Prediction device, method of generating prediction model, and computing device

Abstract

To provide a user interface enabling a user to visually recognize changes in the prediction accuracy of physiological information that may occur in accordance with changes in observation environment of an observed parameter for obtaining physiological information of a subject.SOLUTION: Time series data TS including multiple observed values of an observed parameter that are acquired at different time points from a subject for obtaining physiological information of the subject is received. By inputting the time series data into a prediction model 23, prediction of unobserved values of the observed parameter is performed. A range within which the unobserved values may fall is visualized on an output device. The range is changed in accordance with a time interval between the different time points.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present disclosure relates to an inference device that performs inference on a subject's bioinformation by inputting time-series data including a plurality of observed values ​​obtained from the subject for an observable parameter for obtaining the bioinformation of the subject into an inference model. The present disclosure also relates to a method and a computing device for generating the inference model. [Background technology]

[0002] Inference models generated using machine learning technology to replace or support clinical judgments of medical professionals in clinical settings are known. For example, Non-Patent Document 1 discloses a model that infers specific biological information based on time series data of observed parameters acquired from a living body. Examples of the specific biological information include the future value of the observed parameter (used as a time series prediction model), the probability of in-hospital death (used as a classification model), and the remaining stay time in an intensive care unit (ICU) (used as a regression model). The prediction accuracy of the inference model can be visualized as a confidence interval, for example, as described in Non-Patent Document 2. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] G. Harerimana, et al. "A Multi-Headed Transformer Approach for Predicting the Patient's Clinical Time-Series Variables From Charted Vital Signs", IEEE Access (Volume 10), 105993-106004, October 3, 2022 [Non-Patent Document 2] Y. Du, et al., "Time-series and Machine Learning Scenarios for COVID-19 Infection Prediction", The 2nd International Conference on Biological Engineering and Medical Science, 397-403, June 7, 2022 Summary of the Invention [Problem to be solved by the invention]

[0004] There is a demand for providing a user interface that enables a user to visually recognize changes in the prediction accuracy of a subject's biometric information that may occur in response to changes in the observation environment of the observable parameters used to obtain the biometric information. [Means for solving the problem]

[0005] A first example aspect provided by the present disclosure is an inference device, comprising: an interface that accepts time-series data including a plurality of observation values ​​obtained from a subject at different times for an observable parameter for obtaining biological information of the subject; a processor that infers unobserved values ​​of the observed parameters by inputting the time series data into an inference model and visualizes possible ranges of the unobserved values ​​on an output device; Equipped with The range varies depending on the distance between the different points in time.

[0006] A second aspect provided by the present disclosure is a method for generating an inference model according to the first aspect by a computing device, comprising: obtaining a first observation data set comprising a plurality of observations obtained from a living body for the observed parameter at a first time interval; obtaining a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; storing a difference between an observed value of the observed parameter acquired after the first observation data set and an unobserved value of the observed parameter predicted based on the first observation data set in association with the first time interval; storing a difference between an observed value of the observed parameter acquired after the second observation data set and an unobserved value of the observed parameter predicted based on the second observation data set in association with the second time interval; Machine learning is performed on an inference model so as to estimate a range of possible unobserved values ​​of the observable parameter based on the time intervals during which multiple observed values ​​related to the observable parameter were obtained from the subject.

[0007] A third aspect provided by the present disclosure is a computing device for generating an inference model according to the first aspect, comprising: an interface for receiving a first observation data set including a plurality of observations of the observable parameter obtained from the living body at a first time interval, and a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; a processor that performs machine learning of an inference model based on the first observational data set and the second observational data set; It is equipped with The processor, storing a difference between an observed value of the observed parameter acquired after the first observation data set and an unobserved value of the observed parameter predicted based on the first observation data set in association with the first time interval; storing a difference between an observed value of the observed parameter acquired after the second observation data set and an unobserved value of the observed parameter predicted based on the second observation data set in association with the second time interval; The machine learning is performed such that the range can be inferred based on a time interval during which multiple observations of the observable parameter were obtained from the subject.

[0008] In general, the shorter the time interval at which multiple observed values ​​included in the time series data of the observed parameter are obtained, the higher the prediction accuracy of the unobserved value of the observed parameter inferred by the inference model. On the other hand, for example, armpit temperature is generally measured manually by medical professionals in clinical medical settings. In such cases, the time interval at which the observed value of the observed parameter is obtained may not be constant or may change. In other words, the prediction accuracy of the inferred unobserved value of the observed parameter may change dynamically. However, according to the configurations related to each of the first to third aspects, the range of the unobserved value of the observed parameter inferred based on the input time series data of the observed parameter is visualized, and the range can be dynamically changed according to the time interval at which the observed value was obtained. Therefore, it is possible to provide a user interface that allows a user to visually recognize changes in the prediction accuracy of the unobserved value of the observed parameter that may occur depending on changes in the observation environment of the observed parameter.

[0009] A fourth example aspect provided by the present disclosure is an inference device, comprising: an interface that accepts time-series data including a plurality of observation values ​​obtained from a subject at different times for an observable parameter for obtaining biological information of the subject; a processor that inputs the time series data into an inference model to infer the probability of an event related to the observable parameter occurring, and visualizes the range of the probability on an output device; It is equipped with The range varies depending on the distance between the different points in time.

