Stool image analysis system and method

Abstract

[Technical Field] The present disclosure relates to a system and a method for analyzing a stool image, and more particularly, to a system and a method for analyzing a stool image, which derive a large bowel state of a user by analyzing a stool image of the user using a pre-trained deep learning model.SOLUTION: A system for analyzing a stool image according to an embodiment of the present disclosure may include an input unit for receiving a stool image of a user, a data set generation unit for forming a data set by clustering the stool images in chronological order, an analysis unit for analyzing the data set to derive stool status information or colon status information of the user, and an output unit for outputting the stool status information or the colon status information. According to an embodiment of the present invention, it is possible to predict the endoscopic activity of ulcerative colitis from a camera image of feces without performing colonoscopy.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a stool image analysis system and method, and more particularly to a stool image analysis system and method that derives a user's colon status by analyzing the user's stool image using a pre-trained deep learning model. [Background technology]

[0002] Ulcerative colitis is a chronic inflammatory bowel disease of unknown cause characterized by inflammation limited to the mucosa or submucosa of the large intestine. It invades the large intestine continuously from the rectum, and the pathological changes are not scattered here and there but are all connected.

[0003] Ulcerative colitis is a chronic disease that causes inflammation in the large intestine, and requires continuous drug treatment. If the inflammation cannot be controlled through drug treatment, problems such as hospitalization, bowel resection surgery, and an increased risk of developing colorectal cancer can occur.

[0004] To prevent the onset of diseases such as colon cancer, it is desirable to regularly monitor the health of the colon, especially for patients with ulcerative colitis, who are at high risk of developing colon cancer.

[0005] Meanwhile, conventional methods for monitoring a patient's colon status include measuring the inflammatory activity of the mucosa through a colonoscopy. However, colonoscopy requires hypnotic anesthesia in some cases, is relatively time-consuming and expensive, can cause serious side effects such as perforation and infection, and causes significant discomfort to the patient, making it difficult to perform as frequently as necessary.

[0006] Therefore, there is a need for the development of a colon condition monitoring technology that is inexpensive, time-consuming, and can reduce the discomfort felt by the subject. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Korean Patent No. 10-2268304 Summary of the Invention [Problem to be solved by the invention]

[0008] In order to solve the above-mentioned problems, the present invention aims to provide a stool image analysis system and method that can analyze a user's colon condition using the user's stool image and a deep learning model. [Means for solving the problem]

[0009] In one embodiment of the present invention, a stool image analysis system is provided.

[0010] A stool image analysis system according to one embodiment of the present invention may include an input unit that receives a user's stool image, a dataset generation unit that chronologically clusters the stool images to create a dataset, an analysis unit that analyzes the dataset to derive stool status information or large intestine status information of the user, and an output unit that outputs the stool status information or large intestine status information.

[0011] In a stool image analysis system according to an embodiment of the present invention, the dataset may include stool images from the user for seven days immediately following a colonoscopy.

[0012] The stool image analysis system according to an embodiment of the present invention may further include a pre-processing unit that crops the stool image based on a specific point of the toilet included in the stool image.

[0013] In a stool image analysis system according to one embodiment of the present invention, the pre-processing unit extracts a specific point of the toilet bowl from the stool image and crops the stool image so that any edge of the stool image includes the specific point of the toilet bowl.

[0014] A stool image analysis system according to one embodiment of the present invention further includes a communication unit for transmitting and receiving stool images to and from a user terminal including a display unit, and the display unit may include a guideline for matching with a specific point inside a toilet bowl.

[0015] In one embodiment of the stool image analysis system, the analysis unit may include a deep learning model that has already been trained using the stool image or the dataset as input data and stool condition information or large intestine condition information as output data.

[0016] In one embodiment of the present invention, the stool image analysis system is characterized in that when the large intestine condition information is a discrete value indicating the degree of the user's large intestine condition, the already trained deep learning model is trained using a loss function corresponding to a classification problem.

[0017] In one embodiment of the present invention, the stool image analysis system is characterized in that when the large intestine condition information is a continuous value indicating a specific numerical value, the pre-trained deep learning model is trained using a loss function corresponding to a linear regression problem.

[0018] In one embodiment of the present invention, a method for analyzing stool images is provided.

[0019] A stool image analysis method according to one embodiment of the present invention may include an input step in which an input unit inputs a user's stool image; a dataset generation step in which a dataset construction unit chronologically clusters the stool images to construct a dataset; an analysis step in which an analysis unit analyzes the dataset to derive stool status information or large intestine status information of the user; and an output step in which an output unit outputs the stool status information or large intestine status information.

