Determination device, determination method, and computer program

Abstract

Provided is a determination device comprising a control unit that uses a determination model representing the relationship between vital information obtained after surgery of a digestive organ and complication information including information indicating whether or not a complication occurred within a predetermined period of time after the surgery, and determines, on the basis of the vital information obtained from a determination target person who has undergone surgery of the digestive organ, complication information indicating whether or not the determination target person develops a complication at a predetermined time after the vital information has been obtained.

Description

Determination device, determination method, and computer program

[0001] This application claims priority to Japanese Patent Application No. 2024-090586, filed on June 4, 2024, the contents of which are incorporated herein by reference.

[0002] Various methods have been developed to treat postoperative complications in patients undergoing surgical procedures. One of these methods involves estimating the likelihood of complications and taking appropriate measures for patients with a high likelihood of complications. For example, Patent Literature 1 discloses the following technology. First, a prediction model is constructed based on the background factors of a first sample that developed postoperative complications within a certain period of time and a second sample that did not. Then, according to the constructed prediction model, the risk of a subject developing postoperative complications is determined based on the subject's background factors measured preoperatively.

[0003] Japanese Patent Application Laid-Open No. 2020-057053

[0004] However, the accuracy of conventional assessments has not always been sufficient. The present disclosure has been made in consideration of the above-mentioned circumstances, and provides a technology that enables a more accurate estimation of the likelihood of postoperative complications in a subject who has undergone a surgical procedure.

[0005] One aspect of the present disclosure is a determination device that includes a control unit that uses a determination model that indicates the relationship between vital information obtained after digestive surgery and complication information that includes information indicating whether a complication will occur within a predetermined period after the surgery, based on the vital information obtained after the digestive surgery of the subject, to determine complication information that includes whether the subject will develop a complication at a predetermined timing after the vital information is obtained.

[0006] One aspect of the present disclosure is a determination method including a determination step of determining, based on the vital information obtained after a digestive surgery performed by a subject, complication information including whether the subject will develop a complication at a predetermined timing after the vital information is obtained, using a determination model that indicates the relationship between the vital information obtained after the digestive surgery and complication information including information indicating whether the subject will develop a complication within a predetermined period after the surgery.

[0007] One aspect of the present disclosure is a computer program for causing a computer to function as a judgment device including a control unit that uses a judgment model that indicates the relationship between vital information obtained after digestive surgery and complication information including information indicating whether a complication will occur within a predetermined period after the surgery, based on the vital information obtained after the surgery of a person who has undergone digestive surgery, to judge complication information including whether the person will develop a complication at a predetermined timing after the vital information is obtained.

[0008] As an example, the present disclosure makes it possible to estimate with greater accuracy the likelihood of a subject who has undergone surgical procedures developing postoperative complications.

[0009] 1 is a schematic block diagram showing the system configuration of a determination system 100 of the present disclosure. FIG. 1 is a schematic block diagram showing a specific example of the functional configuration of a terminal device 10. FIG. 1 is a schematic block diagram showing a specific example of the functional configuration of a learning device 20. FIG. 2 is a diagram showing an overview of teacher data. FIG. 3 is a flowchart showing a specific example of the processing of the learning device 20. FIG. 3 is a schematic block diagram showing a specific example of the functional configuration of a determination device 30. FIG. 4 is a flowchart showing a specific example of the processing of the determination device 30. FIG. 5 is a diagram showing an overview of teacher data used in an experiment. FIG. 6 is a diagram showing the contents of a first dataset. FIG. 7 is a diagram showing AUC values ​​of an experiment conducted using a trained model of a dataset of each POD. FIG. 8 is a diagram showing the results of an experiment conducted using a trained model of a second dataset. FIG. 9 is a diagram showing the results of an experiment conducted using a trained model of a second dataset. FIG. 10 is a diagram showing an overview of a third dataset. FIG. 11 is a diagram showing the results of an experiment conducted using a trained model of a third dataset. FIG. 12 is a diagram showing the results of an experiment conducted using a trained model of a third dataset. FIG. 13 is a diagram showing the importance value of which explanatory variable values ​​contributed to improving the accuracy of determination for each complication in an experiment conducted using a trained model of the third dataset. 1 is a diagram showing, by a value called importance, which explanatory variable value contributed to improving the accuracy of judgment in the judgment of each complication in an experiment conducted using a trained model of the third dataset. FIG. 2 is a diagram showing the results of an experiment conducted using a trained model of the fourth dataset. FIG. 3 is a diagram showing, by a value called importance, which explanatory variable value contributed to improving the accuracy of judgment in the judgment of each complication in an experiment conducted using a trained model of the fourth dataset. FIG. 4 is a diagram showing an outline of an example hardware configuration of an information processing device 90 applied to this embodiment. FIG. 5 is a diagram showing a modified example of the judgment device 30. FIG. 6 is a diagram showing a modified example of the judgment device 30.

