JP2018054443A - Information processing device, information processing system and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To maintain the accuracy of determining the existence/nonexistence of the possibility of a specific disease or the quality of a health condition even though an area having a different color exists or noise is carried in a portion of either image of an image of an excretion from a living body or an image of an object area of the living body.SOLUTION: An information processing device includes: an extraction part for extracting a plurality of image areas from an image of an excretion from a living body or an image of an object area of the living body; a generation part for generating color feature amounts representing color information about each of the extracted image areas; a determination part for determining the existence/nonexistence of the possibility of a specific disease or the quality of a health condition by using similarity to a first feature amount representing the feature of color of an excretion or an object area in a group having the specific disease or specific health abnormality and similarity to a second feature amount representing the feature of color of an excretion or an object area in a healthy person or a group having good health to each of the generated color feature amounts; and a majority decision part for deciding a plurality of determination results by the determination part by majority and outputting output information based on a majority decision result.SELECTED DRAWING: Figure 7

Description

  The present invention relates to an information processing apparatus, an information processing system, and a program.

  In Japan, a fecal color card was introduced nationwide in 2012 as a screening for biliary atresia. Patent Document 1 reports the development and operation of a smartphone application (PoopMD) for the purpose of screening biliary atresia in the United States. PoopMD is an application that captures fecal color and automatically determines the possibility of biliary atresia based on the color. As a basis for the discrimination system, a threshold is set for issuing an alert indicating the possibility of illness based on the RGB information of the stool color card used nationwide in Taiwan. This is a mobile version of the stool color card that is automatically discriminated rather than the subjective stool color judgment of the carer.

Franciscovich, Amy, et al. "PoopMD, a Mobile Health Application, Accurately Identifies Infant Acholic Stools." PloS one 10.7 (2015): e0132270

  Depending on how the light hits, bright and dark places appear in the stool image. In Patent Document 1, since the user designates an image area, a part of the image area is very bright or dark depending on the designated place of the image area. There is a possibility that. As described above, there is a possibility that an erroneous determination is made if there is a region with a different color in a part of the image or if noise appears.

  This problem is not limited to the determination of fecal images, but also images of other discharges (eg, urine, tears, etc.) from living organisms (eg, humans, dogs, cats, etc.) , Facial skin, intestine, etc.) can also occur in image determination. In addition, this problem is not only a problem of biliary atresia but also a problem that can occur when determining the possibility of another specific disease or when determining whether the health condition is good or bad.

  The present invention has been made in view of the above problems, and there is a region having a different color or noise in any part of either the image of the discharge from the living body or the image of the target region of the living body. However, it is an object of the present invention to provide an information processing apparatus, an information processing system, and a program capable of maintaining the accuracy of determining whether or not there is a possibility of a specific disease or whether the health condition is good or bad.

  The information processing apparatus according to the first aspect of the present invention includes an extraction unit that extracts a plurality of image regions from an image of discharge from a living body or an image of a target region of a living body, and a color for each of the extracted image regions. A generating unit that generates a color feature amount representing information, and a first feature that represents a color feature of an emission or target region in a group having a specific disease or a specific health abnormality with respect to each of the generated color feature amount The presence of a specific disease or the state of health using the similarity to the quantity and the similarity to the second characteristic quantity representing the color characteristics of the discharge or the target area in a healthy or healthy group A determination unit that determines whether the quality is good or bad, and a majority decision unit that outputs a large number of determination results by the decision unit and outputs output information based on the majority decision result.

  According to this configuration, a plurality of image regions extracted from an image are determined, and a plurality of determination results obtained by the determination are determined in large numbers. For this reason, even if there is a region with a different color in part of the image, or a wrong determination is made on a part of the image region due to noise, it is correct if a correct determination is made in other majority image regions. A determination result can be obtained. Thereby, even if there is a region having a different color or noise appears in a part of the stool image, it is possible to maintain the accuracy of determining whether there is a possibility of a specific disease or whether the health condition is good or bad.

  An information processing apparatus according to a second aspect of the present invention is the information processing apparatus according to the first aspect, wherein the determination unit includes the first feature amount and the color feature amount for each of the color feature amounts. , And the similarity between the second feature quantity and the color feature quantity is used to determine whether or not the determination is possible, and if the determination is not possible, the determination result is determined to be impossible. Each of the determination results can take three options of whether or not there is a possibility of a specific disease, indeterminate, or three choices of good or bad health and undeterminable, the majority voting unit, as a result of the majority decision, When the determination becomes impossible, output information for prompting re-determination after a predetermined time has elapsed is output.

