Medical information processing apparatus, medical information processing method, and program

Abstract

To propose measures for intervention with less burden on a subject.SOLUTION: A medical information processing apparatus is an apparatus for proposing intervention to a subject, and comprises state definition means, cost calculation means, searching means, and output means. The state definition means defines an unavoidable state related to an undesired outcome. The cost calculation means calculates costs for the intervention. The searching means searches for measures based on the cost with the unavoidable state as restraints. The output means outputs at least information related to the measures.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The embodiments disclosed in this specification and the drawings relate to a medical information processing device, a medical information processing method, and a program.

Background Art

[0002] With the development of AI technology, techniques that can not only assist in diagnosis but also recommend treatments are spreading. In the sequential decision-making problem in the ICU (Intensive Care Unit), there are methods to identify dead ends in medical treatment, which mean that the patient will die regardless of the future treatment order, warn of dead ends in advance, or learn decision-making strategies to avoid dead ends. Also, in dynamic diagnosis, there are methods to learn optimal strategies under the constraint of the total cost considering the cost of examinations.

[0003] However, when assisting in determining interventions such as treatment, it is often insufficient to only avoid dead ends or introduce cost limitations. For example, when assisting in selecting daily health behaviors for disease prevention, even if there are cost constraints, the personal hobbies and preferences will be significantly restricted, so there is a high probability that the strategy will be unrealistic.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] One of the problems that the embodiments disclosed in this specification and the drawings seek to solve is to propose an intervention strategy that places less burden on the target person. However, the problems to be solved by the embodiments disclosed in this specification and the drawings are not limited to the above problems. It is also possible to position, as other problems, the problems corresponding to the respective effects of each configuration shown in the embodiments described later.

Means for Solving the Problems

[0006] The medical information processing apparatus according to the embodiment is an apparatus that proposes an intervention for a target, and includes a state definition means, a cost calculation means, a search means, and an output means. The state definition means defines an inescapable state related to an undesirable outcome. The cost calculation means calculates the cost for the intervention. The search means searches for a strategy based on the cost, with the inescapable state as a constraint. The output means outputs at least information related to the strategy.

Brief Description of the Drawings

[0007]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Modes for Carrying Out the Invention

[0008] Hereinafter, embodiments will be described with reference to the drawings. In the present disclosure, regarding the form in which information processing by software is specifically realized using hardware resources, this software can be executed by a program and can be implemented by the program itself or a non-transitory computer-readable medium storing the program. Also, among various data in this specification and the drawings, those mainly used by the information processing apparatus are typically digital data.

[0009] In the present disclosure (specification, drawings, and / or claims), the term "inescapable state" is used. This inescapable state is a term indicating at least a partial concept of an undesirable outcome for a patient, for example, a state in which it is difficult to return to a better state even if a predetermined treatment or a treatment by a predetermined intervention is performed. The inescapable state is a concept including dead ends in medical treatment, for example, the death of a patient or a state in which remission or cure is difficult.

[0010] FIG. 1 is a diagram schematically showing an example of a medical information processing apparatus according to an embodiment. The medical information processing apparatus 1 is an apparatus that proposes a strategy for intervention regarding a target patient or the like. The medical information processing apparatus 1 includes an input / output interface (hereinafter referred to as I / F 10), a storage unit 20, and a processing unit 30. The medical information processing apparatus 1 further includes components necessary for operation, such as control means for controlling each component and power supply means for supplying power to each component.

[0011] I / F 10 is an interface for inputting and outputting for the medical information processing apparatus 1. As an input means, I / F 10 acquires, for example, the current state of the target and information from a database. As an output means, I / F 10 outputs, for example, a strategy for intervention regarding the target.

[0012] That is, the I / F 10 can operate as input means and / or output means, for example, in the medical information processing apparatus 1 or a medical information processing system including the medical information processing apparatus 1.

[0013] The storage unit 20 stores data and the like necessary for the operation of the medical information processing apparatus 1. Further, the storage unit 20 may have a storage area necessary during the operation of the medical information processing apparatus 1 or an area for storing the operation result.

