Information processing apparatus, information processing method, and program

Abstract

To provide an information processing apparatus, an information processing method, and a program that can support efficiently questioning a patient without involvement of a medical worker in order to learn the patient's preferences related to medical examination.SOLUTION: An information processing apparatus of an embodiment has an acquisition unit, an estimation unit, a response determination unit, and an output control unit. The acquisition unit acquires an answer from a user. The estimation unit estimates medical examination information including a plurality of user's preferences related to medical examination on the basis of the answer. The response determination unit determines a response to the answer on the basis of the medical examination information. The output control unit outputs the response through an output interface. The response determination unit further determines at least one of a narrowed response to the user and an approval request for the user as the response.SELECTED DRAWING: Figure 4

Description

Technical Field

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

Background Art

[0002] Shared Decision Making, in which a patient and a medical staff discuss treatment goals, preferences, and values and jointly seek the optimal treatment, has become important. In the utilization of AI (Artificial Intelligence) in such a scenario, support for providing information for the patient and the medical staff to efficiently communicate with each other is more important than support for directly providing medical judgment. However, when there are many preferences, which are matters that the patient is concerned about or interested in regarding medical treatment, a lot of time and effort are required for the medical staff to elicit the preferences during the dialogue with the patient.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The problem to be solved by the embodiments disclosed in this specification and the drawings is to support efficiently asking a patient questions without the involvement of a medical staff in order to know the patient's preferences regarding medical treatment. 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

[0005] The information processing apparatus according to the embodiment includes an acquisition unit, an estimation unit, a response determination unit, and an output control unit. The acquisition unit acquires a response from the user. The estimation unit estimates medical information including a plurality of preferences of the user regarding medical treatment based on the response. The response determination unit determines a response to the response based on the medical information. The output control unit outputs the response via an output interface. Further, the response determination unit determines, as the response, at least one of a narrowing-down response that is a response made to the response of the user in order to narrow down the plurality of preferences, and an approval request to the user for each of the plurality of preferences.

Brief Description of Drawings

[0006]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Embodiments for Carrying Out the Invention

[0007] Hereinafter, an information processing apparatus, an information processing method, and a program according to the embodiments will be described with reference to the drawings.

[0008] [Configuration of Information Processing System] FIG. 1 is a diagram showing an example of the configuration of an information processing system 1 according to an embodiment. The information processing system 1 includes, for example, a terminal device 10, a medical database 20, and an information processing apparatus 100. The terminal device 10, the medical database 20, and the information processing apparatus 100 are communicably connected via, for example, a communication network NW.

[0009] The communication network NW may mean the entire information communication network using telecommunication technologies. For example, the communication network NW includes wireless / wired LANs such as a hospital backbone LAN (Local Area Network), the Internet, as well as a telephone communication line network, an optical fiber communication network, a cable communication network, and a satellite communication network.

[0010] The terminal device 10 is, for example, a terminal device such as a personal computer, a tablet terminal, or a mobile phone, and is used by patients and medical staff. The medical staff is typically a doctor, but may also be a nurse or other person involved in medical treatment.

[0011] For example, a patient may touch-input or voice-input their own answers to questions into the terminal device 10. Also, a medical staff member may verbally ask a patient, listen to the patient's answer to the question, and input the hearing result into the terminal device 10. The questions include matters related to medical treatment.

[0012] In this embodiment, "medical treatment" may include not only treatments such as surgery and medication, but also examinations before or after treatment, or any other medical actions until treatment is reached.

[0013] The terminal device 10 transmits information input by patients or medical staff to the information processing device 100 via the communication network NW, or receives information from the information processing device 100.

[0014] In particular, the terminal device 10, as a response to the patient's answer, adds questions (hereinafter referred to as narrowing-down responses) for narrowing down one or more preferences from a number of matters related to medical treatment based on the information received from the information processing device 100, and displays them as a GUI (Graphical User Interface).

[0015] Preferences are typically, but not limited to, matters that patients worry about or are interested in regarding medical treatment. For example, in addition to or instead of matters that patients worry about or are interested in regarding medical treatment, preferences may be matters that patients feel apply to themselves regarding medical treatment, matters that patients agree with, matters that patients feel have high affinity, matters that are acceptable, matters that seem reasonable, matters that patients have little resistance to, and the like.