  According to this configuration, when it is impossible to determine, the next day is urged to perform an inspection again. It can be determined that there is a possibility or poor health. As a result, it is possible to encourage the observation and judgment over time, and it is possible to promote early detection of a specific disease or deterioration of the health condition.

  An information processing apparatus according to a third aspect of the present invention is the information processing apparatus according to the first or second aspect, wherein the discharge from the living body is feces, and the specific disease is biliary closure The color information includes at least one of hue and saturation, and the determination unit determines whether or not there is a possibility of biliary atresia.

  According to this configuration, as a result of logistic regression analysis with the presence or absence of biliary atresia as an outcome, determination is made using at least one of the most statistically significant hue and saturation information. It is possible to accurately determine whether or not there is a symptom.

  An information processing system according to a fourth aspect of the present invention includes an extraction unit that extracts a plurality of image areas from an image of discharge from a living body or an image of a target area of a living body, and a color for each of the extracted image areas. A generating unit that generates a color feature amount representing information, and a first feature that represents a color feature of an emission or target region in a group having a specific disease or a specific health abnormality with respect to each of the generated color feature amount The presence of a specific disease or the state of health using the similarity to the quantity and the similarity to the second characteristic quantity representing the color characteristics of the discharge or the target area in a healthy or healthy group A determination unit that determines whether the quality is good or bad, and a majority decision unit that outputs a large number of determination results by the decision unit and outputs output information based on the majority decision result.

  According to this configuration, a plurality of image regions extracted from an image are determined, and a plurality of determination results obtained by the determination are determined in large numbers. For this reason, even if there is a region with a different color in part of the image, or a wrong determination is made on a part of the image region due to noise, it is correct if a correct determination is made in other majority image regions. A determination result can be obtained. Thereby, even if there is a region having a different color or noise appears in a part of the stool image, it is possible to maintain the accuracy of determining whether there is a possibility of a specific disease or whether the health condition is good or bad.

  According to a fifth aspect of the present invention, there is provided a program that extracts a plurality of image areas from an image of a discharge from a living body or an image of a target area of the living body, and each of the extracted image areas. A generating unit that generates a color feature amount representing color information; and a first feature that represents a color feature of an emission or target region in a group having a specific disease or a specific health abnormality for each of the generated color feature amounts Using the similarity to the feature amount, and the similarity to the second feature amount representing the color feature of the discharge or target area in a healthy person or a group with good health, the presence or absence of a specific disease or the health This is a program for causing a determination unit that determines whether the state is good or bad and a majority decision unit that outputs a plurality of determination results by the decision unit to output information based on the majority decision result.

  According to this configuration, a plurality of image regions extracted from an image are determined, and a plurality of determination results obtained by the determination are determined in large numbers. For this reason, even if there is a region with a different color in part of the image, or a wrong determination is made on a part of the image region due to noise, it is correct if a correct determination is made in other majority image regions. A determination result can be obtained. Thereby, even if there is a region having a different color or noise appears in a part of the stool image, it is possible to maintain the accuracy of determining whether there is a possibility of a specific disease or whether the health condition is good or bad.

  According to the present invention, a plurality of image regions extracted from an image are determined, and a plurality of determination results obtained by the determination are determined. For this reason, even if there is a region with a different color in part of the image, or a wrong determination is made on a part of the image region due to noise, it is correct if a correct determination is made in other majority image regions. A determination result can be obtained. Thereby, even if there is a region having a different color or noise appears in a part of the stool image, it is possible to maintain the accuracy of determining whether there is a possibility of a specific disease or whether the health condition is good or bad.

It is a table | surface which shows the result of logistic regression analysis. 1 is a schematic configuration diagram of an information processing system according to a first embodiment. It is a schematic block diagram of a terminal device. This is a method for specifying a flight image area in a target application. This is another method for specifying the image area of the flight in the target application. 1 is a schematic configuration diagram of an information processing apparatus according to a first embodiment. It is a functional block diagram of the information processor concerning a 1st embodiment. It is a schematic diagram which shows the acquisition order of a pixel value. It is a figure which shows the outline of a color vector. It is a table | surface showing an eigenvector. It is a functional block diagram of the information processing apparatus which concerns on 2nd Embodiment. It is a functional block diagram of the information processing apparatus which concerns on 3rd Embodiment. It is a table showing the eigenvector which concerns on 3rd Embodiment.

