Information processing device

Abstract

To provide an information processing device that can improve the reliability of therapeutic effect by a treatment method to be output.SOLUTION: An information processing device uses a treatment method output model that outputs a treatment method according to a state of a patient and comprises: a therapeutic effect prediction model building section which builds a therapeutic effect prediction model for comparing the therapeutic effect when the treatment method is used and the therapeutic effect when the treatment method is not used, and including the treatment method output model as a constituent element; and a model adjustment section which generates a prediction model group by deactivating weighting coefficients of the therapeutic effect prediction model on the basis of tendency scores when the treatment method is used and when the treatment method is not used, and adjusts the treatment method output model, which is a constituent element of the therapeutic effect prediction model, so that dispersion of the therapeutic effect output from the prediction model group is within a predetermined range.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present invention relates to an information processing device that uses a model that outputs a treatment method according to the condition of a patient, and in particular to a technique for updating the model. [Background technology]

[0002] Digital therapy (DTx) is an information processing device that uses a model constructed by machine learning of known data and outputs a treatment method according to the patient's condition. Because DTx can collect data in real time after being launched, it is possible to update the model based on the collected data.

[0003] Patent Document 1 discloses that a predicted value of a clinical parameter is obtained from a pair of the most accurate model selected from multiple models according to the predictor and the predictor, and the model is updated according to the obtained predicted value and the measured value of the clinical parameter to improve accuracy. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Special Publication No. 2016-519807 Summary of the Invention [Problem to be solved by the invention]

[0005] However, Patent Document 1 does not consider improving the reliability of the model. That is, even if an updated model outputs a highly effective treatment method, the treatment effect may vary due to a slight change in the predictor.

[0006] SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide an information processing device that can improve the reliability of the therapeutic effect of an output therapeutic method. [Means for solving the problem]

[0007] In order to achieve the above object, the present invention provides an information processing device that uses a treatment method output model that outputs a treatment method in accordance with the condition of a patient, and is characterized by comprising: a treatment effect prediction model construction unit that constructs a treatment effect prediction model that includes the treatment method output model as a component and compares the treatment effect when the treatment method is used with that when it is not used; and a model adjustment unit that generates a group of prediction models by inactivating weighting coefficients of the treatment effect prediction model based on the propensity scores when the treatment method is used and when it is not used, and adjusts the treatment method output model, which is a component of the treatment effect prediction model, so that the variance of the treatment effect output from the group of prediction models is within a predetermined range. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide an information processing device that can improve the reliability of the therapeutic effect of the outputted therapeutic method. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is an overall configuration diagram of an information processing apparatus. [Figure 2] FIG. 1 is a diagram showing an example of a DTx intervention model that outputs a treatment method according to the condition of a patient. [Figure 3] FIG. 2 is a diagram illustrating an example of a functional block according to the first embodiment. [Figure 4] FIG. 1 is a diagram illustrating an example of a processing flow according to the first embodiment. [Figure 5] FIG. 1 is a diagram showing an example of a processing flow for constructing a therapeutic effect prediction model. [Figure 6] FIG. 1 is a diagram showing an example of a treatment effect prediction model. [Figure 7] FIG. 10 is a diagram showing an example of the flow of a process for adjusting a DTx intervention model. [Figure 8] FIG. 10 is a diagram showing an example of a therapeutic effect prediction model used to adjust a DTx intervention model. [Figure 9] FIG. 10 is a diagram showing an example of a result display screen. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment of an information processing device according to the present invention will be described with reference to the accompanying drawings. In the following description and the accompanying drawings, components having the same functional configuration will be designated by the same reference numerals, and redundant description will be omitted. [Example]

[0011] 1 is a diagram showing the hardware configuration of an information processing device 101. The information processing device 101 is configured by a calculation unit 102, a memory 103, a storage unit 104, and a network adapter 105, which are connected via a system bus 106 so as to be capable of transmitting and receiving signals. The information processing device 101 is also connected via a network 109 to an electronic medical record 110 and a digital therapy 111 so as to be capable of transmitting and receiving signals. Furthermore, a display device 107 and an input device 108 are connected to the information processing device 101. Here, "capable of transmitting and receiving signals" refers to a state in which signals can be transmitted and received between each other or from one device to the other, regardless of whether the transmission is electrically or optically wired or wireless.

