Electronic device and method for distinguishing between types of thyrotoxicosis including graves' disease, autoimmune thyroiditis, and subacute thyroiditis on basis of machine-learning

Abstract

A method for distinguishing between types of thyrotoxicosis including Graves' disease, autoimmune thyroiditis, and subacute thyroiditis on the basis of machine learning according to an embodiment of the present invention comprises the steps of: acquiring a basic ultrasound image (gray scale), a color Doppler image, and a power Doppler image of the thyroid of a diagnosis subject; and determining whether the diagnosis subject corresponds to one of Graves' disease, autoimmune thyroiditis, or subacute thyroiditis by using a machine-learning model trained to distinguish between types of thyrotoxicosis on the basis of basic ultrasound images, color Doppler images, and power Doppler images of the thyroids of each of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis patients.

Description

Electronic device and method for differentiating thyrotoxicosis diseases including Graves' disease, autoimmune thyroiditis, and subacute thyroiditis based on machine learning

[0001] The present invention relates to an electronic device and method for diagnosing hyperthyroidism based on machine learning.

[0002] Thyrotoxicosis is a condition characterized by an excessive increase in thyroid hormones in the body, resulting in various clinical symptoms and signs. Causes of thyrotoxicosis include hyperthyroidism, which involves excessive thyroid hormone secretion, and thyroiditis, which occurs when the thyroid gland is damaged by inflammation or other factors, leading to a sudden surge of thyroid hormones.

[0003] Graves' disease, which causes hyperthyroidism, is a common autoimmune disease, affecting approximately 0.5% of the general population. Thyroiditis, on the other hand, can be caused by painless thyroiditis (silent thyroiditis), subacute thyroiditis, or autoimmune thyroiditis (Hashimoto's disease).

[0004] Meanwhile, prompt decisions regarding treatment for patients with thyrotoxicosis are clinically crucial. While antithyroid drugs are essential for Graves' disease, thyroiditis requires conservative treatment to control symptoms. Inappropriate use of antithyroid drugs can actually lead to hypothyroidism.

[0005] Currently, serum thyroid-stimulating hormone receptor antibody (TSH-R-Ab) measurement and thyroid scan are used to differentiate between thyroiditis and Graves' disease in clinical practice.

[0006] However, in the case of TSH-R-Ab, it takes at least 3 days to confirm the results after the test, and although rare, there is a disadvantage in that the results are sometimes inconsistent (3-5% of cases where Graves' disease and thyroiditis cannot be differentiated) depending on the antibody test method.

[0007] On the other hand, thyroid scans are only available at tertiary hospitals or larger hospitals, and cannot be performed on pregnant women or those with multiple thyroid nodules. Previously, thyroid ultrasound was considered of limited utility in patients with thyrotoxicosis. However, recent studies have highlighted the importance of thyroid ultrasound in making immediate treatment decisions, given the limitations of thyroid scans and TSH-R-Ab.

[0008] Doppler ultrasound, a complementary ultrasound technique used in the diagnosis of Graves' disease, is used as an adjunctive method to blood tests for real-time diagnosis. However, its sensitivity is only 84%, limiting its diagnostic utility. Previous studies have shown high inter-observer variability, making it difficult to obtain objective results, and research findings are limited. Furthermore, the currently used color Doppler test is highly dependent on the subjectivity of the examiner.

[0009] Furthermore, while microvascular flow imaging offers the advantage of objectively measuring blood flow in the differential diagnosis of thyrotoxicosis, it lacks the accuracy of conventional Doppler techniques. Furthermore, it exhibits intra-observer and inter-observer variability in its results.

[0010] Therefore, it is necessary to find the best combination to differentiate splenic toxicity by combining basic ultrasound images (gray scale) rather than just one element such as Doppler imaging or microvascular imaging.

[0011] The purpose of the present invention is to provide an electronic device and method for diagnosing hyperthyroidism with higher reliability by minimizing observer variation.

[0012] A method for differentiating thyrotoxicosis, including Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, based on machine learning according to one embodiment of the present invention may include the steps of: acquiring a basic ultrasound image (gray scale), a color Doppler image (color Doppler), and a power Doppler image (power Doppler) of a thyroid of a subject to be diagnosed; and using a machine learning model learned to differentiate thyrotoxicosis based on the basic ultrasound image, the color Doppler image, and the power Doppler image of the thyroid of each patient with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, the step of differentiating whether the subject to be diagnosed corresponds to any one of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis.