  Furthermore, in Patent Document 1, automatic discrimination makes it possible to reduce the subjectivity bias that can occur when an observer compares a paper-based stool color sample with an infant's stool color, but a stool color card is included. Not all other problems have been solved. A problem other than the subjective bias is the false negative of color card # 4. This means that many infants who are finally diagnosed with biliary atresia have the number 4 color defined on the color card as "Need to watch (not an obvious positive color)" This is because there is a reality. The fecal color corresponding to the color of interest is also observed in the early stages of onset of biliary atresia, as in healthy children. It is necessary to observe and judge over time. In light of the fact that surgical intervention within 60 days from the onset of biliary atresia improves prognosis, in order to encourage early detection of biliary atresia, There is a need for a continuous observation of the color, and a system that can detect the color when the color changes after observation over time.

  In addition, even if the disease is not biliary atresia, if it is judged as “Necessary observation (not an obvious positive color)”, it will be observed over time and when the color changes. A system that can detect is needed.

  In each embodiment, as an example, a method for determining whether or not there is a possibility of biliary atresia will be described with stool as a discharge from a living body and biliary atresia as a specific disease. Each embodiment will be described below with reference to the drawings.

  In the first embodiment, 50 samples of color information of fecal color photographs of existing biliary atresia and healthy children are extracted, RGB and HSV values are measured from each sample, and the presence or absence of biliary atresia is used as an outcome. Analysis was carried out. FIG. 1 is a table showing the results of logistic regression analysis. In Figure 1, the crude odds ratio (Crude OR) and 95% confidence interval (Confidence) for each of Red, Green, Blue, Hue, Saturation, and Value. Interval: CI) and p value (p-value).

  As a result of this logistic regression analysis, as shown in FIG. 1, since the p values of Hue and Saturation are the smallest, Hue and Hue (saturation) are factors that predict the presence or absence of biliary atresia. The result was that the value of (saturation) was statistically most significant. For this reason, in the first embodiment, the color information uses HSV values including information on hue (hue) and saturation (saturation). Thereby, the presence or absence of the possibility of biliary atresia can be determined with high accuracy.

FIG. 2 is a schematic configuration diagram of the information processing system according to the first embodiment. The information processing system 10 includes terminal devices 1-1,..., 1-N (N is a natural number) and the information processing device 2.
The terminal devices 1-1,..., 1 -N are capable of wireless communication and can communicate with the information processing device 2 via the communication network 3. The terminal devices 1-1,..., 1-N are multifunctional mobile phones (so-called smartphones), tablet terminals, notebook computers, or the like. The terminal devices 1-1, ..., 1-N according to the first embodiment will be described below as an example of a smartphone. The information processing apparatus 2 will be described below as an example of a server. Hereinafter, the terminal devices 1-1, ..., 1-N are collectively referred to as the terminal device 1.

  FIG. 3 is a schematic configuration diagram of the terminal device. As illustrated in FIG. 3, the terminal device 1 includes a RAM (Random Access Memory) 11, a storage unit 12, a touch panel 13, an imaging unit 14, a display unit 15, a communication unit 16, and a CPU (Central Processing Unit) 17. A RAM (Random Access Memory) 11 temporarily stores information. The storage unit 12 stores various programs. The touch panel 13 receives a user operation on the touch panel. The imaging unit 14 images a stool according to a user operation using the touch panel 13. The display unit 15 displays information according to the control of the CPU 17. The communication unit 16 is a wireless communication circuit and communicates with the information processing apparatus 2 according to the control of the CPU 17.

  The CPU 17 controls the communication unit 16 to download a target application for determining whether or not there is a possibility of biliary atresia from the application distribution server in accordance with a user operation, and downloads and stores the target application in the storage unit 12. CPU17 starts a target application according to a user's operation. Then, the CPU 17 controls the imaging unit 14 to capture the stool according to the user operation in the target application.

  FIG. 4 shows a method for designating a flight image area in the target application. As shown in FIG. 4, the image area B1 of the stool image is specified as shown on the right side of FIG. 4 by photographing within the range of the red frame F1 on the left side of FIG. Then, the communication unit 16 transmits the data in the image area B1 to the information processing device 2. In the information processing apparatus 2, as an example, the image area B1 is divided into nine as indicated by a broken line, and determination is performed for each of the nine divided image areas.

  FIG. 5 shows another method for specifying the image area of the flight in the target application. As shown in FIG. 5, when the stool image is enlarged by swiping or the like, the entire enlarged image B2 is used for the determination. Then, the communication unit 16 transmits the data of the enlarged image B2 to the information processing device 2. In the information processing apparatus 2, as an example, the enlarged image B2 is divided into nine as indicated by broken lines, and determination is performed for each of the nine divided image areas.

