Information processing device, information processing method, and program

Abstract

The system optimally performs predictions about the target individuals while referring to input data, including multivariate time series data. [Solution] The information processing device includes an acquisition unit that acquires input data including multivariate time series data relating to one or more subjects; a structuring unit that generates structured data by structuring the multivariate time series data into a graph; a calculation unit that calculates feature vectors for the one or more subjects by referring to at least the structured data; and a prediction unit that makes predictions relating to the subjects by referring to the feature vectors.

Description

Technical Field

[0004] ,

[0001] The present invention relates to an information processing apparatus, an information processing method, and a program.

Background Art

[0002] There is known a technique called Embedding Propagation (EP) that learns embedding (vectorization) of data, instances, etc. based on a graph structure that represents the relationship between the data and instances (Non-Patent Document 1). Also, there is known a technique for performing outcome prediction such as in-hospital mortality, length of stay, and discharge destination of inpatients using Embedding Propagation (Non-Patent Document 2).

Prior Art Documents

Non-Patent Documents

[0003]

Non-Patent Document 1

Non-Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] Not only in the aforementioned embedding propagation, but in any technology that makes predictions about subjects such as patients, it is common to refer to multiple time series data (also called multivariate time series data). On the other hand, multivariate time series data can include various time series data with different acquisition frequencies. If, for example, missing values ​​are interpolated by time synchronization for each time series data for such multivariate time series data, the original data distribution will be distorted due to the small number of effective values, and as a result, suitable predictions about the subject will be hindered.

[0005] This disclosure has been made in view of the above-mentioned problems, and one exemplary purpose is to provide a technology that can suitably perform predictions about subjects while referring to input data including multivariate time series data. [Means for solving the problem]

[0006] An information processing device relating to an exemplary aspect of this disclosure includes: acquisition means for acquiring input data including multivariate time series data relating to one or more subjects; structuring means for generating structured data by structuring the multivariate time series data into a graph; calculation means for calculating feature vectors of the one or more subjects with reference to at least the structured data; and prediction means for making predictions relating to the subjects with reference to the feature vectors.

[0007] An example of an information processing method relating to this disclosure includes one or more processors acquiring input data including multivariate time series data relating to one or more subjects, generating structured data by graphing the multivariate time series data, calculating feature vectors for the one or more subjects with reference to at least the structured data, and making predictions about the subjects with reference to the feature vectors.

[0008] An exemplary aspect of the present disclosure is a program that causes a computer to function as an information processing device, wherein the computer functions as: an acquisition means for acquiring input data including multivariate time series data relating to one or more subjects; a structuring means for generating structured data by structuring the multivariate time series data into a graph; a calculation means for calculating feature vectors of the one or more subjects by referring to at least the structured data; and a prediction means for making predictions relating to the subjects by referring to the feature vectors. [Effects of the Invention]

[0009] According to an illustrative aspect of this disclosure, one exemplary effect is that predictions regarding the subject can be suitably made while referring to input data including multivariate time series data. [Brief explanation of the drawing]

[0010] [Figure 1] This is a block diagram showing the configuration of the information processing device related to this disclosure. [Figure 2] This is a flowchart showing the flow of the information processing method related to this disclosure. [Figure 3] This is a block diagram showing the configuration of the information processing system related to this disclosure. [Figure 4] This is a diagram illustrating the processing in the information processing device related to this disclosure. [Figure 5] This is a diagram illustrating the processing in the information processing device related to this disclosure. [Figure 6] This is a block diagram showing an example configuration of the information processing device related to this disclosure. [Figure 7] This is a flowchart showing the flow of the information processing method related to this disclosure. [Figure 8] This is a block diagram showing an example configuration of the information processing device related to this disclosure. [Figure 9] This is a flowchart showing the flow of the information processing method related to this disclosure. [Figure 10] This is a diagram illustrating the processing in the information processing device related to this disclosure. [Figure 11] This is a block diagram showing an example configuration of the information processing device related to this disclosure. [Figure 12] This is a block diagram showing an example configuration of the information processing device related to this disclosure. [Figure 13] This is a block diagram showing the configuration of the information processing system related to this disclosure. [Figure 14] This is a block diagram showing the hardware configuration of the information processing device related to this disclosure. [Modes for carrying out the invention]

[0011] The following are examples of embodiments of the present invention. However, the present invention is not limited to the exemplary embodiments shown below, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technical means employed in each of the exemplary embodiments shown below may also be included in the scope of the present invention. Furthermore, embodiments obtained by appropriately omitting some of the technical means employed in each of the exemplary embodiments shown below may also be included in the scope of the present invention. In addition, the effects mentioned in each of the exemplary embodiments shown below are examples of effects that can be expected in that exemplary embodiment and do not define the scope of the present invention. That is, embodiments that do not produce the effects mentioned in each of the exemplary embodiments shown below may also be included in the scope of the present invention.

[0012] [First Embodiment] A first exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. This exemplary embodiment is the basic form for each of the exemplary embodiments described later. The scope of application of each technical means adopted in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technical means adopted in this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical problems occur. Furthermore, each technical means shown in the drawings referenced to explain this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical problems occur.

[0013] (Configuration of Information Processing Apparatus 1) The configuration of the information processing apparatus 1 according to this exemplary embodiment will be described with reference to FIG. 1. FIG. 1 is a block diagram showing the configuration of the information processing apparatus 1. The information processing apparatus 1 may also be referred to as a prediction apparatus, a learning apparatus, or the like. As shown in FIG. 1, the information processing apparatus 1 includes an acquisition unit 11, a structuring unit 12, a calculation unit 13, and a prediction unit 14.

[0014] (Acquisition Unit 11) The acquisition unit 11 acquires input data including multivariate time-series data regarding one or more subjects. Here, the multivariate time-series data may include, as an example, a plurality of time-series data. More specifically, the multivariate time-series data may include time-series data regarding one variable and time-series data regarding another variable. The number of time-series data included in the multivariate time-series data does not limit this exemplary embodiment.

[0015] (Structuring Unit 12) The structuring unit 12 generates structured data by graph-structuring the multivariate time-series data acquired by the acquisition unit 11. Here, "graph-structuring" refers to, as an example, generating structured data in a graph format. The graph format refers to a data format including a plurality of nodes and one or more links (edges) connecting the nodes. The structured data in the graph format may also be referred to as graph-structured data.

[0016] The structuring unit 12 may generate, as an example, a graph having nodes corresponding to each of the plurality of data values included in the multivariate time-series data and edges with weights according to the time differences between the plurality of data values as the structured graph.

