Apparatus and method for analyzing medical examination result images

The apparatus and method leverage artificial neural networks to select and analyze medical test images, addressing the challenges of image analysis by providing accurate and user-friendly results with natural language explanations and digital biomarkers, enhancing diagnostic efficiency and convenience.

WO2024228593A1PCT designated stage expired Publication Date: 2024-11-07SEOUL NAT UNIV HOSPITAL
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
PCT/KR2024/006043
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2024-05-02
Filing Date
2024-05-03
Publication Date
2024-11-07

AI Technical Summary

Technical Problem

Current medical image analysis methods, such as those for electrocardiogram and chest X-ray images, face challenges in accurately selecting and analyzing various types of medical images for high user convenience and efficiency, particularly in providing timely and accurate results.

Method used

An apparatus and method utilizing an artificial neural network-based detection model to automatically select a test area from medical test result images, providing a natural language explanation and digital biomarker test results, with features like confidence scoring and user interface modifications for selecting and resizing candidate boundary boxes, and integrating with large language models for clinical insights.

Benefits of technology

Enables accurate and efficient analysis of medical test results with high user convenience, providing graphical and natural language outputs that enhance diagnostic efficiency and user experience, while allowing for expert consultation and biomarker combination analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure KR2024006043_07112024_PF_FP_ABST
    Figure KR2024006043_07112024_PF_FP_ABST
Patent Text Reader

Abstract

Disclosed are a medical examination result image analysis apparatus and a medical examination result image analysis method performed thereby. The method comprises the steps of: selecting, by a processor, by using an artificial neural network-based detection model, an examination area to be provided for analysis from among medical examination result images displayed on a screen; and analyzing, by the processor, the selected examination area by using an artificial neural network-based analysis model, and providing results of the analysis, wherein the medical examination result images may include one or more of an electrocardiogram image and a chest X-ray image.
Need to check novelty before this filing date? Find Prior Art

Description

Device and method for analyzing medical test result images

[0001] The present specification relates to a device and method for analyzing medical examination result images.

[0002] [Cross-reference to related applications]

[0003] This application claims priority to Republic of Korea Provisional Patent Application No. 10-2023-0058706, filed May 4, 2023, and Republic of Korea Patent Application No. 10-2024-0058514, filed May 2, 2024, the entire contents of which are incorporated herein by reference.

[0004] An electrocardiogram (ECG) is a recording of the electrical activity generated in the myocardium as the heart beats. An ECG is a simple test, with immediate results, and is also useful for ongoing patient monitoring. Chest X-ray images, which reveal information about the thoracic structures and are readily available, are widely used.

[0005] Analysis of medical images, such as electrocardiograms or chest X-rays, plays a vital role in the medical field for purposes such as diagnosing diseases, establishing treatment plans, and monitoring diseases, especially heart-related diseases.

[0006] Meanwhile, advanced computer vision technologies, particularly deep learning-based image analysis models, can contribute to improving the accuracy and efficiency of medical image analysis. However, specialized methods are still required to effectively select and analyze various types of medical images, such as electrocardiograms and chest X-rays, and to provide user-friendly results.

[0007] In embodiments of the present application, a device and method are provided for analyzing a test result image, which automatically selects a test area from a medical test result image such as an electrocardiogram image and / or a chest X-ray image, and then provides a natural language description and / or a diagram of a digital biomarker test result after analysis.

[0008] A method for analyzing a medical examination result image according to one embodiment is a method for analyzing a medical examination result image performed by a medical examination result image analysis device having one or more processors and one or more memories for storing commands executed by the one or more processors, the method comprising: a step performed by the processor, selecting an examination area to be provided for analysis among medical examination result images displayed on a screen using an artificial neural network-based detection model; and a step performed by the processor, analyzing the selected examination area using an artificial neural network-based analysis model to provide an analysis result; wherein the medical examination result image includes one or more of an electrocardiogram image and a chest X-ray image.

[0009] The above detection model can be configured to generate one or more inspection area candidate boundary boxes and output a tuple including one or more of target inspection class information and the candidate boundary box information.

[0010] The above tuple may further include a confidence score (S), which is a quantitative value indicating whether the class information and candidate boundary box information correspond to the target class or target boundary box information.