[0010] A fifth aspect provided by the present disclosure is a method for generating an inference model according to the fourth aspect by a computing device, comprising: obtaining a first observation data set comprising a plurality of observations obtained from a living body for the observed parameter at a first time interval; obtaining a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; storing a difference between a value corresponding to whether or not an event related to the observable parameter has actually occurred since the first observation data set was acquired and a value corresponding to a probability of occurrence of the event predicted based on the first observation data set, in association with the first time interval; storing a difference between a value corresponding to whether or not an event related to the observable parameter has actually occurred since the second observation data set was acquired and a value corresponding to the occurrence probability of the event predicted based on the second observation data set, in association with the second time interval; Machine learning is performed on an inference model so as to infer a possible range of the probability of occurrence of the event based on the time intervals at which multiple observed values ​​related to the observable parameters were obtained from the subject.

[0011] A sixth aspect provided by the present disclosure is a computing device for generating an inference model according to the fourth aspect, comprising: an interface for receiving a first observation data set including a plurality of observations of the observable parameter obtained from the living body at a first time interval, and a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; a processor that performs machine learning of an inference model based on the first observational data set and the second observational data set; Equipped with The processor, storing a difference between a value corresponding to whether or not an event related to the observable parameter has actually occurred since the first observation data set was acquired and a value corresponding to a probability of occurrence of the event predicted based on the first observation data set, in association with the first time interval; storing a difference between a value corresponding to whether or not an event related to the observable parameter has actually occurred since the second observation data set was acquired and a value corresponding to the occurrence probability of the event predicted based on the second observation data set, in association with the second time interval; The machine learning is performed such that the range can be inferred based on a time interval during which multiple observations of the observable parameter were obtained from the subject.

[0012] In general, the shorter the time interval at which multiple observed values ​​included in the time series data of the observable parameter are obtained, the higher the prediction accuracy of the occurrence probability of an event related to the observable parameter inferred by the inference model. On the other hand, for example, armpit temperature is generally measured manually by a medical professional in clinical medical practice. In such a case, the time interval at which the observed values ​​of the observable parameter are obtained may not be constant or may change. In other words, the prediction accuracy of the occurrence probability of the inferred event may change dynamically. However, according to the configurations according to each of the fourth to sixth aspects, the possible range of the occurrence probability of an event related to the observable parameter inferred based on the input time series data of the observable parameter is visualized, and the range can be dynamically changed according to the time interval at which the observed values ​​are obtained. Therefore, it is possible to provide a user interface that allows a user to visually recognize changes in the prediction accuracy of the occurrence probability of an event related to the observable parameter that may occur depending on changes in the observation environment of the observable parameter.

[0013] A seventh example aspect provided by the present disclosure is an inference device, comprising: an interface that accepts time-series data including a plurality of observation values ​​obtained from a subject at different times for an observable parameter for obtaining biological information of the subject; a processor that inputs the time series data into an inference model to infer an observed value of an observable parameter other than the observed parameter, and visualizes a possible range of the observed value on an output device; It is equipped with The range varies depending on the distance between the different points in time.

[0014] An eighth aspect provided by the present disclosure is a method for generating an inference model according to the seventh aspect by a computing device, comprising: obtaining a first observation data set comprising a plurality of observations obtained from a living body for the observed parameter at a first time interval; obtaining a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; storing a difference between an observation value of an observable parameter other than the observable parameter acquired after acquisition of the first observation data set and an observation value of the other observable parameter predicted based on the first observation data set, in association with the first time interval; storing a difference between an observation value of the other observable parameter acquired after acquisition of the second observation data set and an observation value of the other observable parameter predicted based on the second observation data set, in association with the second time interval; Machine learning is performed on an inference model so as to infer a possible range of observed values ​​of another observable parameter based on the time intervals during which multiple observed values ​​related to the observable parameter were obtained from the subject.

[0015] A ninth aspect provided by the present disclosure is a computing device for generating an inference model according to the seventh aspect, comprising: an interface for receiving a first observation data set including a plurality of observations of the observable parameter obtained from the living body at a first time interval, and a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; a processor that performs machine learning of an inference model based on the first observational data set and the second observational data set; It is equipped with The processor, storing a difference between an observation value of an observable parameter other than the observable parameter acquired after acquisition of the first observation data set and an observation value of the other observable parameter predicted based on the first observation data set, in association with the first time interval; storing a difference between an observation value of the other observable parameter acquired after acquisition of the second observation data set and an observation value of the other observable parameter predicted based on the second observation data set, in association with the second time interval; The machine learning is performed such that the range can be inferred based on a time interval during which multiple observations of the observable parameter were obtained from the subject.

[0016] In general, the shorter the time interval at which multiple observed values ​​included in the time series data of the observable parameter are obtained, the higher the prediction accuracy of the observed value of the other observable parameter inferred by the inference model. On the other hand, for example, armpit temperature is generally measured manually by medical professionals in clinical medical settings. In such cases, the time interval at which the observed value of the observable parameter is obtained may not be constant or may change. In other words, the prediction accuracy of the inferred observed value of the other observable parameter may change dynamically. However, according to the configurations according to each of the seventh to ninth aspects, the possible range of the observed value of the other observable parameter inferred based on the input time series data of the observable parameter is visualized, and the range can be dynamically changed according to the time interval at which the observed value of the observable parameter is obtained. Therefore, it is possible to provide a user interface that allows a user to visually recognize changes in the prediction accuracy of the observed value of the other observable parameter that may occur depending on changes in the observation environment of the observable parameter. [Brief description of the drawings]

[0017] [Figure 1] 1 illustrates an example of the functional configuration of an inference system according to one embodiment. [Diagram 2] 2 shows an example of a processing flow performed in the inference system of FIG. 1. [Diagram 3] 2 shows an example of the operation of the inference device of FIG. 1. [Figure 4] 2 illustrates another example of the operation of the inference device of FIG. 1. [Diagram 5] 2 illustrates another example of the operation of the inference device of FIG. 1. [Figure 6] 2 illustrates another example of the operation of the inference device of FIG. 1. [Figure 7] 2 illustrates another example of the operation of the inference device of FIG. 1. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] The embodiments will be described in detail with reference to the accompanying drawings. In each drawing, the scale is appropriately changed so that each element to be described can be recognized.

