Evaluation system, processing device, and computer program

Abstract

To diversify environments for evaluating measurement waveforms of biological parameters acquired from an object person.SOLUTION: An inference device 11 acquires first data D1 corresponding to a measurement waveform of biological parameters of an object person 20, infers the probability of the value of the biological parameters being measured correctly, and generates second data D2 in which the probability is added to the first data D1. A processing device 12 calculates a value score of the measurement waveform by applying a predetermined evaluation function to the second data D2 in which the probability exceeds a threshold, and generates third data D3 in which the value score is added to the second data D2. The value score has a positive correlation for the probability, and a negative correlation for the time elapsed from the measurement of the biological parameters. The processing device 12 acquires an index value showing a risk of a sudden change in the condition of the object person 20 on the basis of at least one piece of the third data D3 in which a higher value score is added. The processing device 12 changes at least one of the threshold and the evaluation function on the basis of the index value.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to a system for evaluating biometric information obtained from a subject, a processing device included in the system, and a computer program executable by a processor installed in the processing device. [Background technology]

[0002] Patent Document 1 discloses a device that infers the probability that the value of a biological parameter will be correctly measured by inputting data corresponding to the measured waveform of the biological parameter obtained from a subject into a machine-learned model. By identifying data that is assigned a probability below a threshold, biological information obtained in a low-reliability environment, such as under the influence of noise, can be excluded from the evaluation target. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2022-019602 Summary of the Invention [Problem to be solved by the invention]

[0004] There is a demand for diversifying the environments in which measurement waveforms of biological parameters obtained from subjects can be evaluated. [Means for solving the problem]

[0005] One exemplary embodiment provided by the present disclosure is an evaluation system, comprising: an inference device that acquires first data corresponding to a measured waveform of a biological parameter of a subject, infers a probability that the value of the biological parameter will be correctly measured, and generates second data in which the first data is assigned the probability; A processing device that calculates a value score of the measured waveform by applying a predetermined evaluation function to the second data in which the probability exceeds a first threshold value, and generates third data in which the value score is assigned to the second data; It is equipped with the evaluation function is defined such that the value score has a positive correlation with the probability and a negative correlation with the time elapsed since the biological parameter was measured; The processing device includes: Obtaining an index value indicating a risk of sudden change in the condition of the subject based on at least one of the third data to which a higher value score is assigned; and At least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold is changed based on the index value.

[0006] One embodiment of the present disclosure provides a processing device communicably connected to an inference device that acquires first data corresponding to a measured waveform of a biological parameter of a subject, infers a probability that a value of the biological parameter will be correctly measured, and generates second data in which the first data is assigned the probability, an interface for accepting the second data; a processor that calculates a value score of the measured waveform by applying a predetermined evaluation function to the second data in which the probability exceeds a first threshold value, and generates third data in which the value score is assigned to the second data; It is equipped with the evaluation function is defined such that the value score has a positive correlation with the probability and a negative correlation with the time elapsed since the biological parameter was measured; The processor, Obtaining an index value indicating a risk of sudden change in the condition of the subject based on at least one of the third data to which a higher value score is assigned; and At least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold is changed based on the index value.

[0007] One embodiment of the present disclosure provides a computer program executable by a processor mounted on a processing device that cooperates with an inference device to acquire first data corresponding to a measured waveform of a biological parameter of a subject, infer a probability that the value of the biological parameter will be correctly measured, and generate second data in which the first data is assigned the probability, comprising: When executed, the processing device: Accepting the second data; Calculating a value score of the measured waveform by applying a predetermined evaluation function to the second data in which the probability exceeds a first threshold value; generating third data in which the value score is assigned to the second data; the evaluation function is defined such that the value score has a positive correlation with the probability and a negative correlation with the time elapsed since the biological parameter was measured; The processing device further comprises: Obtaining an index value indicating a risk of sudden change in the condition of the subject based on at least one of the third data to which a higher value score is assigned; and At least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold is changed based on the index value.