[0010] FIG. 1 is a schematic block diagram showing the system configuration of a determination system 100 of the present disclosure. The determination system 100 is used to determine information regarding the risk of a subject suffering from a postoperative complication (hereinafter referred to as "complication information") based on information regarding the subject (hereinafter referred to as "subject information"). The subject information includes at least postoperative vital sign information of the subject. The subject information may include preoperative information and intraoperative information in addition to postoperative vital sign information. The determination system 100 may be used, for example, to determine complication information using the postoperative vital sign information of the subject as subject information.

[0011] The determination system 100 includes a terminal device 10, a learning device 20, and a determination device 30. The terminal device 10 and the determination device 30 are communicatively connected via a network 70. The learning device 20 and the determination device 30 may also be communicatively connected via the network 70. The network 70 may be a network using wireless communication or a network using wired communication. The network 70 may be configured using, for example, the Internet or a local area network (LAN). The network 70 may also be configured by combining multiple networks.

[0012] 2 is a schematic block diagram showing a specific example of the functional configuration of the terminal device 10. The terminal device 10 is configured using information devices such as a smartphone, tablet, personal computer, dedicated device, etc. The terminal device 10 includes a communication unit 11, an input unit 12, an output unit 13, a storage unit 14, and a control unit 15.

[0013] The communication unit 11 is a communication device. The communication unit 11 may be configured as, for example, a network interface. The communication unit 11 communicates data with other devices via the network 70 in accordance with the control of the control unit 15. The communication unit 11 may be a device that performs wireless communication or a device that performs wired communication.

[0014] The input unit 12 is configured using existing input devices such as a keyboard, a pointing device (mouse, tablet, etc.), buttons, a touch panel, etc. The input unit 12 is operated by a user when inputting user instructions to the terminal device 10. The input unit 12 may be an interface for connecting the input device to the terminal device 10. In this case, the input unit 12 inputs an input signal generated in the input device in response to a user input to the terminal device 10. The input unit 12 may be configured using a microphone and a voice recognition device. In this case, the input unit 12 acquires an acoustic signal generated by the user's speech, performs voice recognition on the words spoken by the user, and inputs character string information of the recognition result to the terminal device 10. The voice recognition process may be performed by the control unit 15. The input unit 12 may be configured in any way as long as it is capable of inputting user instructions to the terminal device 10.

[0015] The output unit 13 outputs information in a form that can be recognized by the user. The output unit 13 may be, for example, an image display device such as a liquid crystal display or an organic EL (Electro Luminescence) display. The output unit 13 may be an interface for connecting an image display device to the terminal device 10. In this case, the output unit 13 generates a video signal for displaying image data and outputs the video signal to the image display device connected to the output unit 13. The output unit 13 may be a device for outputting sound, such as a speaker. The output unit 13 may be an interface for connecting an audio output device, such as a speaker or headphones, to the terminal device 10. In this case, the output unit 13 generates an audio signal for reproducing audio data and outputs the audio signal to the audio output device connected to the output unit 13. The output unit 13 may be configured as a touch panel integrated with the input unit 12.

[0016] The storage unit 14 is configured using a storage device such as a magnetic hard disk drive or a semiconductor storage device. The storage unit 14 stores data used by the control unit 15. The storage unit 14 stores data required when the control unit 15 performs processing.

[0017] The control unit 15 is configured using a processor such as a CPU (Central Processing Unit) and a memory (main storage device). The control unit 15 functions when the processor executes a program. Note that all or part of the functions of the control unit 15 may be realized using hardware such as an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array). The program may be recorded on a computer-readable recording medium. Examples of computer-readable recording media include portable media such as a flexible disk, a magneto-optical disk, a ROM, a CD-ROM, and a semiconductor storage device (e.g., an SSD: Solid State Drive), as well as storage devices such as a hard disk or semiconductor storage device built into a computer system. The program may be transmitted via a telecommunications line.

[0018] The control unit 15 may execute, for example, an application installed on its own device (the terminal device 10). A specific example of such an application is an application provided to the terminal device 10 as a dedicated application for the determination system 100. Another specific example of such an application is a web browser application. Such an application may be pre-installed on the terminal device 10 or may be downloaded each time the determination process is executed. For example, when implemented as a web browser application, the terminal device 10 may download and execute the application from a device specified by the web server (for example, the web server itself or another server) in response to the terminal device 10 connecting to the specific web server. The control unit 15 operates according to the program of the application being executed.

[0019] The control unit 15 controls the terminal device 10 in response to user operations and information received from the determination device 30. For example, the control unit 15 transmits information input by users of the determination subject or their related parties (e.g., relatives, doctors, nurses, etc.) operating the input unit 12 to the determination device 30 using the communication unit 11. For example, when the communication unit 11 receives information transmitted from the determination device 30 via the network 70, the control unit 15 generates screen data based on the received information and displays the screen data on the output unit 13. Such screen data includes images and text indicating the information transmitted from the determination device 30. For example, when the communication unit 11 receives information transmitted from the determination device 30 via the network 70, the control unit 15 generates audio data based on the received information and outputs the audio data from the output unit 13.