[0019] 1 illustrates an example of a functional configuration of an inference system 1 according to an embodiment. The inference system 1 includes an inference device 2.

[0020] The inference device 2 according to this example is a device that predicts the value of armpit temperature after (in the future) a certain point in time based on time series data TS of armpit temperature values ​​acquired from a subject S before that point in time. The armpit temperature value is an example of an observed parameter. Other examples of observed parameters include respiratory rate obtained by visually observing chest movement, Glasgow Coma Scale (GCS) obtained by visually confirming the state of consciousness, urine volume measured in a urine collection bag, index values ​​obtained through blood tests (red blood cell count, white blood cell count, blood lactate level, arterial blood oxygen partial pressure, etc.), and blood pressure obtained by non-invasive measurement. The future value of the observed parameter is an example of biological information.

[0021] As used herein, the term "time series of an observable parameter" refers to the change over time in the values ​​of an observable parameter obtained at multiple time points, which may or may not be spaced apart by a regular interval.

[0022] The inference device 2 includes an input interface 21. The input interface 21 is configured as a hardware interface capable of receiving the above-mentioned time-series data TS. The time-series data TS may be input from an appropriate sensor or measuring device attached to the subject S, or may be input via an appropriate input device and user operation. The time-series data TS may be in the form of analog data or digital data, depending on the specifications of the input source. When the time-series data TS is in the form of analog data, the input interface 21 includes an appropriate conversion circuit including an A / D converter.

[0023] The inference device 2 includes a processor 22 and an inference model 23. The inference model 23 is a computer algorithm that can be executed by the processor 22. The inference model 23 is configured to output, as an inference result, a range in which unobserved values ​​of the armpit temperature of the subject S in the future may be, in response to input of time series data TS.

[0024] The inference system 1 includes an output device 3. Examples of the output device 3 include a display device and a printing device. The inference device 2 has an output interface 24. The processor 22 is configured to input the time series data TS accepted by the input interface 21 to the inference model 23, and output, from the output interface 24, control data CT that causes the output device 3 to visualize the inference result output from the inference model 23.

[0025] The output interface 24 is configured as a hardware interface capable of outputting the control data CT. The control data CT may be in the form of analog data or digital data, depending on the specifications of the output device 3. When the control data CT is in the form of analog data, the output interface 24 includes an appropriate conversion circuit including a D / A converter.

[0026] 2 illustrates a first time series data TS1 including a plurality of armpit temperature observation values ​​obtained from a subject S in a first time interval T1. When this first time series data TS1 is input to the inference model 23, a first range RG1 in which unobserved armpit temperature values ​​following the first time series data TS1 may be assumed is inferred and is provided for visualization by the output device 3.

[0027] The inference device 2 in this embodiment is configured to change the possible range of unobserved values ​​of armpit temperature visualized in the output device 3 depending on the time interval between multiple observation values ​​contained in the time series data received by the input interface 21.

[0028] In the second time series data TS2 shown in Fig. 2, a plurality of armpit temperature observation values ​​are acquired in a second time interval T2. The second time interval T2 is longer than the first time interval T1. The second time series data TS2 may be acquired from the same subject as the first time series data TS1, or may be acquired from a different subject.

[0029] This second time series data TS2 is input to the inference model 23, whereby a second range RG2 in which unobserved values ​​of the armpit temperature following the second time series data TS2 may be assumed is inferred and is provided for visualization by the output device 3. The second range RG2 is larger than the first range RG1.

[0030] 1, the inference system 1 includes an inference model generation device 4. An inference model 23 of the inference device 2 is generated through machine learning executed by the inference model generation device 4. The inference model generation device 4 includes an input interface 41, a processor 42, and an output interface 43. The inference model generation device 4 is an example of a computing device.

[0031] With reference to FIG. 2, a method for generating an inference model 23 capable of implementing the above-mentioned operations by the inference model generation device 4 will be described.

[0032] A first observation data set OB1 is received by the input interface 41, the data set comprising a number of observations of armpit temperature taken from a living subject over a first time interval T1.

[0033] The processor 42 is configured to predict, based on a plurality of armpit temperature observations taken from the living body during a first time interval T1, an armpit temperature observation taken after that observation.

[0034] The processor 42 stores the difference between the observed armpit temperature value acquired after the first observation data set OB1 and the unobserved armpit temperature value predicted based on the first observation data set OB1 in association with the first time interval T1. The stored information constitutes a part of the training data TR.

[0035] Similarly, a second observation data set OB2 is received by the input interface 41, the data set including a plurality of observations of the armpit temperature taken from the living subject over a second time interval T2. The second time interval T2 is different from the first time interval T1. In this example, the second time interval T2 is longer than the first time interval T1.