[0016] For example, when "medical treatment" is "differential diagnosis of cold", several symptoms such as sore throat, cough, phlegm, and fever are asked of the patient as part of the medical history. In such a case, among the multiple symptoms asked, the symptoms that the patient answers as applying to themselves (in other words, the symptoms that the patient is aware of) become preferences. Thus, preferences may be matters selected motivated by various emotions or inner feelings of patients such as worry, interest, curiosity, and preference.

[0017] Furthermore, the terminal device 10 displays a user-customized summary report as a GUI based on the patient's answers to a plurality of questions including narrowing-down responses.

[0018] Figures 2 and 3 are diagrams showing an example of the GUI screen of the terminal device 10. Figure 2 shows an example in which further refined responses are displayed for the answers given by the patient to questions regarding a certain medical treatment.

[0019] As shown in the figure, for example, it is assumed that a question such as "Please tell me the current degree of numbness in your hand" is asked. In such a case, the patient answers the question by operating an object (operation button) such as the one shown at B1, which allows the patient to select the degree of numbness step by step. In the example shown in the figure, the patient has answered "There is a little".

[0020] In this embodiment, in response to such an answer, a further refined response such as "Currently, which causes more anxiety, family matters or money matters?" is made. B2 represents an object for selecting "family matters", and B3 represents an object for selecting "money matters". In the example shown in the figure, the patient has answered "money matters".

[0021] And when such a series of dialogues between the patient's answer and the response thereto is made, a link to a user-customized summary report is displayed based on the content of the series of dialogues. B4 represents an object (operation button) for accessing the summary report.

[0022] For example, when the patient selects the object B4, a summary report as shown in Figure 3 is displayed. The summary report shows preferences suggested to the patient, such as "matters regarding the burden of treatment", "matters regarding the content and effects of treatment", "matters regarding symptoms / side effects / post - sequelae", "matters regarding family relationships", and "matters regarding the relationship with medical staff".

[0023] On the summary report, for each of the plurality of suggested preferences, an object B5 that allows the patient to confirm the preference and an object B6 that allows the rejection of the preference are displayed. Further, on the summary report, an object B7 that allows the patient to confirm all the suggested preferences at once is also displayed.

[0024] For example, when the patient operates objects B5 to B7, the results are fed back. As a result, subsequent questions for the patient can be made closer to questions regarding more appropriate preferences (for example, preferences that make the patient more worried, more interested, or feel more applicable to themselves). The detailed algorithms for the filtered responses and the summary report will be described later.

[0025] Returning to the description of FIG. 1, the medical database 20 is a database that stores patient attribute information, examination data, and the like. The medical database 20 transmits patient attribute information, examination data, etc. to the information processing device 100 via the communication network NW. Further, the medical database 20 may store the data transmitted from the information processing device 100. The medical database 20 may be, for example, a general-purpose server or a cloud server.

[0026] The information processing device 100 receives information from the terminal device 10 and the medical database 20 via the communication network NW and processes the received information. For example, the information processing device 100 generates the above-described filtered responses and summary reports. Then, the information processing device 100 transmits the processed information to the terminal device 10 and the medical database 20 via the communication network NW. In addition to or instead of transmitting the processed information to the terminal device 10, the information processing device 100 may transmit it to a dedicated terminal of medical staff installed in the hospital.

[0027] The information processing apparatus 100 may be a single device, or may be a system in which a plurality of devices connected via a communication network NW cooperate with each other. That is, the information processing apparatus 100 may be realized by a plurality of computers (processors) included in a distributed computing system or a cloud computing system. Further, the information processing apparatus 100 does not necessarily have to be a separate device different from the terminal device 10, and may be a device integrated with the terminal device 10.

[0028] [Configuration of Information Processing Apparatus] FIG. 4 is a diagram showing an example of the configuration of the information processing apparatus 100 according to the embodiment. The information processing apparatus 100 includes, for example, a communication interface 111, an input interface 112, an output interface 113, a memory 114, and a processing circuit 120.

[0029] The communication interface 111 communicates with an external device via the communication network NW. The external devices include, for example, the terminal device 10 and the medical database 20. The communication interface 111 includes, for example, a NIC (Network Interface Card), an antenna for wireless communication, and the like.

[0030] The input interface 112 receives various input operations from an operator, converts the received input operations into electrical signals, and outputs them to the processing circuit 120. For example, the input interface 112 includes a mouse, a keyboard, a trackball, a switch, a button, a joystick, a touch panel, and the like. The input interface 112 may be a user interface that receives voice input such as a microphone. When the input interface 112 is a touch panel, the input interface 112 may also serve as a display function of the display 113a included in the output interface 113 described later.