  FIG. 6 is a schematic configuration diagram of the information processing apparatus according to the first embodiment. As illustrated in FIG. 6, the information processing apparatus 2 stores a RAM 21, a storage unit 22, a communication unit 23, and a CPU 24. The RAM 21 temporarily stores information. The storage unit 22 stores various programs. The communication unit 23 is a wireless communication circuit and communicates with the terminal device 1 according to the control of the CPU 24. The CPU 24 executes each function of the functional block diagram shown in FIG. 7 by reading the program from the storage unit 22 and executing it.

  FIG. 7 is a functional block diagram of the information processing apparatus according to the first embodiment. As illustrated in FIG. 7, the CPU 24 functions as a first feature amount determination unit 241, a second feature amount determination unit 242, an extraction unit 243, a generation unit 244, a determination unit 245, and a majority decision unit 246.

  The first feature amount determination unit 241 determines a first feature amount that represents a color feature of the discharge or target region in a group having a specific disease or a specific health abnormality. In the first embodiment, as an example, the first feature amount determination unit 241 determines a first feature amount that represents a color characteristic of stool in a patient with biliary atresia.

  The second feature amount determination unit 242 determines a second feature amount that represents the color feature of the discharged matter or the target region in a healthy person or a group with good health. In the first embodiment, as an example, the second feature amount determination unit 242 determines a second feature amount that represents the fecal color feature of a healthy child.

  Specific processing will be described below. In the first embodiment, a graphic method (CLAFIC method) is used as an example of machine learning. In addition, stool color photographs of 30 cases of biliary atresia and 34 cases of healthy children are used. Obtain a numerical vector based on HSV information as an observation value from one stool color photograph for each case in advance, and then optimize the threshold that indicates how many eigenvectors of the autocorrelation matrix are to be used based on divided cross-validation. Do.

  FIG. 8 is a schematic diagram illustrating the order of obtaining pixel values. FIG. 9 is a diagram showing an outline of a color vector. FIG. 10 is a table showing eigenvectors. Subsequently, for each stool color photograph of 30 cases of biliary atresia and 34 cases of healthy children, an image region R showing stool is extracted as shown in FIG. Then, with respect to the extracted image, pixel values are extracted by HSV by skipping a predetermined number of pixels in the direction indicated by the arrow A1 in FIG. Then, as shown in FIG. 9, for each case, a color vector is created in which a collection of H values, a collection of S values, and a collection of V values are arranged in order. For example, if the extracted image is a total of 150084 pixel numerical data of 396 × 379, if every 132 pixel values are extracted, the data related to the H value, S value, and V value will be 1137 each. Therefore, the color vector in this case is 1 row 3411 column data.

  The i-th observed value (color vector) of the group of children with biliary atresia is expressed by the following equation.

X1_ {i} = (x1_ {i, 1}, x1_ {i, 2}, ..., x1_ {i, 3411}) ^T (1)

  However, X1_ {i} is normalized to norm 1. Here, if j is an index, x1_ {i, j} represents a pixel value, and T represents transposition. When the number of observed values (number of cases) in the group of children with biliary atresia is N1 (here, 30 as an example), the autocorrelation matrix G1 is calculated first. The autocorrelation matrix G1 is expressed by the following equation (2).

  Here, T represents transposition. The autocorrelation matrix G1 is expanded by the eigenvalue λ1 and the normal eigenvector v1 of the autocorrelation matrix G1 as follows:

G1 = λ1_1 ・ v1_1 ・ v1_1 ^T + λ1_2 ・ v1_2 ・ v1_2 ^T + ・ ・ ・ + λ1_3411 ・ v1_3411 ・ v1_3411 ^T

  Based on the threshold given by the cross-validation method, the eigenvector v1_1 for the maximum eigenvalue λ1_1 to the eigenvector v1_c1 for the c1th largest eigenvalue λ1_c1 are adopted, and the following eigenvector matrix C1 is created (where 1 ≦ c1 ≦ 3411) .

  C1: = (v1_1, v1_2, ..., v1_c1)

  These calculations are performed by the first feature amount determination unit 241. Subsequently, the i-th observed value of the group of healthy children (color vector) is expressed by the following equation.

X0_ {i}: = (x0_ {i, 1}, x0_ {i, 2}, ..., x0_ {i, 3411}) ^T (3)

  However, it is assumed that X0_ {i} is normalized to norm 1. Here, x0_ {i, j} represents a pixel value, and T represents transposition. When the number of observed values (number of cases) in the group of healthy children is N0 (34 as an example here), the autocorrelation matrix G0 is calculated first. The autocorrelation matrix G0 is expressed by the following equation (4).

  Here, T represents transposition. Autocorrelation matrix G0 is expanded as follows by eigenvalue λ0 and normal eigenvector v0 of autocorrelation matrix G0.