[0014] That is, the storage unit 20 can operate as storage means, for example, in the medical information processing apparatus 1 or a medical information processing system including the medical information processing apparatus 1.

[0015] The processing unit 30 includes a processing circuit that executes various information processes in the medical information processing apparatus 1. The processing circuit may be a general-purpose processor or a dedicated circuit. In this embodiment, the processing unit 30 operates as processing means and can at least execute means such as a state definition means 300, a cost calculation means 302, and a search means 304.

[0016] The state definition means 300 is means for defining an inescapable state, which is an undesirable outcome, based on the state of the target. The state definition means 300, for example, acquires a state including a disease state and a morbid state of the target via the I / F 10, and defines an inescapable state in this disease state and morbid state. The state definition means 300 can define an inescapable state, for example, by referring to a database that associates a disease or illness acquired via the I / F 10 with an undesirable outcome as a result of intervening in these diseases or illnesses.

[0017] Also, as another example, the state definition means 300 can also receive a specification of an undesirable outcome from a target who is a user, such as a medical professional, via the I / F 10. In this case, the state definition means 300 can also define this input undesirable outcome as an unavoidable state.

[0018] For example, for a patient with a circulatory disease during hospitalization, "drug treatment", "catheter treatment", or "surgical treatment" is applicable, but it is desired to avoid "death", which is an undesirable outcome with only "drug treatment". In this case, regardless of the future treatment order, if "death" cannot be avoided with only the combination of "drug treatment", that is, if "death" cannot be avoided without any intervention of "catheter treatment" or "surgical treatment", such a state is defined as an unavoidable state.

[0019] For example, for a healthy person who has not yet developed a disease state, "performing healthy behaviors (improving diet and exercise habits)", "taking only over-the-counter drugs", or "visiting a medical institution" is applicable, but it is desired to avoid "onset of a disease", which is an undesirable outcome with only "performing healthy behaviors" and / or "taking over-the-counter drugs". In this case, if "onset of a disease" cannot be avoided with only the combination of "performing healthy behaviors" and / or "taking over-the-counter drugs" regardless of the future treatment order, such a state is defined as an unavoidable state.

[0020] As can be seen from the above, the unavoidable state, the type of intervention, etc. may be different for each target.

[0021] The state definition means 300 can define the rewards and values regarding the environment related to the intervention in the state of the target. The state definition means 300 can, for example, define the rewards and values for the intervention in the environment by reinforcement learning. The search means 304 described later can search for a strategy that does not lead to an unavoidable state based on this reward and value.

[0022] The state definition means 300 learns, by reinforcement learning, which strategy is advantageous in the environment for the target state. Hereinafter, the intervention (action) is denoted as a, the state as s, and the strategy as π.

[0023] The state definition means 300 defines the following inescapable states under the condition that if the probability of reaching a desirable state is λ or more in the intervention a in the state s, the probability of selecting the strategy π (performing the intervention a in the state s) is 1 - λ.

[0024] For example, when the terminal state reached by continuing the intervention according to the strategy is not an inescapable state, the state definition means 300 may give a reward indicated by a value larger than that when the terminal state is an inescapable state.

[0025] As a non-limiting example, when the terminal state reached by continuing the intervention according to the strategy is an inescapable state, the state definition means 300 gives -1, and when continuing the intervention according to the strategy shows a transition different from the inescapable state, gives 0 to the value function Q * D (s, a) can be defined. Also, for example, when the terminal state reached by continuing the intervention according to the strategy is not an inescapable state, the state definition means 300 gives +1, and when showing a different transition, gives 0 to the value function Q * R (s, a) can be defined.

[0026] The state definition means 300 uses these value functions Q * D (s, a), Q * R By setting appropriate thresholds for (s, a) and Q(s, a), a criterion can be defined to determine whether the intervention according to the strategy π falls into an inescapable state.

[0027] The cost calculation means 302 calculates the cost for exploring the policy π. The cost is determined based on an index indicating, for example, the magnitude of economic, time, or mental burden.

[0028] The economic cost indicates the magnitude of the financial burden associated with the intervention. For example, the amount borne by the subject, the price of the equipment itself required for the intervention, etc. are applicable.