[0031] Note that in this specification, the input interface 112 is not limited to only those equipped with physical operation components such as a mouse and a keyboard. For example, an electric signal processing circuit that receives an electric signal corresponding to an input operation from an external input device provided separately from the apparatus and outputs this electric signal to the control circuit is also included in the examples of the input interface 112.

[0032] The output interface 113 includes, for example, a display 113a and a speaker 113b. The display 113a displays various kinds of information. For example, the display 113a displays an image generated by the processing circuit 120 and a GUI or the like for receiving various input operations from the operator. For example, the display 113a is an LCD (Liquid Crystal Display), a CRT (Cathode Ray Tube) display, an organic EL (Electro Luminescence) display, or the like. The speaker 113b outputs the information input from the processing circuit 120 as sound.

[0033] The memory 114 is realized, for example, by a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, a hard disk, or an optical disk. These non-transitory storage media may be realized by other storage devices connected via a communication network NW such as a NAS (Network Attached Storage) or an external storage server device. Further, the memory 114 may include non-transitory storage media such as a ROM (Read Only Memory) or a register. The memory 114 stores a program executed by the hardware processor of the processing circuit 120 and various calculation results by the processing circuit 120.

[0034] The processing circuit 120 includes, for example, an acquisition function 121, an estimation function 122, a response determination function 123, and an output control function 124. The response determination function 123 includes a summary report generation function 123A and a narrowing-down response generation function 123B.

[0035] The processing circuit 120 realizes these functions, for example, by a hardware processor (computer) executing a program stored in the memory 114 (storage circuit). The acquisition function 121 is an example of an "acquisition unit", the estimation function 122 is an example of an "estimation unit", the response determination function 123 is an example of a "response determination unit", and the output control function 124 is an example of an "output control unit".

[0036] The hardware processor in the processing circuit 120 means circuitry 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) or a complex programmable logic device (CPLD), a field programmable gate array (FPGA)). Instead of storing a program in the memory 114, it may be configured to directly incorporate the program into the circuit of the hardware processor. In this case, the hardware processor realizes its functions by reading and executing the program incorporated in the circuit. The above program may be stored in the memory 114 in advance, or may be stored in a non-transitory storage medium such as a DVD or a CD-ROM, and may be installed from the non-transitory storage medium into the memory 114 when the non-transitory storage medium is attached to a drive device (not shown) of the information processing apparatus 100. The hardware processor is not limited to being configured as a single circuit, and may be configured as one hardware processor by combining a plurality of independent circuits to realize each function. Also, a plurality of components may be integrated into one hardware processor to realize each function.

[0037] [Processing Flow of Information Processing Apparatus] Next, a series of processes performed by the processing circuit 120 of the information processing apparatus 100 will be described while referring to the flowchart. FIG. 5 is a flowchart showing the flow of a series of processes of the processing circuit 120 according to the embodiment.

[0038] First, when a question regarding medical treatment is asked of a patient, the acquisition function 121 acquires the patient's answer to the question (step S100).

[0039] For example, the acquisition function 121 may acquire the patient's answer from the terminal device 10 via the communication interface 111. Also, when a medical staff member such as the patient's attending physician inputs the patient's answer to the input interface 112, the acquisition function 121 may acquire the patient's answer from the input interface 112. Further, when the patient's answer is stored in the memory 114, the acquisition function 121 may acquire the patient's answer from the memory 114.

[0040] When the acquisition function 121 acquires the patient's answer, it may acquire the patient's attribute information from the medical database 20 via the communication interface 111. The patient's attribute information may include, for example, age, gender, weight, blood type, vital signs, concurrent diseases, predicted complications, epidemiological information, patient life function indicators, main disease state, general condition, heredity, past history, family history, treatment history, lifestyle, disease organ state, state other than the disease organ, tumor state, and the like.

[0041] Next, the estimation function 122 estimates one or more preferences that the patient is concerned about, interested in, or feels apply to himself / herself regarding medical treatment (step S102). The preference is an example of "medical treatment information".

[0042] Figures 6 and 7 are diagrams showing an example of preferences. As shown in the illustration, the preferences have a hierarchical structure (are hierarchical). The matters to be questioned to the patient as the above-mentioned narrowing-down response may be all preferences, for example, as shown by Item1 in the figure. On the other hand, the matters to be included in the above-mentioned summary report may be the lowest-level preferences, for example, as shown by Item2 in the figure.