[0013] The method may further include: acquiring learning data including basic ultrasound images, color Doppler images, and power Doppler images of the thyroid gland of patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis; learning patterns of increased blood flow for each of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis using the learning data; and learning to differentiate Graves' disease, autoimmune thyroiditis, and subacute thyroiditis through differences in patterns of increased blood flow between Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, thereby generating the machine learning model.

[0014] The step of learning the pattern of increased blood flow may include the step of setting a region of interest for each image frame using learning data for patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis; and the step of identifying the proportion of color pixels among all gray scale pixels of the region of interest.

[0015] In one embodiment of the present invention, an electronic device for differentiating thyrotoxicosis including Graves' disease, autoimmune thyroiditis, and subacute thyroiditis based on machine learning may include a processor for acquiring a basic ultrasound image (gray scale), a color Doppler image (color Doppler), and a power Doppler image (power Doppler) of a thyroid of a diagnosis target, and using a machine learning model learned to differentiate thyrotoxicosis based on the basic ultrasound image, the color Doppler image, and the power Doppler image of the thyroid of each patient with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis to determine whether the diagnosis target corresponds to any one of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis.

[0016] The above processor can obtain learning data including basic ultrasound images, color Doppler images, and output Doppler images of the thyroid gland of patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, learn patterns of increased blood flow for each of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis using the learning data, and learn to differentiate Graves' disease, autoimmune thyroiditis, and subacute thyroiditis through differences in patterns of increased blood flow between Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, thereby generating the machine learning model.

[0017] The above processor can set a region of interest for each image frame using learning data for patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, and identify the ratio of color pixels among all grayscale pixels of the region of interest.

[0018] According to one embodiment of the present invention, hyperthyroidism can be diagnosed more objectively by excluding the subjectivity of the observer.

[0019] According to one embodiment of the present invention, by using various types of ultrasound images, the accuracy of differentiation of hyperthyroidism can be further improved.

[0020] According to one embodiment of the present invention, machine learning can be used to minimize inter-observer variation and to differentiate thyrotoxicosis from a pattern of increased blood flow that is difficult for humans to distinguish.

[0021] According to one embodiment of the present invention, it is expected that it will be helpful to observers (medical staff) with little experience in thyroid diseases.

[0022] Figure 1 is a schematic diagram illustrating a hyperthyroidism disease diagnosis system according to one embodiment of the present invention.

[0023] FIG. 2 is a block diagram illustrating the configuration of an electronic device according to one embodiment of the present invention.

[0024] FIG. 3 is a diagram illustrating an operation flow chart of an electronic device according to one embodiment of the present invention.

[0025] FIG. 4 is a diagram illustrating a process of learning a machine learning model for diagnosing hyperthyroidism according to one embodiment of the present invention.

[0026] Figure 5 is a flowchart illustrating an operation of learning a machine learning model for diagnosing hyperthyroidism according to one embodiment of the present invention.

[0027] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The detailed description set forth below, together with the accompanying drawings, is intended to explain exemplary embodiments of the present invention and is not intended to represent the only embodiments in which the present invention may be practiced. In the drawings, portions irrelevant to the description may be omitted for clarity in describing the present invention, and the same reference numerals may be used throughout the specification for identical or similar components.

[0028] Figure 1 is a schematic diagram illustrating a hyperthyroidism disease diagnosis system according to one embodiment of the present invention.

[0029] The system (1) for diagnosing hyperthyroidism of FIG. 1 (hereinafter also referred to as system (1)) includes an electronic device (100) and an ultrasonic device (200). According to one embodiment of the present invention, the electronic device (100) is a device that uses an ultrasonic image (10) to diagnose hyperthyroidism and provides diagnostic information (20), and may be implemented as a computer, a server, a smart phone, a tablet PC, a smart pad, a laptop, etc.