[0029] The time cost is the magnitude of the time burden associated with the intervention. For example, the time spent by the subject on the intervention, the time spent by the medical staff performing the intervention, etc. are applicable.

[0030] The mental cost is the magnitude of the mental burden associated with the intervention. For example, the pain and fear associated with the intervention are applicable.

[0031] The cost calculation means 302 can calculate the cost, for example, by multiplying a predetermined constant or performing a predetermined conversion on an index of economic cost, more specifically, the price indicating the amount of cost that can be incurred. The same applies to time and mental costs. The cost calculation means 302 calculates the cost for determining the policy π by performing a predetermined conversion on the index indicating the cost.

[0032] The cost calculation means 302 can calculate the cost using, for example, the cost C(e) defined as the economic cost. The cost calculation means 302 can calculate the cost using, for example, the cost C(τ) defined as the time cost. The cost calculation means 302 can calculate the cost using, for example, the cost C(m) defined as the mental cost.

[0033] Not limited to this, the cost may be a combination of these. The cost calculation means 302 can calculate the cost, for example, using the cost C(e, τ) including economic and time costs, the cost C(τ, m) including time and mental costs, the cost C(m, e) including mental and economic costs, and the cost C(e, τ, m) including economic, time, and mental costs. Also, the cost may be defined based on other indicators.

[0034] In this way, multiple types of costs can be integrated or the cost can be set as a single value. Also, similar to the inescapable state, the calculation method and value of the cost can be set differently for each target.

[0035] The search means 304 searches for a strategy π that does not lead to an inescapable state based on the value defined by the state definition means 300, the reward, and the cost C calculated by the cost calculation means 302.

[0036] For example, when the search means 304 selects a strategy π defined for an intervention, it may search for a strategy in which the difference between the cost calculated by the cost calculation means 302 and the target value is smaller than that of other strategies in a situation where the probability that the terminal state becomes an inescapable state is equal to or less than a first threshold. The search means 304 executes the search for the strategy π according to, for example, the following formula.

Equation

[0037] Here, B is the target cost, and P DE (π) is the probability of falling into an inescapable state when the strategy π is executed, and δ D is the first threshold. According to this formula (1), the search means 304 searches for a strategy π with a cost close to the target value on the condition that the probability of becoming an inescapable state is lower than the first threshold. * to search for.

[0038] In the present disclosure, throughout the description, it is assumed that the cost (e.g., target cost B, cost C defined as a function, etc.) increases as the positive value becomes larger, that is, the cost (load) increases. Without being limited to this, a definition by a function similar to the present disclosure may be provided. The medical information processing apparatus 1 may, for example, use a negative value for the cost and appropriately rewrite the function of the present embodiment accordingly for processing.

[0039] In Equation (1), B can also be set to 0.

Equation

[0040] According to this equation, the search means 304 searches for the policy π that minimizes the cost. * That is, the medical information processing apparatus 1 can present a policy (intervention) with a smaller burden on the subject.

[0041] Further, the search means 304 may search for a policy in which the difference between the cost calculated by the cost calculation means 302 and the target value is smaller than that of other policies, for example, in a situation where the probability that the end state becomes an inescapable state is equal to or greater than a second threshold value when a policy π defined for the intervention is selected. The search means 304 executes a search for a policy π according to, for example, the following equation.

Equation

[0042] Here, δ D' is the second threshold value. According to this equation (3), the search means 304 searches for a policy π with a cost higher than the target value on the condition that the probability of becoming an inescapable state is lower than the first threshold value and higher than the second threshold value. Similar to Equation (2), the search means 304 may search so that the cost is minimized. *

Equation

[0043] By defining a measure using the second threshold value, the medical information processing apparatus 1 can output, for example, a setting of the probability of being in an inescapable state. More specifically, the medical information processing apparatus 1 can propose the behavior of the subject by presenting an intervention such that the probability of being in an inescapable state is 80% or more according to the measure.

[0044] In this way, the search means 304 can search for a measure that satisfies at least one of the conditions that the probability of falling into an inescapable state in the environment, that is, the probability of an undesirable outcome occurring, is equal to or less than the first threshold value or equal to or greater than the second threshold value.