[0011] The above screen is a screen for capturing a medical examination result image, and the method may include a step of calculating a distance (D) between the center of the capturing screen and the centers of a plurality of candidate boundary boxes; and a step of selecting one of the plurality of candidate boundary boxes as an examination area of ​​the medical examination result image based on the confidence score (S) and the distance (D).

[0012] The method includes a step of selecting one of a plurality of candidate boundary boxes as an examination area of ​​the medical examination result image based on a result value (S') of a function [f(S, D)] having a confidence score (S) and a distance (D) as variables, wherein the function [f(S, D)] can be selected as one of S, 1 / D, and S / D.

[0013] The above distance (D) can be calculated by one or more of the Euclidean distance, the Manhattan distance, and the Chebyshev distance.

[0014] The above class information may include an index value of the target inspection, and the above candidate boundary box information may include a number or a combination of numbers that can define the boundary of the box.

[0015] The method may include a step of changing the area of ​​a plurality of candidate boundary boxes based on a user drag or resize input.

[0016] The above method may include a step of providing analysis results corresponding to a target inspection using an analysis model assigned to each target inspection class.

[0017] The method may further include providing one or more of an examination date, an examination time, patient identification information, and a patient information note in addition to the examination area.

[0018] The above method may further include a step of providing at least one of a test name, a waiting indication, and a completion notification when providing the analysis results.

[0019] The method may include a step of providing a natural language description as an analysis result in the analysis result providing step, and generating the natural language description based on a biomarker value corresponding to the analysis result.

[0020] The above natural language description may be a rule-based description or a large language model-based description.

[0021] The above rule-based description can be automatically output when a given biomarker value corresponding to the analysis result is within a set range.

[0022] The above large language model-based explanation can be input into a large language model and then output by generating a prompt based on a biomarker value corresponding to the analysis result.

[0023] The above large language model-based descriptive text can be generated based on preset rules for one or more pairs of clinical significance and metric-based risk of biomarkers.

[0024] The above prompt may further include additional prompts for describing responses to related requests.

[0025] The step of providing the above analysis results may include a step of providing a visual representation of the digital biomarker test results.

[0026] The above digital biomarker test results may include one or more of the name of the digital biomarker, the digital biomarker value, and the distribution of the value.

[0027] The above digital biomarker test results may include one or more of a type of heart rhythm and a probability of that type of heart rhythm.

[0028] The above digital biomarker test results may further include left heart failure rate.

[0029] The above digital biomarker test result may further provide an electrocardiogram thumbnail image, and the method may further include a step of providing an electrocardiogram original image in response to a request for output of the thumbnail image.

[0030] The method further includes a step of requesting advice from medical professionals regarding medical test results, wherein the step of requesting advice includes the steps of: providing a list of medical professional friends; indicating whether the medical professional friends in the list are available for advice based on the stored work schedules; and transmitting a request for advice to the requested medical professional friend in response to the input of the request for advice and receiving a response regarding whether the medical professional friend is available for advice; wherein the work schedule is electronically updateable, and when an update of the work schedule is automatically detected by the processor, the status of availability for advice can be updated.

[0031] The method may further include the step of providing an activation indicator indicating whether the medical friend is available for consultation.

[0032] The step of providing a user interface for modifying the above work schedule may further be included.

[0033] The above advice request input may include one or more of patient information and advice request items.

[0034] The step of providing the above analysis may include: a step of matching the input first test result with a preset first test biomarker pool; a step of additionally identifying another second test biomarker if the matching result satisfies a predefined condition; a step of analyzing the first test biomarker and the second test biomarker and combining the results; and a step of outputting the combined result.

[0035] The first test is an electrocardiogram, the second test is a chest X-ray, and the preset first test or second test biomarker pool may include biomarkers for one or more of left ventricular dysfunction, right ventricular dysfunction, pulmonary hypertension, pulmonary edema, pericardial effusion, aortic dissection, aortic dilatation, (left / right) atrial hypertrophy, (left / right) ventricular hypertrophy, coronary artery calcium deposition / elevated calcium score, coronary artery disease, hypertrophic / dilated / stress-induced cardiomyopathy, and valvular stenosis / insufficiency / replacement (AV, MV, TV, PV).