[0020] In a stool image analysis method according to an embodiment of the present invention, the dataset may include stool images from the user for seven days immediately following a colonoscopy.

[0021] The stool image analysis method according to an embodiment may further include a preprocessing step in which a preprocessing unit crops the stool image based on a specific point of the toilet included in the stool image.

[0022] In a stool image analysis method according to one embodiment of the present invention, the preprocessing step may further include a reference extraction step in which the preprocessing unit extracts a specific point of the toilet from the stool image, and a cropping step in which the preprocessing unit crops the stool image so that any edge of the stool image includes the specific point of the toilet.

[0023] The stool image analysis method according to one embodiment of the present invention may further include an image transmitting / receiving step in which a communication unit transmits and receives a stool image to and from a user terminal including a display unit, and a guideline display step in which the display unit displays a guideline for matching with a specific point inside a toilet bowl.

[0024] The stool image analysis method according to one embodiment of the present invention may further include a learning step of training a deep learning model using the stool image or the dataset as input data and the stool condition information or the large intestine condition information as output data.

[0025] In a stool image analysis method according to one embodiment of the present invention, the learning step is characterized in that, when the large intestine condition information is a discrete value indicating the degree of the user's large intestine condition, the deep learning model is trained using a loss function corresponding to a classification problem.

[0026] In a stool image analysis method according to one embodiment of the present invention, the learning step is characterized in that, when the large intestine condition information is a continuous value indicating a specific numerical value, the deep learning model is trained using a loss function corresponding to a linear regression problem.

[0027] As one embodiment of the present invention, a computer-readable recording medium having a program for implementing the above-described method recorded thereon is provided. [Effects of the Invention]

[0028] According to one embodiment of the present invention, the endoscopic activity of ulcerative colitis can be predicted from camera images of stool without performing a colonoscopy.

[0029] Furthermore, according to one embodiment of the present invention, the state of the user's large intestine can be monitored by simply taking images of stool at home without going to a hospital.

[0030] Furthermore, according to one embodiment of the present invention, there is an advantage that even users who live in places where movement is inconvenient or where hospitals are not easily accessible can monitor the condition of their colon.

[0031] The effects obtained by the present disclosure are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those having ordinary skill in the art to which the present disclosure pertains from the following description. [Brief explanation of the drawings]

[0032] [Figure 1]1 is a block diagram of a stool image analysis system according to one embodiment of the present invention. [Figure 2] 1 is a diagram illustrating an example of a pre-processing process performed by a pre-processing unit according to an embodiment of the present invention. [Figure 3] 10 illustrates a display of a user terminal including a guideline displayed through a communication unit according to an embodiment of the present invention. [Figure 4] 1 is an example of training data for a deep learning model according to an embodiment of the present invention. [Figure 5] This shows the structure of a deep learning model that trains using a loss function that corresponds to a classification problem. [Figure 6] This shows the structure of a deep learning model that uses a loss function corresponding to a linear regression problem (mean square error). [Figure 7] FIG. 10 is a diagram illustrating an example of a process for evaluating a user's colon condition using a trained deep learning model with a loss function corresponding to a classification problem. [Figure 8] FIG. 10 is a diagram illustrating an example of a process for evaluating a user's colon condition using deep learning that has been trained using a loss function corresponding to a linear regression problem. [Figure 9] 1 is a flowchart of a stool image analysis method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0033] Hereinafter, with reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily carry out the present invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. In order to clearly explain the present invention in the drawings, parts that are not relevant to the description will be omitted, and similar parts will be designated by similar reference numerals throughout the specification.

[0034] The terms used in this specification will be briefly explained to more specifically describe the present invention.

[0035] The terms used in this invention are generally selected as widely used as possible while taking into consideration the functions in this invention, but these may change depending on the intentions of engineers in this field, precedents, the emergence of new technologies, etc. In addition, in certain cases, the applicant may arbitrarily select terms, and in such cases, the meanings thereof will be described in detail in the relevant description of the invention. Therefore, the terms used in this invention should be defined based on the meanings of the terms and the overall content of this invention, rather than simply the names of the terms.