[0045] The medical information processing apparatus 1 presents and outputs (for example, displays via a display) one or more measures specified by the search means 304 to the subject or a medical staff member or the like via the I / F 10. The I / F 10 can also output a plurality of candidates including escapable and inescapable measures.

[0046] FIG. 2 is a flowchart showing an example of the processing of the medical information processing apparatus 1 according to an embodiment.

[0047] The state definition means 300 defines an inescapable state (S100).

[0048] The medical information processing apparatus 1 sets an objective function for each subject (S102). This objective function is a function to be minimized / maximized in the optimization in Expressions (1) to (4).

[0049] The state definition means 300 learns a measure in the above objective function (S104). This learning may be executed as reinforcement learning for optimizing the value, reward, etc. of the measure in the environment.

[0050] The above processing can be performed as a type of constrained optimization problem. For example, regarding the constraint conditions, this constrained optimization problem can be solved by performing an operation such as giving a large negative reward when the constraint conditions are not satisfied. In the processing of S104, the medical information processing device 1 can, for example, define the constraint conditions as rewards, thereby enabling inference as an optimization problem of reinforcement learning using rewards.

[0051] The above processing is given as a non-limiting example, and the forms in the present disclosure do not exclude optimization methods other than this example. That is, it should be noted that the medical information processing device 1 can achieve the same effect by processing an appropriate constrained optimization problem using the above constraints.

[0052] If the learning is not completed (S106: NO), the model is optimized by repeating the processing of S104. Also, in some cases, as indicated by the dashed line, repeated learning may be executed starting from the processing of setting the objective function (S102).

[0053] After the optimization of the reinforcement learning model of the policy for each target environment (which may include rewards, values, actions, etc.) is completed (S106: YES), the medical information processing device 1 executes exploration using this model.

[0054] The exploration means 304 explores the policy π based on the model optimized by the state definition means 300 and the cost calculated by the cost calculation means 302 (S108). The exploration means 304 can, for example, execute the exploration of the policy π by simulation under the condition of the cost calculated by the cost calculation means 302 in the model optimized by the state definition means 300.

[0055] The medical information processing device 1 outputs, via the I / F 10, the strategy π searched by the search means 304 and / or the intervention related to the strategy π (for example, a proposal for the action to be taken by the target) (S110). A user (for example, the target or a medical staff member) can view the output result. Further, in one embodiment, the user can also request the medical information processing device 1 to perform a re-search process via the I / F 10 based on the output result.

[0056] The I / F 10 as the output means can output at least the information related to the strategy searched by the search means 304. The medical information processing device 1 can display the result on a display which is the I / F 10. The medical information processing device 1 can display, by the output means, information such as a strategy and the state transition when the strategy is executed.

[0057] When the search is completed (S112: YES), the medical information processing device 1 completes the process.

[0058] When receiving a re-search request from the user, the medical information processing device 1 determines that the search is not completed (S112: NO) and can repeatedly execute the process from the I / F 10 8. Also, in some cases, as indicated by the dashed line, it can also repeat from the process of setting the objective function (S102).

[0059] Further, the medical information processing device 1 can also search for strategies of a plurality of patterns with different conditions in order to present interventions of a plurality of patterns, not at the request of the user. In this case, the simulation can be repeated while changing the conditions until an appropriate number of strategies are obtained.

[0060] The medical information processing device 1 can search for, for example, a strategy with low cost and not falling into an unavoidable state, and can also search for a strategy with a possibility of falling into an unavoidable state of 80% or more.

[0061] FIG. 3 is a diagram showing an example of output according to an embodiment. As shown in this figure, for example, the medical information processing apparatus 1 indicates a flag indicating whether it is unavoidable, a plan based on the explored measure (user behavior as an intervention result), a final deadline when the plan is continued and becomes unavoidable, and the probability of falling into an unavoidable state when the plan is continued.

[0062] More specifically, the limit indicates, for example, a grace period until it becomes unavoidable when the plan is executed. If the user does not change to another appropriate plan during this period, it can indicate a period in which the possibility of falling into an unavoidable state increases. In other words, it can indicate a period in which the possibility of falling into an unavoidable state can be reduced by the user changing to another appropriate plan during this period.