[0036] If the ECG biomarker is STEMI / ACS, the combined chest radiography biomarker may be Aortic dissection. If the ECG biomarker is LV Dysfunction or Pulmonary edema, the combined chest radiography biomarker may be LV Dysfunction or Pulmonary edema. If the ECG biomarker is RV Dysfunction or Pulmonary hypertension, the combined chest radiography biomarker may be RV Dysfunction or Pulmonary hypertension or Pulmonary Embolism. If the ECG biomarker is Large Pericardial Effusion, the combined chest radiography biomarker may be Large Pericardial Effusion.

[0037] A medical examination result analysis device according to one embodiment comprises one or more processors; and one or more memories storing instructions executed by the one or more processors, and is a device for analyzing a medical examination result image by the above-described method, wherein the one or more processors are configured to: provide a first UI screen for acquiring a medical examination result image, and provide a second UI screen for displaying one or more of a boundary box and a target class for selecting an examination area from the acquired medical examination result image.

[0038] In one aspect, according to embodiments of the present application, an examination area can be accurately and effectively acquired from a medical examination result image, such as an electrocardiogram image and / or a chest X-ray image, and provided for analysis.

[0039] In another aspect, according to embodiments of the present application, medical image analysis results can be provided with high user convenience by automatically selecting and analyzing the acquired inspection area and then providing a diagram of a natural language description and / or digital biomarker inspection results.

[0040] The effects of the present application are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description of the claims.

[0041] To more clearly explain the exemplary embodiments of the present application, the drawings required for the description of the embodiments are briefly introduced below. It should be understood that the drawings below are for the purpose of illustrating the embodiments of this specification and are not intended to be limiting. Furthermore, for clarity of explanation, some elements may be depicted with various modifications, such as exaggeration or omission, in the drawings below.

[0042] Figure 1 is a screen showing the configuration of an inspection result analysis device in one embodiment of the present application.

[0043] FIG. 2A is a screen illustrating an initial input user interface of an inspection result analysis device in one embodiment of the present application.

[0044] FIG. 2b is a screen illustrating selection of an electrocardiogram image inspection area of ​​a test result analysis device in one embodiment of the present application.

[0045] FIG. 2c is another example screen of selecting an electrocardiogram image inspection area of ​​an inspection result analysis device in one embodiment of the present application.

[0046] FIG. 2d is a screen illustrating a change in area by drag input of a boundary box of an inspection result analysis device in one embodiment of the present application.

[0047] FIG. 2e is a screen illustrating selection of a chest X-ray image examination area of ​​an examination result analysis device in one embodiment of the present application.

[0048] FIG. 2f is a screen illustrating a user interface of a standby screen after analysis of a test result analysis device begins in one embodiment of the present application.

[0049] FIG. 2g is a screen illustrating a user interface of an analysis completion screen of an inspection result analysis device in one embodiment of the present application.

[0050] FIG. 2h is a screen illustrating a natural language description output screen of an inspection result analysis device in one embodiment of the present application.

[0051] FIG. 2i is another example screen of a natural language description output screen of an inspection result analysis device in one embodiment of the present application.

[0052] FIG. 2J is a screen illustrating a digital biomarker test result image of a test result analysis device in one embodiment of the present application.

[0053] FIG. 2K is another example screen of a digital biomarker test result image of a test result analysis device in one embodiment of the present application.

[0054] FIG. 2l is an example screen of other output information images of the inspection result analysis device in one embodiment of the present application.

[0055] FIGS. 3A to 3E are screens illustrating a user interface of an advisory function of an inspection result analysis device in one embodiment of the present application.

[0056] FIGS. 4A and 4B are exemplary screens in which, in one embodiment of the present application, the test result analysis device further instructs the evaluation of chest radiography biomarkers after the evaluation of electrocardiogram biomarkers.

[0057] Hereinafter, some embodiments of the present application will be described in detail with reference to exemplary drawings. When designating components in each drawing, identical components may be given the same reference numerals, as much as possible, even if they are depicted in different drawings. Furthermore, in describing the present embodiments, detailed descriptions of related known structures or functions may be omitted if it is determined that such detailed descriptions may obscure the gist of the present technical concept.