[0043] The questions asked as the narrowing-down response may directly correspond to the preferences or may indirectly correspond to them. Also, the summary report includes content corresponding to the ratings for each of a plurality of preferences according to the ratings.

[0044] The rating is an index that quantifies the degree to which a patient is worried about each preference, the degree to which a patient shows interest, the degree to which a patient feels it applies, etc. The estimation function 122 estimates the rating for each preference in order to estimate the preference. The summary report does not necessarily have to be expressed only in natural language (character string), and may include, for example, a graph quantitatively showing the rating. The rating is another example of "medical information". Also, the rating is an example of "predetermined index".

[0045] Furthermore, a label (Positive in the figure) indicating that the patient has given an affirmative answer in the past may be assigned to the preference. In the optimization of matrix factorization described later, label propagation is performed from the lower-level preferences to the upper-level preferences.

[0046] Figure 8 is a diagram for explaining a method of estimating the rating for each preference. As shown in the illustration, for example, the estimation function 122 estimates the rating for the preferences of the target patient using a matrix factorization that has been optimized in advance (previously learned).

[0047] Matrix factorization is a method of expressing patients (users) and preferences as feature vectors of the same length (dimensionality reduction), and expressing ratings by the inner product of the feature vectors. The estimation of ratings to which matrix factorization is applied can be represented, for example, by formulas (1) to (3).

[0048] [Number]

[0049] [Number]

[0050] [Number]

[0051] Here, u i represents the latent feature vector of the i-th patient (user). x i represents the observed feature quantity of the i-th patient (user). p j represents the latent feature vector of the j-th preference. v j represents the index of the j-th preference. r i,j represents the rating of the i-th patient (user) for the j-th preference.

[0052] When meta-information of patients (users) and preferences can be used, a function (such as fully-connected layers) that converts the meta-information into a feature vector can be used instead of embedding to address the Cold-Start problem. Here, assuming that meta-information of patients (users) (observable features of attributes such as patient age and gender) can be utilized, these meta-information are converted into feature vectors using a fully-connected layer. Also, since preferences have a hierarchical structure, existing methods that consider the hierarchical structure may be used when learning feature vectors by embedding.

[0053] To optimize matrix factorization (to learn matrix factorization), a loss function based on Bayesian Personalized Ranking is utilized.

[0054] Bayesian Personalized Ranking is a method used for implicit feedback such as click history and purchase history (cases where users do not explicitly give ratings), and it only evaluates the ranking relationship of preference pairs.

[0055] The loss function based on Bayesian Personalized Ranking can be represented, for example, by mathematical formulas (4) to (6).

[0056]

Number

[0057]

Number

[0058]

Number

[0059] Here, Θ represents the parameter to be optimized. λ θ represents the coefficient of the regularization term. P represents the set of all preferences. P u + represents the set of preferences evaluated as positive examples by the user. P u x>y represents the set of pairs of preferences compared by the user.

[0060] As shown in Equation (5), when the feedback is only positive examples, unevaluated preferences by the user are sampled as negative examples (negative sampling). When considering the hierarchical structure of preferences, as shown in Equation (6), as pairwise feedback, the calculation of the loss may be limited to items at the same hierarchical level only.

[0061] As described above in FIG. 7, when the preferences are labeled, label propagation may be performed to learn the relationality of the hierarchical structure of the preferences. That is, based on the labels of the lower-level preferences, the labels of the higher-level preferences may be determined. For example, assume that a positive label is given to a lower-level preference of "Regarding insurance application for medical expenses". In this case, positive labels are also given to "Economic burden" and "Matters regarding the burden of treatment", which are higher-level preferences of "Regarding insurance application for medical expenses". Furthermore, the labels of the higher-level preferences may be determined according to the ratio of the labels of the lower-level preferences. For this, for example, Graded Implicit Feedback etc. may be utilized.

[0062] Returning to the description of the flowchart. Next, the response determination function 123 determines the next response to be given to the patient for the patient's response based on the estimated preferences and the patient's attribute information (step S104).

[0063] FIG. 9 is a diagram for explaining a method for determining a response to be given to a patient. As shown in the figure, for example, the response determination function 123 may determine a response using DQN (Deep Q-Network), which is a representative method of reinforcement learning. The DQN shown in equations (7) to (9) may be used.