[0030] According to one embodiment of the present invention, an ultrasonic device (200) is a device that transmits ultrasonic waves to a thyroid examination site through a transducer and then receives the returning ultrasonic waves. The ultrasonic device (200) can operate in multiple modes, for example, a basic mode, B mode (brightness mode), and a Doppler mode that utilizes the Doppler effect.

[0031] At this time, the Doppler mode includes a color Doppler mode that can express the direction of blood flow with color and the speed of blood flow with brightness, and a power Doppler mode that can express blood flow more sensitively than the color Doppler mode.

[0032] According to one embodiment of the present invention, an electronic device (100) can be connected to an ultrasonic device (200) via wired or wireless communication to receive ultrasonic information and visualize it in real time to generate an ultrasonic image (10). However, the present invention is not limited thereto, and a previously generated ultrasonic image (10) can be received from an external device. The ultrasonic image (10) may be a moving image composed of a plurality of image frames.

[0033] Accordingly, the electronic device (100) can obtain a basic ultrasound image (11) in the basic mode, and a color Doppler image (12) and an output Doppler image (13) in the Doppler mode.

[0034] Meanwhile, as described above, in the past, thyrotoxicosis disease was differentiated using ultrasound images according to a single mode, but the differential diagnosis was limited due to the advantages and disadvantages of each mode.

[0035] In the present invention, a method is proposed to overcome and integrate the limitations of existing technologies by using a machine learning algorithm that has been trained to differentiate hyperthyroidism using not only basic ultrasound images (11) but also color Doppler images (12) and output Doppler images (13).

[0036] Hereinafter, the configuration and operation of an electronic device according to one embodiment of the present invention will be specifically described with reference to the drawings.

[0037] Figure 2 is a block diagram illustrating the configuration of an electronic device according to one embodiment of the present invention.

[0038] An electronic device (100) according to one embodiment of the present invention includes an input unit (110), a communication unit (120), a display unit (130), a storage unit (140), and a processor (150).

[0039] The input unit (110) generates input data in response to a user input of the electronic device (100). For example, the user input may be a user input that initiates the operation of the electronic device (100), a user input for uploading an ultrasound image (10), a user input for generating a machine learning model for diagnosing hyperthyroidism, etc. In addition, any user input required for diagnosing hyperthyroidism may be applied without limitation.

[0040] The input unit (110) includes at least one input means. The input unit (110) may include a keyboard, a key pad, a dome switch, a touch panel, a touch key, a mouse, a menu button, etc.

[0041] The communication unit (120) performs communication with external devices such as an ultrasonic device (200) and a server to transmit and receive ultrasonic images (10), diagnostic information (20), etc.

[0042] To this end, the communication unit (120) can perform wireless communication such as 5G (5th generation communication), LTE-A (long term evolution-advanced), LTE (long term evolution), Wi-Fi (wireless fidelity), Bluetooth, or wired communication such as LAN (local area network), WAN (Wide Area Network), and power line communication.

[0043] The display unit (130) displays display data according to the operation of the electronic device (100). The display unit (130) can display a screen for displaying an ultrasound image (10), a screen for displaying diagnostic information (20), a screen for receiving user input, etc.

[0044] The display unit (130) includes a liquid crystal display (LCD), a light emitting diode (LED) display, an organic light emitting diode (OLED) display, a micro electro mechanical systems (MEMS) display, and an electronic paper display. The display unit (130) may be implemented as a touch screen by being combined with the input unit (110).

[0045] The storage unit (140) stores the operation programs of the electronic device (100). The storage unit (140) includes non-volatile storage that can preserve data (information) regardless of whether power is supplied, and volatile memory that cannot preserve data if power is not supplied and into which data to be processed by the processor (150) is loaded. The storage includes flash memory, hard-disc drive (HDD), solid-state drive (SSD), read-only memory (ROM), etc., and the memory includes buffer, random access memory (RAM), etc.

[0046] The storage unit (140) can store ultrasound images (10), diagnostic information (20), machine learning models (30), etc. The storage unit (140) can store computational programs, etc. required in the process of generating ultrasound images (10), generating machine learning models (30), etc.

[0047] The processor (150) can control at least one other component (e.g., hardware or software component) of the electronic device (100) by executing software such as a program, and can perform various data processing or operations.