[0008] According to the configurations of the above-mentioned embodiments, at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value is changed, so that even if the measurement value of the biological parameter of the subject is the same, the value can be changed according to the risk of a sudden change in the condition of the subject. For example, when the index value shows a change corresponding to an increase in the risk of a sudden change in the condition, a change is made to increase the sensitivity of abnormality detection and decrease the specificity. Specifically, at least one of a change to decrease the value of the reliability of the measurement value and a change to increase the value of the freshness of the measurement value is made. On the other hand, when the index value shows a change corresponding to a decrease in the risk of a sudden change in the condition, a change is made to decrease the sensitivity of abnormality detection and increase the specificity. Specifically, at least one of a change to increase the value of the reliability of the measurement value and a change to decrease the value of the freshness of the measurement value is made. Therefore, it is possible to diversify the environment in which the measurement waveform of the biological parameter obtained from the subject is evaluated. [Brief description of the drawings]

[0009] [Figure 1] 2 illustrates an example of a functional configuration of an evaluation system according to an embodiment. [Diagram 2] 2 illustrates the time course of the probability of a normal measurement assigned by the inference device of FIG. 1. [Diagram 3] 2 shows an example of third data generated by the processing device of FIG. 1. [Figure 4] FIG. 2 is a diagram for explaining a change in the setting of the evaluation model of FIG. 1. [Diagram 5] FIG. 2 is a diagram for explaining a change in the setting of the evaluation model of FIG. 1. [Figure 6] FIG. 2 is a diagram for explaining a change in the setting of the evaluation model of FIG. 1. [Figure 7] FIG. 2 is a diagram for explaining a change in the setting of the evaluation model of FIG. 1. [Figure 8] 2 illustrates an example of a screen displayed by the display device of FIG. 1. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Exemplary embodiments are described in detail below with reference to the accompanying drawings.

[0011] 1 illustrates a functional configuration of an evaluation system 10 according to an embodiment. The evaluation system 10 has a configuration for evaluating biological information acquired from a subject 20.

[0012] The evaluation system 10 includes an inference device 11. The inference device 11 has an input interface 111. The input interface 111 is configured as a hardware interface that receives first data D1 corresponding to a measured waveform of a biological parameter of a subject 20. The first data D1 includes information indicating the time point when the measured waveform was acquired.

[0013] The first data D1 may be received directly from the sensor 30 attached to the subject 20, or may be received via a monitor device (not shown) to which a plurality of types of sensors are connected. The first data D1 may be in the form of analog data or digital data depending on the specifications of the output source. When the first data D1 is in the form of analog data, the input interface 111 includes an appropriate conversion circuit including an A / D converter.

[0014] The term "biological parameter of a subject" as used herein refers to a measurable value among various indicators that represent the physiological state of the subject. The measured value of a biological parameter is an example of biological information.

[0015] The inference device 11 includes an inference model 112. The inference model 112 is a computer algorithm configured to output, as an inference result, the probability that the value of a biological parameter will be correctly measured based on the measurement waveform of the biological parameter corresponding to the first data D1. The probability takes a value between 0 and 1. The value 0 corresponds to 0%. The value 1 corresponds to 100%. The level of the probability depends on the measurement environment of the biological parameter, the condition of the subject, etc.

[0016] The inference model 112 can be generated by performing machine learning using training data or through an appropriate statistical method. The training data is provided as a combination of measurement waveforms of biological parameters acquired under different measurement environments and subject conditions, and annotations indicating whether the biological parameters have been calculated correctly. Examples of algorithms used in machine learning include neural networks, decision trees, random forests, support vector machines, Bayesian estimation, and Gaussian processes. Examples of statistical methods include a method using the coefficient of variation.

[0017] The inference device 11 includes a processor 113 and an output interface 114. The processor 113 is configured to input the first data D1 received by the input interface 111 to the inference model 112, and output the second data D2 generated by assigning the probability output from the inference model 112 to the first data D1 from the output interface 114. In other words, the output interface 114 is configured as a hardware interface capable of outputting the second data D2.

[0018] The processor 113 may calculate the value of the biological parameter based on the first data D1 corresponding to the measured waveform of the biological parameter. In this case, the calculated value may be included in the second data D2 as the measured value of the biological parameter. Depending on the specifications of the output source of the first data D1, the measured value of the biological parameter may be included in the first data D1.

[0019] The evaluation system 10 includes a processing device 12. The processing device 12 is equipped with an input interface 121. The input interface 121 is configured as a hardware interface that accepts second data D2.

[0020] The processing device 12 includes an evaluation model 122. The evaluation model 122 is a computer algorithm configured to calculate a value score of the measurement waveform of the biological parameter related to the first data D1 by applying a predetermined evaluation function to the second data D2. In this example, the value score ranges from 0 to 100.