[0036] The processor 42 performs an interpolation such that the time interval of the second observation data set OB2 is resampled at the first time interval T1. The processor 42 then calculates the difference between the observed armpit temperature values ​​acquired after the second observation data set OB2 and the unobserved armpit temperature values ​​predicted based on the resampled second observation data set OB2. The processor 42 then stores the difference in association with the second time interval T2. The stored information constitutes a part of the training data TR.

[0037] Further, a third observation data set OB3 may be received by the input interface 41, the third observation data set OB3 comprising a plurality of observations of the armpit temperature taken from the living body during a third time interval T3. The third time interval T3 is different from the first time interval T1 and the second time interval T2. In this example, the third time interval T3 is longer than the second time interval T2.

[0038] The processor 42 performs an interpolation such that the time interval of the third observation data set OB3 is resampled at the first time interval T1. The processor 42 then calculates the difference between the observed armpit temperature values ​​acquired after the third observation data set OB3 and the unobserved armpit temperature values ​​predicted based on the resampled third observation data set OB3. The processor 42 then stores the difference in association with the third time interval T3. The stored information constitutes a part of the training data TR.

[0039] Furthermore, a similar process may be performed on an observation data set including a plurality of armpit temperature observations taken at different time intervals to generate training data TR.

[0040] The processor 42 executes machine learning of the inference model 23 using the training data TR generated as described above. Specifically, the neural network is made to learn the influence of the time interval at which multiple observed values ​​of armpit temperature were obtained from the subject on the prediction accuracy of the unobserved value of armpit temperature. As a result, the inference model 23 is configured to infer the possible range of the unobserved value of armpit temperature using the time interval at which multiple observed values ​​included in the input armpit temperature time series data TS were obtained as a feature.

[0041] In general, the shorter the time interval at which multiple observed values ​​included in the time series data TS of armpit temperature are obtained, the higher the prediction accuracy of the unobserved value of armpit temperature inferred by the inference model 23. On the other hand, the time interval at which observed values ​​are obtained for an observed parameter such as armpit temperature may not be constant or may change. In other words, the prediction accuracy of the inferred unobserved value of armpit temperature may change dynamically. However, according to the configuration of this embodiment, the range of the unobserved value of armpit temperature inferred based on the input time series data TS of armpit temperature is visualized, and the range can be dynamically changed according to the time interval at which the observed value was obtained. Therefore, it is possible to provide a user interface that allows the user to visually recognize the change in prediction accuracy of the unobserved value of armpit temperature that may occur depending on the change in the observation environment of the armpit temperature.

[0042] In addition, the term "possible range" as used in this specification broadly means "the difference between an inferred value and an actually observed value," and the "possible range" may include the statistical standard error of the difference, the difference between the maximum and minimum values ​​of the difference, the quartile of the difference, the percentile of the difference, the confidence interval of the difference, and the standard deviation of the difference.

[0043] In this example, the observation data set including the observations with larger time intervals may be generated by decimating a portion of the observation data set including the observations with smaller time intervals, e.g., the second observation data set OB2 and / or the third observation data set OB3 may be generated by decimating a portion of the observations included in the first observation data set OB1.

[0044] With this configuration, learning data TR for generating an inference model 23 that can adapt even when the time intervals at which multiple observation values ​​contained in the time series data TS are obtained change can be efficiently generated from a limited number of observation data sets.

[0045] As illustrated in FIG. 3, the range RG of the possible unobserved values ​​of the armpit temperature inferred by the inference model 23 is triangular and is defined by a first boundary line B1, a second boundary line B2, and a third boundary line B3. The first boundary line B1 extends in the vertical direction and corresponds to the final time point te of the period for which prediction is possible. The second boundary line B2 corresponds to the change over time of the maximum value that the unobserved value of the armpit temperature can take from the time point ts when the latest observation value included in the time series data TS is obtained to the time point te. The third boundary line B3 corresponds to the change over time of the minimum value that the unobserved value of the armpit temperature can take from the time point ts to the time point te. In other words, the range RG represents the change over time of the range of the possible unobserved values ​​of the armpit temperature.

[0046] The inference model 23 may be configured to infer the range of possible unobserved values ​​of the armpit temperature as well as the time-dependent change in the most likely value of the unobserved value. The most likely value is an example of a representative value. The processor 22 may be configured to cause the output device 3 to visualize a line L indicating the time-dependent change in the most likely value in addition to the range RG.

[0047] In addition to or instead of the maximum likelihood value, other representative values ​​may be provided for visualization, such as the mean, median, quartiles, confidence limits, etc.

[0048] The inference model 23 may be configured to infer a probability distribution of unobserved values ​​at the prediction time. In this case, the processor 22 may be configured to visualize the inferred probability distribution, as illustrated in Fig. 4. In this example, the visualization is performed using a color map in which values ​​with a high probability that the unobserved value is possible are displayed in darker colors. However, the visualization may be performed in a manner such as a violin plot.

[0049] At least one of the second boundary line B2 and the third boundary line B3 that divide the range RG may be formed by a plurality of straight lines with different gradients. Alternatively, at least one of the second boundary line B2 and the third boundary line B3 may include a nonlinear portion. Therefore, the line L visualized within the range RG may also include a plurality of straight lines with different gradients or a nonlinear portion depending on the shapes of the second boundary line B2 and the third boundary line B3.