[0058]

[0059] Definition of terms

[0060] In the specification, where "includes," "has," and "consists of" are used, other parts may be added, unless "only" is used. When a component is expressed in the singular, it may include cases where it includes the plural, unless otherwise explicitly stated.

[0061] Additionally, terms such as first, second, A, B, (a), (b), etc. may be used to describe components of the present application. Unless otherwise specified, these terms are only intended to distinguish the components from other components, and do not limit the nature, order, sequence, or number of the components.

[0062] In this specification, terms such as "unit," "module," "device," or "system" are intended to refer not only to hardware but also to a combination of software driven by the hardware. For example, hardware may be a data processing device including a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), or another processor. Furthermore, software may refer to a running process, object, executable, thread of execution, program, etc.

[0063] In this specification, a boundary box or candidate boundary box refers to an area used to select an inspection area to be analyzed on a screen containing an inspection result image. It is referred to as a boundary box because the box's boundaries are defined by the bounding box, which is the boundary of the area.

[0064] In this specification, the inspection area means an analysis target area selected from a screen including an inspection result image.

[0065] In this specification, a class means a display index of a target test, and the index may be expressed as a number, for example.

[0066] As used herein, medical examination images include, but are not limited to, electrocardiogram images and / or chest X-ray images.

[0067] In this specification, acquisition of a medical examination image may include acquisition by a photographing device of a user terminal or acquisition from a storage device of the user terminal.

[0068] In this specification, biomarker means a digital biomarker.

[0069]

[0070] Description of Exemplary Embodiments

[0071] Figure 1 is a screen showing the configuration of an inspection result analysis device in one embodiment of the present application.

[0072] Referring to FIG. 1, a device (200) of one embodiment includes one or more processors (210) and one or more memories (220) that store instructions executed by the one or more processors (210).

[0073] The device acquires a medical examination image, which is a result of a medical examination, selects an examination area to be analyzed from the acquired medical examination image, analyzes it, and provides the analysis result.

[0074] Specifically, the device may include a first UI screen including, for example, a capture button for obtaining a medical examination image, and a second UI screen displaying one or more of a boundary box and a target class for selecting an examination area from the obtained medical examination image.

[0075] The processor of the above device can provide analysis results from medical examination result images as follows.

[0076] In one embodiment, the processor may perform a step of selecting an examination area to be provided for analysis among medical examination result images displayed on a screen using an artificial neural network-based detection model, and a step performed by the processor of analyzing the selected examination area using an artificial neural network-based analysis model to provide an analysis result.

[0077] Fig. 2a is a screen illustrating an initial input user interface of a test result analysis device, Fig. 2b is a screen illustrating selection of an electrocardiogram image inspection area of ​​a test result analysis device, Fig. 2c is another screen illustrating selection of an electrocardiogram image inspection area of ​​a test result analysis device, Fig. 2d is a screen illustrating change of an area by drag input of a boundary box of a test result analysis device, and Fig. 2e is a screen illustrating selection of an examination area of ​​a chest X-ray image of a test result analysis device.

[0078] As can be seen in Fig. 2a, when the user presses the capture button or the analysis button on the initial input user interface screen displayed on the user terminal display, a new medical examination result image such as an electrocardiogram image and / or a chest X-ray image can be captured, or a recent medical examination result image stored in a storage device of the user terminal can be called up and displayed.

[0079] As shown in Figures 2b and 2c, an inspection area to be provided for analysis from a captured or called medical examination result image is selected using an artificial neural network-based detection model. The detection model generates one or more candidate inspection area bounding boxes and outputs a tuple containing information about the target examination class and one or more of the candidate boundary boxes.

[0080] The above class information includes index values ​​of target examinations such as electrocardiogram and / or chest X-ray.

[0081] The above candidate boundary box information includes a numerical value or a combination of numerical values ​​that can define the boundaries of the box. For example, the target examination area. For example, the target examination electrocardiogram image area and / or the target examination chest X-ray image may include the coordinate values ​​of the four corners of the rectangle of the boundary box, or a combination of numerical values ​​in another format that can define the same.

[0082] In one embodiment, in order to select a test area from a plurality of candidate boundary boxes, the tuple may further include a confidence score (S), which is a quantitative value indicating whether the class information and the candidate boundary box information correspond to the target class or the target boundary box information.