[0020] For example, the person being assessed or a person related to the person may input information used as an explanatory variable in the assessment system 100 by operating the input unit 12. An example of such information is subject information of the person being assessed. For example, vital signs information of the person being assessed may be input as subject information. The output unit 13 may display an input field for information that needs to be input, or may output the name of the information that needs to be input by voice. The user may input each value. Furthermore, the information used as an explanatory variable may be input by another device (for example, an electronic medical record). Furthermore, the terminal device 10 may be implemented using an electronic medical record terminal.

[0021] 3 is a schematic block diagram showing a specific example of the functional configuration of the learning device 20. The learning device 20 is configured using an information processing device such as a personal computer or a server device. The learning device 20 includes a communication unit 21, a storage unit 22, and a control unit 23.

[0022] The communication unit 21 is a communication device. The communication unit 21 may be configured as, for example, a network interface. The communication unit 21 communicates data with other devices via the network 70 in accordance with the control of the control unit 23. The communication unit 21 may be a device that performs wireless communication or a device that performs wired communication.

[0023] The storage unit 22 is configured using a storage device such as a magnetic hard disk drive or a semiconductor storage device. The storage unit 22 stores data used by the control unit 23. The storage unit 22 may function as, for example, a teacher data storage unit 221, a preprocessed teacher data storage unit 222, and a trained model storage unit 223.

[0024] The training data storage unit 221 stores training data used in the learning process executed by the learning device 20. The training data stored in the training data storage unit 221 is information obtained from people who have previously undergone surgery. The target surgery is, for example, a digestive surgery. More specifically, it is a stomach surgery. A specific example of a disease that can cause surgery is cancer. For example, surgery for stomach cancer may be the target. Other specific examples of target surgeries include subtotal esophagectomy, total gastrectomy, distal gastrectomy, right hemicolectomy, low anterior resection, liver resection, and pancreaticoduodenectomy. Target complications include anastomotic failure, pancreatic fistula, intraperitoneal abscess, mediastinal abscess, postoperative bleeding (anastomotic bleeding, intraperitoneal bleeding), intestinal obstruction, peristalsis disorder, and pneumonia.

[0025] FIG. 4 is a diagram showing an outline of the training data. The training data includes, for example, vital signs at various times after surgery for a person who has previously undergone surgery. PODn (n is an integer from 1 to N) indicates the number of days that have passed since surgery. For example, POD1 indicates the first day after surgery. A predetermined number is used for N. For example, the value of N may be the number of days that will result in an evaluation of "no complications occurring" if no complications have occurred by POD(N+1). As a specific example, N may be 10.

[0026] Vital information is information related to a person's pulse (heart rate), respiration, blood pressure, or body temperature. The vital information may further include information regarding the level of consciousness and urine volume. More specifically, the vital information may include body temperature, systolic blood pressure, diastolic blood pressure, pulse rate, respiratory rate, and SpO2 (oxygen saturation). Furthermore, the vital information may be further classified according to the timing of acquisition throughout the day. For example, vital information acquired in the morning, vital information acquired during the day, and vital information acquired at night or before bedtime may each be used as values ​​for different vital information items. The timing may be divided into three equally spaced time periods, such as midnight to 8:00, 8:00 to 16:00, and 16:00 to 24:00, or into four equally spaced time periods, such as midnight to 6:00, 6:00 to 12:00, 12:00 to 18:00, and 18:00 to 24:00, or may be divided in other ways.

[0027] The complication information in Figure 4 includes at least information indicating whether a complication has occurred. The complication information may further include information indicating the timing of the complication. This timing may be expressed, for example, as the number of days elapsed since surgery, or information indicating a time period may be added. The complication information may further include information indicating the type of complication that has occurred.

[0028] The training data may further include postoperative information other than preoperative information, intraoperative information, and vital sign information. Specific examples of preoperative information include the patient's age, sex, height, weight, BMI, ASA risk, comorbidities, preoperative disease stage, blood test information on the day before surgery, vital sign information on the day before surgery (body temperature, blood pressure, pulse rate, respiratory rate, SpO2, etc.), and vital sign information on the morning of the day of surgery. Specific examples of intraoperative information include the date of surgery, the surgical procedure used, the approach (e.g., laparotomy or laparoscopy), the time required for surgery, the amount of intraoperative blood loss, whether or not the gallbladder was removed, whether or not the spleen was removed, whether or not other concomitant resections were performed, whether or not intraoperative blood transfusions were performed, and the degree of cure. Specific examples of postoperative information other than vital sign information include blood test information on the first and third days after surgery, drain biochemistry information, etc.

[0029] The preprocessed teacher data storage unit 222 stores preprocessed teacher data obtained by performing preprocessing on the teacher data storage unit 221. By performing the preprocessing, a dataset to be used in the learning process is generated. The first dataset, second dataset, third dataset, and fourth dataset will be described below.