[0017] (Calculation Unit 13) The calculation unit 13 calculates feature vectors (also called feature quantities or feature vectors) for one or more subjects, referring at least to the structured data generated by the structuring unit 12. The specific processing performed by the calculation unit 13 is not limited to this exemplary embodiment, but as an example, The acquisition unit 11 references the attribute data of the one or more subjects included in the input data it has acquired to generate a graph (also called a property graph or patient graph, etc.) that includes multiple patients, including the one or more subjects, as nodes. • Calculate the feature vectors of the one or more subjects by referring to the generated graph. The following process is performed. Here, the structured data may be referenced in the generation of the graphs (property graph, patient graph), and the structured data may be referenced in the calculation of the feature vectors.

[0018] Furthermore, the calculation unit 13 may be configured to calculate the feature vectors of one or more subjects by performing embedding propagation with reference to the graphs (property graph, patient graph). However, this example is not limited to the exemplary embodiment described herein.

[0019] (Prediction Unit 14) The prediction unit 14 makes predictions about the subject by referring to the feature vector calculated by the calculation unit 13. For example, the prediction unit 14 may be configured to perform processes such as regression analysis or classification by referring to the feature vector calculated by the calculation unit 13, and then use the results of these processes to make predictions about the subject. Alternatively, for example, the prediction unit 14 may perform outcome predictions such as in-hospital mortality rate, length of hospital stay, and discharge destination for one or more subjects by referring to the feature vector. However, these examples are not limited to illustrative embodiments.

[0020] (Effects of Information Processing Device 1) As described above, in the information processing device 1, • Obtain input data that includes multivariate time series data for one or more subjects, • Structured data is generated by structuring the aforementioned multivariate time series data into a graph. - Calculate the feature vectors of the one or more subjects by referring to the structured data, • Predictions regarding the subject are made by referring to the feature vectors mentioned above. The above configuration is adopted. According to the above configuration, structured data is generated by structuring the multivariate time series data into a graph, feature vectors for one or more subjects are calculated by referring to at least the structured data, and predictions about the subjects are made by referring to the feature vectors. For this reason, predictions about subjects can be suitably performed while referring to input data including multivariate time series data.

[0021] (Information processing method S1 flow) Next, the flow of the information processing method S1 according to this exemplary embodiment will be explained with reference to Figure 2. Figure 2 is a flowchart showing the flow of the information processing method S1. As shown in Figure 2, the information processing method S1 includes a step (processing) S11 for acquiring input data, a step (processing) S12 for generating structured data, a step (processing) S13 for calculating feature vectors, and a step (processing) S14 for performing prediction.

[0022] (Step S11) In step S11, the acquisition unit 11 acquires input data that includes multivariate time series data relating to one or more subjects. The specific processing performed by the acquisition unit 11 has been described above and will not be explained here.

[0023] (Step S12) Next, in step S12, the structuring unit 12 generates structured data by structuring the multivariate time series data acquired by the acquisition unit 11 in step S11 into a graph. The specific processing performed by the structuring unit 12 has been described above, so it will not be explained here.

[0024] (Step S13) Next, in step S13, the calculation unit 13 calculates the feature vectors of the one or more subjects, referring at least to the structured data generated by the structuring unit 12 in step S12. The specific processing by the calculation unit 13 has been described above and will not be explained here.

[0025] (Step S14) Next, in step S14, the prediction unit 14 makes a prediction about the subject by referring to the feature vector calculated by the calculation unit 13. The specific processing performed by the prediction unit 14 has been described above, so it will not be explained here.

[0026] (Effects of information processing method S1) As described above, in the information processing method S1, • Obtain input data that includes multivariate time series data for one or more subjects, • Structured data is generated by structuring the aforementioned multivariate time series data into a graph. - Calculate the feature vectors of the one or more subjects by referring to the structured data, • Predictions regarding the subject are made by referring to the feature vectors mentioned above. This configuration is employed. According to the above configuration, the same effect as that of the information processing device 1 is achieved.

[0027] [Second Embodiment] A second exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same function as those described in the above-described exemplary embodiment are denoted by the same reference numerals, and their descriptions are omitted as appropriate. The scope of application of each technology adopted in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology adopted in this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical problems arise. Furthermore, each technology shown in the drawings referenced to describe this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical problems arise.

[0028] (Configuration of Information Processing System 100A) The configuration of the information processing system 100A according to this exemplary embodiment will be described with reference to Figure 3. Figure 3 is a block diagram showing the configuration of the information processing system 100A. As shown in Figure 3, the information processing system 100A comprises an information processing device 1A and a patient data management device 50 connected to the information processing device 1A via a network N. Here, the specific configuration of the network N is not limited to this exemplary embodiment, but as an example, a wireless LAN (Local Area Network), a wired LAN, a WAN (Wide Area Network), a public telephone network, a mobile data communication network, or a combination of these networks can be used.

[0029] In this exemplary embodiment, a patient data management device 50 is given as an example of a configuration that provides input data including multivariate time series data relating to one or more subjects, as described later. However, this is not limited to this exemplary embodiment, and other devices may be used as the configuration that provides the input data.

[0030] (Patient data management device 50) The patient data management device 50 manages data including multivariate time-series data relating to one or more subjects. For example, the patient data management device 50 manages data including multivariate time-series data relating to one or more subjects. • Multivariate time series data for one or more subjects • Attribute data for one or more subjects The system manages the following. Specific examples of multivariate time series data and attribute data will be described later. As an example, this data may be included in the electronic medical records of one or more subjects (patients), and the patient data management device 50 may be configured to function as an electronic medical record management device. The above data managed by the patient data management device 50 is referenced by the information processing device 1A as input data IN, which will be described later.

[0031] (Configuration of Information Processing Device 1A) Next, the configuration of the information processing device 1A according to this exemplary embodiment will be described with reference to Figure 3. Figure 3 is a block diagram showing the configuration of the information processing device 1A. As shown in Figure 3, the information processing device 1A includes a control unit 10A, a storage unit 20A, a communication unit 30, and an input / output unit 40. The information processing device 1A may also be referred to as a prediction device or a learning device.

[0032] (Communications Section 30) The communication unit 30 communicates with external devices of the information processing device 1A via the network N. For example, the communication unit 30 transmits data supplied from the control unit 10A to an external device, or supplies data received from an external device to the control unit 10A. More specifically, the communication unit 30 receives data from the patient data management device 50, • Multivariate time series data for one or more subjects TD • Attribute data AD for one or more subjects The data is acquired and stored in the storage unit 20A.