[0036] Throughout the specification, when a part is said to "include" a certain component, this does not mean that it excludes other components, but that it may further include other components, unless otherwise specified. Furthermore, terms such as "module" and "unit" used in this specification refer to a unit that processes at least one function or operation, and this can be embodied in hardware or software, or a combination of hardware and software. Furthermore, throughout the specification, when a part is said to be "connected" to another part, this includes not only being "directly connected" but also being connected "with other elements interposed therebetween."

[0037] The present invention will now be described in detail with reference to the accompanying drawings.

[0038] FIG. 1 is a block diagram of a stool image analysis system according to one embodiment of the present invention.

[0039] Referring to FIG. 1, a stool image analysis system according to one embodiment of the present invention may include an input unit 100, a data set generation unit 200, an analysis unit 300, and an output unit 400.

[0040] The input unit 100 may input a user's feces image, which is a toilet image containing the user's feces obtained through a user terminal.

[0041] The data set generating unit 200 can generate a data set by clustering stool images in chronological order.

[0042] According to an embodiment, the dataset may include stool images collected from the user for seven days immediately after the user's colonoscopy, which may include information related to changes in stool over the seven days or information related to the number of endoscopies.

[0043] The analysis unit 300 may analyze the data set to derive stool status information or large intestine status information of the user.

[0044] Here, the stool status information may include color information or texture information of the stool. For example, the color information of the stool may include yellow, brown, red, or dark red, and the texture information may include, but is not limited to, diarrhea, regular, very hard, hard, slightly hard, normal, slightly watery, watery, or very watery.

[0045] The colon status information may also indicate a colon-related disease and the degree of progression of the disease, including, but not limited to, information such as early stage colon cancer, mid-stage colon cancer, late stage colon cancer, mild ulcerative colitis, moderate ulcerative colitis, severe ulcerative colitis, or critical condition of ulcerative colitis.

[0046] The output unit 400 may output stool status information or large intestine status information. According to an embodiment, the output unit 400 may output the stool status information or large intestine status information to a display unit included in a user terminal.

[0047] In other words, the stool image analysis system according to one embodiment of the present invention can be expanded or applied to all medical procedures related to colonoscopy (e.g., disease diagnosis and analysis, treatment, examination, bowel cleansing for colonoscopy, etc.).

[0048] Meanwhile, the stool image analysis system according to an embodiment of the present invention may further include a pre-processing unit 500. Hereinafter, the pre-processing unit 500 according to an embodiment of the present invention will be described with reference to FIGS.

[0049] FIG. 2 shows an example of a pre-processing process performed by the pre-processing unit 500 according to an embodiment of the present invention.

[0050] Referring to FIG. 2, the pre-processing unit 500 according to an embodiment of the present invention may crop a stool image based on a specific point on the toilet included in the stool image.

[0051] Here, cropping is a photo editing technique that refers to trimming a photo to fit a desired size.

[0052] According to an embodiment, the pre-processing unit 500 extracts a specific point of the toilet bowl from the stool image and crops the stool image so that any edge of the stool image includes the specific point of the toilet bowl.

[0053] Here, the specific point on the toilet bowl may refer to the inner oval of the toilet seat or the edge of the water present inside the toilet bowl, among the components of the toilet bowl.

[0054] Figure 2(a) is an image of a stool image cropped based on the inner oval shape of the toilet seat, and Figure 2(b) is an image of a stool image cropped based on the edge of the water inside the toilet seat.

[0055] The fecal images taken by each user using their mobile phone camera vary greatly (distance, angle, etc.) depending on the user, the location, and the time of the photograph. Therefore, if the colon condition is analyzed using data that has not undergone preprocessing, there is a high probability of errors in the analysis results.

[0056] However, according to the pre-processing unit 500 of the embodiment of the present invention, it is possible to reduce errors that may occur in the analysis results by reducing differences between data that may occur due to the high degree of freedom in photography during the data acquisition process.

[0057] FIG. 3 shows a display of a user terminal including a guideline displayed through a communication unit 600 according to an embodiment of the present invention.

[0058] The stool image analysis system according to an embodiment of the present invention may further include a communication unit 600. The communication unit 600 is configured to communicate with an external device such as a user terminal.

[0059] According to an embodiment, the communication unit 600 may be connected to an external electronic device based on a network implemented through wired communication and / or wireless communication.

[0060] Here, the wired communication may include at least one of communication methods such as Ethernet, optical network, USB (Universal Serial Bus), and Thunder Bolt.

