver function assessment system

[0112] 100: Liver function assessment model learning device

[0113] 110: High-level fire unit

[0114] 111: Standard collection unit

[0115] 112: Initial generation section

[0116] 113: Answer generation section

[0117] 114: Algorithm Construction Section

[0118] 120: Video Acquisition Unit

[0119] 130: Clinical Collection Department

[0120] 140: Model Construction Section

[0121] 200: Liver function evaluation device

[0122] 210: Acquisition Section

[0123] 220: Extraction section

[0124] 230: Output section

[0125] 300: Database

Claims

1. Regarding the training method for the liver function evaluation model, A step of refining the segmentation algorithm based on feedback information; A step of collecting image data of multiple tissues of the liver based on the above-mentioned segmentation algorithm; A step of collecting clinical data for the plurality of tissues mentioned above; and A step of constructing an evaluation model that outputs liver function prediction data using the above image data and the above clinical data as inputs, A learning method including 2. In Paragraph 1, The above image data is, A learning method generated based on magnetic resonance imaging data based on reading set acid.

3. In Paragraph 1, The above plurality of organizations are, A learning method comprising at least one of the liver parenchyma, hepatobiliary tract, hepatic portal vein, kidney, and spleen.

4. In Paragraph 2, The step of collecting image data for the above plurality of tissues is, A learning method comprising generating analysis data by performing an analysis on at least one of signal strength, volume, and texture for each of the plurality of tissues from segmentation data obtained through the segmentation algorithm, and collecting image data that reflects the analysis data.

5. In Paragraph 4, The above signal strength is, A learning method that is analyzed based on at least one of the degree of absorption, degree of excretion, and degree of contrast enhancement of the above-mentioned acetic acid.

6. In Paragraph 1, The step of enhancing the above partitioning algorithm is, Steps for collecting standard medical data; A step of generating initial mask data by inputting the above standard medical data into an initial segmentation algorithm; A step of generating correct answer mask data by modifying the initial mask data based on feedback information based on expert evaluation of the initial mask data; and A step of constructing an advanced partitioning algorithm by tuning the initial partitioning algorithm using the above correct answer mask data, A learning method that includes 7. In a liver function evaluation model learning device, An advancement unit that refines the segmentation algorithm based on feedback information; An image acquisition unit that collects image data of multiple tissues of the liver based on the above-mentioned segmentation algorithm; A clinical data collection unit for collecting clinical data for the plurality of tissues mentioned above; and A model building unit that constructs an evaluation model that outputs liver function prediction data using the above image data and the above clinical data as inputs, A learning device including 8. In Paragraph 7, The above image data is, A learning device generated based on magnetic resonance imaging data based on reading set acid.

9. In Paragraph 7, The above plurality of organizations are, A learning device comprising at least one of the liver parenchyma, hepatobiliary tract, hepatobiliary portal vein, kidney, and spleen.

10. In Paragraph 8, The above image acquisition unit is, A learning device that generates analysis data by performing an analysis on at least one of signal strength, volume, and texture for each of the plurality of tissues from segmentation data obtained through the segmentation algorithm, and collects image data that reflects the analysis data.

11. In Paragraph 10, The above signal strength is, A learning device that is analyzed based on at least one of the degree of absorption, degree of excretion, and degree of contrast enhancement of the above-mentioned acetic acid.

12. In Paragraph 7, The above-mentioned advanced unit is, Standard collection unit for collecting standard medical data; An initial generation unit that generates initial mask data by inputting the above standard medical data into an initial segmentation algorithm; A correct answer generation unit that generates correct answer mask data by modifying the initial mask data based on feedback information based on expert evaluation of the initial mask data; and An algorithm construction unit that builds an advanced partitioning algorithm by tuning the initial partitioning algorithm using the above correct answer mask data, A learning device that includes 13. Regarding liver function evaluation methods, Step of acquiring medical imaging data for a subject; A step of inputting the above medical image data into an advanced segmentation algorithm based on feedback information to extract image data for multiple tissues of the liver; and A step of inputting the above image data into a pre-trained evaluation model to output liver function prediction data, An evaluation method including 14. In a liver function evaluation device, An acquisition unit for acquiring medical image data of a subject; An extraction unit that inputs the above medical image data into an advanced segmentation algorithm based on feedback information to extract image data for multiple tissues of the liver; and An output unit that outputs liver function prediction data by inputting the above image data into a pre-trained evaluation model, An evaluation device including 15. A computer-readable recording medium having a program for executing the method of any one of paragraphs 1 through 6 and paragraph 13 on a computer.

Images