[0064]

Number

[0065]

Number

[0066]

Number

[0067] s his represents a vector obtained by encoding the dialogue history with the patient. s σ represents the confidence interval of the preference. Q(s,a) represents the action value function. s represents the state. a represents the action (Ask Question or Recommend).

[0068] As the state s for using the Q value as the input to the neural network of the DQN, a vector combining the dialogue history and the estimated confidence of the preference, which is considered to have a strong influence on the determination of questions and recommendations, is used.

[0069] The encoding of the dialogue history is, for example, when the content of a dialogue of up to 10 turns is expressed with questions as ±1 (+ for success, - for failure), recommendations as ±2 (+ for success, - for failure), and un-reached turns as 0, s his is a vector represented as [+1, -1, -2, +1, +2, 0, 0, 0, 0, 0]. Here, a turn means a series of dialogues in which a question is asked to the patient and the patient answers it.

[0070] The confidence interval of the preference means the uncertainty of the rating for the preference during the interaction.

[0071] Equation (10) represents the loss function used for the learning of DQN. As in the general DQN, the TD error based on the Bellman equation may be used for the loss function.

[0072]

Number

[0073] s j represents the state at step j. a j represents the action at step j. r j represents the reward at step j. d j represents a flag indicating whether step j is the end (d j ∈ {0, 1}). γ represents the discount rate.

[0074] The reward r used in reinforcement learning may be set to any of the following, for example.

[0075] r suc : Approval of the summary report (Strongly Positive Reward) r part : Partial approval of the summary report (Positive or Negative Reward depending on a ratio of rejection) r rej : Rejection of the summary report (Strongly Negative Reward) r quit : End of the conversation (Strongly Negative Reward) r turn : Cost consumed per turn (Slightly Negative Reward)

[0076] The summary report generation function 123A included in the response determination function 123 generates a summary report as a response to the patient, and the filtered response generation function 123B included in the response determination function 123 generates a filtered response as a response to the patient.

[0077] Figure 10 is a diagram for explaining a method of generating a summary report and a filtered response. The summary report generation function 123A and the filtered response generation function 123B generate a summary report and a filtered response while adapting to the answers from the patient obtained during the dialogue. Online learning is performed for this.

[0078] Figure 11 is a diagram showing an example of an online learning algorithm. P cand ask represents the preference of the question candidate. P cand rec represents the preference of the recommendation candidate. P a represents the a-th preference. P a child is P a represents the set of preferences corresponding to the children of P. A i ,b i represents the parameter of the Contextual Bandit.

[0079] The algorithm shown in Figure 11 is an algorithm for balancing exploration and exploitation of selections in an interactive recommendation system based on Linear Thompson Sampling (LinTS), which is a type of Contextual Bandit Algorithm. While estimating the user's preference, on the other hand, considering the uncertainty at the initial stage of the dialogue, this algorithm is characterized by sampling the user's feature vector from a certain probability distribution.

[0080] For example, as shown in the 3rd to 4th lines, by using the feature vector of the user sampled from a multi-dimensional normal distribution having the mean vector and covariance matrix updated during the conversation based on LinTS, a rating adapted to the response from the user during the conversation is estimated.

[0081] As shown in the 1st line, the feature vector specified by the estimation function 122 is set as the initial value of the mean vector of the multi-dimensional normal distribution. As a result, even at the initial stage of the conversation, it is possible to estimate with a certain degree of accuracy relying on the observed features of the user.

[0082] The process shown in the 7th to 11th lines is the process when asking the user two preferences pairwise.

[0083] The process shown in the 10th line is the process for reflecting the feedback on the pairwise question in the Contextual Bandit. A reward is set for the difference in the feature vectors of the two preferences.

[0084] In the question about preferences, it is desirable to start by asking about the upper-level preferences first and then dig deeper into the lower-level preferences for the preferences the user is interested in. Therefore, the preference P cand ask of the question candidate is set with the upper-level preference as the initial value, and the lower-level preference is added to the candidate according to the response from the user (11th line).

[0085] The process shown in the 12th to 21st lines is the process when recommending a summary report composed of multiple preferences to the user.

[0086] Since feedback can be obtained simultaneously for multiple preferences, an absolute reward is set for each preference. Here, since it is assumed that only the lowest level of preferences is included in the summary report, the preference P cand recSets the lowest preference as the initial value and removes the recommended preferences from the candidates (line 19).