[0021] The significance of the value score will be explained with reference to FIG. 2. FIG. 2 illustrates the change over time in the probability that a value estimated for each value of a biological parameter obtained from a subject 20 has been correctly measured (hereinafter referred to as "probability of correct measurement"). Thus, the horizontal axis represents the passage of time, and the vertical axis represents the probability. In this example, the biological parameter is respiration rate. Although not shown in the graph, one measured value of respiration rate is associated with each plot.

[0022] The normal measurement probability can be used, for example, in so-called data cleansing. A respiration rate measurement value to which a low probability value is assigned can be considered to be more likely to have been obtained under relatively poor measurement conditions. Factors that deteriorate the measurement conditions include the subject's body movement, sensor misalignment, and failure. By setting an appropriate threshold value Pt for the normal measurement probability and not adopting a measurement value to which a probability value below the threshold value is assigned, a highly reliable respiration rate measurement value can be selected. The threshold value Pt is an example of a first threshold value.

[0023] In Figure 2, the probability of normal measurement assigned to the respiration rate value measured at time t1 is higher than the probability of normal measurement assigned to the respiration rate value measured at time t2. If we focus only on the probability values, it can be said that the respiration rate value measured at time t1 is more reliable than the respiration rate value measured at time t2. However, it can also be said that the respiration rate measured later in time better reflects the subject's current condition and is therefore more reliable.

[0024] In this embodiment, the value score is calculated based on the guideline that "the longer the elapsed time since the measurement, the lower the reliability of the measurement value." In other words, the value score represents the reliability of the measurement value of the respiration rate with the evaluation based on the guideline taken into account. Therefore, the above evaluation function applied to the second data D2 by the evaluation model 122 is determined so that the value score calculated as a result has a positive correlation with the probability of normal measurement and a negative correlation with the elapsed time from the time of measurement of the respiration rate.

[0025] 1, the processing device 12 includes a processor 123. The processor 123 is configured to input the second data D2 received by the input interface 121 to an evaluation model 122, and generate third data D3 by assigning a value score output from the evaluation model 122 to the second data D2.

[0026] Fig. 3 shows an example of the third data D3 generated by the processor 123. A normal measurement probability and a value score are calculated for each of the respiratory rates measured at seven time points. In Fig. 3, the third data D3 including a higher value score is shown at the top. As a result, the measurement dates and times are not arranged in chronological order, and the normal measurement probabilities are not arranged in ascending order.

[0027] The processor 123 is configured to acquire an index value indicating a risk of sudden change in the condition of the subject 20 based on a respiration rate value corresponding to at least one third data D3 to which a higher value score has been assigned. As an example, the index value is acquired based on a respiration rate value corresponding to the third data D3 to which the highest value score has been assigned. When multiple respiration rate values ​​corresponding to multiple third data D3 are referred to, the index value is acquired based on a representative value of the multiple respiration rates. Examples of the representative value include an average value, a median value, and a maximum value.

[0028] The determination of the index value based on the respiratory rate value may be based on, for example, the following rules: A higher index value represents a higher risk of a change in condition. 12-20 times per minute: 0 9-11 times per minute: 1 21-24 times per minute: 2 8 or less per minute or 25 or more per minute: 3

[0029] The processor 123 is configured to change at least one of the degree of positive correlation and the degree of negative correlation in the above evaluation function based on the identified index value.

[0030] As illustrated in FIG. 4, the aforementioned evaluation function can be expressed as a triangle TR having a first side S1 and a second side S2. The first side S1 extends parallel to the coordinate axis representing the probability of a normal measurement. The second side S2 extends parallel to the coordinate axis representing time. The shading inside the triangle TR represents the value score. Darker areas correspond to higher value scores. Measurements outside the triangle TR are not included in the calculation of the value score.

[0031] When the positive correlation of the value score in the evaluation function with the probability of normal measurement is linear, the process of changing the degree of correlation corresponds to an operation of changing the length of the first side S1 of the triangle TR. In the example shown in FIG. 5, the first side S1 is changed to be longer than in the example shown in FIG. 4. This operation corresponds to lowering the value of the reliability of the measurement value. In the example shown in FIG. 6, the first side S1 is changed to be shorter than in the example shown in FIG. 4. This operation corresponds to raising the value of the reliability of the measurement value.