[0063] For example, when the probability of becoming unavoidable is higher than a predetermined value, the medical information processing apparatus 1 can display a "!" mark as having a high probability of becoming unavoidable, or can change other outputs, for example, the color of the row. When the probability of becoming unavoidable is high, the medical information processing apparatus 1 can present an appropriate measure (plan) to the user by performing an output that clearly indicates that the probability of becoming unavoidable is high according to the plan.

[0064] For example, in the process of S108, the search means 304 can search for a plan with a high probability of becoming unavoidable. The search means 304 may search for a measure with a high probability of becoming unavoidable and the length of the grace period by repeatedly simulating actions according to each measure in reinforcement learning. Also, this learning may be executed at the learning stage in S104.

[0065] The plan may be, for example, something like an exercise habit or a smoking habit as shown in FIG. 3, or may indicate an action such as taking medicine, undergoing surgical treatment, or receiving a diagnosis as described above.

[0066] In FIG. 3, for example, when the action of "exercise & smoking cessation once a month" is continued, it can be presented that the probability of reaching an inevitable state is 80% after 6 months.

[0067] FIG. 4 is a diagram showing an example of output according to an embodiment. As shown in FIG. 4, it is also possible to output a graph display with the burden and the avoidability as axes. For example, in this case, by aligning the cursor with the triangles and circles shown as pointers in the graph, or by clicking, tapping, etc., the plan can be displayed. In addition to the graph, a table showing the pointer and the plan may be output separately.

[0068] By outputting as in FIG. 4, the user can visually obtain how much burden is acceptable.

[0069] FIG. 5 is a diagram showing an example of output according to an embodiment. As shown in FIG. 5, it is also possible to output a graph display with time and avoidability as axes. The medical information processing apparatus 1 can create and display, for example, a graph showing the avoidable probability along the time series for each plan. In addition, for a plan that becomes an inevitable state, it is also possible to explicitly present the time at which it becomes inevitable.

[0070] By outputting as in FIG. 5, the user can visually obtain how much burden and how much time margin there is.

[0071] Not limited to the output from FIG. 3 to FIG. 5, the medical information processing apparatus 1 can, for example, also output a graph in a three-dimensional display with time, burden, and avoidability as three axes, or can output a graph or a table with other parameters as axes.

[0072] As described above, according to the present embodiment, the medical information processing apparatus 1 can appropriately present an intervention that reduces the user's cost.

[0073] In the above-described embodiment, the input interface, which is a part of I / F 10, can be realized by a trackball, a switch button, a mouse, a keyboard, a touch pad for performing an input operation by touching an operation surface, a touch screen in which a display screen and the touch pad are integrated, a non-contact input circuit using an optical sensor, an audio input circuit, etc. for performing various settings and the like. The input interface is connected to the control circuit, converts the input operation received from the operator into an electrical signal, and outputs it to the control circuit. Note that in this specification, the input interface is not limited to those provided with physical operation components such as a mouse and a keyboard. For example, an electrical signal processing circuit that receives an electrical signal corresponding to an input operation from an external input device provided separately from the apparatus and outputs this electrical signal to the control circuit is also included in the example of the input interface.

[0074] In the above-described embodiment, the output interface, which is a part of I / F 10, can be configured as an arbitrary interface such as a display for outputting a result as a display, a speaker for outputting a result as audio, a printer for printing and outputting a result, or a communication interface for outputting information to a network.

[0075] Also, in the above-described embodiment, each processing function of the information processing function is recorded in a storage circuit in the form of a program executable by a computer. The processing circuit can include a processor. For example, the processing circuit reads out and executes a program from the storage circuit to realize the functions corresponding to the respective programs. In other words, the processing circuit in the state of having read out each program has each function shown in the processing circuit illustrated in the drawings. Although the drawings have described that each processing function is realized by a single processor, a processing circuit may be configured by combining a plurality of independent processors, and each processor may execute a program to realize a function. Also, although the drawings have described that a single storage circuit stores programs corresponding to the respective processing functions, a plurality of storage circuits may be distributed and arranged, and the processing circuit may be configured to read out corresponding programs from the respective storage circuits.