[0030] The first dataset is used in a learning process to acquire information about the possibility of developing complications after (X+1) days as a dependent variable by using vital signs information from the day of surgery until X days after surgery as an explanatory variable. For example, the training data of people who developed complications after (X+1) days uses the value "onset" as the dependent variable, and the training data of people who did not develop complications until N+1 days uses the value "no onset" as the dependent variable.

[0031] The second dataset is used in a learning process to acquire information regarding the possibility of developing complications on the (X+1)th day as a dependent variable by using vital signs information from the day of surgery to X days after surgery as an explanatory variable. For example, even if the data is of a person who developed complications on the fourth day, if the person did not develop complications on the third day, the dependent variable of that data may be treated as "no onset." However, the dataset may be configured so that data of people who had already developed complications prior to the number of days used as an explanatory variable is not included.

[0032] The third dataset is used in a learning process to acquire information about the possibility of developing each complication after (X+1) days as a dependent variable by using vital signs information from the day of surgery until X days after surgery as an explanatory variable. For example, the training data of a person who developed a complication of pancreatic fistula after (X+1) days uses the value "pancreatic fistula" as the dependent variable, and the training data of a person who did not develop a complication until N+1 days uses the value "no onset" as the dependent variable.

[0033] The first to third data sets described above may have the following training data, for example: However, the value of the objective variable differs depending on the type of data set (first to third data sets).

[0034] - A combination of vital information up to POD 0 and complication information for people who developed complications on or after day 1 and for people who did not develop complications by day (N+1). - A combination of vital information from POD 0 to 1 and complication information for people who developed complications on or after day 2 and for people who did not develop complications by day (N+1). ... - A combination of vital information from POD 0 to M and complication information for people who developed complications on or after day (M+1) and for people who did not develop complications by day (N+1). Note that the value of M can be any value from 1 to N.

[0035] The fourth dataset is used in a learning process to acquire, as an explanatory variable, information regarding the possibility of each complication developing in the next adjacent time segment by using vital sign information for the nth time segment on the Xth day after surgery. For example, even if the data is of a person who developed a complication on the fourth day, if the complication had not developed by the nth time segment on the third day, the objective variable for that data may be treated as "no onset." However, data of people who had already developed a complication before the time segment used as the explanatory variable may be configured not to be included in the dataset. The fourth dataset may include, for example, the following training data:

[0036] A combination of vital information in the first hourly division of POD0 and complication information of people who developed complications after the second hourly division of POD0 and people who did not develop complications by the (N+1) day. A combination of vital information in the third hourly division of POD0 and complication information of people who developed complications after the first hourly division of POD1 and people who did not develop complications by the (N+1) day. A combination of vital information in the third hourly division of POD0 and complication information of people who developed complications after the (M+1) day and people who did not develop complications by the (N+1) day.

[0037] In this case, the training data for "onset" is limited to people who developed the disease within one time segment, so the training data contains a large number of "no onset" cases and a small number of "onset" cases. In particular, when trying to create datasets for each type of complication, such as the fourth dataset, the small number of data cases becomes even more pronounced.

[0038] Therefore, preprocessing may be performed to adjust the complication incidence rate in the dataset to a predetermined value (e.g., 10%, 30%, etc.). Such adjustment may be performed, for example, by randomly excluding training data with "no incidence." In the following explanation, such preprocessing is referred to as "Down sample." For example, preprocessing performed to adjust the complication incidence rate in the dataset to 10% is referred to as "Down sample 1," and preprocessing performed to adjust the complication incidence rate in the dataset to 30% is referred to as "Down sample 2."

[0039] The trained model storage unit 223 stores trained models obtained by a learning process using the teacher data stored in the teacher data storage unit 221 and the preprocessed teacher data stored in the preprocessed teacher data storage unit 222.

[0040] The control unit 23 is configured using a processor such as a CPU and a memory. The control unit 23 functions as an information control unit 231, a preprocessing control unit 232, and a learning control unit 233 by the processor executing a program. Note that all or part of the functions of the control unit 23 may be realized using hardware such as an ASIC, a PLD, or an FPGA. The above program may be recorded on a computer-readable recording medium. Examples of computer-readable recording media include portable media such as a flexible disk, a magneto-optical disk, a ROM, a CD-ROM, and a semiconductor storage device (e.g., an SSD), as well as storage devices such as a hard disk or semiconductor storage device built into a computer system. The above program may be transmitted via a telecommunications line.

[0041] The information control unit 231 controls the input and output of information. For example, the information control unit 231 acquires teacher data from another device (an information processing device or a storage medium) and records it in the teacher data storage unit 221. For example, the information control unit 231 transmits the trained model stored in the trained model storage unit 223 to another device (for example, the determination device 30).

[0042] The preprocessing control unit 232 generates preprocessed teacher data by performing a predetermined preprocessing on the teacher data. The preprocessing control unit 232 may generate, as the preprocessed teacher data, one or more of the first to fourth datasets described above, for example.