[0048] A processor (150) according to one embodiment of the present invention acquires a basic ultrasound image, a color Doppler image, and an output Doppler image of the thyroid gland of a diagnosis target, and uses a machine learning model (30) learned to differentiate thyrotoxicosis disease based on the basic ultrasound image, the color Doppler image, and the output Doppler image of the thyroid gland of each patient with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis to determine whether the diagnosis target corresponds to any one of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis.

[0049] At this time, the processor (150) may learn a machine learning model (30) or may receive and store a previously learned and generated model from the outside and use it, but is not limited to either. The process of the processor (150) generating a machine learning model (30) is described with reference to FIGS. 4 and 5.

[0050] Meanwhile, the processor (150) may perform at least a portion of the data analysis, processing, and result information generation for performing the above operations using at least one of a machine learning, neural network, or deep learning algorithm as a rule-based or artificial intelligence (AI) algorithm. Examples of the neural network may include models such as a CNN (Convolutional Neural Network), a DNN (Deep Neural Network), and an RNN (Recurrent Neural Network).

[0051] FIG. 3 is a diagram illustrating an operation flow chart of an electronic device according to one embodiment of the present invention.

[0052] According to one embodiment of the present invention, the processor (150) can obtain a basic ultrasound image, a color Doppler image, and an output Doppler image of the thyroid gland of the diagnosis target (S10).

[0053] A diagnostic subject according to one embodiment of the present invention refers to a patient suspected of having hyperthyroidism.

[0054] As described above, the processor (150) can generate the ultrasound images based on information received in real time from the ultrasound device (200), or can receive previously generated images.

[0055] According to one embodiment of the present invention, the processor (150) can determine whether the diagnosis subject corresponds to any one of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis by using a machine learning model learned to differentiate thyrotoxicosis disease based on basic ultrasound images, color Doppler images, and output Doppler images of the thyroid gland of patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis (S20).

[0056] The processor (150) inputs a basic ultrasound image, a color Doppler image, and an output Doppler image of the thyroid gland of the subject of diagnosis into a machine learning model, and as a result thereof, can output diagnostic information (20) including information that determines whether the subject of diagnosis corresponds to any one of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis.

[0057] According to one embodiment of the present invention, the diagnostic information (20) is information about the disease of the diagnosis target, and may include the type or probability of the disease, and may include information about the blood flow increase pattern described below.

[0058] Meanwhile, the processor (150) can determine whether the ultrasound image of the diagnostic target corresponds to one of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis by comparing the ultrasound image with the ultrasound image pattern for each thyrotoxicosis disease.

[0059] According to one embodiment of the present invention, hyperthyroidism can be diagnosed more objectively by excluding the subjectivity of the observer.

[0060] According to one embodiment of the present invention, by using various types of ultrasound images, the accuracy of differentiation of hyperthyroidism can be further improved.

[0061] Hereinafter, with reference to FIGS. 4 and 5, a process of learning and generating a machine learning model (30) by a processor (150) according to one embodiment of the present invention will be described in detail.

[0062] FIG. 4 is a diagram illustrating a process of learning a machine learning model for diagnosing hyperthyroidism according to one embodiment of the present invention.

[0063] According to one embodiment of the present invention, the electronic device (100) can use an ultrasound image (410) including a basic ultrasound image, a color Doppler image, and an output Doppler image of the thyroid of a patient with Graves' disease, an ultrasound image (420) including a basic ultrasound image, a color Doppler image, and an output Doppler image of the thyroid of a patient with autoimmune thyroiditis, and an ultrasound image (430) including a basic ultrasound image, a color Doppler image, and an output Doppler image of the thyroid of a patient with subacute thyroiditis to generate a machine learning model (30).

[0064] Figure 5 is a flowchart illustrating an operation of learning a machine learning model for diagnosing hyperthyroidism according to one embodiment of the present invention.

[0065] According to one embodiment of the present invention, the processor (150) can obtain learning data including ultrasound images (410), ultrasound images (420), and ultrasound images (430) of the thyroid gland of patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis (S510).

[0066] Specifically, patients with hyperthyroidism can be screened through a questionnaire including clinical signs and physical examinations, medical charts, and medical histories, and subjects can be selected by checking the patient's thyroid blood test values ​​and thyroid autoantibody levels.