[0033] (Input / output section 40) The input / output unit 40 is configured to include at least one of the following input / output devices: a keyboard, mouse, display, printer, touch panel, etc. Alternatively, the input / output unit 40 may be configured to have input / output devices such as a keyboard, mouse, display, printer, touch panel, etc. connected to it. In this configuration, the input / output unit 40 receives various types of information from the connected input device to the information processing device 1A. The input / output unit 40 also outputs various types of information to the connected output device under the control of the control unit 10A. An interface such as USB (Universal Serial Bus) can be used as the input / output unit 40.

[0034] (Storage unit 20A) The storage unit 20A stores various data referenced by the control unit 10A, as well as various data generated by the control unit 10A. For example, the storage unit 20A stores: • Input data IN • Structured Data Group SDG • Property Graph PG Feature vector group FVG • Output information OUT • Predictive Model PM These are stored there.

[0035] Here, the input data IN is obtained from the patient data management device 50. • Multivariate time series data for one or more subjects TD • Attribute data AD for one or more subjects It includes [the following]. Multivariate time series data (TD) is sometimes simply referred to as time series data (TD).

[0036] Multivariate time series data (TD) includes multiple time series data. For example, these multiple time series data include measured values ​​(data values) of multiple data items relating to one or more subjects (patients). For instance, it may include time series data such as changes in body temperature and heart rate during hospitalization for one or more subjects (patients).

[0037] To give a more specific example, the multivariate time series data TD for subject 1 is: • Measurement data of heart rate (HR) of subject 1 (also simply referred to as HR data) • Platelet measurement data (also simply referred to as Platelets data) for Subject 1 • Measurement data of arterial blood carbon dioxide partial pressure (PaCO2) of subject 1 (also simply referred to as PaCO2 data) • Data on the measurement of aspartate aminotransferase (AST) in subject 1 (also simply referred to as AST data). It may include the following. Also, the multivariate time series data TD for subject 2 is • HR data for participant 2 • Platelets data for participant 2 • PaCO2 data for subject 2 • AST data for subject 2 These may include, for example, data measured at different times or frequencies for each subject and data item.

[0038] Attribute data AD is data that shows the attributes of each subject, and includes, for example, age, gender, and disease name.

[0039] The structured data group SDG is data generated by the structuring unit 12 described later, and includes structured data SD for one or more subjects (patients). Specific examples of structured data SD will be described later.

[0040] The property graph PG is a graph generated by the calculation unit 13 (described later) by referring to the structured data set SDG. The feature vector set FVG contains one or more feature vectors FV calculated by the calculation unit 13 by referring to the property graph PG. Feature vectors FV are sometimes also called feature quantities FV or feature vectors FV. Specific examples of property graph PG and feature vectors FV will be described later.

[0041] The output information OUT includes the prediction results from the prediction unit 14, which will be described later. Specific examples of the output information OUT will be described later. The prediction model PM is a model used for prediction by the prediction unit 14. For example, it is a model that takes one or more feature vectors FV calculated by the calculation unit 13 as input and performs outcome prediction for the target person. Specific examples of the prediction model PM will be described later.

[0042] The prediction model PM is a model used by the prediction unit 14, which will be described later, and is learned by the learning unit 15, for example. A specific example of the prediction model PM will be described later.

[0043] (Control Unit 10A) As shown in Figure 3, the control unit 10A includes an acquisition unit 11, a structuring unit 12, a calculation unit 13, a prediction unit 14, and a learning unit 15.

[0044] (Acquisition part 11) The acquisition unit 11 acquires input data IN, which includes multivariate time series data TD for one or more subjects. Specific examples of multivariate time series data TD have been described above, so a redundant explanation will be omitted.

[0045] (structuring part 12) The structuring unit 12 generates structured data SD by graph structuring the multivariate time series data TD acquired by the acquisition unit 11. Here, "graph structuring" refers, for example, to generating structured data in graph format, similar to the exemplary embodiment 1. Graph format refers to a data format that includes multiple nodes and one or more links (edges) connecting the nodes.

[0046] Here, the structuring unit 12 may generate a directed graph containing directed edges as structured data SD, or an undirected graph containing undirected edges. Furthermore, some attribute value may be attached to each node or edge. Structured data in graph format is sometimes referred to as graph-structured data.

[0047] Figure 4 is a diagram illustrating an example of processing by the structuring unit 12. The upper part of Figure 4 shows multivariate time series data TD for a certain subject that the structuring unit 12 references. As shown in the upper part of Figure 4, the multivariate time series data TD includes, as an example, time series data for each of the data items mentioned above (HR data, Platelets data, PaCO2 data, AST data). Also, as shown in the upper part of Figure 4, these data items were measured at different timings or frequencies for each data item.

[0048] The structuring unit 12 generates structured data SD from such multivariate time series data TD, as shown in the lower part of Figure 4 as an example. As shown in the lower part of Figure 4, although the structured data SD generated by the structuring unit 12 includes the concept of the passage of time, time synchronization between data items is not necessarily maintained. Because the structured data SD generated by the structuring unit 12 has a very flexible data structure, it is possible to represent the multivariate time series data TD as a graph that has only significant values ​​as nodes, without taking each data item, as an example. The structuring unit 12 may generate a structured graph having nodes corresponding to each of the multiple data values ​​included in the multivariate time series data TD, and edges weighted according to the time difference (difference in measurement time) between the multiple data values. More specifically, in the example shown in the lower part of Figure 4, each edge included in the structured data SD may be weighted according to the time difference (at the time of measurement) between the data values. A more specific example of processing by the structuring unit 12 will be described later.

[0049] (Calculation part 13) The calculation unit 13 calculates the feature vector FV of one or more subjects by referring at least to the structured data SD generated by the structuring unit 12. Figure 5 is a diagram illustrating an example of processing by the calculation unit 13.

[0050] As illustrated in Figure 5, the calculation unit 13, as an example, • Attribute data AD of one or more subjects (patients) included in the input data IN acquired by the acquisition unit 11 • Structured data SD for each subject generated by the structuring unit 12 Refer to this and generate a property graph (patient graph) PG. Here, a property graph PG is, for example, • Multiple nodes of the same type, each having one or more attribute values, • One or more links connecting the aforementioned multiple nodes This is a graph composed of the following. More specifically, each node is associated with each patient, and the one or more attribute values ​​mentioned above may include attribute values ​​included in the attribute data AD.

[0051] The calculation unit 13 then calculates the feature vector FV for each of the one or more subjects by referring to the generated property graph (patient graph) PG. Here, the specific example of calculating the feature vector FV by referring to the property graph PG is not limited to this exemplary embodiment, but as an example, it is performed by embedding propagation.