[0076] In the above description, an example has been described in which the "processor" reads and executes a program corresponding to each function from the storage circuit. However, the embodiment is not limited to this. The term "processor" can mean, for example, a circuit such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an application specific integrated circuit (ASIC), a programmable logic device (for example, a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)). When the processor is, for example, a CPU, the processor realizes its function by reading and executing a program stored in the storage circuit. On the other hand, when the processor is an ASIC, instead of storing a program in the storage circuit, the function is directly incorporated as a logic circuit in the circuit of the processor. Note that each processor of the present embodiment is not limited to being configured as a single circuit for each processor, and a plurality of independent circuits may be combined to be configured as one processor to realize its function. Further, a plurality of components in the drawings may be integrated into one processor to realize its function.

[0077] According to at least one of the embodiments described above, it is possible to propose an intervention with low cost, that is, with a low burden on the target patient.

[0078] Although some embodiments have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, changes, and combinations of embodiments can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, as well as in the invention described in the claims and the equivalent scope thereof.

Description of Reference Numerals

[0079] 1: Medical information processing apparatus, 10: I / F, 20: Storage unit, 30: Processing unit, 300: State definition means, 302: Cost calculation means, 304: Search means

Claims

1. An apparatus for proposing an intervention for a subject, comprising: state definition means for defining an inescapable state related to an undesirable outcome; cost calculation means for calculating the cost of the intervention; search means for searching for a policy based on the cost, with the inescapable state as a constraint; output means for outputting at least information related to the policy; A medical information processing apparatus comprising the above.

2. The search means: searches for a policy that optimizes the cost, with the inescapable state as a constraint. The medical information processing apparatus according to Claim 1.

3. The state definition means: defines the undesirable outcome based on the disease state of the subject. The medical information processing apparatus according to Claim 1.

4. The state definition means: defines the undesirable outcome by accessing a database that stores the disease state and the undesirable outcome in association with each other. The medical information processing apparatus according to Claim 3.

5. input means for receiving an input from a user, further comprising: The state definition means: defines the state input via the input means as the undesirable outcome. The medical information processing apparatus according to Claim 1.

6. The state definition means: for the terminal state after continuously executing the intervention in the environment, defines a larger value than when the terminal state is the undesirable outcome as the reward when the terminal state is not the undesirable outcome, defines a value, and learns the intervention that maximizes the value by reinforcement learning. The medical information processing apparatus according to Claim 1.

7. The search means: searches for the policy that satisfies at least one of the conditions: the probability of occurrence of the undesirable outcome in the environment is less than or equal to a first threshold value, or the probability of occurrence of the undesirable outcome is greater than or equal to a second threshold value. The medical information processing apparatus according to Claim 6.

8. The search means: searches for the policy such that the difference between the cost calculated by the cost calculation means and the target value is smaller than that of other policies in a situation where the probability that the terminal state becomes the undesirable outcome is less than or equal to the first threshold value when the policy defined for the intervention is selected. The medical information processing apparatus according to Claim 7.

9. The search means: searches for the policy with the target value set to 0. The medical information processing apparatus according to Claim 8.

10. ​ The search means is when selecting the strategy defined for the intervention, in a situation where the probability that the end state becomes the undesirable outcome is equal to or greater than the second threshold value, the search means searches for the strategy in which the difference between the cost calculated by the cost calculation means and the target value is smaller than that of other strategies. The medical information processing apparatus according to any one of claims 7 to 9.

11. The search means is searching for strategies in a plurality of patterns with changed conditions, The output means is outputting information related to the strategies in the plurality of patterns. The medical information processing apparatus according to any one of claims 7 to 9.

12. The output means is outputting information related to the strategy and the state transition when the strategy is executed. The medical information processing apparatus according to claim 1.

13. A method for a computer to propose an intervention for a target, the method comprising: a state definition means for defining an inescapable state related to an undesirable outcome; a cost calculation means for calculating a cost for the intervention; a search means for searching for a strategy based on the cost, with the inescapable state as a constraint; an output means for outputting the strategy; A program for causing the above to be executed.

Images