G0 = λ0_1 ・ v0_1 ・ v0_1 ^T + λ0_2 ・ v0_2 ・ v0_2 ^T + ・ ・ ・ + λ0_3411 ・ v0_3411 ・ v0_3411 ^T

  Finally, based on the threshold given by the cross-validation method, the eigenvector v0_1 for the maximum eigenvalue λ0_1 to the eigenvector v0_c0 for the c0th largest eigenvalue λ0_c0 are adopted, and the following eigenvector matrix C0 is created (where 1 ≦ c0 ≦ 3411).

  C0: = (v0_1, v0_2, ..., v0_c0)

  These calculations are performed by the second feature amount determination unit 242.

  The first feature quantity determination unit 241 stores the eigenvectors v1_1, v1_2,..., V1_c1 of the autocorrelation matrix G1 in the storage unit 22 as the first feature quantity representing the fecal color feature of the child with biliary atresia. To do. Here, the eigenvector v1_1 is the largest eigenvector, the eigenvector v1_2 is the second largest eigenvector orthogonal to the eigenvector v1_1, and the eigenvector v1_c1 is the c1 largest eigenvector orthogonal to all other eigenvectors.

  The second feature quantity determination unit 242 stores the eigenvectors v0_1, v0_2,..., V0_c0 of the autocorrelation matrix G0 in the storage unit 22 as the second feature quantity representing the fecal color feature of the healthy child. Here, the eigenvector v0_1 is the largest eigenvector, the eigenvector v0_2 is the second largest eigenvector orthogonal to the eigenvector v0_c0, and the eigenvector v0_c0 is the c0th largest eigenvector orthogonal to all other eigenvectors.

  Thus, as shown in FIG. 10, the storage unit 22 stores eigenvectors v1_1, v1_2,..., V1_c1 of the autocorrelation matrix G1 and eigenvectors v0_1, v0_2,. Remembered. In the following description, it is assumed that the thresholds c1 and c0 are 3.

  The extraction unit 243 extracts a plurality of image areas from the image of the discharge from the living body or the image of the target area of the living body. In the first embodiment, the discharge from the living body is feces, and the extraction unit 243 extracts nine image areas from the fecal image (for example, the image area B1 or the enlarged image B2) as an example.

  The generation unit 244 generates a color feature amount representing color information for each extracted image region. At this time, the generation unit 244 extracts a pixel value by HSV by skipping a predetermined number of pixels in the direction indicated by the arrow A1 in FIG. 8 for each of the extracted image regions. Then, as illustrated in FIG. 9, the generation unit 244 generates, as a color feature amount, a color vector x in which a collection of H values, a collection of S values, and a collection of V values are arranged in order.

  The determination unit 245 compares each generated color feature amount with the first feature amount representing the color feature of the discharge or target region in the group having a specific disease or specific health abnormality, and a healthy person Alternatively, the similarity with the second feature value representing the color characteristics of the emission or the target area in the group with good health is calculated, and the presence or absence of the possibility of a specific disease or the quality of the health condition is calculated using these similarities. Determine. In the first embodiment, the specific disease is biliary atresia, and the determination unit 245 determines whether or not there is a possibility of biliary atresia.

Specifically, for example, the determination unit 245 calculates the norm square value of the vector (first norm) when the color vector x is projected onto the projection matrix calculated from the eigenvectors v1_1, v1_2,..., V1_c1 of the autocorrelation matrix G1. (Referred to as the square value). The larger this amount, the smaller the sum of the distances between x and the vector X1_ {i} (referred to as the first sum). Therefore, this sum is an example of the similarity between the color vector x and the first feature amount.
In addition, the determination unit 245 calculates a norm square value of a vector (referred to as a second norm square value) when the color vector x is projected onto a projection matrix calculated from the eigenvectors v0_1, v0_2,. ). The larger this amount, the smaller the sum of distances between x and the vector X0_ {i} (referred to as the second sum). Therefore, this sum is an example of the similarity between the color vector x and the second feature amount.

  Then, the determination unit 245 compares the first norm square value and the second norm square value to determine that they belong to the larger group. That is, the determination unit 245 determines that there is a possibility of biliary atresia if the first norm squared value is larger, whereas if the second norm squared value is larger, the determining unit 245 determines biliary atresia. It is determined that there is no possibility.

  The majority decision unit 246 never outputs a plurality of determination results by the decision unit 245 and outputs output information based on the majority decision result. This output information is transmitted from the communication unit 23 to the terminal device 1. Thereby, the terminal device 1 can display the presence or absence of the possibility of biliary atresia. As a result, all 30 cases of learning data of children with biliary atresia were correctly identified.