[0052] In the embedding propagation performed by the calculation unit 13, the features of each node included in the property graph PG are learned based on the graph structure of the property graph PG. In other words, in embedding propagation, the method of embedding each node included in the property graph PG into the feature space (the method of vectorization, the method of generating the feature vector FV) is learned (unsupervised learning) based on the graph structure of the property graph PG. The relationships between nodes in the property graph PG are carried over directly in embedding propagation, and the relationships between instances (between nodes) are maintained in the learned embedding data. Embedding propagation can represent combinations of different representation formats such as categories, floats, free text, and images (in other words, multimodal data) in a single consistent embedding space (feature space). Furthermore, embedding propagation can generate more useful embeddings for missing values ​​than simple imputation methods.

[0053] (Prediction Unit 14) The prediction unit 14 makes predictions about the subject (patient) by referring to the feature vector FV calculated by the calculation unit 13. As an example, the prediction unit 14 inputs the feature vector FV calculated by the calculation unit 13 to a trained prediction model PM and makes predictions about the subject (patient) using the output of the prediction model PM.

[0054] The prediction unit 14 may, for example, be configured to perform processes such as regression analysis and classification using a prediction model PM that references the feature vector FV calculated by the calculation unit 13, and then use the results of these processes to make predictions about the subject. Alternatively, the prediction unit 14 may, for example, perform outcome predictions such as in-hospital mortality, length of hospital stay, and discharge destination for one or more subjects using a prediction model PM that references the feature vector FV. These prediction results may include information that supports decision-making by users (doctors, healthcare professionals, etc.). Therefore, the prediction unit 14 may be described as performing outcome predictions about the subject (patient) in order to support decision-making by users (doctors, healthcare professionals, etc.).

[0055] (Learning Section 15) The learning unit 15 trains the prediction model PM used by the prediction unit 14. For example, the learning unit 15 uses training data, which includes feature vectors FV and the correct labels assigned to the feature vectors FV, ​​to train the prediction model PM.

[0056] As described above, in the information processing device 1A, • Obtain input data IN which includes multivariate time series data TD for one or more subjects. • Structured data SD is generated by structuring the aforementioned multivariate time series data TD into a graph. - Calculate the feature vector FV of the one or more subjects by referring at least to the structured data SD, • Predictions regarding the subject are made by referring to the feature vector FV. The above configuration is adopted. According to the above configuration, structured data is generated by structuring the multivariate time series data into a graph, feature vectors for one or more subjects are calculated by referring to at least the structured data, and predictions about the subjects are made by referring to the feature vectors. For this reason, predictions about subjects can be suitably performed while referring to input data including multivariate time series data.

[0057] In particular, in the medical field, time-series data tends to be irregularly sampled and very sparse. If time synchronization is performed for each time-series data point and missing value interpolation is carried out as in conventional techniques, the original data distribution may be distorted due to the small number of valid values. In the information processing device 1A configured as described above, the multivariate time-series data is graphed and then referenced in the calculation of the feature vector FV, thereby suppressing such problems of data distribution distortion.

[0058] Furthermore, in the information processing device 1A, as an example, the feature vector FV is calculated using embedding propagation. As described above, the structured data SD obtained by graphing the multivariate time series data can be suitably referenced in embedding propagation as one of the multimodal data.

[0059] Thus, the information processing device 1A can suitably perform multimodal data processing including multivariate time series data and suitably perform predictions regarding the target individuals.

[0060] (Specific configuration example 1) Below, with reference to Figure 6, a more specific configuration example 1 of the information processing device 1A will be described. This example is a configuration example for the learning phase of the information processing device 1A. However, this is not an exhaustive example.

[0061] As shown in Figure 6, first, multivariate time series data TD for one or more patients is supplied from the patient data DB (storage unit 20A) to the time series data graph structuring unit 12 (the structuring unit 12 described above). The time series data graph structuring unit 12 generates structured data SD for each patient by structuring the multivariate time series data TD for each patient into a graph, and supplies the generated structured data SD to the calculation unit 13.

[0062] In this example, the calculation unit 13 includes a patient-to-patient graph construction unit (patient graph generation unit) 131, a patient data encoding unit 132, and a graph patient feature vector calculation unit 133. The structured data SD for each patient generated by the time-series data graph structuring unit 12 is supplied to the patient data encoding unit 132.

[0063] On the other hand, the patient-to-patient graph construction unit 131 obtains attribute data AD relating to one or more patients from the patient data DB (storage unit 20A), and generates a first patient-to-patient graph (first patient graph, first property graph) PG1 by referring to the obtained attribute data AD. Here, the patient graph PG1 is, as an example, • Multiple nodes of the same type, each having one or more attribute values, • One or more links connecting the aforementioned multiple nodes This is a graph composed of the following. More specifically, each node is associated with each patient, and the one or more attribute values ​​may include attribute values ​​included in the attribute data AD. The first patient graph PG1 generated by the patient-to-patient graph construction unit 131 is supplied to the patient data encoding unit 132.

[0064] The specific process for generating patient graphs by the patient graph construction unit 131 is not limited to this example, but as an example, a configuration may be used in which patient attribute data AD is used to construct a patient graph with edges drawn between similar patients using kNN clustering.

[0065] The patient data encoding unit 132 encodes the structured data SD for each patient generated by the structuring unit 12 and the first patient graph PG1 generated by the inter-patient graph construction unit 131. The specific configuration of the encoding unit 132 is not limited to this example, but a configuration capable of concoding graph data, such as a GNN (Graph Neural Network) or GCN (Graph Convolutional Network), is employed. In the encoded patient graph, each node is associated with encoded attribute values ​​and encoded structured data SD. The encoded patient graph is also referred to as the second patient graph PG2 or the second property graph PG2.

[0066] The graph patient feature vector calculation unit 133 calculates the feature vector FV for each patient by referring to the second patient graph PG2 generated by the patient data encoding unit 132. As an example, the graph patient feature vector calculation unit 133 calculates the feature vector FV for each patient by performing the embedding propagation described above.

[0067] The patient data encoding unit 132 and the graph patient feature vector calculation unit 133 are sometimes collectively referred to as the feature vector calculation unit. The feature vector calculation unit can be described as having a configuration that calculates feature vectors for one or more patients by referring to the structured data SD and the patient graph (first patient graph PG1 or second patient graph PG2).

[0068] The patient outcome prediction unit 14 (the prediction unit 14 described above) refers to the feature vectors FV of one or more patients and performs outcome prediction for those patients using the prediction model PM. The prediction results from the prediction unit 14 are supplied to the learning unit 15.

[0069] The learning unit 15 performs machine learning on the prediction model PM by referring to the correct labels for the one or more subjects and the prediction results from the prediction unit 14. More specifically, the learning unit 15 updates the parameters of the prediction model PM so that the prediction results from the prediction unit 14 approach the correct labels. The updated parameters are stored in the storage unit 20A.