[0087] The update of the parameters of the Contextual Bandit from lines 22 to 26 is the same as the algorithm of LinTS.

[0088] Lines 8 and 13 are the parts that determine the preferences to be questioned and the preferences to be recommended using the estimated rating r(∼). i,j There are various ways to determine these preferences.

[0089] Typically, among multiple preferences, the preferences may be adopted in descending order of the estimated rating r(∼). i,j If preferences with a small rating are also considered important, the preferences may be adopted in descending order of the absolute value of the estimated rating r(∼). i,j

[0090] Also, for pairwise questions, it is not always optimal to compare the top two preferences of the estimated rating r(∼). Other elements besides the estimated rating r(∼) may be considered. i,j i,j

[0091] This algorithm can also be extended when the same user uses it multiple times. In this case, considering that the user's values change over time, uncertainty may be introduced into the user's feature vector again. Specifically, while setting the previous user's feature vector as the initial value in line 1, uncertainty may be introduced again by adjusting the value of the covariance matrix A i -1 or the hyperparameter σ that controls uncertainty.

[0092] ​​​FIG. 12 is a diagram for explaining a method of generating a summary report. As shown in the figure, the summary report generation function 123A may preferentially include in the summary report the preferences with higher ratings among the lowest-level preferences. Also, the summary report generation function 123A may include in the summary report the preferences with lower ratings.

[0093] Furthermore, the summary report generation function 123A may also determine, based on the ratings, the categories to be included in the summary report according to the hierarchical structure. For example, as shown in ItemX, when the rating of the preference in the upper layer is lower than the ratings of other preferences in the same upper layer, the preference in the upper layer with a lower rating and the preferences existing below this preference in the upper layer may be excluded from the summary report as one category.

[0094] As described with reference to FIG. 3, the summary report can be approved / rejected for all preferences or for each preference. Furthermore, the summary report may be approved / rejected for each category, or the rejected preferences may be modifiable by the user.

[0095] FIG. 13 is a diagram for explaining a method of generating a filtered response. When the question to be asked to the patient as the filtered response is "comparison of two preferences", the filtered response generation function 123B may select any two preferences, such as ItemY and ItemZ in the figure. The filtered response generation function 123B may select two preferences with the same hierarchical level or two preferences with the same parent preference in consideration of the hierarchical structure.

[0096] In addition to "comparison of two preferences", the filtering response generation function 123B may also use, as the filtering response, "Yes / No for one preference", "rating for one preference", "reordering of multiple preferences", "multiple selections from multiple preferences", etc.

[0097] Returning to the description of the flowchart. Next, the output control function 124 outputs the determined response (summary report or filtering response) (step S106).

[0098] For example, the output control function 124 transmits the determined response to the terminal device 10 via the communication interface 111. As a result, a filtering response as shown in FIG. 2 or a summary report as shown in FIG. 3 is displayed on the screen of the terminal device 10. Also, the output control function 124 may cause the determined response to be displayed on the display 113a. Thereby, the processing of this flowchart is completed.

[0099] FIG. 14 is a diagram showing another example of the GUI screen of the terminal device 10. In the summary report displayed on the GUI screen, it may be displayed so as to distinguish whether the content is based on a specific answer of the patient himself / herself or is an estimate. For example, it may be displayed with an icon, and when the mouse is over the corresponding location, the corresponding specific answer of the patient himself / herself may be displayed.

[0100] According to the embodiment described above, the processing circuit 120 of the information processing apparatus 100 acquires a patient's response and estimates the patient's preferences and ratings regarding medical treatment based on the response. The processing circuit 120 determines a response to the patient's response based on the preferences and ratings. Specifically, the processing circuit 120 determines at least one of a narrowing-down response, which is a response made to the patient's response to narrow down a plurality of preferences, and a summary report including an approval request to the patient for each of the plurality of preferences, as the response to the patient's response. Then, the processing circuit 120 transmits the response to the patient's response to the terminal device 10 via the communication interface 111 or causes it to be displayed on the display 113a. Thereby, it is possible to assist in efficiently questioning the patient without the involvement of medical staff in order to know the patient's preferences regarding medical treatment. In particular, even when there is a large amount of information to be collected by questioning the patient, it is possible to efficiently question the patient without the involvement of medical staff.