[0032] If the positive correlation of the value score in the evaluation function with the probability of normal measurement is nonlinear, in addition to or instead of the process of changing the length of the first side S1, a process of changing the profile of the correlation characteristics can be performed.

[0033] When the negative correlation of the value score in the evaluation function with respect to the elapsed time from the measurement time point is linear, the process of changing the degree of the correlation corresponds to an operation of changing the length of the second side S2 of the triangle TR. In the example shown in FIG. 5, the second side S2 is changed to be shorter than in the example shown in FIG. 4. This operation corresponds to increasing the value of the freshness of the measurement value. In the example shown in FIG. 6, the second side S2 is changed to be longer than in the example shown in FIG. 4. This operation corresponds to decreasing the value of the freshness of the measurement value.

[0034] If the negative correlation of the value score in the evaluation function with the probability of a normal measurement is nonlinear, in addition to or instead of changing the length of the second side S2, a process can be performed to change the profile of the correlation characteristics.

[0035] Processor 123 is configured to monitor changes in an index value related to the risk of a sudden change in condition. When the index value exhibits a change corresponding to an increase in the risk, processor 123 modifies the evaluation function to increase the sensitivity of anomaly detection and decrease the specificity. Specifically, at least one of a modification to decrease the value of the reliability of the measurement value and a modification to increase the value of the freshness of the measurement value is performed. In the example shown in FIG. 5, both modifications are performed.

[0036] In situations where there is a high risk of a sudden change in the subject's condition, the reliability of the measured values ​​is generally judged by medical professionals themselves, so by widening the allowable range of reliability and increasing the amount of data, priority is given to preventing the subject's dangerous signs from being overlooked. On the other hand, by reducing the amount of data to exclude measurements that have been taken a long time ago, the impact of measurements that are not in line with the subject's current condition on the calculation of the value score is suppressed.

[0037] When the index value shows a change corresponding to a decrease in the risk of a sudden change in condition, the processor 123 changes the evaluation function to lower the sensitivity of anomaly detection and increase the specificity. Specifically, at least one of a change to increase the value of the reliability of the measurement value and a change to decrease the value of the freshness of the measurement value is made. In the example shown in FIG. 6, both changes are made.

[0038] In situations where the risk of a sudden change in the subject's condition is low, the restrictions on the time elapsed since measurement are relaxed while the restrictions on reliability are tightened, thereby enabling gentle monitoring of the subject while reducing the burden on medical staff caused by frequent alerts.

[0039] In other words, even if the same measurement value is obtained from a subject, the value can be changed according to the risk of sudden change in the subject's condition. Therefore, it is possible to diversify the environment for evaluating the measurement waveforms of biological parameters obtained from a subject.

[0040] As illustrated in Fig. 5, the processor 123 is configured to be able to change the threshold value Pt related to data cleansing based on the index value related to the risk of sudden change in the condition of the subject 20. When the evaluation function is changed to lower the value of the reliability of the measurement value as in this example, measurement values ​​with low reliability (probability of normal measurement) may also affect the calculation of the value score. In such a case, the threshold value Pt is changed to lower it (to the threshold value Pt' in the illustrated example), so that the amount of data that can be used to calculate the value score can be increased.

[0041] The threshold value Pt may be changed alone without changing the evaluation function. For example, the threshold value Pt may be lowered when the index value shows a change corresponding to an increase in the risk of a sudden change in condition, and the threshold value Pt may be increased when the index value shows a change corresponding to a decrease in the risk of a sudden change in condition.

[0042] The processor 123 may be configured to exclude the second data D2 relating to a particular time point if the consistency between the second data D2 generated for a respiration rate measured at the particular time point and the second data D2 generated for a respiration rate measured before the particular time point and / or a respiration rate measured after the particular time point falls below a threshold value. The threshold value is an example of a second threshold value.

[0043] Specifically, the processor 123 determines that the consistency is low when the difference between the value of the normal measurement probability assigned to the respiratory rate measured at a specific time point and the values ​​of the normal measurement probability assigned to the respiratory rates measured before and after that time is large. An algorithm that makes a determination based on the difference between a moving average value of a specific number of measured values ​​and a specific measured value, an algorithm based on deep metric learning, or the like may be used to determine the consistency.