[0070] (Specific example of processing 1) Next, with reference to Figure 7, a more specific example of processing 1 for the information processing device 1A will be described. This example corresponds to the configuration example 1 described above and is a configuration example for the learning phase of the information processing device 1A. However, this is not limited to this example. Figure 7 is a flowchart showing the processing flow related to this example.

[0071] (Step S11) First, in step S11, the acquisition unit 11 acquires input data IN, which includes multivariate time series data TD for one or more patients. The acquired input data IN is referenced by the time series data graph structuring unit 12 (structuring unit 12) and the calculation unit 13.

[0072] (Step S12) Next, in step S12, the time series data graph structuring unit 12 (structuring unit 12) generates structured data SD for each patient by structuring the multivariate time series data TD for each patient into a graph.

[0073] (Step S131) Next, in step S131, the patient graph construction unit 131 refers to the attribute data AD of each patient included in the input data IN and generates a graph (first patient graph PG1) based on the similarity of the patients.

[0074] (Step S132) Next, in step S132, the patient data encoding unit 132 defines encoders corresponding to the modality of the data to be referenced. As an example, the patient data encoding unit 132 defines encoders corresponding to attribute data AD and structured data SD. Then, the patient data encoding unit 132 generates a second patient graph PG2 by encoding the attribute data AD and structured data SD using the defined encoders.

[0075] (Step S1331) Next, in step S1331, the graph patient feature vector calculation unit 133 trains the encoder by performing embedding propagation. This training may be repeated multiple times. Then, the graph patient feature vector calculation unit 133 updates the second patient graph PG2 using the trained encoder.

[0076] (Step S1332) Next, in step S1332, the graph patient feature vector calculation unit 133 calculates feature vectors for one or more patients by referring to the updated second patient graph PG2.

[0077] (Step S14) Next, in step S14, the patient outcome prediction unit 14 (prediction unit 14) refers to the feature vector FV of one or more patients and performs outcome prediction for those patients using the prediction model PM.

[0078] (Step S15) Next, in step S15, the prediction model PM is trained using machine learning by referring to the correct labels for the one or more subjects and the prediction results from the prediction unit 14. The parameters of the trained prediction model PM are stored in the storage unit 20A.

[0079] (Specific configuration example 2 (at the time of inference)) Next, with reference to Figure 8, a more specific configuration example 2 of the information processing device 1A will be described. This example is a configuration example for the inference phase of the information processing device 1A. However, this is not an exhaustive example.

[0080] As shown in Figure 8, this configuration example differs from the above-described Configuration Example 1 in that it does not include a learning unit 15, but the other configurations are the same as those of Configuration Example 1. In this example, the time series data graph structuring unit 12 (structuring unit 12) refers to the time series data TD relating to the patient to be predicted. In this example, the inter-patient graph construction unit 131 refers to the attribute data AD relating to the patient to be predicted. The patient outcome prediction unit 14 (prediction unit 14) then uses the above-described trained prediction model PM to perform outcome prediction for the patient to be predicted.

[0081] (Specific example of processing 2) Next, with reference to Figure 9, a more specific processing example 2 of the information processing device 1A will be described. This example corresponds to the configuration example 2 described above and is a configuration example for the inference phase of the information processing device 1A. However, this is not limited to this example. Figure 9 is a flowchart showing the processing flow related to this example.

[0082] (Step S11) First, in step S11, the acquisition unit 11 acquires input data IN, which includes multivariate time series data TD related to new patients (patients to be predicted). The acquired input data IN is referenced by the time series data graph structuring unit 12 (structuring unit 12) and the calculation unit 13.

[0083] (Steps S12~S1332) Since the processing in steps S12 to S1332 is the same as in processing example 1, redundant explanations will be omitted.

[0084] (Step S14) In step S14, the patient outcome prediction unit 14 (prediction unit 14) refers to the feature vector FV of one or more patients and performs outcome prediction for those patients using the trained prediction model PM.

[0085] Figure 10 shows an example of a prediction performed in step S14. In the example shown in Figure 10, the patient outcome prediction unit 14 performs regression analysis by referring to the feature vector FV to predict the length of hospital stay for a particular patient. In the example shown in Figure 10, classification is performed by referring to the feature vector FV to predict whether or not an ICU is necessary for other patients. These predictions are examples of outcome predictions for one or more patients.

[0086] More specifically, the information processing device 1A receives an instruction (query) from the user via the input / output unit 40 to predict the length of hospital stay for a certain patient. Based on this instruction, it refers to the attribute data AD and time series data TD of the patient and performs embedding propagation through the process described above. Then, the patient outcome prediction unit 14 makes a prediction regarding the length of hospital stay for the patient through regression analysis referencing the patient's feature vector. In the example shown in Figure 10, "30 days" is derived as the predicted length of hospital stay for the patient. The information processing device 1A may also visually present the prediction result to the user via the input / output unit 40. For example, "The estimated length of hospital stay for patient A is 30 days." Output information such as OUT may be displayed on the display of the input / output unit 40.

[0087] As another example, the information processing device 1A receives an instruction (query) from the user via the input / output unit 40 to predict whether a certain patient needs to be admitted to the ICU. Based on this instruction, it refers to the attribute data AD and time-series data TD of the patient and performs embedding propagation through the process described above. Then, the patient outcome prediction unit 14 makes a prediction regarding the need for ICU for the patient based on a class classification that refers to the feature vector of the patient. In the example shown in Figure 10, "False (not needed)" is derived as the prediction regarding the need for ICU for the patient. The information processing device 1A may also visually present the prediction result to the user via the input / output unit 40. For example, "We predict that ICU admission will not be necessary for patient B." Output information such as OUT may be displayed on the display of the input / output unit 40.

[0088] (Specific configuration example 3 (during learning)) Below, with reference to Figure 11, a more specific configuration example 3 of the information processing device 1A will be described. This example is a configuration example for the learning phase of the information processing device 1A. However, this is not an exhaustive example.

[0089] As shown in Figure 11, the configuration in this example differs from that of Configuration Example 1 described above in the data flow related to the calculation unit 13, but is otherwise the same as Configuration Example 1. Below, we will mainly explain the differences from Configuration Example 1, and explanations that overlap with Configuration Example 1 may be omitted.

[0090] As shown in Figure 11, in this example, first, multivariate time series data TD for one or more patients is supplied from the patient data DB (storage unit 20A) to the time series data graph structuring unit 12 (the structuring unit 12 described above). The time series data graph structuring unit 12 generates structured data SD for each patient by structuring the multivariate time series data TD for each patient into a graph, and supplies the generated structured data SD to the inter-patient graph construction unit 131.