  In addition, all 34 examples of learning data of healthy children were correctly determined. That is, the specificity according to the present embodiment is 100%. Here, the specificity is the probability that a negative one is correctly determined as negative. On the other hand, PoopMD reported in Non-Patent Document 1 has determined that 3 cases of learning data of healthy subjects are intermediate in 3 cases, and the specificity is 89 (= 24/27 × 100)%. Therefore, the information processing apparatus 2 according to the present embodiment is superior in that the specificity is higher than the PoopMD reported in Non-Patent Document 1.

  In particular, regarding the color of color card # 4, which is difficult to distinguish with the existing stool color cards used nationwide in Japan, the stool color information of cases of biliary atresia with similar colours, By reading the stool color information of healthy children, it was possible to correctly determine the stool color of difficult tones.

  However, it may be difficult to judge at a certain point in time for a stool color corresponding to color card # 4. On the other hand, in the first embodiment, the determination unit 245, for each color feature amount, similarity between the first feature amount and the color feature amount, and similarity between the second feature amount and the color feature amount. Is used to determine whether or not the determination is possible. If the determination is not possible, it is determined that the determination is impossible. Specifically, for example, when the larger one of the first norm square value and the second norm square value is equal to or less than a predetermined first set value, it is determined as “undecidable”. If it exceeds the first set value and is less than or equal to the predetermined second set value, it is determined that “it will be again tomorrow”, and if it exceeds the second set value, whether or not there is a possibility of biliary atresia It is also possible to determine. Here, the second set value is larger than the first set value.

  When the majority decision unit 246 determines that determination is impossible, the majority decision unit 246 outputs output information for prompting redetermination after a predetermined time has elapsed. For example, the majority decision unit 246 outputs output information for displaying on the terminal device 1 an alert that prompts the user to redetermine the next day's inspection value. This output information is transmitted from the communication unit 23 to the terminal device 1. As a result, the terminal device 1 can display an alert that prompts the user to redetermine the next day's inspection value.

  As a result, if it is difficult to make a determination at one point in time using a stool color corresponding to the color card No. 4, the next day is urged to perform another inspection. For this reason, the stool color changes to be more whitish the next day, and it can be determined that there is a possibility of biliary atresia when the stool color corresponds to the stool color of the color card 3 or less. As a result, after the stool color corresponding to the color card No. 4 can be encouraged to make a continuous observation and judgment as to whether or not the stool color becomes lighter, and promote early detection of biliary atresia be able to.

  As described above, in the information processing apparatus 2 according to the first embodiment, the extraction unit 243 extracts a plurality of image regions from the fecal image. Then, the generation unit 244 generates a color feature amount representing color information for each of the extracted image regions. Then, the determination unit 245 determines, for each of the generated color feature amounts, the similarity to the first feature amount representing the stool color feature in the biliary atresia group and the stool color feature in the healthy child group. The possibility of biliary atresia is determined using the similarity with the second feature value to be represented. The majority decision unit 246 never outputs a plurality of determination results, and outputs output information based on the majority decision results.

  With this configuration, a plurality of image regions extracted from the fecal image are determined, and a plurality of determination results obtained by the determination are determined in large numbers. For this reason, even if there is an area with a different color in part of the stool image or noise is present and some of the image areas are erroneously determined, if the correct determination is made in the other image areas, A correct determination result can be obtained. Thereby, even if a part of the stool image has a region having a different color or noise appears, it is possible to maintain the accuracy of the determination of the presence or absence of biliary atresia.

  In the first embodiment, HSV values including information on hue (hue) and saturation (saturation) are used as color information. However, the present invention is not limited to this. The color information may be any one of Hue and Saturation, may use the HV value, or may use the SV value.

  In the first embodiment, the HSV value including Hue and Saturation information is used as a preferred aspect of the color information based on the result of the logistic regression analysis. May be used as color information.

  In the first embodiment, the majority decision unit 246 decides all of the determination results, but the present invention is not limited to this. There may never be many judgment results other than the judgment impossible.

  In the first embodiment, each of the determination results can be selected from the three options of whether or not biliary atresia may be possible and which cannot be determined. However, the determination result is not limited to this. When judging whether the health condition is good or bad, there are three options: good or bad health condition and inability to judge.