[0044] 7, when comparing the second data D2 surrounded by the frame F1 with the second data D2 surrounded by the frame F2, it is found that the latter has a large variation between the measured values ​​and low consistency. In such a case, the processor 123 calculates the value score by excluding the second data D2 surrounded by the frame F2, even if the time elapsed since the measurement is short and the probability of normal measurement is high.

[0045] The processor 123 may be configured to change the threshold value related to the consistency based on the index value indicating the risk of sudden change in the condition of the subject 20. Specifically, when the index value shows a change corresponding to an increase in the risk of sudden change in the condition, the threshold value is changed to be lowered. This prioritizes preventing the subject's dangerous signs from being overlooked, and even measured values ​​with relatively large variances can be used to calculate the value score. On the other hand, when the index value shows a change corresponding to a decrease in the risk of sudden change in the condition, the threshold value is changed to be higher. This reliably excludes measured values ​​with relatively large variances from the calculation of the value score, and reduces the burden on medical professionals caused by frequent alerts.

[0046] The above-mentioned decrease in consistency may occur due to a sudden change in the value of the normal measurement probability. Such a sudden change in the probability value is likely due to noise. By excluding measurements obtained under such circumstances from the calculation of the value score, it is possible to prevent a situation in which a measurement that is actually unreliable receives a high evaluation. By making it possible to change the consistency threshold for achieving this effect according to the risk of a sudden change in the subject's condition, it is possible to further diversify the environment for evaluating the measurement waveforms of biological parameters obtained from the subject.

[0047] As illustrated in Fig. 1, the evaluation system 10 includes a display device 13. The processing device 12 is provided with an output interface 124 configured as a hardware interface. The processor 123 is configured to output, from the output interface 124, display control data DC that causes the display device 13 to display an image illustrated in Fig. 4.

[0048] That is, in addition to the time-dependent change in the probability of normal measurement of the respiratory rate, a straight line indicating the threshold value Pt for data cleansing and a triangle TR indicating the characteristics of the evaluation function can be displayed on the display device 13.

[0049] The display control data DC may be in the form of analog data or digital data depending on the specifications of the display device 13. When the display control data DC is output in the form of analog data, the output interface 124 includes an appropriate conversion circuit including a D / A converter. This description is similarly applied to other data that can be output by the output interface 124, which will be described later.

[0050] 1, the evaluation system 10 may include a user interface 14. The user interface 14 is configured to receive an instruction to change at least one of the threshold Pt, the length of the first side S1, and the length of the second side S2 as the settings of the evaluation model 122, and to output instruction data IS corresponding to the instruction. The user interface 14 may be provided as a part of the inference device 11 or the processing device 12, or may be provided as a part of a device independent of these.

[0051] The instruction data IS is received by the input interface 121 of the processing device 12. The instruction data IS may be provided in the form of analog data or in the form of digital data, depending on the specifications of the user interface 14. When the instruction data IS is provided in the form of analog data, the input interface 121 has an appropriate conversion circuit including an A / D converter. This description is similarly applied to other data that can be received by the input interface 121, which will be described later.

[0052] Based on the instruction data IS received by the input interface 121, the processor 123 of the processing device 12 outputs, from the output interface 124, display control data DC that causes the display device 13 to visualize changes relating to at least one of the threshold value Pt and the evaluation function described with reference to Figures 4 to 6.

[0053] According to such a configuration, it is possible to visually confirm the result of the change made by the processor 123 to at least one of the threshold value Pt and the evaluation function based on the index value indicating the risk of a sudden change in the condition of the subject 20. In addition, it is possible to visually support the input of an instruction to further change at least one of the threshold value Pt and the evaluation function made by the user as necessary.

[0054] As illustrated in FIG. 1, the user interface 14 can be configured to accept feedback regarding the validity of the results of a change made by the processor 123 to at least one of the threshold Pt and the evaluation function based on an index value indicating the risk of a sudden change in the condition of the subject 20, and to output feedback data FB corresponding to the feedback.

[0055] For example, as illustrated in Fig. 8, the display device 13 may display a first screen 131. The first screen 131 includes a Confirm button image 131a and a Reject button image 131b in addition to the images illustrated in Fig. 4. When the Confirm button image 131a is set to a selected state, feedback is given that the presented change result is valid. When the Reject button image 131b is set to a selected state, feedback is given that the presented change result is invalid.

[0056] Each button image can be selected using an input device such as a mouse. If the display device 13 is a touch panel type, each button image can be selected through a touch operation by the user. In other words, the user interface 14 can be provided as a part of the display device 13.