[0012] The calculation unit 102 is a device that controls the operation of each component, and is specifically a CPU (Central Processing Unit) or an MPU (Micro Processor Unit). The calculation unit 102 loads programs stored in the storage unit 104 and data required for program execution into the memory 103, executes them, and performs various processes on them. The memory 103 stores programs executed by the calculation unit 102 and intermediate progress of calculation processes. The storage unit 104 is a device that stores programs executed by the calculation unit 102 and data required for program execution, and is specifically a HDD (Hard Disk Drive) or SSD (Solid State Drive). The network adapter 105 is used to connect the information processing device 101 to a network 109 such as a LAN, a telephone line, or the Internet. Various data handled by the calculation unit 102 may be transmitted to and received from outside the information processing device 101 via the network 109, such as a LAN (Local Area Network).

[0013] The display device 107 is a device that displays the processing results, etc., of the information processing device 101, and is specifically a liquid crystal display, etc. The input device 108 is an operation device with which an operator issues operation instructions to the information processing device 101, and is specifically a keyboard, mouse, touch panel, etc. The mouse may be another pointing device such as a trackpad or trackball.

[0014] The electronic medical record 110 stores medical data related to a patient, such as data related to the patient's condition and the therapeutic effect of a certain treatment method. The digital therapy 111 has a treatment method output model that outputs a treatment method according to the patient's condition and profile.

[0015] An example of a DTx intervention model, which is a treatment method output model, is explained using Figure 2. The DTx intervention model is a model that outputs a treatment method sd(i) according to the patient profile x(i,j) and patient condition c(i,j), and is a neural network constructed by machine learning of known data. Note that i is an index that identifies the patient, and j is an index that represents the type of patient condition or patient profile, where the patient condition is calorie intake and amount of exercise, and the patient profile is gender, age, etc. The treatment method output from the DTx intervention model is, for example, the type of medication.

[0016] The functional blocks of the first embodiment will be described with reference to Fig. 3. These functional blocks may be configured with dedicated hardware, or may be configured with software that runs on the calculation unit 102. In the following description, a case will be described in which the functional blocks of the first embodiment are configured with software. In the first embodiment, a therapeutic effect prediction model construction unit 301 and a DTx intervention model adjustment unit 302 are provided. Each unit will be described below.

[0017] The therapeutic effect prediction model construction unit 301 constructs a therapeutic effect prediction model that compares the therapeutic effect when the therapeutic method output from the DTx intervention model is used with the therapeutic effect when it is not used.

[0018] The DTx intervention model adjustment unit 302 adjusts the DTx intervention model so as to reduce the variance of the treatment effect when the treatment method output from the DTx intervention model is used.

[0019] An example of the flow of processing executed in the first embodiment will be described step by step with reference to FIG.

[0020] (S401) The therapeutic effect prediction model construction unit 301 constructs a therapeutic effect prediction model.

[0021] An example of the flow of the process executed in S401 will be described step by step with reference to FIG.

[0022] (S501) The therapeutic effect prediction model construction unit 301 acquires the DTx intervention model. For example, the DTx intervention model is transmitted from the digital therapy 111.

[0023] (S502) The treatment effect prediction model construction unit 301 constructs the treatment effect prediction model illustrated in Figure 6 by connecting the DTx intervention model acquired in S501 with three networks: the DTx network, the standard care network, and the shared network. The DTx network outputs a predicted value of the treatment effect when the treatment method output from the DTx intervention model is used. The standard care network outputs a predicted value of the treatment effect when the treatment method output from the DTx intervention model is not used. The shared network outputs parameters shared by the DTx network and the standard care network according to the patient profile x(i,j) and patient condition c(i,j).