[0101] (Other Embodiments) Hereinafter, other embodiments will be described. In the above-described embodiment, it has been described that ratings for the preferences of the target patient are estimated using pre-optimized (pre-learned) matrix factorization, but the present invention is not limited to this. For example, instead of matrix factorization, preferences and their ratings may be estimated using a large language model such as ChatGPT (Chat Generative Pre-trained Transformer). Also, in the above-described embodiment, it has been described that a response to the patient is determined using reinforcement learning (e.g., DQN), but the present invention is not limited to this. For example, instead of reinforcement learning, a response to the patient may be determined using a large language model.

[0102] When a large language model is used, the information output by the large language model is subjected to natural language processing and provided to the patient in a state where it is converted into plain text (character string) recognizable by the patient. Next, preprocessing is performed to convert the input of plain text from the patient into an input format for natural language processing and input it to the estimation function 122. Then, natural language processing is performed in the estimation function 122. In this case, the natural language processing model may be obtained by additionally training a general-purpose model with the dataset actually input to or output from the estimation function 122.

[0103] Although several embodiments have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes 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 its equivalent scope.

Explanation of Reference Numerals

[0104] 10... Terminal device, 20... Medical database, 100... Information processing device, 111... Communication interface, 112... Input interface, 113... Output interface, 113a... Display, 114... Memory, 120... Processing circuit, 121... Acquisition function, 122... Estimation function, 123... Response determination function, 123A... Summary report generation function, 123B... Narrowed-down response generation function, 124... Output control function

Claims

1. An acquisition unit that acquires a user's response; An estimation unit that estimates medical information including a plurality of preferences of the user regarding medical treatment based on the response; A response determination unit that determines a response to the response based on the medical information; An output control unit that outputs the response via an output interface, and includes: The response determination unit determines at least one of a narrowing-down response that is a response made to the user's response in order to narrow down the plurality of preferences, and an approval request to the user for each of the plurality of preferences, as the response. An information processing apparatus.

2. When a series of dialogues in which the user answers the response is repeated, the response determination unit determines the response to the most recent response based on the plurality of responses obtained in the repeated dialogues. The information processing apparatus according to claim 1.

3. The response determination unit determines the response to the most recent response based on, in addition to the plurality of responses, the attributes of the user. The information processing apparatus according to claim 2.

4. The response determination unit determines the response to the most recent response based on the attributes of the user and at least one of the user's response to the narrowing-down response and the user's response to the approval request to the user, as either the narrowing-down response or the approval request to the user. The information processing apparatus according to claim 3.

5. The response determination unit uses, as the response, a summary report in which the plurality of preferences are arranged and an object for selecting whether to approve some or all of the plurality of preferences is arranged. The information processing apparatus according to claim 1 or 2.

6. The acquisition unit acquires the user's operation on the object arranged in the summary report as the user's response to the approval request to the user. The information processing apparatus according to claim 5.

7. When estimating the medical information, the estimation unit calculates a predetermined index for each of the plurality of preferences. The response determination unit selects items to be included in the summary report from among the plurality of preferences based on the respective predetermined indexes of the plurality of preferences. The information processing apparatus according to claim 5.

8. The plurality of preferences has a hierarchical structure, and the response determination unit selects, based on the respective predetermined indicators of the plurality of preferences and the hierarchical structure, the preferences to be included in the summary report from among the plurality of preferences. The information processing apparatus according to claim 7.

9. The plurality of preferences has a hierarchical structure, and when the narrowing-down response includes comparing at least two preferences among the plurality of preferences, the response determination unit selects the two preferences based on the hierarchical structure. The information processing apparatus according to claim 1 or 2.

10. An information processing method using a computer, comprising: obtaining a response of a user; estimating medical information including a plurality of preferences of the user regarding medical treatment based on the response; determining a response to the response based on the medical information; outputting the response via an output interface; determining, as the response, at least one of a narrowing-down response, which is a response made to the response of the user to narrow down the plurality of preferences, and an approval request to the user for each of the plurality of preferences; An information processing method including the above.

11. A program for causing a computer to execute, comprising: obtaining a response of a user; estimating medical information including a plurality of preferences of the user regarding medical treatment based on the response; determining a response to the response based on the medical information; outputting the response via an output interface; determining, as the response, at least one of a narrowing-down response, which is a response made to the response of the user to narrow down the plurality of preferences, and an approval request to the user for each of the plurality of preferences; A program including the above.

Images