[0057] When the reject button image 131b is selected, the second screen 132 is displayed on the display device 13. The second screen 132 includes a plurality of radio button images 132a. The plurality of radio button images 132a are provided to receive more specific feedback. By selecting one of the plurality of radio button images 132a, the user can input the reason why the change result provided through the first screen 131 is considered to be inappropriate.

[0058] There may be cases where the relationship between the risk of sudden changes in condition and the value of the measurement value that is initially adopted by the evaluation system 10 does not conform to the user's preferences or the guidelines of the medical institution to which the user belongs. With the above-mentioned configuration, it is possible to provide room for improvement to the evaluation model 122 by allowing feedback related to such actual conditions. This makes it possible to further increase the diversity of the environment for evaluating the measurement waveforms of the biological parameters obtained from the subject.

[0059] The second screen 132 further includes a back button image 132b and a confirm button image 132c. When the back button image 132b is selected, the display device 13 redisplays the first screen 131. When the confirm button image 132c is selected, the display device 13 displays a third screen 133. The third screen 133 includes a proposal to re-change at least one of the evaluation functions of the threshold value Pt based on the above feedback.

[0060] As an example, when the second screen 132 provides feedback that the anomaly detection is "too sensitive," the processor 123 proposes changes to the settings of the evaluation model 122 as illustrated in FIG. 6. Specifically, a change to increase the value of the reliability and a change to decrease the value of the freshness are proposed. The user responds to the suggestions by inputting whether or not each suggestion is necessary. The third screen 133 includes a confirm button image 133a and a reject button image 133b. When the confirm button image 133a is selected, the proposed change to the settings of the evaluation model 122 is validated. When the reject button image 131b is selected, the display device 13 redisplays the second screen 132.

[0061] According to such a configuration, the evaluation system 10 can be caused to change the relationship between the risk of sudden change in condition and the value of the measurement value that is initially adopted by the evaluation system 10 so as to bring it closer to the preference of the user or the guidelines of the medical institution to which the user belongs. This can further increase the diversity of the environment for evaluating the measurement waveforms of the biological parameters obtained from the subject.

[0062] The third screen 133 may include a simulation image 133c of the result of changing the settings of the evaluation model 122 based on the user's response to the suggestion. With this configuration, it is possible to visually confirm how the response to the suggestion may change the settings of the evaluation model 122, so that the user can efficiently change the settings as desired.

[0063] 1, the evaluation system 10 may include a notification device 15. The processing device 12 is configured to output notification control data NC, which causes the notification device 15 to perform a notification when a setting change is made to the evaluation model 122, from an output interface 124. The notification device 15 may be a mobile device that can be carried by a medical worker, or may be a terminal device that is installed at a specific location and can be shared by multiple medical workers. The notification may be made through at least one of a visual notification, an audible notification, and a tactile notification.

[0064] A change in the settings of the evaluation model 122 in the processing device 12 has a significant impact on medical practice, and so is something that should be made known to medical professionals. With this configuration, the fact that the settings have been changed can be efficiently made known to those who were not directly involved in the work of changing the settings.

[0065] The evaluation function of the evaluation model 122 may be provided by an inference model generated through machine learning. Examples of algorithms used in machine learning include neural networks, decision trees, random forests, support vector machines, Bayesian estimation, and Gaussian processes. As training data, the results of changes made by the user (which settings were selected or rejected) described with reference to FIG. 8 may be used.

[0066] 1, the evaluation system 10 may include a re-learning device 16. The re-learning device 16 is a device for re-learning an evaluation model 122 generated through machine learning. The re-learning device 16 is communicatively connected to the processing device 12.

[0067] In this case, the processor 123 of the processing device 12 is configured to generate fourth data D4 by adding information corresponding to the feedback data FB received by the input interface 121 to the third data D3. The fourth data D4 is output from the output interface 124 to the re-learning device 16.

[0068] The re-learning device 16 holds the same evaluation model M as that implemented in the processing device 12. The fourth data D4 includes information (annotation) about the value score assigned to a specific measured waveform and its validity. Therefore, by using the fourth data D4 as teacher data, the re-learning device 16 can perform re-learning on the evaluation model M to adjust the evaluation function so that a more valid value score can be assigned. The processing device 12 updates the evaluation model 122 by receiving the re-learned evaluation model M from the re-learning device 16.