[0083] In addition, the distance (D) between the center of the shooting screen of the user terminal and the center of the plurality of candidate boundary boxes is calculated, and based on the confidence score (S) and the distance (D), one of the plurality of candidate boundary boxes can be selected as the inspection area of ​​the medical examination result image.

[0084] More specifically, the method selects one of a plurality of candidate boundary boxes as an examination area of ​​the medical examination result image based on a result value (S') of a function [f(S, D)] that has a confidence score (S) and a distance (D) as variables, wherein the function [f(S, D)] may be one selected from among S, 1 / D, and S / D, and corresponds to all types of functions in which S is maximized and D is minimized when the output of the f function is maximum or minimum.

[0085] In one embodiment, the distance (D) can be calculated by one or more of the Euclidean distance, the Manhattan distance, and the Chebyshev distance.

[0086] In one embodiment, a boundary box that minimizes or maximizes the aforementioned confidence score (S) and / or the result value (S') of the function [f(S, D)] may be selected as an inspection area among candidate boundary boxes.

[0087] As shown in FIG. 2d, one embodiment of the present application can change the area of ​​the plurality of candidate boundary boxes based on a user drag or resize input. For the user drag, the drag points can be indicated, for example, at the four corners of the boundary box.

[0088] Figure 2e illustrates selecting an examination area from a chest X-ray image rather than an electrocardiogram image. The same detection algorithm can be used for selecting classes and selecting an examination area from a boundary box for a chest X-ray image as for an electrocardiogram image.

[0089] In the selected test area above, one or more of the following may be further included: test date, test time, patient identification information such as a barcode or QR code, patient information note (information about the patient other than the patient identification information), which may be stored on the device and provided for later analysis.

[0090] Next, the analysis steps are described.

[0091] In one embodiment, analysis can be initiated by cropping the area selected as the inspection area.

[0092] In one embodiment, analysis results corresponding to a target examination may be provided using an analysis model assigned to each target examination class, such as an electrocardiogram and / or chest X-ray.

[0093] Figure 2f is a screen illustrating a user interface of a standby screen after analysis of a test result analysis device has started, and Figure 2g is a screen illustrating a user interface of a screen after analysis of a test result analysis device has ended.

[0094] As illustrated in FIGS. 2f and 2g, when providing the analysis results, one or more of the test name, waiting indication, and completion notification may be further provided.

[0095] For example, a waiting indicator could be an animated indicator indicating that the test is waiting. Upon completion of the test, the animation could stop, an alarm sound could be played, and an alarm notification window could be created.

[0096] Meanwhile, Fig. 2h is a screen showing an example of a natural language description output screen of a test result analysis device, and Fig. 2i is another example screen showing a natural language description output screen of a test result analysis device.

[0097] In one embodiment, when providing analysis results, a natural language description may be provided. The natural language description may be generated based on the biomarker values ​​corresponding to the analysis results.

[0098] As an example, the natural language description may be a rule-based description or a large language model-based description.

[0099] The above rule-based description is automatically output when a given biomarker value corresponding to the analysis result is within a set range, and the range condition may be a single or complex condition.

[0100] The above large language model-based explanation may be a text that generates a prompt based on a biomarker value corresponding to the analysis result, inputs it into a large language model, and then outputs it.

[0101] In one embodiment, generating a prompt based on the biomarker value may include generating one or more pairs of clinical significance and metric-based risk of the biomarker according to preset rules.

[0102] Additionally, in one embodiment, the prompt may further include an additional prompt for describing a response to the associated request.

[0103] Meanwhile, FIG. 2j is a screen illustrating a digital biomarker test result image of the test result analysis device, FIG. 2k is another screen illustrating a digital biomarker test result image of the test result analysis device, and FIG. 2l is an example screen illustrating other output information images of the test result analysis device.

[0104] As illustrated in FIGS. 2J to 2L, the digital biomarker test results can be graphically provided when providing the analysis results.

[0105] The above digital biomarker test results may include one or more of the name of the digital biomarker, the digital biomarker value, and distribution information of the value, such as a confidence interval.