[0024] (S503) The treatment effect prediction model construction unit 301 inputs the patient profile x(i, j) and patient condition c(i, j) of the collected data into the treatment effect prediction model and outputs a predicted value of the treatment effect. The predicted value of the treatment effect is output for both cases where the treatment method output from the DTx intervention model is used and where it is not used. Here, in response to heterogeneity in patient background between cases where the treatment method output from the DTx intervention model is used and where it is not used, such as the patient group using the treatment method output from the DTx intervention model being younger, the weighting coefficient of the treatment effect prediction model is inactivated based on the propensity scores for both cases, and the treatment effect prediction model is constructed while correcting for heterogeneity in patient background. The propensity score p(i) is calculated, for example, using the following equation.

[0025]

number

[0026] Here, a1, a2, ..., b1, b2, ... are coefficients calculated from the collected data.

[0027] The weighting coefficient that is inactivated based on the propensity score p(i) is expressed as follows:

[0028]

number

[0029] where r(l) ss is the weight coefficient of the lth layer of the shared network, r(l) i,d is the weight coefficient of the l layer of the DTx network, r(l) i,s is the weight coefficient of the l layer of the standard treatment network, and γ is the offset hyperparameter, which is usually set to 1.

[0030] The therapeutic effect prediction model illustrated in FIG. 6 is expressed by the following equation:

[0031]

number

[0032] where s~(x) is the shared network, sd~(x) is the DTx intervention model, Yd~ is the DTx network, Ys~ is the standard care network, and f(·) is the activation function.

[0033] (S504) The treatment effect prediction model construction unit 301 adjusts the coefficients of the treatment effect prediction model based on a loss function related to the error of the predicted value of the treatment effect contained in the collected data. That is, the coefficients are adjusted so that the output of the loss function becomes smaller. For example, a squared error function is used as the loss function, and for example, a stochastic gradient method is used to adjust the coefficients.

[0034] (S505) The therapeutic effect prediction model construction unit 301 determines whether a termination condition is satisfied. If the termination condition is satisfied, the process flow ends. If the termination condition is not satisfied, the process returns to S503. The termination condition may be, for example, a predetermined number of iterations, a threshold for the output of the loss function, or a threshold for the amount of change in the output of the loss function.

[0035] A treatment effect prediction model in which coefficients are adjusted according to newly collected data is constructed through the processing flow illustrated in Figure 5. Returning to the description of Figure 4,

[0036] (S402) The DTx intervention model adjustment unit 302 adjusts the DTx intervention model so as to reduce the variance of the treatment effect when the treatment method output from the DTx intervention model is used.

[0037] An example of the flow of the process executed in S402 will be described step by step with reference to FIG.

[0038] (S701) The DTx intervention model adjustment unit 302 deletes the standard treatment network from the treatment effect prediction model constructed in S401 and constructs a treatment effect prediction model to be used for adjusting the DTx intervention model. Fig. 8 shows an example of a treatment effect prediction model to be used for adjusting the DTx intervention model.

[0039] (S702) The DTx intervention model adjustment unit 302 generates a prediction model group by inactivating the weighting coefficients of the treatment effect prediction model used to adjust the DTx intervention model, based on the propensity scores for when the treatment method output from the DTx intervention model is used and when it is not used, calculated by the treatment effect prediction model construction unit 301. The propensity score is calculated using Equation 1, and the weighting coefficients to be inactivated are shown in Equation 2.

[0040] (S703) The DTx intervention model adjustment unit 302 inputs the patient profile x(i,j) and patient condition c(i,j) of the newly collected data into each of the prediction models generated in S702 and outputs the predicted value of the treatment effect.

[0041] (S704) The DTx intervention model adjustment unit 302 adjusts the coefficients of the DTx intervention model based on the loss function related to the variance of the predicted values ​​output in S703. That is, the coefficients of the DTx intervention model are adjusted so that the variance of the predicted values ​​becomes smaller. For example, a squared error function is used as the loss function, and for example, a stochastic gradient method or the like is used to adjust the coefficients.