[0069] Even with such a configuration, changes that bring the validity of the value score closer to the user's preferences or the guidelines of the medical institution to which the user belongs can be reflected in the operation of the evaluation model 122. Although not shown in the figure, the re-learning device 16 can be shared by multiple processing devices 12. In this case, the same evaluation model 122 can be shared between multiple evaluation systems 10 that share preferences and guidelines regarding the validity of the value score.

[0070] Each of the processor 113 of the inference device 11 and the processor 123 of the processing device 12 having the various functions described above can be realized by a general-purpose microprocessor operating in cooperation with a general-purpose memory. Examples of the general-purpose microprocessor include a CPU, an MPU, and a GPU. Examples of the general-purpose memory include a ROM and a RAM. In this case, the ROM can store a computer program that executes the above-mentioned processing. The ROM is an example of a non-transitory computer-readable medium that stores a computer program. The general-purpose microprocessor specifies at least a part of the computer program stored on the ROM, expands it on the RAM, and executes the above-mentioned processing in cooperation with the RAM. The computer program may be pre-installed in the general-purpose memory, or may be downloaded from an external server device via a communication network and then installed in the general-purpose memory. In this case, the external server device is an example of a non-transitory computer-readable medium that stores a computer program.

[0071] Each of processor 113 of inference device 11 and processor 123 of processing device 12 may be realized by a dedicated integrated circuit capable of executing the above computer program, such as a microcontroller, an ASIC, or an FPGA. In this case, the above computer program is pre-installed in a storage element included in the dedicated integrated circuit. The storage element is an example of a computer-readable medium storing a computer program. Each of processor 113 of inference device 11 and processor 123 of processing device 12 may also be realized by a combination of a general-purpose microprocessor and a dedicated integrated circuit.

[0072] The configurations described above are merely examples for facilitating understanding of the present disclosure. Each configuration example may be appropriately modified or combined with other configuration examples without departing from the spirit of the present disclosure.

[0073] In Fig. 1, reasoning device 11, processing device 12, display device 13, notification device 15, and re-learning device 16 are illustrated as independent devices. However, at least two of these devices may be implemented as different functional modules in a single device. For example, if reasoning device 11 and processing device 12 are provided as a single device, output interface 114 of reasoning device 11 and input interface 121 of processing device 12 may be implemented as software interfaces rather than hardware interfaces.

[0074] The inference model 112 does not necessarily have to be installed in the inference device 11. Although not shown in the figure, the inference device 11 may be communicatively connected to an external server device via a communication network. In this case, the inference model 112 may be installed in the external server device.

[0075] The evaluation model 122 does not necessarily have to be mounted on the processing device 12. Although not shown in the figure, the processing device 12 may be communicably connected to an external server device via a communication network. In this case, the evaluation model 122 may be mounted on the external server device.

[0076] The configurations listed below also form part of this disclosure. Item 1: an inference device that acquires first data corresponding to a measured waveform of a biological parameter of a subject, infers a probability that the value of the biological parameter will be correctly measured, and generates second data in which the first data is assigned the probability; A processing device that calculates a value score of the measured waveform by applying a predetermined evaluation function to the second data in which the probability exceeds a first threshold value, and generates third data in which the value score is assigned to the second data; It is equipped with the evaluation function is defined such that the value score has a positive correlation with the probability and a negative correlation with the time elapsed since the biological parameter was measured; The processing device includes: Obtaining an index value indicating a risk of sudden change in the condition of the subject based on at least one of the third data to which a higher value score is assigned; and changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value based on the index value; Rating system. Item 2: The processing device includes: If the consistency between the second data generated for the biological parameter measured at a particular time point and the second data generated for at least one of the biological parameter measured before the particular time point and the biological parameter measured after the particular time point is below a second threshold, the second data related to the particular time point is excluded; and Varying the second threshold based on the index value. The evaluation system described in item 1. Item 3: The processing device includes: visualizing the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value on a display device; receiving, from a user interface, an instruction to change at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value; 3. The evaluation system according to item 1 or 2. Item 4: the processing device receives, from a user interface, feedback on a result of changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value; 4. The evaluation system according to any one of items 1 to 3. Item 5: the processing device proposes re-changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value based on the feedback. The evaluation system described in item 4. Item 6: The processing device causes a display device to present a simulation of the result of the re-modification. The evaluation system described in item 5. Item 7: the processing device causes a notification device to perform a notification when at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value is changed; 7. The evaluation system according to any one of items 1 to 6. Item 8: The evaluation function is provided by an inference model generated through machine learning. 8. The evaluation system according to any one of items 1 to 7. Item 9: The processing device includes: receiving feedback from a user interface regarding a result of changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold; and generating fourth data in which information corresponding to the feedback is added to the third data; and a device for retraining the inference model based on the fourth data. The evaluation system described in item 8. [Explanation of symbols]