(Second Embodiment)
Next, the second embodiment will be described. FIG. 11 is a functional block diagram of the information processing apparatus according to the second embodiment. As shown in FIG. 11, compared with the configuration of FIG. 7, the second embodiment has a configuration in which a division number determination unit 247 and a second majority decision unit 248 are added.
The division number determination unit 247 determines a plurality of image regions to be divided by the extraction unit 243. For example, the division number determination unit 247 determines a plurality of image regions to be divided using random numbers. Here, a case where the number of image areas to be divided is set to 9, 16, 25 will be described as an example.

  In the case where the extraction unit 243 is divided into nine, the generation unit 244, the determination unit 245, and the majority decision unit 246 sequentially execute processing, and the majority decision unit 246 stores the first majority decision result in the storage unit 22. Similarly, when the extraction unit 243 is divided into 16, the generation unit 244, the determination unit 245, and the majority decision unit 246 sequentially execute processing, and the majority decision unit 246 stores the second majority decision result in the storage unit 22. Similarly, when the extraction unit 243 is divided into 25, the generation unit 244, the determination unit 245, and the majority decision unit 246 sequentially execute processing, and the majority decision unit 246 stores the third majority decision result in the storage unit 22.

  Next, the second majority voting unit 248 reads out the three majority decisions from the first majority decision result to the third majority decision result from the storage unit 22, makes a majority decision, and outputs output information based on the majority decision result. This output information includes information for prompting re-determination after the passage of a predetermined time or whether or not there is a possibility of biliary atresia. Thus, by preparing three image division numbers and further voting the majority results of each, it is possible to prevent determination errors caused by the number of image divisions and improve the robustness of the determination results. .

(Third embodiment)
Subsequently, a third embodiment will be described. FIG. 12 is a functional block diagram of an information processing apparatus according to the third embodiment. As shown in FIG. 12, compared with the configuration of FIG. 7, the third embodiment is increased to three determination units 245-1, 245-2, and 245-3, and majority decision units 246-1, 246-2, The configuration is such that the number is increased to three of 246-3 and a second majority decision unit 249 is added.

  FIG. 13 is a table showing eigenvectors according to the third embodiment. In the first embodiment, one image region R is extracted for one case and the eigenvector is determined, but in the third embodiment, three regions R1, R2, and R3 are extracted for one case, respectively. An eigenvector is determined for each region. As a result, as shown in the table of FIG. 13, the storage unit 22 stores v1_ {1,1}, v1_ {2,1},..., V1_ {c1_1, 1} is stored, and v0_ {1,1}, v0_ {2,1},..., V0_ {c0_1,1} are stored as eigenvectors of the normal child region R1. Further, as shown in the table of FIG. 13, v1_ {1,2}, v1_ {2,2},..., V1_ {c1_2,2 are stored in the storage unit 22 as eigenvectors of the region R2 of the child with biliary atresia. } Is stored, and v0_ {1,2}, v0_ {2,2},..., V0_ {c0_2,2} are stored as eigenvectors of the normal child region R2. Furthermore, as shown in the table of FIG. 13, the storage unit 22 includes v1_ {1,3}, v1_ {2,3},..., V1_ {c1_3,3 as eigenvectors of the region R3 of the child with biliary atresia. } Is stored, and v0_ {1,3}, v0_ {2,3},..., V0_ {c0_3,3} are stored as eigenvectors of the region R3 of the healthy child.

  The determination unit 245-1 refers to the storage unit 22, and the eigenvectors v1_ {1,1}, v1_ {2,1}, ..., v1_ {c1_1,1} of the region R1 of the child with biliary atresia are healthy. Using eigenvectors v0_ {1,1}, v0_ {2,1}, ..., v0_ {c0_1,1} of the child's region R1, determine the possibility of biliary atresia for each color feature . Then, the majority decision unit 246-1 decides a plurality of determination results by the determination unit 245-1.

  Further, the determination unit 245-2 refers to the storage unit 22, and the eigenvectors v1_ {1,2}, v1_ {2,2}, ..., v1_ {c1_2,2} of the region R2 of the child with biliary atresia, Using eigenvectors v0_ {1,2}, v0_ {2,2}, ..., v0_ {c0_2,2} in the region R2 of a healthy child, determine the possibility of biliary atresia for each color feature To do. Then, the majority decision unit 246-2 decides the majority of the plurality of determination results by the determination unit 245-2.

  Further, the determination unit 245-3 refers to the storage unit 22, and the eigenvectors v1_ {1,3}, v1_ {2,3}, ..., v1_ {c1_3,3} of the region R3 of the child with biliary atresia, Using eigenvectors v0_ {1,3}, v0_ {2,3}, ..., v0_ {c0_3,3} in the region R3 of healthy children, determine the possibility of biliary atresia for each color feature To do. The majority decision unit 246-3 decides the majority of the plurality of determination results by the determination unit 245-3.