[0061] In addition, the wireless communication may include at least one of communication methods such as LTE (long-term evolution), LTE-A (LTE Advance), 5G (5th Generation) mobile communication, CDMA (code division multiple access), WCDMA (wideband CDMA), UMTS (universal mobile telecommunications system), WiBro (Wireless Broadband), GSM (Global System for Mobile Communications), DMA (Time Division Multiple Access), WiFi (Wi-Fi), WiFi Direct, Bluetooth, NFC (near field communication), and Zigbee.

[0062] Referring to FIG. 3, a communication unit 600 according to an embodiment of the present invention can transmit and receive feces images to and from a user terminal 10 including a display unit 11, and the display unit 11 may include a guideline X for matching with a specific point inside a toilet bowl.

[0063] FIG. 4 is an example of training data for a deep learning model according to one embodiment of the present invention.

[0064] Figure 4(a) shows the input data for training the deep learning model, and Figure 4(b) shows the user's endoscopic video images corresponding to each stool image in Figure 4(a).

[0065] According to an embodiment, the analysis unit 300 may include a deep learning model that has already been trained using the stool image or the dataset as input data and stool condition information or large intestine condition information as output data.

[0066] According to an embodiment, the deep learning model may be trained to output specific output data that matches specific input data when specific input data is input. That is, the deep learning model may be trained after matching stool images and associated stool or large intestine state information with input data and output data, respectively.

[0067] According to an embodiment, the stool image refers to an image obtained through a video capturing device such as a camera included in a user terminal 10, and may refer to an image of stool present on a toilet bowl, for example, as shown in Figure 4(a).

[0068] 4(a) and 4(b), it can be seen that the state of a user's stool is related to the state of the user's large intestine. Therefore, by analyzing the state of the stool, information on the health state of the large intestine can be obtained.

[0069] According to an embodiment, the stool condition information may include discrete values ​​such as Myo score, stool frequency, bloody stool, Mayo endoscopic subscore, PGA, UCEIS, Vascular score, Bleeding score, and Erosion / Ulcer score, and continuous values ​​such as WBC, Hb, Platelet, ESR, serum albumin, and CRP, and may be used as output data for training the deep learning model.

[0070] For example, the output data can be configured as shown in Table 1 or Table 2 below. [Table 1] [Table 2]

[0071] Figure 5 shows the structure of a deep learning model that uses a loss function corresponding to a classification problem for training, and Figure 6 shows the structure of a deep learning model that uses a loss function corresponding to a linear regression problem (mean square error) for training.

[0072] Referring to Figure 5, the deep learning model according to an embodiment of the present invention is characterized in that when the colon condition information is a discontinuous value indicating the degree of the user's colon condition, the already trained deep learning model performs training using a loss function corresponding to a classification problem (cross-entropy).

[0073] Here, the discrete values ​​may include, but are not limited to, information indicating the degree of a condition related to the large intestine (for example, the degree of ulcerative colitis).

[0074] Referring to FIG. 6, the deep learning model according to an embodiment of the present invention is characterized in that when the colon condition information is a continuous value indicating a specific numerical value, the already trained deep learning model performs training using a loss function corresponding to a linear regression problem.

[0075] Here, the continuous value may include, but is not limited to, information indicating a precise numerical value related to the health of the user's colon.

[0076] FIG. 7 is a diagram illustrating an example of a process for evaluating a user's colon condition using a trained deep learning model with a loss function corresponding to a classification problem.

[0077] When data consisting of stool condition images over a certain period of time is input as one data set into a deep learning model that has been trained using a loss function corresponding to a classification problem according to one embodiment of the present invention, output data corresponding to each stool condition image (corresponding to Probabilities 1 to 5 in FIG. 7) is obtained. In this case, the output data corresponds to discrete values, and the presence or absence of a specific disease can be evaluated based on the value of the output data.

[0078] For example, if the value of each output data satisfies a specific criterion (in FIG. 7, if the Probability exceeds 0.5, it is evaluated as abnormal (displayed in red)), it can be evaluated as abnormal.

[0079] In addition, according to the embodiment, it is also possible to evaluate whether or not a specific disease is present based on whether or not the average value of multiple output data satisfies a specific criterion, or based on the number of output data among the multiple output data that satisfy a specific criterion.

[0080] For example, it is possible to derive an evaluation of whether a particular disease is normal or abnormal based on the number of probabilities of 0.5 or more among Probabilities 1 to 5 in FIG.

[0081] FIG. 8 is a diagram illustrating an example of a process for evaluating the state of a user's large intestine using deep learning that has been trained using a loss function corresponding to a linear regression problem.