[0077] 10: Evaluation system, 11: Inference device, 12: Processing device, 121: Input interface, 122: Evaluation model, 123: Processor, 13: Display device, 14: User interface, 15: Notification device, 16: Re-learning device, 20: Subject, D1: First data, D2: Second data, D3: Third data, D4: Fourth data

Claims

1. an inference device that acquires first data corresponding to a measured waveform of a biological parameter of a subject, infers a probability that the value of the biological parameter will be correctly measured, and generates second data in which the first data is assigned the probability; A processing device that calculates a value score of the measured waveform by applying a predetermined evaluation function to the second data in which the probability exceeds a first threshold value, and generates third data in which the value score is assigned to the second data; Equipped with the evaluation function is defined such that the value score has a positive correlation with the probability and a negative correlation with the time elapsed since the biological parameter was measured; The processing device includes: Obtaining an index value indicating a risk of sudden change in the condition of the subject based on at least one of the third data to which a higher value score is assigned; and changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value based on the index value; Rating system.

2. The processing device includes: If the consistency between the second data generated for the biological parameter measured at a particular time point and the second data generated for at least one of the biological parameter measured before the particular time point and the biological parameter measured after the particular time point is below a second threshold, the second data related to the particular time point is excluded; and Varying the second threshold based on the index value. The evaluation system according to claim 1 .

3. The processing device includes: visualizing the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value on a display device; receiving, from a user interface, an instruction to change at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value; The evaluation system according to claim 1 .

4. the processing device receives, from a user interface, feedback on a result of changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value; The evaluation system according to claim 1 .

5. the processing device proposes re-changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value based on the feedback. The evaluation system according to claim 4.

6. The processing device causes a display device to present a simulation of the result of the re-modification. The evaluation system according to claim 5 .

7. the processing device causes a notification device to perform a notification when at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value is changed; The evaluation system according to claim 1 .

8. The evaluation function is provided by an inference model generated through machine learning. The evaluation system according to claim 1 .

9. The processing device includes: receiving feedback from a user interface regarding a result of changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold; and generating fourth data in which information corresponding to the feedback is added to the third data; and a device for retraining the inference model based on the fourth data. The evaluation system according to claim 8.

10. A processing device communicably connected to an inference device that acquires first data corresponding to a measured waveform of a biological parameter of a subject, infers a probability that the value of the biological parameter will be correctly measured, and generates second data in which the first data is assigned the probability, an interface for accepting the second data; a processor that calculates a value score of the measured waveform by applying a predetermined evaluation function to the second data in which the probability exceeds a first threshold value, and generates third data in which the value score is assigned to the second data; Equipped with the evaluation function is defined such that the value score has a positive correlation with the probability and a negative correlation with the time elapsed since the biological parameter was measured; The processor, Obtaining an index value indicating a risk of sudden change in the condition of the subject based on at least one of the third data to which a higher value score is assigned; and changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value based on the index value; Processing unit.

11. A computer program executable by a processor mounted on a processing device that cooperates with an inference device to acquire first data corresponding to a measured waveform of a biological parameter of a subject, infer a probability that the value of the biological parameter will be correctly measured, and generate second data in which the probability is assigned to the first data, When executed, the processing device: Accepting the second data; Calculating a value score of the measured waveform by applying a predetermined evaluation function to the second data in which the probability exceeds a first threshold value; generating third data in which the value score is assigned to the second data; the evaluation function is defined such that the value score has a positive correlation with the probability and a negative correlation with the time elapsed since the biological parameter was measured; The processing device further comprises: Obtaining an index value indicating a risk of sudden change in the condition of the subject based on at least one of the third data to which a higher value score is assigned; and changing at least one of the degree of positive correlation in the evaluation function, the degree of negative correlation in the evaluation function, and the first threshold value based on the index value; Computer program.

Images