[0043] The learning control unit 233 performs a learning process using the preprocessed training data (dataset) stored in the preprocessed training data storage unit 222. Specific examples of such a learning process include supervised learning for classification, such as support vector machines, random forests, and neural networks. The learning control unit 233 generates a trained model for outputting complication information based on input vital sign information for a predetermined postoperative period, for example, by performing supervised learning. The input vital sign information corresponds to explanatory variables of the dataset used for learning. The output complication information corresponds to objective variables of the dataset used for learning. The learning control unit 233 records the generated trained model in the trained model storage unit 223. The trained model obtained by the learning control unit 233 may be transmitted to the determination device 30 and recorded in the determination model storage unit 321 of the determination device 30.

[0044] 5 is a flowchart showing a specific example of processing by the learning device 20. First, the information control unit 231 acquires training data (step S101). The training data may be, for example, input by a user, acquired via communication from another information device, or acquired from a recording medium connected to the learning device 20. The preprocessing control unit 232 generates a dataset (preprocessed training data) by performing a predetermined preprocessing on the training data (step S102). The learning control unit 233 executes a learning process using the preprocessed training data and records a trained model in the trained model storage unit 223 (step S103).

[0045] 6 is a schematic block diagram showing a specific example of the functional configuration of the determination device 30. The determination device 30 is configured using an information processing device such as a personal computer or a server device. The determination device 30 includes a communication unit 31, a storage unit 32, and a control unit 33.

[0046] The communication unit 31 is a communication device. The communication unit 31 may be configured as, for example, a network interface. The communication unit 31 communicates data with other devices via the network 70 in accordance with the control of the control unit 33. The communication unit 31 may be a device that performs wireless communication or a device that performs wired communication.

[0047] The storage unit 32 is configured using a storage device such as a magnetic hard disk drive or a semiconductor storage device. The storage unit 32 stores data used by the control unit 33. The storage unit 32 may function as a determination model storage unit 321, for example.

[0048] The judgment model storage unit 321 stores a judgment model used by the judgment unit 332 when performing the judgment process. The judgment model may be configured using information of a trained model generated in advance by a learning process, for example. Such a learning process may be executed by another device (e.g., the learning device 20) or by the device itself (the judgment device 30). The judgment model does not necessarily have to be generated by a learning process. The judgment model may be configured using, for example, a lookup table that associates vital information and complication information according to the rules of any of the first to fourth datasets described above, or may be configured in another manner.

[0049] The control unit 33 is configured using a processor such as a CPU and a memory. The control unit 33 functions as an information control unit 331 and a determination unit 332 by the processor executing a program. Note that all or part of the functions of the control unit 33 may be realized using hardware such as an ASIC, PLD, or FPGA. The above program may be recorded on a computer-readable recording medium. Examples of computer-readable recording media include portable media such as flexible disks, magneto-optical disks, ROMs, CD-ROMs, and semiconductor storage devices (e.g., SSDs), as well as storage devices such as hard disks and semiconductor storage devices built into a computer system. The above program may be transmitted via a telecommunications line.

[0050] The information control unit 331 acquires vital information corresponding to the explanatory variables of the person being assessed from another device, such as the terminal device 10. At this time, the information control unit 331 acquires at least vital information corresponding to the explanatory variables of the assessment model used by the assessment unit 332. For example, if the assessment model used by the assessment unit 332 is a trained model obtained by a learning process using any of the first to third datasets, vital information obtained from the time after surgery until that day (vital information for POD0 to PODn) is acquired. Furthermore, if the assessment model used by the assessment unit 332 is a trained model obtained by a learning process using the fourth dataset, vital information obtained during that time segment on that day (vital information for PODn) is acquired.

[0051] The information control unit 331 transmits information indicating the determination result obtained by the determination unit 332 to another device such as the terminal device 10. Such exchange of information between the information control unit 331 and another device may be performed by communication using the communication unit 31, for example.

[0052] The determination unit 332 performs a determination process using the determination model stored in the determination model storage unit 321 and the vital information of the person to be determined. The vital information used in the determination process may be a combination of vital information obtained from the day of the surgery (POD0 vital information) up to the day or time segment to be determined (for example, the day of the determination or the most recent time segment). The determination process determines complication information according to the objective variable of the determination model used.

[0053] For example, when a trained model obtained by a training process using the first dataset is used, information indicating the possibility of a complication occurring on or after the day after the day on which the vital information of the subject used as the explanatory variable is obtained (the day indicated by the POD) is obtained is obtained as the objective variable. This information may be expressed as a binary value (will occur, will not occur) or as a numerical value such as the possibility of occurrence or the possibility of not occurrence.

[0054] For example, when a trained model obtained by a training process using the second dataset is used, information indicating the possibility of a complication occurring on the day after the day on which the vital signs of the subject used as the explanatory variable are obtained (the day indicated by the POD) is obtained is obtained as the objective variable. This information may be expressed as a binary value (will occur, will not occur) or as a numerical value such as the possibility of occurrence or the possibility of not occurrence.