  The second majority decision unit 249 outputs more output results based on the majority decision result than the majority decision result obtained by the majority decision units 246-1, 246-2, and 246-3. This output information includes information for prompting re-determination after the passage of a predetermined time or whether or not there is a possibility of biliary atresia. This prepares three image areas for generating eigenvectors of children with biliary atresia and eigenvectors of healthy children, and prevents decision errors caused by selection of image areas by further deciding the majority of the results for each. And the robustness of the determination result can be improved.

  In the third embodiment, three image regions for generating eigenvectors of children with biliary atresia and eigenvectors of healthy children are selected. However, the present invention is not limited to this, and the number of image regions may be two or four or more. Good.

  Note that a system including a plurality of devices may process each process of the information processing device according to each embodiment in a distributed manner by the plurality of devices. Moreover, the terminal device 1 may process part or all of the processing of the information processing apparatus according to each embodiment.

Further, a program for executing each process of the information processing apparatus according to each embodiment is recorded on another computer-readable recording medium, and the program recorded on the recording medium is read into a computer system, and the processor By executing, the above-described various processes related to the information processing apparatus may be performed.
Moreover, each part of the information processing apparatus according to each embodiment may be implemented by hardware.

  As described above, the present invention is not limited to the above-described embodiments as they are, and can be embodied by modifying the components without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

Terminal device 1-1, ..., 1-N
11 RAM
DESCRIPTION OF SYMBOLS 12 Memory | storage part 13 Touch panel 14 Imaging part 15 Display part 16 Communication part 17 CPU
2 Information processing device 21 RAM
22 storage unit 23 communication unit 24 CPU
241 First feature value determination unit 242 Second feature value determination unit 243 Extraction unit 244 Generation unit 245, 245-1, 245-2, 245-3 Determination unit 246, 246-1, 246-2, 246-3 Majority decision unit 247 Division number decision unit 248, 249 Second majority decision unit 3 Communication network

Claims (5)

An extraction unit that extracts a plurality of image areas from an image of discharge from a living body or an image of a target area of a living body;
For each of the extracted image regions, a generation unit that generates a color feature amount representing color information;
For each of the generated color features, similarity to the first feature representing the color features of the discharge or target area in the group having a specific disease or specific health abnormality, and healthy or good health A determination unit that determines whether or not there is a possibility of a specific disease or whether a health condition is good or bad by using a similarity with a second feature value representing a color characteristic of an emission or a target area in the group;
A majority decision unit that outputs a plurality of decision results by the decision unit, and output information based on the majority decision result;
An information processing apparatus comprising: For each of the color feature amounts, the determination unit performs determination using the similarity between the first feature amount and the color feature amount and the similarity between the second feature amount and the color feature amount. It is determined whether or not it is possible, and if the determination is not possible, it is determined that the determination result cannot be determined,
For each of the determination results, there may be three options of whether or not there is a possibility of a specific disease, indeterminate, or three options of poor or bad health and undecidable,
The information processing apparatus according to claim 1, wherein the majority decision unit outputs output information for prompting re-determination after a lapse of a predetermined time when the decision becomes impossible as a result of the majority decision. The discharge from the living body is feces,
The specific disease is biliary atresia;
The color information includes at least one of hue and saturation,
The information processing apparatus according to claim 1, wherein the determination unit determines whether or not there is a possibility of biliary atresia. An extraction unit that extracts a plurality of image areas from an image of discharge from a living body or an image of a target area of a living body;
For each of the extracted image regions, a generation unit that generates a color feature amount representing color information;
For each of the generated color features, similarity to the first feature representing the color features of the discharge or target area in the group having a specific disease or specific health abnormality, and healthy or good health A determination unit that determines whether or not there is a possibility of a specific disease or whether a health condition is good or bad by using a similarity with a second feature value representing a color characteristic of an emission or a target area in the group;
A majority decision unit that outputs a plurality of decision results by the decision unit, and output information based on the majority decision result;
An information processing system comprising: Computer
An extraction unit that extracts a plurality of image areas from an image of discharge from a living body or an image of a target area of a living body;
For each of the extracted image regions, a generation unit that generates a color feature amount representing color information;
For each of the generated color features, similarity to the first feature representing the color features of the discharge or target area in the group having a specific disease or specific health abnormality, and healthy or good health A determination unit that determines whether or not there is a possibility of a specific disease or whether a health condition is good or bad by using a similarity with a second feature value representing a color characteristic of an emission or a target area in the group;
A program for causing a majority decision unit that outputs a plurality of determination results by the decision unit to output information based on the majority decision result.