[0050] As another example, if the time interval at which the observed values ​​of armpit temperature included in the time series data TS are obtained exceeds a threshold, at least one unobserved value of armpit temperature may be interpolated between the observed values. The threshold for the time interval may be, for example, the shortest time interval assumed in the observation data set set used in the above-mentioned machine learning. In other words, the number of interpolated values ​​may be determined so that the time interval between each observed value and the adjacent interpolated value, and the time interval between the interpolated values ​​(in the case of multiple interpolated values) match the shortest time interval assumed in the observation data set set used in the above-mentioned machine learning.

[0051] In this case, a possible range of the interpolated value can be inferred. The inference model 23 can be configured to infer a possible range of the unobserved value of the armpit temperature based on the range. An example of a specific operation of the inference model 23 configured in this manner will be described with reference to Fig. 5. The time series data TS received by the input interface 21 includes an observation value v1 of the armpit temperature acquired at time t1 and an observation value v2 of the armpit temperature acquired at time t2.

[0052] The processor 22 is configured to generate a plurality of interpolated data sets in different ways based on the time series data TS. In this example, a first interpolated data set IT1, a second interpolated data set IT2, and a third interpolated data set IT3 are generated.

[0053] The first interpolated data set IT1 contains three interpolated armpit temperature values ​​for the period between time t1 and time t2. The interpolation method is based on the Last Observation Carried Forward (LOCF) method, in which the interpolation is performed such that the observed value at a certain time point is maintained until the next observed value is obtained.

[0054] The second interpolated data set IT2 is obtained by interpolating three armpit temperature values ​​for the period between time t1 and time t2. The interpolation method is based on linear interpolation, where a value corresponding to an observation at any time point lying on a straight line connecting two observations is interpolated.

[0055] The third interpolated data set IT3 is an interpolation of three axillary temperature values ​​in the time period between time t1 and time t2. The interpolation method is based on spline interpolation, in which a value corresponding to an observation at any time point is interpolated, which lies on a polynomial curve connecting two observations.

[0056] As long as the interpolation methods are different, the number of multiple interpolated data sets generated by the processor 22 can be determined arbitrarily. Note that the expression "different interpolation methods" used in this specification is not intended to refer only to cases where the types of the interpolation methods are different. As an example, an interpolated data set generated by spline interpolation using a cubic polynomial and an interpolated data set generated by spline interpolation using a quintic polynomial are interpreted as "different interpolation methods" even though they are the same type of "spline interpolation". As another example, multiple interpolated data sets using the same type of interpolation method but with different numbers of values ​​to be interpolated in the same period are also interpreted as "different interpolation methods".

[0057] Next, the processor 22 generates integrated data IG by integrating the first interpolation data set IT1, the second interpolation data set IT2, and the third interpolation data set IT3. Furthermore, the processor 22 identifies a range IR defined by a maximum value and a minimum value among the multiple interpolated values ​​included in the integrated data IG.

[0058] The processor 22 determines a specific representative value within the specified range IR as the interpolated value. Examples of the representative value include a maximum value, a minimum value, an average value, a median value, and a mode value. This provides a data set to which inference by the inference model 23 generated through the above-mentioned machine learning can be applied. The processor 22 infers the range RG within which the armpit temperature value can be taken based on the above-mentioned method, and visualizes it on the output device 3.

[0059] With this configuration, even when it is difficult to obtain time-series data including a sufficient number of observed values ​​for the observable parameters, the above-mentioned user interface can be maintained.

[0060] Next, another configuration example of the inference device 2 will be described with reference to Fig. 6. Only the differences from the configuration described with reference to Fig. 1 to Fig. 5 will be described. The above description is applicable to other matters.

[0061] In this example, the inference model 23 is configured as a classification model that outputs, as an inference result, the occurrence probability of an event that requires an antipyretic analgesic, in response to input of time-series data TS including multiple observed values ​​of armpit temperature acquired from the subject S. The event that requires an antipyretic analgesic is an example of an event related to the observed parameter.

[0062] Other examples of the probability of occurrence of an event related to an observable parameter inferred by a classification model include the probability of occurrence of an event requiring respiratory management using a ventilator for an input of time series data including intermittently observed observations such as respiratory rate, heart rate, arterial blood oxygen saturation, etc.; the probability of developing acute kidney injury for an input of time series data including regularly measured urine volume; and the probability of in-hospital mortality for an input of time series data including intermittently observed observations such as respiratory rate, heart rate, blood pressure, etc.

[0063] The occurrence probability of an event related to an observable parameter may be inferred only at a specific time point, or may form time-series data including a plurality of inferred values ​​as illustrated in Fig. 6. The inferred occurrence probability value may or may not be used for visualization processing.

[0064] As illustrated in FIG. 6, the processor 22 visualizes on the output device 3 the range RG of the probability of occurrence of an event requiring an antipyretic analgesic, which is inferred by the inference model 23. The black circle in the figure represents the maximum likelihood value of the occurrence probability. The maximum likelihood value is an example of a representative value. Other examples of representative values ​​include the mean, median, quartile, and confidence limit value. The display of the representative value may be omitted.

[0065] The inference model 23 that performs such operations is generated by the inference model generating device 4 in a manner similar to that described with reference to FIG.

[0066] A first observation data set OB1 is received by the input interface 41, the data set comprising a number of observations of armpit temperature taken from a living subject over a first time interval T1.

[0067] The processor 42 is configured to infer, based on a plurality of observations of armpit temperature acquired from the living subject over a first time interval T1, a probability of an event occurring since that acquisition that requires an antipyretic and analgesic drug.