[0055] For example, when a trained model obtained by a training process using the third dataset is used, information indicating the possibility of each complication occurring on or after the day after the day on which the vital information of the subject used as the explanatory variable was obtained (the day indicated by the POD) is obtained is obtained as the objective variable. This information may be expressed as a binary value (will occur, will not occur) for each type of complication, or may be expressed as a numerical value such as the possibility of each type of complication occurring or not occurring.

[0056] For example, when a trained model obtained by a training process using the fourth dataset is used, information indicating the possibility of a complication occurring in the time segment following the time segment in which the vital information of the subject used as the explanatory variable was obtained is obtained as the objective variable for each complication. This information may be expressed as a binary value (will occur, will not occur) or as a numerical value such as the possibility of occurrence or the possibility of not occurrence.

[0057] 7 is a flowchart showing a specific example of the processing of the determination device 30. First, the information control unit 331 acquires vital sign information used as explanatory variables from the terminal device 10 (step S201). The determination unit 332 performs a determination process using at least the vital sign information (step S202). The determination unit 332 transmits information indicating the determination result to the terminal device 10 (step S203).

[0058] Next, an experiment was conducted to construct a trained model and measure its accuracy using actually obtained information as training data, and this experiment will be described. Figure 8 shows an overview of the training data used in the experiment. The training data used was training data obtained from 4,139 patients who underwent gastric cancer surgery between 2013 and 2019. Of these, 782 patients, accounting for 18.9% of the total, developed complications. The combination of the PODX value and the number of patients shown in Figure 8 indicates the number of days since surgery (POD) at which complications developed. For example, a POD2 of 54 indicates that the training data includes training data for 54 patients who developed complications on the second day after surgery. In the experiment, a random forest was used as the learning model, and two values, "with complications (developed)" and "without complications (no complications)," were used as the objective variables.

[0059] FIG. 9 is a diagram showing the contents of the first dataset. The first dataset was generated by preprocessing using the training data shown in FIG. 8 . The total number of training data up to POD 0 is 4,131. That is, since there were eight patients who developed complications on the day of surgery out of the total 4,139 patients, the number of training data in the POD 0 dataset (onset of complications after POD 1) used as the first dataset was 4,131, of which 774 patients developed complications. Similarly, the dataset up to POD 1 (onset or no onset after the second day after surgery), the dataset up to POD 2 (onset or no onset after the third day after surgery), ..., and the dataset up to POD 5 (onset or no onset after the sixth day after surgery) were generated.

[0060] AUC was calculated as an index representing the accuracy of a trained model obtained by performing a training process using each data set. The trained model was generated for each dataset in which the "X" value in PODX matches. For example, a trained model obtained by performing a training process using a dataset up to POD 0 outputs complication information (objective variable) indicating whether or not the subject will develop complications on postoperative day 1 or later when given the vital signs (explanatory variables) of the subject up to POD 0 as input. Similarly, a trained model obtained by performing a training process using a dataset up to POD 1 (training data on the day of surgery and up to day 1) outputs complication information (objective variable) indicating whether or not the subject will develop complications on postoperative day 2 or later when given the vital signs (explanatory variables) of the subject up to POD 1 as input.

[0061] Figure 10 shows the AUC values ​​of experiments conducted using trained models for each POD dataset. It can be seen that the longer the period during which training data is used as a dataset (the larger the "X" value in PODX), the higher the accuracy. Meanwhile, an AUC of at least 0.65 was obtained in all judgments using trained models for each dataset, indicating a correlation between the explanatory variables and objective variables used in these trained models.

[0062] 11 and 12 are diagrams showing the results of an experiment conducted using a trained model of the second dataset. The second dataset was generated by preprocessing using the training data shown in FIG. 8. AUC was calculated as an index representing the accuracy of the trained model obtained by performing a training process using each data set. Unlike the first dataset, the trained model using the second dataset was trained using a dataset of all PODX (e.g., X = 0 to 5). Therefore, the resulting trained model performs a judgment process using POD vital information up to the day of assessment as the explanatory variable and outputting a value indicating whether or not complications will develop the day after assessment as the objective variable.

[0063] As shown in Figure 11, an AUC of 0.68 or more was obtained, indicating a correlation between the explanatory variables and the objective variable used in this trained model. Figure 12 is a diagram showing the importance of each feature used as an explanatory variable. In Figure 12, MDI (Mean Decrease in Impurity) is shown as one index of importance. As shown in Figure 12, blood test findings and intraoperative information were also found to have a high correlation with the objective variable as explanatory variables, and vital signs information also showed a high correlation.

[0064] FIG. 13 is a diagram showing an outline of the third dataset. The third dataset was generated by preprocessing using the training data shown in FIG. 8. FIG. 13 shows the number of training data items for each complication and each POD as the third dataset. For example, it shows that there are 68 training data items for patients who developed anastomotic leakage on PODs 0 to 10, 64 training data items for patients who developed anastomotic leakage on POD 2 or later, and 51 training data items for patients who developed anastomotic leakage on POD 4 or later.