[0067] For each subject, an ultrasound examination can be performed on the thyroid gland in the three modes described above (basic mode, color Doppler mode, and power Doppler mode), and ultrasound images can be acquired. Each ultrasound image can be acquired in video format.

[0068] Alternatively, the processor (150) can search for patients with diagnostic codes such as hyperthyroidism, Graves' disease, autoimmune thyroiditis, and subacute thyroiditis through a hospital server, etc., and collect ultrasound images provided for each patient.

[0069] According to one embodiment of the present invention, the processor (150) can learn the blood flow increase pattern for each of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis using learning data (S520).

[0070] To learn the pattern of increased blood flow, the processor (150) can first set a region of interest (ROI) using learning data for patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis. In this case, since each ultrasound image (410-430) may be composed of multiple image frames, the processor (150) can set a region of interest for each image frame.

[0071] According to one embodiment of the present invention, the processor (150) can identify the proportion of color pixels among the total gray scale pixels of the region of interest.

[0072] According to one embodiment of the present invention, the processor (150) can learn to differentiate Graves' disease, autoimmune thyroiditis, and subacute thyroiditis through differences in blood flow increase patterns between Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, thereby generating a machine learning model (30) (S530).

[0073] Artificial neural networks used in machine learning typically analyze images by matching patterns with each disease. Using machine learning, the limitations of real-time Doppler imaging, which can lead to mutations, can be overcome by analyzing the patterns themselves.

[0074] According to one embodiment of the present invention, machine learning can be used to minimize inter-observer variation and to differentiate thyrotoxicosis from a pattern of increased blood flow that is difficult for humans to distinguish.

[0075] According to one embodiment of the present invention, it is expected that it will be helpful to observers (medical staff) with little experience in thyroid diseases.

Claims

1. A method for differentiating hyperthyroidism, including Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, based on machine learning. A step of acquiring basic ultrasound images, color Doppler images, and power Doppler images of the thyroid gland of the subject of diagnosis; A method comprising the step of determining whether the diagnosis subject corresponds to any one of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis by using a machine learning model trained to differentiate thyrotoxicosis based on basic ultrasound images, color Doppler images, and power Doppler images of the thyroid gland of patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis.

2. In paragraph 1, A step of acquiring training data including baseline ultrasound images, color Doppler images and output Doppler images of the thyroid gland of patients with Graves' disease, autoimmune thyroiditis and subacute thyroiditis; A step of learning an increased blood flow pattern for each of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis using the above learning data; A method further comprising the step of generating the machine learning model by learning to differentiate between Graves' disease, autoimmune thyroiditis and subacute thyroiditis through differences in the pattern of increased blood flow between Graves' disease, autoimmune thyroiditis and subacute thyroiditis.

3. In paragraph 2, The steps to learn the pattern of increased blood flow are: A step of setting a region of interest for each image frame using learning data for patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis; A method comprising the step of identifying the proportion of color pixels among the total gray scale pixels of the region of interest.

4. In an electronic device for differentiating hyperthyroidism diseases including Graves' disease, autoimmune thyroiditis, and subacute thyroiditis based on machine learning, Acquire basic ultrasound images (gray scale), color Doppler images, and power Doppler images of the thyroid gland of the subject of diagnosis. An electronic device comprising a processor for determining whether the subject of the diagnosis corresponds to any one of Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, using a machine learning model trained to differentiate thyrotoxicosis based on basic ultrasound images, color Doppler images, and output Doppler images of the thyroid gland of patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis.

5. In paragraph 4, The above processor, Acquire training data including baseline ultrasound images, color Doppler images, and power Doppler images of the thyroid gland in patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis. Using the above learning data, we learn the pattern of increased blood flow for Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, respectively. An electronic device that generates the machine learning model by learning to differentiate between Graves' disease, autoimmune thyroiditis and subacute thyroiditis through differences in patterns of increased blood flow between the three conditions.

6. In paragraph 4, The above processor, Using learning data from patients with Graves' disease, autoimmune thyroiditis, and subacute thyroiditis, we set up regions of interest for each image frame. An electronic device that identifies the proportion of color pixels among the total gray scale pixels in the region of interest.

Images