[0106] As an example, the digital biomarker test result may include one or more of the following: a type of heart rhythm and a probability of that type of heart rhythm. For example, the digital biomarker test result may include a left ventricular fibrillation rate.

[0107] Additionally, in one example, the digital biomarker test result may further provide an electrocardiogram thumbnail image, and in response to a request for output of the thumbnail image, an electrocardiogram original image may further be provided.

[0108] Figures 3a to 3e are screens illustrating the user interface of the advisory function of the inspection result analysis device.

[0109] As illustrated in FIGS. 3A to 3E, the medical examination result image analysis method of one embodiment of the present application may further include a step of requesting mutual advice between medical professionals regarding the medical examination results.

[0110] The step of requesting the above-mentioned advice may include a step of providing a list of medical friends, a step of indicating whether the medical friends in the list are available for advice based on the saved work schedules, a step of transmitting a request for advice to the requested medical friend in response to the input of the request for advice and receiving a response regarding whether the request for advice is available, and an activation display indicating whether the medical friend is available for advice may be further provided.

[0111] Here, the schedule can be electronically updated, and the processor can automatically detect updates to the schedule and update the advisory status. Additionally, a user interface for modifying the schedule can be further provided.

[0112] The above advice request input may include one or more of patient information and advice request items.

[0113] Figures 4a and 4b are exemplary screens where the test result analysis device further instructs chest radiography biomarker evaluation after electrocardiogram biomarker evaluation.

[0114] As illustrated in FIGS. 4A and 4B, the processor of the test result analysis device may match the input first test result (electrocardiogram or chest X-ray) when providing the analysis with a preset first test biomarker pool, and if the matching result satisfies a predefined condition, instruct to additionally check another second test (chest X-ray or electrocardiogram) biomarker, combine the first test (electrocardiogram or chest X-ray) biomarker and the second test (chest X-ray or electrocardiogram) biomarker results, and output the combined result.

[0115] The above preset first or second test biomarker pool may include biomarkers for one or more of left ventricular dysfunction, right ventricular dysfunction, pulmonary hypertension, pulmonary edema, pericardial effusion, aortic dissection, aortic dilatation, (left / right) atrial hypertrophy, (left / right) ventricular hypertrophy, coronary artery calcium deposition / elevated calcium score, coronary artery disease, hypertrophic / dilated / stress-induced cardiomyopathy, and valvular stenosis / insufficiency / replacement (AV, MV, TV, PV).

[0116] For example, if the ECG biomarker is STEMI / ACS, the combined chest radiography biomarker may be Aortic dissection.

[0117] If the ECG biomarker is LV Dysfunction or Pulmonary edema, the combined chest radiography biomarker may also be LV Dysfunction or Pulmonary edema.

[0118] If the ECG biomarker is RV Dysfunction or Pulmonary Hypertension, the combined chest radiography biomarker may be RV Dysfunction or Pulmonary Hypertension or Pulmonary Embolism.

[0119] If the ECG biomarker is Large Pericardial Effusion, the combined chest radiography biomarker may also be Large Pericardial Effusion.

[0120] The operations of the devices according to the embodiments described above may be implemented at least partially by a computer, or as a computer program, and recorded on a computer-readable recording medium. For example, the operations may be implemented together with a program product comprising a computer-readable medium containing program code, which may be executed by a processor to perform any or all of the described steps, operations, or processes.

[0121] The computer may be any computing device, such as a desktop computer, laptop computer, notebook computer, smartphone, or the like, or may be integrated into any such device. The computer is a device having one or more alternative and special purpose processors, memory, storage space, and networking components (either wireless or wired). The computer may run an operating system, such as, for example, an operating system compatible with Microsoft's Windows, Apple's OS X or iOS, a Linux distribution, or Google's Android OS.

[0122] The computer-readable recording medium includes all types of recording devices that store data that can be read by a computer. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc. In addition, the computer-readable recording medium may be distributed across network-connected computer systems, so that computer-readable codes can be stored and executed in a distributed manner. In addition, functional programs, codes, and code segments for implementing the present embodiment will be readily understood by those skilled in the art to which the present embodiment pertains.

[0123] While the present invention has been described above with reference to the embodiments illustrated in the drawings, these are merely exemplary, and those skilled in the art will appreciate that various modifications and variations of the embodiments are possible. However, such modifications should be considered within the technical protection scope of the present invention. Therefore, the true technical protection scope of the present invention should be determined by the technical spirit of the appended claims.

