Electronic medical record creation support device, electronic medical record creation support method, and program

Abstract

To provide an electronic medical chart creation support device, an electronic medical chart creation support method, and a program for supporting creation of an electronic medical chart intended by a user.SOLUTION: An electronic medical chart creation support device 10 according to the present invention comprises: data acquisition means 151 that acquires time-series data based on a medical examination of a subject; a selection receiving unit 152 that receives a user's selection from a plurality of input items; an inference target data acquisition unit 153 that, on the basis of the timing when the selection receiving unit 152 receives the user's selection of the item and the selected selection item, acquires inference target data obtained by extracting data corresponding to a predetermined section from the time-series data; an inference unit 155 that performs inference processing on the inference target data and infers input candidates for the selection item; and a display control unit 158 that displays the input candidates on a display 14 on the basis of a result of the inference performed by the inference unit 155.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to an electronic medical record creation support device, an electronic medical record creation support method, and a program for assisting in the creation of electronic medical records.

Background Art

[0002] At the site of medical examinations, there are cases where the doctor's line of sight is directed towards the personal computer for data input into the electronic medical record, resulting in insufficient communication with the subject. In response to such problems, efforts have also begun to automate electronic medical record input using AI technology (Non-Patent Document 1). In addition, a technique for generating a medical document by combining the analysis results of input voice data and diagnostic target image data has been proposed (Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Non-Patent Documents

[0004]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, it is difficult to input the information intended by the user into an electronic medical record composed of a plurality of input items from the vast amount of information based on medical examinations.

[0006] In view of the above circumstances, the present invention aims to provide an electronic medical record creation support device, an electronic medical record creation support method, and a program that support the creation of electronic medical records as intended by the user. [Means for solving the problem]

[0007] An electronic medical record creation support device according to one embodiment of the present invention is an electronic medical record creation support device that assists in the creation of an electronic medical record consisting of a plurality of input items, and the examination of the subject Collected inside A data acquisition method for obtaining time-series data, and multiple input items To select one of the following as an option: User instructions A selection reception unit that accepts selections, and the selection reception unit accepts user Instructions Timing of receiving The data for a predetermined interval determined according to the selected item, including the previous time-series data, is used as the data to be inferred. The system is characterized by comprising: an inference target data acquisition unit that extracts data from time-series data; an inference unit that performs inference processing on the inference target data and infers input candidates for selected items; and a display control unit that displays the input candidates on a display unit based on the inference results from the inference unit. [Effects of the Invention]

[0008] According to the present invention, it is possible to support the creation of electronic medical records as intended by the user. [Brief explanation of the drawing]

[0009] [Figure 1] A system configuration diagram including an electronic medical record creation support device according to the first embodiment. [Figure 2] A functional configuration diagram of an electronic medical record creation support device according to the first embodiment. [Figure 3] A flowchart illustrating the processing procedure of an electronic medical record creation support device in the first embodiment. [Figure 4] A flowchart showing the processing procedure of an electronic medical record creation support device in a modified example of the first embodiment. [Figure 5] An example of an electronic medical record input screen in the first embodiment. [Figure 6]An example of the acquisition rules for the data to be inferred in the first embodiment. [Figure 7] An example of the conversation content, the output of the first inference unit, and the output of the second inference unit in the first embodiment. [Figure 8] An example of how input candidates are displayed in the first embodiment. [Modes for carrying out the invention]

[0010] Hereinafter, exemplary embodiments for carrying out the present invention will be described in detail with reference to the drawings. However, the dimensions, materials, shapes, and relative arrangements of components described in the following embodiments are arbitrary and can be changed according to the configuration of the device to which the present invention is applied or various conditions. In addition, the same reference numerals are used between drawings to indicate elements that are identical or functionally similar.

[0011] [First Embodiment] The electronic medical record creation support device in this embodiment infers the input content of the electronic medical record from time-series data such as conversation data collected during a medical examination, according to the input items selected by the user creating the electronic medical record, such as a physician, and displays it on the display unit as an input candidate. For example, if the device receives a selection of an input item corresponding to medication from among multiple input items, it obtains the information necessary for medication from the time-series data, which is audio data. Note that time-series data also includes data based on audio data, such as text data transcribed from audio data. When acquiring audio data, the electronic medical record creation support device infers information related to medication (drug name / dosage / medication period, etc.) from the physician's statements in the audio data and displays it on the display unit as an input candidate. Furthermore, if the device receives a selection of an input item corresponding to symptoms from the user, such as a physician, it infers information related to the symptoms (chief complaint, etc.) from the audio data or video data.