[0091] On the other hand, the patient-to-patient graph construction unit 131 obtains attribute data AD relating to one or more patients from the patient data DB (storage unit 20A), and generates a first patient-to-patient graph (first patient graph, first property graph) PG1 by referring to the obtained attribute data AD and structured data SD.

[0092] The patient data encoding unit 132 encodes the structured data SD for each patient generated by the structuring unit 12 and the first patient graph PG1 generated by the inter-patient graph construction unit 131. In this example as well, the encoded patient graph is also referred to as the second patient graph PG2 or the second property graph PG2.

[0093] The graph patient feature vector calculation unit 133, the patient outcome prediction unit 14 (prediction unit 14), and the learning unit 15 are the same as in Configuration Example 1, so redundant explanations will be omitted.

[0094] (Specific configuration example 4 (at the time of inference)) Next, with reference to Figure 12, a more specific configuration example 4 of the information processing device 1A will be described. This example is a configuration example for the inference phase of the information processing device 1A. However, this is not an exhaustive example.

[0095] As shown in Figure 12, this configuration example differs from the above-described configuration example 3 in that it does not include a learning unit 15, but the other configurations are the same as those of the above-described configuration example 3. In this example, the time series data graph structuring unit 12 (structuring unit 12) refers to the time series data TD relating to the patient to be predicted. In this example, the inter-patient graph construction unit 131 refers to the attribute data AD relating to the patient to be predicted. The patient outcome prediction unit 14 (prediction unit 14) then uses the above-described trained prediction model PM to perform outcome prediction for the patient to be predicted.

[0096] (Additional notes regarding the structural section 12) Regarding the structuring unit 12 (time-series data graph structuring unit 12), a more specific processing example will be described. As described above, the structuring unit 12 generates graph-structured data SD (hereinafter also simply referred to as a graph) which includes multiple nodes and one or more links (edges) connecting the nodes. Here, as an example, each node represents a measurement event for each time period, and each node is associated with a sensor value and sensor type. On the other hand, the edges are set according to certain rules, as an example. Here, as such rules, • Create edges between sensor value nodes that are within a certain time window. • Draw weighted edges using weights inversely proportional to the measurement interval. Based on domain knowledge in the medical field, edges may be connected or removed depending on the type of sensor (for example, sensor measurements related to the circulatory system may be grouped together and connected as an edge, while edges are not connected to other systems). Furthermore, the structuring unit 12 may be configured to generate the above graph in a data-driven manner using a predetermined algorithm (for example, the RAINDROP algorithm).

[0097] [Third Embodiment] A third exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same function as those described in the above-described exemplary embodiments are denoted by the same reference numerals, and their descriptions are omitted as appropriate. The scope of application of each technology adopted in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology adopted in this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical hindrance occurs. Furthermore, each technology shown in the drawings referenced to describe this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical hindrance occurs.

[0098] (Configuration of Information Processing System 100B) The configuration of the information processing system 100B according to this exemplary embodiment will be described with reference to Figure 13. Figure 13 is a block diagram showing the configuration of the information processing system 100B. As shown in Figure 13, the information processing system 100B comprises an information processing device 1A, a patient data management device 50 connected to the information processing device 1A via a network N, and an in-hospital management device 60. The information processing device 1A and the patient data management device 50 are the same as in exemplary embodiment 2 and have already been described, so redundant explanations will be omitted.

[0099] The in-hospital management device 60 manages hospital beds and ICUs (optimizing usage schedules), as well as managing inventory of pharmaceuticals and other supplies and making ordering suggestions.

[0100] The information processing device 1A performs outcome predictions for one or more patients by executing the processes described in the exemplary embodiment 2, and the in-hospital management device 60 manages hospital beds, ICUs, pharmaceuticals, etc., related to the one or more patients by referring to the outcome predictions.

[0101] For example, if the information processing device 1A predicts that the use of an ICU is unnecessary as an outcome prediction for a certain patient, and the in-hospital management device 60 acquires this prediction result, the in-hospital management device 60 may optimize the bed and ICU utilization schedule based on the prediction result. The in-hospital management device 60 may then visually present output information based on the optimization results to the user (a doctor or medical professional). Furthermore, such presentation may include advice to support the user's decision-making (for example, a suggestion such as, "There is availability in the ICU, so how about moving patient C to the ICU?") in the output information. Furthermore, as an example, if the information processing device 1A predicts the risk of pressure ulcer development as an outcome prediction for a certain patient, and the in-hospital management device 60 acquires the prediction result, the in-hospital management device 60 may control the system to optimize the pressure distribution of the pressure ulcer prevention air mattress based on the prediction result. Alternatively, as an example, if the information processing device 1A predicts the risk of pneumonia as an outcome prediction for a certain patient, and the in-hospital management device 60 acquires the prediction result, the in-hospital management device 60 may control the adjustment of the electric bed angle to optimize it based on the prediction result.

[0102] [Examples of implementation using software] Some or all of the functions of the information processing devices 1,1A (hereinafter also referred to as "each of the above devices") may be implemented by hardware such as integrated circuits (IC chips) or by software.

[0103] In the latter case, each of the above devices is implemented, for example, by a computer that executes instructions for a program, which is software that realizes each function. An example of such a computer (hereinafter referred to as Computer C) is shown in Figure 14. Figure 14 is a block diagram showing the hardware configuration of Computer C, which functions as each of the above devices.

[0104] Computer C comprises at least one processor C1 and at least one memory C2. Memory C2 stores a program P that causes computer C to operate as each of the above-mentioned devices. In computer C, processor C1 reads program P from memory C2 and executes it, thereby realizing each of the above-mentioned devices.

[0105] For processor C1, for example, a CPU (Central Processing Unit), GPU (Graphic Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating Point Number Processing Unit), PPU (Physics Processing Unit), TPU (Tensor Processing Unit), quantum processor, microcontroller, or a combination thereof can be used. For memory C2, for example, flash memory, HDD (Hard Disk Drive), SSD (Solid State Drive), or a combination thereof can be used.

[0106] Computer C may also be equipped with RAM (Random Access Memory) for loading program P at runtime and for temporarily storing various data. Furthermore, computer C may be equipped with communication interfaces for sending and receiving data with other devices. Additionally, computer C may be equipped with input / output interfaces for connecting input / output devices such as keyboards, mice, displays, and printers.