[0082] When data consisting of stool condition images over a certain period of time is input into a trained deep learning model using a loss function corresponding to a linear regression problem according to one embodiment of the present invention, output data corresponding to each stool condition image (corresponding to values ​​1 to 5 in Figure 8) is obtained.

[0083] At this time, the output data corresponds to a continuous value, and after undergoing a grading process, the output data is used as a criterion for determining whether or not a particular disease is present. That is, the output data (value) can be classified into 2 to N grades according to the clinician's judgment.

[0084] For example, the output data of deep learning that has been trained using a loss function corresponding to a linear regression problem can be classified into grades 0 to 2, as shown in Figure 8.

[0085] In addition, according to an embodiment, it is also possible to evaluate whether or not a specific disease is present based on whether or not the grades of multiple output data satisfy a specific criterion, or based on the number of grades of multiple output data that satisfy a specific criterion.

[0086] The above-described system-related content can be applied to the method according to an embodiment of the present invention, and therefore, the same content as the above-described system-related content will not be described below.

[0087] FIG. 9 is a flowchart of a stool image analysis method according to one embodiment of the present invention.

[0088] Referring to FIG. 9, a stool image analysis method according to one embodiment of the present invention may include an input step S100, a data set generation step S200, an analysis step S300, and an output step S400.

[0089] In the input step S100, the input unit 100 may input an image of the user's feces.

[0090] In the data set generating step S200, the data set generating unit 200 may cluster the stool images in chronological order to form a data set.

[0091] According to an embodiment, the dataset may include stool images from the user for seven days immediately following a colonoscopy.

[0092] In the analysis step S300, the analysis unit 300 can analyze the data set to derive stool condition information or large intestine condition information of the user.

[0093] In the output step S400, the output unit 400 can output the stool condition information or the large intestine condition information.

[0094] The stool image analysis method according to one embodiment of the present invention may further include a pre-processing step.

[0095] In the preprocessing step, the preprocessing unit 500 may crop the stool image based on a specific point of the toilet included in the stool image.

[0096] According to an embodiment, the pre-processing step may further include a reference extraction step and a cropping step.

[0097] In the reference extraction step, the pre-processing unit 500 can extract a specific point of the toilet bowl from the stool image.

[0098] In the cropping step, the pre-processing unit 500 can crop the stool image so that any edge of the stool image includes a specific point on the toilet bowl.

[0099] The stool image analysis method according to an embodiment of the present invention may further include a transmitting / receiving step and a guideline display step.

[0100] In the transmitting / receiving step, the communication unit 600 can transmit and receive the stool image to and from the user terminal 10 including the display unit 11 .

[0101] In the guideline display step, the display unit 11 can display a guideline X for matching with a specific point inside the toilet bowl.

[0102] The stool image analysis method according to one embodiment of the present invention may further include a learning step.

[0103] The learning step can train a deep learning model using stool images or the dataset as input data and the stool condition information or large intestine condition information as output data.

[0104] According to an embodiment, the learning step is characterized in that when the colon condition information is a discontinuous value indicating the degree of the user's colon condition, the deep learning model is trained using a loss function corresponding to a classification problem.

[0105] According to an embodiment, the learning step is characterized in that, when the colon condition information is a continuous value indicating a specific numerical value, the deep learning model is trained using a loss function corresponding to a linear regression problem.

[0106] Meanwhile, the above-described method can be written as a computer-executable program and can be implemented on a general-purpose digital computer that runs the program using a computer-readable recording medium. Furthermore, the data structure used in the above-described method can be recorded on a computer-readable recording medium by various means. The computer-readable recording medium includes storage media such as magnetic storage media (e.g., ROM, RAM, USB, floppy disk, hard disk, etc.) and optically readable media (e.g., CD-ROM, DVD, etc.).

[0107] The above description of the present invention is for illustrative purposes only, and those skilled in the art will understand that the present invention can be easily modified into other specific forms without changing the technical spirit or essential features of the present invention. Therefore, the above-described embodiments should be understood to be illustrative in all respects and not limiting. For example, each component described as a single component may be implemented in a distributed form, and similarly, components described as distributed may be implemented in a combined form.