[0068] The processor 42 stores, in association with the first time interval T1, a difference between a value corresponding to whether or not an event requiring an antipyretic analgesic has actually occurred since the first observation data set OB1 was acquired and the occurrence probability of the event inferred based on the first observation data set OB1. As an example, the value corresponding to the occurrence of the event is set to 100%, and the value corresponding to the absence of the event is set to 0%. The stored information constitutes a part of the training data TR.

[0069] Similarly, a second observation data set OB2 is received by the input interface 41, the data set including a plurality of observations taken from the living subject over a second time interval T2 for the armpit temperature.

[0070] The processor 42 performs an interpolation so that the time interval of the second observation data set OB2 is resampled at the first time interval T1. Furthermore, the processor 42 calculates a difference between a value corresponding to whether or not an event requiring an antipyretic analgesic has actually occurred since the acquisition of the second observation data set OB2 and the occurrence probability of the event inferred based on the resampled second observation data set OB2. The processor 42 then stores the difference in association with the second time interval T2. As an example, the value corresponding to the occurrence of the event is set to 100%, and the value corresponding to the absence of the event is set to 0%. The stored information constitutes a part of the training data TR.

[0071] Additionally, a third observation data set OB3 may be received by the input interface 41, the data set comprising a plurality of observations taken from the living subject over a third time interval T3 of armpit temperature.

[0072] The processor 42 performs an interpolation so that the time interval of the third observation data set OB3 is resampled at the first time interval T1. Furthermore, the processor 42 calculates a difference between a value corresponding to whether or not an event requiring an antipyretic analgesic has actually occurred since the acquisition of the third observation data set OB3 and the occurrence probability of the event inferred based on the resampled third observation data set OB3. The processor 42 then stores the difference in association with the third time interval T3. As an example, the value corresponding to the event occurring is set to 100%, and the value corresponding to the event not occurring is set to 0%. The stored information constitutes a part of the training data TR.

[0073] If the difference can be defined, the value corresponding to whether or not an event has actually occurred and the inferred probability of occurrence of the event can take on any suitable value other than a percentage.

[0074] Furthermore, a similar process may be performed on an observation data set including a plurality of armpit temperature observations taken at different time intervals to generate training data TR.

[0075] The processor 42 executes machine learning of the inference model 23 using the training data TR generated as described above. Specifically, the neural network is made to learn the influence of the time interval at which multiple observed values ​​of armpit temperature were acquired from the subject on the prediction accuracy of the occurrence probability of an event requiring an antipyretic analgesic. As a result, the inference model 23 is configured to infer the possible range of the occurrence probability of an event requiring an antipyretic analgesic, using the time interval at which multiple observed values ​​included in the input time series data TS of armpit temperature were acquired as a feature.

[0076] In general, the shorter the time interval at which the multiple observed values ​​included in the time series data TS of the armpit temperature are acquired, the higher the prediction accuracy of the occurrence probability of an event that requires an antipyretic analgesic inferred by the inference model 23. On the other hand, the time interval at which the observed values ​​are acquired for an observed parameter such as armpit temperature may not be constant or may change. In other words, the prediction accuracy of the inferred occurrence probability may change dynamically. However, according to the configuration of this embodiment, the possible range of the occurrence probability of an event that requires an antipyretic analgesic, which is inferred based on the input time series data TS of the armpit temperature, is visualized, and the range can be dynamically changed according to the time interval at which the observed values ​​are acquired. Therefore, it is possible to provide a user interface that allows the user to visually recognize the change in the prediction accuracy of the occurrence probability that may occur depending on the change in the observation environment of the armpit temperature.

[0077] Next, a further example of a configuration of the inference device 2 will be described with reference to Fig. 7. Only the differences from the configuration described with reference to Fig. 1 to Fig. 5 will be described. The above description is applicable to other matters.

[0078] In this example, the inference model 23 is configured as a regression model that outputs an observation value of the deep body temperature of the subject S as an inference result in response to an input of time series data TS including multiple observation values ​​of armpit temperature acquired from the subject S. The deep body temperature is an example of another observable parameter.

[0079] Other examples of observed parameters whose observed values ​​are inferred by regression models include cardiac index for input of time series data including intermittently observed heart rate, non-invasive blood pressure, and other observed values; blood creatinine for input of time series data including periodically measured urine volume; and remaining stay in an intensive care unit (ICU) for input of time series data including periodically measured blood test results and intermittently observed body temperature, respiratory rate, heart rate, and blood pressure.

[0080] The observed values ​​of the other observable parameters may be inferred only at a particular time point, or may form a time series including multiple inferred values ​​as illustrated in Figure 7. The inferred observed values ​​may or may not be subjected to visualization processing.

[0081] As illustrated in FIG. 7, the processor 22 visualizes on the output device 3 the range RG of possible intracranial pressure observation values ​​inferred by the inference model 23. The black circle in the figure represents the most likely value of the observation value. The most likely value is an example of a representative value. Other examples of representative values ​​include the mean, median, quartile, and confidence limit value. The display of the representative value may be omitted.

[0082] The inference model 23 that performs such operations is generated by the inference model generating device 4 in a manner similar to that described with reference to FIG.

[0083] A first observation data set OB1 is received by the input interface 41, the data set comprising a number of observations of armpit temperature taken from a living subject over a first time interval T1.

[0084] The processor 42 is configured to infer, based on a plurality of armpit temperature observations taken from the living body during a first time interval T1, subsequent core body temperature observations.