[0065] The training data of patients who developed symptoms after POD 2 is used with the complication information of the objective variable as "complications present," and the training data of patients who did not develop symptoms after POD 2 is used with the complication information of the objective variable as "no complications." These are used to train a judgment model (trained model) for determining the objective variable using explanatory variables (vital information, etc.) obtained up to POD 1.

[0066] The training data of patients who developed symptoms after POD 4 is used with the complication information of the objective variable set to "complications present," and the training data of patients who did not develop symptoms after POD 4 is used with the complication information of the objective variable set to "no complications." These are used to train a judgment model (trained model) for determining the objective variable using explanatory variables (vital information, etc.) obtained up to POD 3.

[0067] 14 and 15 show the results of an experiment using the trained model of the third dataset. FIG. 14 shows the accuracy of determining whether complications will occur on postoperative day 2 or later using vital signs up to POD 1. FIG. 15 shows the accuracy of determining whether complications will occur on postoperative day 4 or later using vital signs up to POD 3. As can be seen from these results, high AUC values ​​were obtained overall, indicating a correlation between the explanatory variables and objective variables used in these trained models. In particular, high AUCs were obtained for pancreatic fistula, intraperitoneal / mediastinal abscess, and intraperitoneal surgery-related conditions, indicating a high correlation between the explanatory variables and objective variables.

[0068] 16 and 17 are diagrams showing, as a value called importance, which explanatory variable values ​​contributed to improving the accuracy of the assessment of each complication in an experiment conducted using the trained model of the third dataset. FIG. 16 shows the importance when determining whether a complication will occur on or after the second postoperative day using vital signs information up to POD 1. FIG. 17 shows the importance when determining whether a complication will occur on or after the fourth postoperative day using vital signs information up to POD 3. For example, among the vital signs information, it can be seen that urine volume and pulse rate are effective explanatory variables. Apart from the vital signs information, it can be seen that information such as blood test findings, drainage AMY, surgery time, and bleeding volume are effective explanatory variables. The values ​​of the explanatory variables to be used specifically may be selected depending on the importance of the explanatory variable values ​​for each complication.

[0069] Figure 18 shows the results of an experiment conducted using a trained model of the fourth dataset. The fourth dataset was generated by performing preprocessing using the training data shown in Figure 8. A fourth dataset was also generated by performing down sample 1 and down sample 2. The fourth dataset, for which down sample was not performed, is shown as "All."

[0070] A trained model obtained by performing a learning process using each data set was used to calculate the AUC as an index of accuracy. In the trained model using the fourth dataset, a determination process is performed in which the POD vital information for the most recent time segment to be determined is used as the explanatory variable, and a value indicating whether or not a complication will occur in the next time segment is output as the objective variable. As shown in Figure 18, a sufficiently accurate determination process is achieved even without downsampling preprocessing, and it can be seen that there is a high correlation between the explanatory variables and the objective variable. It can also be seen that downsampling improves the accuracy of the determination process for all complication types.

[0071] FIG. 19 is a diagram showing, in an experiment conducted using the trained model of the fourth dataset, which explanatory variable values ​​contributed to improving the accuracy of the assessment of each complication, as a value called importance. From FIG. 19 , it can be seen that vital information is effective as an explanatory variable. Furthermore, it can be seen that, among vital information, values ​​such as pulse rate, body temperature, MaxSpO2, and blood pressure are effective as explanatory variables. Apart from vital information, it can be seen that information such as blood test findings, surgery time, and blood loss are effective as explanatory variables. The values ​​of the explanatory variables to be used specifically may be selected depending on the importance of the explanatory variable values ​​for each complication.

[0072] In the judgment system 100 configured in this manner, by using subject information (e.g., postoperative vital signs information) of the person (subject to be judged) who is being judged for the possibility of developing a complication in the future, it is possible to more accurately judge whether the person to be judged will subsequently develop a complication.

[0073] FIG. 20 is a diagram illustrating an outline of an example hardware configuration of an information processing device 90 applied to this embodiment. The information processing device 90 includes a processor 91, a main storage device 92, a communication interface 93, an auxiliary storage device 94, an input / output interface 95, and an internal bus 96. The processor 91, the main storage device 92, the communication interface 93, the auxiliary storage device 94, and the input / output interface 95 are communicably connected to each other via the internal bus 96. The information processing device 90 may be applied to, for example, the learning device 20 and the determination device 30. In this case, for example, the communication units 21 and 31 may be configured using the communication interface 93. For example, the memory units 22 and 32 may be configured using the auxiliary storage device 94. Furthermore, the control units 23 and 33 may be configured using the processor 91 and the main storage device 92.