[0107] Furthermore, program P can be recorded on a non-temporary, tangible recording medium M that is readable by computer C. Such a recording medium M could be, for example, tape, disk, card, semiconductor memory, or programmable logic circuitry. Computer C can acquire program P via such a recording medium M. Program P can also be transmitted via a transmission medium. Such a transmission medium could be, for example, a communication network or broadcast waves. Computer C can also acquire program P via such a transmission medium.

[0108] [Additional Note A] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims.

[0109] (Note A1) An acquisition means for acquiring input data including multivariate time series data relating to one or more subjects, A structuring means for generating structured data by structuring the aforementioned multivariate time series data into a graph, A computation means for calculating the feature vectors of one or more subjects with reference to at least the structured data, Prediction means that make predictions about the subject by referring to the feature vector An information processing device equipped with the following features.

[0110] (Appendix A2) The aforementioned input data includes attribute data of one or more subjects, The aforementioned calculation means is A patient graph generation means that generates a patient graph that includes one or more patients as nodes, by referring to the attribute data, A feature vector calculation means that calculates feature vectors for one or more subjects by referring to the structured data and the patient graph. It is equipped with The information processing device described in Appendix A1.

[0111] (Note A3) The feature vector calculation means is Encoded data is generated by encoding the structured data and the data included in the patient graph. The feature vectors of the one or more subjects are calculated by referring to the encoded data. The information processing device described in Appendix A2.

[0112] (Note A4) The aforementioned input data includes attribute data of one or more subjects, The aforementioned calculation means is A patient graph generation means generates a patient graph that includes one or more patients as nodes, by referring to the attribute data and the structured data, A feature vector calculation means that calculates feature vectors for one or more subjects by referring to the patient graph, It is equipped with The information processing device described in Appendix A1.

[0113] (Note A5) The structuring means is, A node corresponding to each of the multiple data values ​​included in the aforementioned multivariate time series data, Edges weighted according to the time difference between the aforementioned multiple data values ​​and A graph having the above is generated as the structured data. An information processing device as described in any one of the appendices A2 to A4.

[0114] (Note A6) The feature vector calculation means calculates the feature vectors of one or more subjects by performing embedding propagation that references at least the patient graph. The information processing device described in Appendix A5.

[0115] (Note A7) The prediction means performs outcome predictions regarding the subject in order to support the user's decision-making. An information processing device as described in any one of the appendices A1 to A6.

[0116] (Note A8) A learning means that causes the prediction means to perform machine learning by referring to training data which includes feature vectors and the correct labels attached to those feature vectors. It also has An information processing device as described in any one of the appendices A1 to A7.

[0117] [Additional Notes B] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims.

[0118] (Note B1) At least one processor performs an acquisition process to acquire input data, which includes multivariate time series data relating to one or more subjects, The at least one processor performs a structuring process that generates structured data by structuring the multivariate time series data into a graph, The at least one processor performs a computation process that calculates feature vectors for one or more subjects by referring to at least the structured data, The at least one processor performs a prediction process that makes a prediction about the subject by referring to the feature vector. An information processing method that includes this.

[0119] (Note B2) The aforementioned input data includes attribute data of one or more subjects, In the calculation process described above, the at least one processor, A patient graph generation process that generates a patient graph containing multiple patients, including the one or more subjects, as nodes, by referring to the attribute data, A feature vector calculation process that calculates feature vectors for one or more subjects by referring to the structured data and the patient graph. Execute The information processing method described in Appendix B1.

[0120] (Note B3) In the feature vector calculation process, the at least one processor is: Encoded data is generated by encoding the structured data and the data included in the patient graph. The feature vectors of the one or more subjects are calculated by referring to the encoded data. The information processing method described in Appendix B2.

[0121] (Note B4) The aforementioned input data includes attribute data of one or more subjects, In the calculation process described above, the at least one processor, A patient graph generation process that generates a patient graph that includes multiple patients, including the one or more subjects, as nodes, by referring to the attribute data and the structured data, A feature vector calculation process that calculates feature vectors for one or more subjects by referring to the patient graph, and Execute The information processing method described in Appendix B1.

[0122] (Note B5) In the structuring process, the at least one processor, A node corresponding to each of the multiple data values ​​included in the aforementioned multivariate time series data, Edges weighted according to the time difference between the aforementioned multiple data values ​​and A graph having the above is generated as the structured data. The information processing method described in any one of the appendices B2 to B4.

[0123] (Note B6) In the feature vector calculation process, the at least one processor calculates the feature vectors of the one or more subjects by performing embedding propagation that references at least the patient graph. The information processing method described in Appendix B5.

[0124] (Note B7) In the prediction process, the at least one processor performs outcome predictions regarding the subject in order to support the user's decision-making. The information processing method described in any one of the appendices B1 to B6.

[0125] (Note B8) The at least one processor performs a learning process that uses machine learning to perform the prediction process by referring to training data which includes feature vectors and the correct labels attached to those feature vectors. It also includes The information processing method described in any one of the appendices B1 through B7.

[0126] [Additional Note C] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims.

[0127] (Note C1) A program that makes a computer function as an information processing device. The aforementioned computer, An acquisition means for acquiring input data including multivariate time series data relating to one or more subjects, A structuring means for generating structured data by structuring the aforementioned multivariate time series data into a graph, A computation means for calculating the feature vectors of one or more subjects with reference to at least the structured data, Prediction means that make predictions about the subject by referring to the feature vector An information processing program that functions as such.

[0128] (Note C2) The aforementioned input data includes attribute data of one or more subjects, The aforementioned calculation means is A patient graph generation process that generates a patient graph containing multiple patients, including the one or more subjects, as nodes, by referring to the attribute data, A feature vector calculation process that calculates feature vectors for one or more subjects by referring to the structured data and the patient graph. Execute The information processing program described in Appendix C1.

[0129] (Note C3) The feature vector calculation means is Encoded data is generated by encoding the structured data and the data included in the patient graph. The feature vectors of the one or more subjects are calculated by referring to the encoded data. The information processing program described in Appendix C2.

[0130] (Note C4) The aforementioned input data includes attribute data of one or more subjects, The aforementioned calculation means is A patient graph generation process that generates a patient graph that includes multiple patients, including the one or more subjects, as nodes, by referring to the attribute data and the structured data, A feature vector calculation process that calculates feature vectors for one or more subjects by referring to the patient graph, and Execute The information processing program described in Appendix C1.

[0131] (Note C5) The structuring means is, A node corresponding to each of the multiple data values ​​included in the aforementioned multivariate time series data, Edges weighted according to the time difference between the aforementioned multiple data values ​​and A graph having the above is generated as the structured data. An information processing program described in any one of the appendices C2 to C4.

[0132] (Appendix C6) The feature vector calculation means calculates the feature vectors of one or more subjects by performing embedding propagation that references at least the patient graph. The information processing program described in Appendix C5.