[0085] The processor 42 stores the difference between the core body temperature observations acquired after the first observation data set OB1 was acquired and the inferred observations based on the first observation data set OB1 in association with the first time interval T1. The stored information constitutes a part of the training data TR.

[0086] Similarly, a second observation data set OB2 is received by the input interface 41, the data set including a plurality of observations taken from the living subject over a second time interval T2 for the armpit temperature.

[0087] The processor 42 performs an interpolation such that the time interval of the second observation data set OB2 is resampled at the first time interval T1. The processor 42 then calculates the difference between the core body temperature observations acquired after the acquisition of the second observation data set OB2 and the inferred observations based on the resampled second observation data set OB2. The processor 42 then stores the difference in association with the second time interval T2. The stored information constitutes part of the training data TR.

[0088] Additionally, a third observation data set OB3 may be received by the input interface 41, the data set comprising a plurality of observations taken from the living subject over a third time interval T3 of armpit temperature.

[0089] The processor 42 performs an interpolation such that the time intervals of the third observation data set OB3 are resampled at the first time interval T1. The processor 42 then calculates the difference between the core body temperature observations acquired after the acquisition of the third observation data set OB3 and the inferred observations based on the resampled third observation data set OB3. The processor 42 then stores the difference in association with the third time interval T3. The stored information constitutes part of the training data TR.

[0090] Furthermore, a similar process may be performed on an observation data set including a plurality of armpit temperature observations taken at different time intervals to generate training data TR.

[0091] The processor 42 executes machine learning of the inference model 23 using the training data TR generated as described above. Specifically, the neural network is made to learn the influence of the time interval at which multiple armpit temperature observations were obtained from the subject on the prediction accuracy of the deep body temperature observation. As a result, the inference model 23 is configured to infer the possible range of the deep body temperature observation value using the time interval at which multiple observation values ​​included in the input armpit temperature time series data TS were obtained as a feature.

[0092] In general, the shorter the time interval at which the multiple observation values ​​included in the time series data TS of the armpit temperature are acquired, the higher the prediction accuracy of the observation value of the deep body temperature inferred by the inference model 23. On the other hand, the time interval at which the observation value is acquired for an observed parameter such as the armpit temperature may not be constant or may change. In other words, the prediction accuracy of the inferred observation value may change dynamically. However, according to the configuration of this embodiment, the range of the deep body temperature observation value inferred based on the input time series data TS of the armpit temperature is visualized, and the range can be dynamically changed according to the time interval at which the observation value is acquired. Therefore, it is possible to provide a user interface that allows the user to visually recognize the change in prediction accuracy of the observation value of the deep body temperature that may occur depending on the change in the observation environment of the armpit temperature.

[0093] Each of the processor 22 of the inference device 2 and the processor 42 of the inference model generating device 4 having the various functions described above can be realized by a general-purpose microprocessor that operates 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-transient 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 via a communication network and then installed in the general-purpose memory. In this case, the external server is an example of a non-transient computer-readable medium that stores a computer program.

[0094] Each of the processors 22 and 42 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 non-transitory computer-readable medium that stores a computer program. Each of the processors 22 and 42 may also be realized by a combination of a general-purpose microprocessor and a dedicated integrated circuit.

[0095] The configurations referred to above are merely examples for facilitating understanding of the present disclosure. Each configuration example may be appropriately modified or combined with other configuration examples within the scope of the present disclosure.

[0096] In the above embodiment, the inference device 2 and the inference model generation device 4 are described as devices independent of each other. In this case, the input interface 21 and the output interface 43 are configured as hardware interfaces. However, both devices may be realized as different functional modules in a single device. In this case, both interfaces may be realized as software interfaces.

[0097] The inference model 23 does not necessarily have to be installed in the inference device 2. Although not shown in the figure, the inference device 2 may be communicatively connected to an external server device via a communication network. In this case, the inference model 23 may be installed in the external server device.

[0098] In the above embodiment, the inference model generating device 4 generates the inference model 23 through machine learning using a neural network. However, the inference model 23 may be generated through other machine learning algorithms. Examples of other machine learning algorithms include linear regression, logistic regression, decision tree, random forest, and support vector machine. [Explanation of symbols]

[0099] 2: inference device, 21: input interface, 22: processor, 23: inference model, 3: output device, 4: inference model generation device, 41: input interface, 42: processor, OB1: first observation data set, OB2: second observation data set, RG: range, S: subject, T1: first time interval, T2: second time interval, TR: learning data, TS: time series data

Claims

1. an interface that accepts time-series data including a plurality of observation values ​​obtained from a subject at different times for an observable parameter for obtaining biological information of the subject; a processor that infers unobserved values ​​of the observed parameters by inputting the time series data into an inference model and visualizes possible ranges of the unobserved values ​​on an output device; It is equipped with The range varies depending on the interval between the different points in time. Reasoning device.

2. an interface that accepts time-series data including a plurality of observation values ​​obtained from a subject at different times for an observable parameter for obtaining biological information of the subject; a processor that inputs the time series data into an inference model to infer the probability of an event related to the observable parameter occurring, and visualizes the range of the probability on an output device; It is equipped with The range varies depending on the interval between the different points in time. Reasoning device.

3. an interface that accepts time-series data including a plurality of observation values ​​obtained from a subject at different times for an observable parameter for obtaining biological information of the subject; a processor that inputs the time series data into an inference model to infer an observed value of an observable parameter other than the observed parameter, and visualizes a possible range of the observed value on an output device; It is equipped with The range varies depending on the interval between the different points in time. Reasoning device.