[0042] (S705) The DTx intervention model adjustment unit 302 determines whether or not a termination condition is satisfied. If the termination condition is satisfied, the processing flow ends, and if not satisfied, the processing returns to S703. The termination condition is, for example, a predetermined number of iterations, a threshold for the output of the loss function, or a threshold for the amount of change in the output of the loss function.

[0043] A treatment effect prediction model in which coefficients are adjusted according to newly collected data is constructed through the processing flow illustrated in Fig. 7. Returning to the description of Fig. 4,

[0044] (S403) The model is updated by the DTx intervention model adjusted by the DTx intervention model adjustment unit 302.

[0045] The processing flow described above can improve the reliability of the therapeutic effect of the treatment method output from the DTx intervention model. The therapeutic effect obtained by the updated model may be displayed on the display device 107.

[0046] An example of a result display screen displayed on the display device 107 will be described with reference to Fig. 9. The result display screen illustrated in Fig. 9 has a model update start button 901 and a result display section 902.

[0047] A model update start button 901 is pressed when new data is collected and the DTx intervention model is to be updated.

[0048] The result display section 902 displays the therapeutic effect obtained by the updated DTx intervention model. The therapeutic effect is displayed in a graph format, for example, with the vertical axis representing frequency and the horizontal axis representing effect amount. The therapeutic effect obtained by the DTx intervention model before the update and the standard treatment may also be displayed together. By displaying both the DTx intervention model before and after the update and the therapeutic effect obtained by the standard treatment, the effect of the model update becomes clear.

[0049] The horizontal axis of the graph shown in Figure 9 is HbA1c, which indicates the proportion of glycated hemoglobin, and it is shown that the treatment method output from the DTx intervention model was able to reduce HbA1c more than standard treatment. Furthermore, the distribution of treatment effects obtained by the updated DTx intervention model is narrower than before the update, indicating that variability in treatment effects has been reduced and reliability has been improved.

[0050] The above describes the embodiments of the present invention. The present invention is not limited to the above embodiments, and the components can be modified and embodied without departing from the spirit of the invention. Furthermore, multiple components disclosed in the above embodiments may be combined as appropriate. Furthermore, some components may be omitted from all the components shown in the above embodiments. [Explanation of symbols]

[0051] 101: Information processing device, 102: Calculation unit, 103: Memory, 104: Storage unit, 105: Network adapter, 106: System bus, 107: Display device, 108: Input device, 109: Network, 110: Electronic medical record, 111: Digital therapy, 301: Treatment effect prediction model construction unit, 302: DTx intervention model adjustment unit, 901: Model update start button, 902: Result display unit

Claims

1. An information processing device using a treatment method output model that outputs a treatment method according to a patient's condition, a therapeutic effect prediction model construction unit that constructs a therapeutic effect prediction model including the therapeutic method output model as a component, and compares the therapeutic effect when the therapeutic method is used with the therapeutic effect when the therapeutic method is not used; an information processing device comprising: a model adjustment unit that generates a group of prediction models by inactivating weighting coefficients of the treatment effect prediction model based on the propensity scores when the treatment method is used and when it is not used; and that adjusts the treatment method output model, which is a component of the treatment effect prediction model, so that the variance of the treatment effect output from the group of prediction models falls within a predetermined range.

2. 2. The information processing device according to claim 1, The information processing device is characterized in that the model adjustment unit displays the variance of the treatment effect when using the treatment method output from the adjusted treatment method output model.

3. 3. The information processing device according to claim 2, The information processing device is characterized in that the model adjustment unit further displays the variance of the treatment effect output from the treatment effect prediction model.

4. 2. The information processing device according to claim 1, The information processing device is characterized in that the model adjustment unit adjusts the treatment method output model when new data is collected.

5. 2. The information processing device according to claim 1, The information processing device is characterized in that the therapeutic effect prediction model construction unit inactivates weighting coefficients of the therapeutic effect prediction model based on propensity scores when the therapeutic method is used and when it is not used.

Images