[0133] (Note C7) The prediction means performs outcome predictions regarding the subject in order to support the user's decision-making. An information processing program described in any one of the appendices C1 to C6.

[0134] (Note C8) The aforementioned computer, A learning means that causes the prediction means to perform machine learning by referring to training data which includes feature vectors and the correct labels attached to those feature vectors. To make it function even better An information processing program described in any one of the appendices C1 through C7.

[0135] [Additional Note D] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims.

[0136] (Note D1) It comprises at least one processor, and the at least one processor is An acquisition process that acquires input data including multivariate time series data for one or more subjects, A structuring process that generates structured data by structuring the aforementioned multivariate time series data into a graph, A computation process that calculates feature vectors for one or more subjects by referring to the structured data, A prediction process that makes predictions about the subject by referring to the feature vector: An information processing device that performs the following actions.

[0137] The information processing device may also include memory. Furthermore, the memory may store a program that causes at least one processor to execute each of the aforementioned processes.

[0138] (Note D2) The aforementioned input data includes attribute data of one or more subjects, In the calculation process described above, the at least one processor, A patient graph generation process that generates a patient graph containing multiple patients, including the one or more subjects, as nodes, by referring to the attribute data, A feature vector calculation process that calculates feature vectors for one or more subjects by referring to the structured data and the patient graph. Execute The information processing device described in Appendix D1.

[0139] (Note D3) In the feature vector calculation process, the at least one processor is: Encoded data is generated by encoding the structured data and the data included in the patient graph. The feature vectors of the one or more subjects are calculated by referring to the encoded data. The information processing device described in Appendix D2.

[0140] (Note D4) The aforementioned input data includes attribute data of one or more subjects, In the calculation process described above, the at least one processor, A patient graph generation process that generates a patient graph that includes multiple patients, including the one or more subjects, as nodes, by referring to the attribute data and the structured data, A feature vector calculation process that calculates feature vectors for one or more subjects by referring to the patient graph, and Execute The information processing device described in Appendix D1.

[0141] (Note D5) In the structuring process, the at least one processor, A node corresponding to each of the multiple data values ​​included in the aforementioned multivariate time series data, Edges weighted according to the time difference between the aforementioned multiple data values ​​and A graph having the above is generated as the structured data. An information processing device as described in any one of the appendices D2 to D4.

[0142] (Note D6) In the feature vector calculation process, the at least one processor calculates the feature vectors of the one or more subjects by performing embedding propagation that references at least the patient graph. The information processing device described in Appendix D5.

[0143] (Note D7) In the prediction process, the at least one processor performs outcome predictions regarding the subject in order to support the user's decision-making. An information processing device as described in any one of the appendices D1 to D6.

[0144] (Note D8) The aforementioned at least one processor, A learning process that uses machine learning to perform the prediction process by referring to training data that includes feature vectors and the correct labels assigned to those feature vectors. Perform further An information processing device as described in any one of the appendices D1 to D7.

[0145] [Additional Note E] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims.

[0146] (Note E1) A program that makes a computer function as an information processing device. To the aforementioned computer, An acquisition process that acquires input data including multivariate time series data for one or more subjects, A structuring process that generates structured data by structuring the aforementioned multivariate time series data into a graph, A computation process that calculates feature vectors for one or more subjects by referring to the structured data, A prediction process that makes predictions about the subject by referring to the feature vector: A non-temporary recording medium that stores an information processing program that executes that program. [Explanation of symbols]

[0147] 1,1A ···Information Processing Device 11 ... Acquisition unit (acquisition means) 12...Structuring part (structuring means) 13...Calculation unit (calculation means) 14. Prediction unit (prediction means) 15 ···Learning Department (Learning Methods) 131 ···Patient graph construction unit (Patient graph generation means) 132 ···Patient data encoding unit (feature vector calculation means) 133 ···Graph Patient Feature Vector Calculation Unit (Feature Vector Calculation Means)

Claims

1. An acquisition means for acquiring input data including multivariate time series data relating to one or more subjects, A structuring means for generating structured data by structuring the aforementioned multivariate time series data into a graph, A calculation means for calculating the feature vectors of one or more subjects by referring to the structured data, Prediction means that make predictions about the subject by referring to the feature vector An information processing device equipped with the following features.

2. The aforementioned input data includes attribute data of one or more subjects. The aforementioned calculation means is A patient graph generation means that generates a patient graph that includes one or more patients as nodes, by referring to the attribute data, A feature vector calculation means that calculates feature vectors for one or more subjects by referring to the structured data and the patient graph. It is equipped with The information processing apparatus according to claim 1.

3. The feature vector calculation means is Encoded data is generated by encoding the structured data and the data included in the patient graph. The feature vectors of the one or more subjects are calculated by referring to the encoded data. The information processing apparatus according to claim 2.

4. The aforementioned input data includes attribute data of one or more subjects. The aforementioned calculation means is A patient graph generation means generates a patient graph that includes one or more patients as nodes, by referring to the attribute data and the structured data, A feature vector calculation means that calculates feature vectors for one or more subjects by referring to the patient graph. It is equipped with The information processing apparatus according to claim 1.

5. The structuring means is, A node corresponding to each of the multiple data values ​​included in the aforementioned multivariate time series data, Edges weighted according to the time difference between the aforementioned multiple data values ​​and A graph having the above is generated as the structured data. The information processing apparatus according to any one of claims 2 to 4.

6. The feature vector calculation means calculates the feature vectors of one or more subjects by performing embedding propagation that references at least the patient graph. The information processing apparatus according to claim 5.

7. The prediction means performs outcome predictions regarding the subject in order to support the user's decision-making. The information processing apparatus according to claim 6.

8. A learning means that causes the prediction means to perform machine learning by referring to training data which includes feature vectors and the correct labels attached to those feature vectors. It also has The information processing apparatus according to any one of claims 1 to 4.

9. One or more processors, Obtain input data that includes multivariate time series data for one or more subjects, The aforementioned multivariate time series data is graphed to generate structured data, The process involves calculating the feature vectors of one or more subjects by referring at least to the aforementioned structured data, By referring to the aforementioned feature vector, a prediction is made regarding the subject. An information processing method that includes this.

10. A program that makes a computer function as an information processing device. The aforementioned computer, An acquisition means for acquiring input data including multivariate time series data relating to one or more subjects, A structuring means for generating structured data by structuring the aforementioned multivariate time series data into a graph, A calculation means for calculating the feature vectors of one or more subjects by referring to the structured data, Prediction means that make predictions about the subject by referring to the feature vector A program that makes it function as such.

Images