4. the processor causes the representative value of the range to be visualized on the output device. An inference device according to any one of claims 1 to 3.

5. The processor causes the distribution of the inference results to be visualized on the output device. An inference device according to any one of claims 1 to 3.

6. The inference model infers the range using the interval as a feature. An inference device according to any one of claims 1 to 3.

7. The inference model infers the range based on a range of possible values ​​that are interpolated for the plurality of observed values ​​during a time period corresponding to the interval. An inference device according to any one of claims 1 to 3.

8. A method for generating the inference model according to claim 1 by a computing device, comprising: obtaining a first observation data set comprising a plurality of observations obtained from a living body for the observed parameter at a first time interval; obtaining a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; storing a difference between an observed value of the observed parameter acquired after the first observation data set and an unobserved value of the observed parameter predicted based on the first observation data set in association with the first time interval; storing a difference between an observed value of the observed parameter acquired after the second observation data set and an unobserved value of the observed parameter predicted based on the second observation data set in association with the second time interval; performing machine learning of an inference model so as to estimate a possible range of unobserved values ​​of the observable parameter based on a time interval during which a plurality of observed values ​​related to the observable parameter are obtained from a subject; How to generate an inference model.

9. A method for generating the inference model according to claim 2 by a computing device, comprising: obtaining a first observation data set comprising a plurality of observations obtained from a living body for the observed parameter at a first time interval; obtaining a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; storing a difference between a value corresponding to whether or not an event related to the observable parameter has actually occurred since the first observation data set was acquired and a value corresponding to a probability of occurrence of the event predicted based on the first observation data set, in association with the first time interval; storing a difference between a value corresponding to whether or not an event related to the observable parameter has actually occurred since the second observation data set was acquired and a value corresponding to a probability of occurrence of the event predicted based on the second observation data set, in association with the second time interval; performing machine learning of an inference model so as to infer a possible range of the probability of occurrence of the event based on a time interval during which a plurality of observed values ​​related to the observable parameter are obtained from the subject; How to generate an inference model.

10. A method for generating the inference model according to claim 3 by a computing device, comprising: obtaining a first observation data set comprising a plurality of observations obtained from a living body for the observed parameter at a first time interval; obtaining a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; storing a difference between an observation value of an observable parameter other than the observable parameter acquired after acquisition of the first observation data set and an observation value of the other observable parameter predicted based on the first observation data set, in association with the first time interval; storing a difference between an observation value of the other observable parameter acquired after acquisition of the second observation data set and an observation value of the other observable parameter predicted based on the second observation data set, in association with the second time interval; performing machine learning of an inference model so as to infer a possible range of observed values ​​of the other observable parameter based on a time interval during which a plurality of observed values ​​related to the observable parameter were obtained from the subject; How to generate an inference model.

11. the second observation data set is generated by thinning out a portion of the observation values ​​included in the first observation data set. A method for generating an inference model according to any one of claims 8 to 10.

12. A computing device for generating the inference model according to claim 1, comprising: an interface for receiving a first observation data set including a plurality of observations of the observable parameter obtained from the living body at a first time interval, and a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; a processor that performs machine learning of an inference model based on the first observational data set and the second observational data set; It is equipped with The processor, storing a difference between an observed value of the observed parameter acquired after the first observation data set and an unobserved value of the observed parameter predicted based on the first observation data set in association with the first time interval; storing a difference between an observed value of the observed parameter acquired after the second observation data set and an unobserved value of the observed parameter predicted based on the second observation data set in association with the second time interval; performing the machine learning so as to infer the range based on a time interval during which a plurality of observations related to the observable parameter are obtained from the subject; Computing device.

13. A computing device for generating the inference model according to claim 2, comprising: an interface for receiving a first observation data set including a plurality of observations of the observable parameter obtained from the living body at a first time interval, and a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; a processor that performs machine learning of an inference model based on the first observational data set and the second observational data set; It is equipped with The processor, storing a difference between a value corresponding to whether or not an event related to the observable parameter has actually occurred since the first observation data set was acquired and a value corresponding to a probability of occurrence of the event predicted based on the first observation data set, in association with the first time interval; storing a difference between a value corresponding to whether or not an event related to the observable parameter has actually occurred since the second observation data set was acquired and a value corresponding to a probability of occurrence of the event predicted based on the second observation data set, in association with the second time interval; performing the machine learning so as to infer the range based on a time interval during which a plurality of observations related to the observable parameter are obtained from the subject; Computing device.

14. A computing device for generating the inference model according to claim 3, comprising: an interface for receiving a first observation data set including a plurality of observations of the observable parameter obtained from the living body at a first time interval, and a second observation data set including a plurality of observations of the observable parameter obtained from the living body at a second time interval different from the first time interval; a processor that performs machine learning of an inference model based on the first observational data set and the second observational data set; It is equipped with The processor, storing a difference between an observation value of an observable parameter other than the observable parameter acquired after acquisition of the first observation data set and an observation value of the other observable parameter predicted based on the first observation data set, in association with the first time interval; storing a difference between an observation value of the other observable parameter acquired after acquisition of the second observation data set and an observation value of the other observable parameter predicted based on the second observation data set, in association with the second time interval; performing the machine learning so as to infer the range based on a time interval during which a plurality of observations related to the observable parameter are obtained from the subject; Computing device.

Images