[0124] A device and method for analyzing a medical examination result image according to one embodiment automatically selects and analyzes an examination area using an artificial neural network, and can be used in the digital medical industry.

Claims

1. A medical examination result image analysis method performed by a medical examination result image analysis device having one or more processors and one or more memories storing instructions executed by the one or more processors, A step of selecting an examination area to be provided for analysis among the medical examination result images displayed on the screen by using an artificial neural network-based detection model, performed by the above processor; and A step of performing the above-described processor and analyzing the selected inspection area using an artificial neural network-based analysis model to provide an analysis result; A method for analyzing a medical test result image, wherein the medical test result image includes at least one of an electrocardiogram image and a chest X-ray image.

2. In paragraph 1, A method for analyzing a medical examination result image, wherein the detection model is configured to generate one or more inspection area candidate boundary boxes and output a tuple including one or more of target examination class information and the candidate boundary box information.

3. In paragraph 2, A method for analyzing medical examination result images, wherein the above tuple further includes a confidence score (S), which is a quantitative value indicating whether the class information and candidate boundary box information correspond to the target class or target boundary box information.

4. In paragraph 3, The above screen is a screen for capturing medical test results images. The method comprises the steps of calculating the distance (D) between the center of the shooting screen and the centers of a plurality of candidate boundary boxes; and A method for analyzing a medical examination result image, comprising: a step of selecting one of a plurality of candidate boundary boxes as an examination area of ​​the medical examination result image based on the confidence score (S) and the distance (D).

5. In paragraph 4, The method includes a step of selecting one of a plurality of candidate boundary boxes as an examination area of ​​the medical examination result image based on a result value (S') of a function [f(S, D)] having a confidence score (S) and a distance (D) as variables. A method for analyzing medical examination result images, wherein the above function [f(S, D)] is selected as one of S, 1 / D, and S / D.

6. In paragraph 4, A method for analyzing medical examination result images, wherein the above distance (D) is calculated by at least one of the Euclidean distance, the Manhattan distance, and the Chebyshev distance.

7. In paragraph 2, The above class information includes the index value of the target inspection, A method for analyzing an image of a medical examination result, wherein the above candidate boundary box information includes a number or a combination of numbers that can define the boundary of the box.

8. In paragraph 1, A method for analyzing medical examination result images, the method comprising a step of changing the area of ​​a plurality of candidate boundary boxes based on a user drag or resize input.

9. In paragraph 2, A method for analyzing medical examination result images, the method comprising a step of providing analysis results corresponding to a target examination using an analysis model assigned to each target examination class.

10. In paragraph 1, A method for analyzing medical examination result images, wherein the method further comprises a step of providing at least one of an examination date, an examination time, patient identification information, and a patient information note in addition to an examination area.

11. In paragraph 1, A method for analyzing medical test result images, wherein the method further includes a step of providing at least one of a test name, a waiting indication, and a completion notification when providing the analysis results.

12. In paragraph 1, A method for analyzing an image of a medical test result, wherein the method comprises a step of providing a natural language description as an analysis result in the analysis result providing step, and generating the natural language description based on a biomarker value corresponding to the analysis result.

13. In paragraph 12, A method for analyzing medical examination result images, wherein the above natural language description is a rule-based description or a large language model-based description.

14. In paragraph 13, A method for analyzing medical test result images, wherein the above rule-based description is automatically output when a set biomarker value corresponding to the analysis result is within a set range.

15. In paragraph 13, A method for analyzing medical test result images, wherein the above large language model-based explanation generates a prompt based on a biomarker value corresponding to the analysis result, inputs the prompt into a large language model, and then outputs the prompt.

16. In paragraph 15, A method for analyzing medical test result images, wherein the large language model-based description is generated according to preset rules, wherein one or more pairs of clinical meaning and measurement-based risk of biomarkers are generated.

17. In paragraph 15, A method for analyzing medical examination result images, wherein the above prompt further includes an additional prompt for describing a response to the related request content.

18. In paragraph 1, A method for analyzing medical test result images, wherein the step of providing the above analysis results includes a step of providing a digital biomarker test result by visualizing it.