[0074] (Modification) In the present embodiment, the terminal device 10 and the determination device 30 are configured as separate devices, but they may also be configured as an integrated device. Fig. 21 is a diagram showing a modification of the determination device 30 configured in this manner. The determination device 30 shown in Fig. 21 includes an input unit 34 and an output unit 35. The input unit 34 and the output unit 35 of the determination device 30 shown in Fig. 21 function in the same manner as the input unit 12 and the output unit 13 of the terminal device 10, respectively. The control unit 33 operates in response to an operation on the input unit 34, performs a determination process using the input subject information, and outputs the information using the output unit 35.

[0075] In this embodiment, the learning device 20 and the determination device 30 are configured as separate devices, but they may also be configured as an integrated device. FIG. 22 is a diagram showing a modified example of the determination device 30 configured in this manner. The memory unit 32 of the determination device 30 shown in FIG. 22 also functions as a teacher data memory unit 322 and a preprocessed teacher data memory unit 323. The control unit 33 of the determination device 30 shown in FIG. 22 also functions as a preprocessing control unit 333 and a learning control unit 334. The teacher data memory unit 322 and the preprocessed teacher data memory unit 323 function in the same way as the teacher data memory unit 221 and the preprocessed teacher data memory unit 222 of the learning device 20, respectively. The preprocessing control unit 333 functions in the same way as the preprocessing control unit 232 of the learning device 20. The learning control unit 334 functions in the same way as the learning control unit 233 of the learning device 20.

[0076] The learning device 20 may be implemented using a plurality of information processing devices. For example, the learning device 20 may be implemented using a device such as a cloud. For example, in the learning device 20, the memory unit 22 and the control unit 23 may be implemented in different information processing devices. For example, the memory unit 22 of the learning device 20 may be distributed and implemented across a plurality of information processing devices. The determination device 30 may be implemented using a plurality of information processing devices. For example, the determination device 30 may be implemented using a device such as a cloud. For example, in the determination device 30, the memory unit 32 and the control unit 33 may be implemented in different information processing devices. For example, the memory unit 32 of the determination device 30 may be distributed and implemented across a plurality of information processing devices.

[0077] The application running on the terminal device 10 may be an application that provides a user with information obtained by performing processing using the determination result of the complication information. Such an application may be, for example, an application that performs processing using an API (Programming Interface) provided by the determination device 30. In this case, instead of information indicating the information itself (complication information) obtained from the determination device 30, the output unit 13 of the terminal device 10 may output other information (e.g., the time when the patient can be discharged from the hospital) obtained by processing using the complication information.

[0078] A trained model using the second dataset may be generated for each POD value by performing a training process for each POD value. For example, a trained model for determining whether or not a complication will occur on postoperative day 2 may be generated by performing a training process using data up to POD 1, and a trained model for determining whether or not a complication will occur on postoperative day 4 may be generated as a model different from the trained model up to POD 1 by performing a training process using data up to POD 3. The same applies to the fourth dataset.

[0079] This embodiment has been described above in detail with reference to the drawings, but the specific configuration is not limited to this embodiment, and includes designs within the scope of the gist of this disclosure.

[0080] According to the present disclosure, as an example, it is possible to estimate with higher accuracy the likelihood of postoperative complications occurring in a subject who has undergone a surgical procedure.

[0081] REFERENCE SIGNS LIST 100... Determination system, 10... Terminal device, 11... Communication unit, 12... Input unit, 13... Output unit, 14... Storage unit, 15... Control unit, 20... Learning device, 21... Communication unit, 22... Storage unit, 221... Teacher data storage unit, 222... Preprocessed teacher data storage unit, 223... Learned model storage unit, 23... Control unit, 231... Information control unit, 232... Preprocessing control unit, 233... Learning control unit, 30... Determination device, 31... Communication unit, 32... Storage unit, 321... Determination model storage unit, 33... Control unit, 331... Information control unit, 332... Determination unit

Claims

1. A judgment device comprising a control unit that uses a judgment model that shows the relationship between vital information obtained after digestive surgery and complication information including information indicating whether a complication has occurred within a specified period after the surgery, and that judges complication information including whether the subject will develop a complication at a specified time after the vital information is obtained based on the vital information obtained after the digestive surgery of the subject.

2. The determination device according to claim 1, wherein the vital information includes at least one of body temperature, blood pressure, and pulse rate.

3. A determination method comprising a determination step of using a determination model showing the relationship between vital information obtained after digestive surgery and complication information including information indicating whether a complication has occurred within a specified period after the surgery, to determine complication information including whether the subject will develop a complication at a specified time after the vital information is obtained, based on the vital information obtained after the digestive surgery of the subject.

4. A computer program for causing a computer to function as a judgment device that includes a control unit that uses a judgment model that shows the relationship between vital information obtained after digestive surgery and complication information that includes information indicating whether a complication has occurred within a specified period after the surgery to judge complication information including whether the subject will develop a complication at a specified time after the vital information is obtained, based on the vital information obtained after the surgery of the subject who has undergone digestive surgery.

Images