[0108] The scope of the present invention is indicated by the claims set forth below rather than by the above detailed description, and all modifications and variations that fall within the meaning and scope of the claims and their equivalents should be construed as being included within the scope of the present invention. [Explanation of symbols]

[0109] X: Guidelines 10: User terminal 11: Display 100: Input section 200: Dataset configuration section 300:Analysis Department 400: Output section 500: Pre-processing section 600: Communications Department

Claims

1. an input unit for inputting a user's feces image; a pre-processing unit for cropping the stool image based on the edge of the water present inside the toilet bowl; a dataset generator that chronologically clusters the cropped stool images to generate a dataset; an analysis unit that uses the dataset as input data and analyzes it through a pre-trained deep learning model to derive stool status information or large intestine status information of the user, wherein the stool status information or large intestine status information includes information indicating the degree of progression of a disease; an output unit that outputs stool status information or large intestine status information of the user; a user terminal including a display unit and a communication unit for transmitting and receiving stool images; the display unit includes a guideline for matching with a specific point inside the toilet bowl; When the stool status information or the large intestine status information is a discrete value including Mayo score or UCEIS, The already trained deep learning model is The stool condition information or large intestine condition information is used as output data, and training is performed using a loss function corresponding to the classification problem, and data of stool condition images over a certain period of time as a single data set is input, and Probability 1 to 5, which are output data corresponding to each stool condition image, are obtained, and the maximum value (max) or minimum value (min) is output through voting, The analysis unit The present invention is characterized in that it evaluates whether or not a specific disease is present based on whether or not an average value of a plurality of output data satisfies a specific criterion, or based on the number of output data among the plurality of output data that satisfy a specific criterion. When the stool status information or the large intestine status information is a continuous value including WBC, Hb, Platelet, ESR, serum albumin, and CRP, The already trained deep learning model is The stool condition information or the large intestine condition information is used as output data, and the information is trained through a loss function corresponding to a linear regression problem. Data is input as a single data set of stool condition images for a certain period of time, and one of the data is output based on voting after going through a ranking process; The analysis unit The present invention is characterized in that it evaluates whether a specific disease is present based on whether a plurality of output data grades, which are divided into 2 to N (N is a natural number), meet a specific standard, or based on the number of grades among the plurality of output data grades that meet a specific standard. Stool image analysis system.

2. The data set is 10. The stool image analysis system of claim 1, comprising stool images for seven days immediately following the user's colonoscopy.

3. an input step in which an input unit inputs a user's stool image; a pre-processing step in which a pre-processing unit crops the stool image based on the edge of water present inside the toilet bowl; a dataset generation step in which a dataset construction unit constructs a dataset by clustering the cropped stool images in chronological order; an analysis step in which an analysis unit uses the dataset as input data and analyzes it through a deep learning model that has already been trained to derive stool status information or large intestine status information of the user, wherein the stool status information or large intestine status information includes information indicating the degree of progression of a disease; an output step in which an output unit outputs the stool condition information or the large intestine condition information of the user; a transmitting / receiving step in which the communication unit transmits and receives the stool image to and from a user terminal including a display unit; a guideline display step in which the display unit displays a guideline for matching with a specific point inside the toilet bowl, When the stool status information or the large intestine status information is a discrete value including Mayo score or UCEIS, The already trained deep learning model is The stool condition information or large intestine condition information is used as output data, and training is performed using a loss function corresponding to the classification problem, and data of stool condition images over a certain period of time as a single data set is input, and Probability 1 to 5, which are output data corresponding to each stool condition image, are obtained, and the maximum value (max) or minimum value (min) is output through voting, The analysis unit The present invention is characterized in that it evaluates whether or not a specific disease is present based on whether or not an average value of a plurality of output data satisfies a specific criterion, or based on the number of output data among the plurality of output data that satisfy a specific criterion. When the stool status information or the large intestine status information is a continuous value including WBC, Hb, Platelet, ESR, serum albumin, and CRP, The already trained deep learning model is The stool condition information or the large intestine condition information is used as output data, and the information is trained through a loss function corresponding to a linear regression problem. Data is input as a single data set of stool condition images for a certain period of time, and one of the data is output based on voting after going through a ranking process; The analysis unit The present invention is characterized in that it evaluates whether a specific disease is present based on whether a plurality of output data grades, which are divided into 2 to N (N is a natural number), meet a specific standard, or based on the number of grades among the plurality of output data grades that meet a specific standard. Stool image analysis method.

4. The data set is 4. The method of claim 3, including stool images for seven days immediately following the user's colonoscopy.

5. A computer-readable recording medium having a program recorded thereon for implementing the method according to any one of claims 3 and 4.

Images