19. In paragraph 18, A method for analyzing an image of a medical test result, wherein the digital biomarker test result includes at least one of a name of a digital biomarker, a digital biomarker value, and a distribution of the value.

20. In paragraph 18, A method for analyzing an image of a medical test result, wherein the digital biomarker test result includes at least one of a type of heart rhythm and a probability of the type of heart rhythm.

21. In paragraph 18, A method for analyzing medical test result images, wherein the above digital biomarker test results further include a left heart failure rate.

22. In paragraph 18, The above digital biomarker test results further provide an electrocardiogram thumbnail image, A method for analyzing a medical examination result image, wherein the method further comprises a step of providing an original electrocardiogram image in response to a request for output of the thumbnail image.

23. In paragraph 1, The method further includes a step of requesting advice from medical professionals regarding the results of medical tests, wherein the step of requesting advice comprises: Steps to provide a list of medical friends; A step of indicating whether a medical friend in the list is available for consultation based on the saved work schedule; and In response to a request for advice input, a step of transmitting a request for advice to a medical professional friend whose advice has been requested and receiving a response as to whether or not to provide advice is included; The above schedule can be updated electronically, A method for analyzing medical examination result images, wherein when an update of the work schedule is automatically detected by the processor, the status of availability for consultation is updated.

24. In paragraph 23, A method for analyzing medical examination result images, wherein the method further comprises a step of providing an activation indicator indicating whether a medical professional friend is available for consultation.

25. In paragraph 23, A method for analyzing medical examination result images, further comprising the step of providing a user interface for modifying the above work schedule.

26. In paragraph 23, A method for analyzing medical examination result images, wherein the above-mentioned advice request input includes at least one of patient information and advice request items.

27. In paragraph 1, The steps of providing the above analysis are: A step of matching the input first test result with a preset first test biomarker pool; If the matching result meets the predefined conditions, a step of additionally checking other second test biomarkers; A step of analyzing the first test biomarker and the second test biomarker and combining the results; and A method for analyzing medical examination result images, comprising: a step of outputting the combined results.

28. In paragraph 27, The first test is an electrocardiogram, and the second test is a chest X-ray. A method for analyzing an image of medical test results, wherein the preset first or second test biomarker pool includes a biomarker for one or more of left ventricular dysfunction, right ventricular dysfunction, pulmonary hypertension, pulmonary edema, pericardial effusion, aortic dissection, aortic dilatation, (left / right) atrial hypertrophy, (left / right) ventricular hypertrophy, coronary artery calcium deposition / elevated calcium score, coronary artery disease, hypertrophic / dilated / stress-induced cardiomyopathy, and valvular stenosis / insufficiency / replacement (AV, MV, TV, PV).

29. In paragraph 28, If the ECG biomarker is STEMI / ACS, the combined chest radiography biomarker is Aortic dissection, If the electrocardiogram biomarker is LV Dysfunction or Pulmonary edema, the combined chest radiography biomarker is LV Dysfunction or Pulmonary edema, If the electrocardiogram biomarker is RV Dysfunction or Pulmonary hypertension, the combined chest radiography biomarker is RV Dysfunction or Pulmonary hypertension or Pulmonary Embolism, A method for analyzing medical test results images in which the electrocardiogram biomarker is Large Pericardial Effusion and the combined chest radiography biomarker is Large Pericardial Effusion.

30. One or more processors; and A device for analyzing a medical examination result image by the method of any one of claims 1 to 29, comprising one or more memories for storing instructions executed by one or more processors, One or more of the above processors: Provides a first UI screen for obtaining medical test result images, A medical examination result image analysis device configured to provide a second UI screen displaying at least one of a boundary box and a target class for selecting an examination area from the obtained medical examination result image.

Citation Information

Patent Citations

  • Apparatus and method for analyzing medical test result image

    KR1020240161770A

  • Evaluation support system and evaluation support method for supporting evaluation of circulatory system condition

    JP7079553B2

  • AI-based ECG reading system

    KR102142841B1

  • Skin disease care System for user specific real time service

    KR102165699B1

  • Automatic opinion generation system for livestock diagnosis using veterinary experience knowledge and professional information

    KR102202810B1