[0012] The system configuration, including the electronic medical record creation support device, will be explained using Figure 1. This system consists of an electronic medical record creation support device 10, a time-series data generation device 20, and a network 30 connecting them.

[0013] The electronic medical record creation support device 10 receives editing such as display and input of electronic medical record information and saves the editing results. Further, it cooperates with various time-series data generation devices 20 in the examination room via the network 30, acquires time-series data, and displays inferred input candidates for input items with respect to the acquired time-series data. Here, the data generation device 20 and the electronic medical record creation support device may be configured as one unit. Further, the display control of the input candidates in the electronic creation support device 10 may be not only for the display connected to the electronic medical record creation support device 10 but also for display processing on other displays or the like via the network.

[0014] The time-series data generation device 20 collects various time-series data in the examination room. Further, it appropriately transmits the acquired time-series data to the electronic medical record creation support device 10 via the network 30. The time-series data generation device 20 includes measuring instruments such as a microphone 21, a camera 22, a stethoscope 23, and a tongue depressor container 24, and generates time-series data from its own sensor information and the like. Note that there may be a plurality of the same type of devices. Further, the time-series data generation device 20 may combine event information such as the timing when data is acquired with various sensor information and transmit it to the electronic medical record creation support device 10.

[0015] The network 30 cooperates the electronic medical record creation support device 10 and the time-series data generation device 20.

[0016] FIG. 2 is a functional configuration diagram of the electronic medical record creation support device 10 according to the first embodiment of the present invention. The electronic medical record creation support device 10 includes a communication IF 11, a ROM 12, a storage unit 13, a display unit 14, and a control unit 15 as its functional configurations.

[0017] The communication IF 11 is realized by a LAN card or the like and controls communication between an external device (for example, the time-series data generation device 20) and the electronic medical record creation support device 10.

[0018] ROM12 is implemented using non-volatile memory, etc., and stores various programs and other data.

[0019] The memory unit 13 is implemented using volatile memory or an SSD / HDD, and stores various information such as inference rules and inference models used in the inference unit described later.

[0020] The display unit 14 consists of a display or the like and displays various information to the user. Note that the display unit 14 is not essential to the configuration of the electronic medical record creation support device 10, and may be implemented by other devices connected to the electronic medical record creation support device 10.

[0021] The control unit 15 is implemented by a CPU (Central Processing Unit) or the like, and provides overall control over the various processes performed by the electronic medical record creation support device 10.

[0022] The control unit 15 has a functional configuration that includes a data acquisition unit 151 for acquiring time-series data, and a selection acceptance unit 152 for accepting user selections from a plurality of input candidates that constitute the electronic medical record. The control unit 15 also includes an inference target data acquisition unit 153 that acquires inference target data by extracting data corresponding to a predetermined interval from the time-series data based on the timing at which the selection acceptance unit 152 accepts a selection from the user for an input item and the selected item, an inference unit 155 that infers input candidates for the selected item from the acquired inference target data, and a display control unit 158 ​​that displays the input candidates on the display unit.

[0023] The functions of each part of the control unit 15 will be explained below, along with the flowchart shown in Figure 3. Note that here, a configuration is shown where a single device encompasses all the functional components of the electronic medical record creation support device 10; however, these components may also be implemented as a system composed of separate or partially independent devices.

[0024] Figure 3 is a flowchart showing the processing procedure executed by the control unit 15 of the electronic medical record creation support device 10. The data acquisition unit 151 in the control unit 15 starts acquiring time-series data from the time-series data generation device 20 when the consultation begins, and sequentially stores the acquired time-series data in the storage unit 13, associating it with the time information of acquisition.

[0025] In step S301, the selection reception unit 152 receives a selection operation for the input item while the electronic medical record creation support device 10 is in the process of creating the electronic medical record. If the selection reception unit 152 receives a selection operation for an input item for the electronic medical record, it proceeds to step S302.

[0026] In step S302, the selection reception unit 152 acquires information on the selected input item selected by the user and information on the timing at which the selection was received by the user, and transmits this information to the inference target data acquisition unit 153. Here, the timing information refers to time information that allows for temporal agreement or sequential relationship with time-series data. Furthermore, the selection reception unit 152 transmits the acquired information on the selected item to the inference unit 155 and proceeds to the next step.

[0027] In step S303, the inference data acquisition unit 153 acquires inference target data by extracting data corresponding to a predetermined interval from the time series data acquired from the data acquisition unit 151, based on the timing information of the user's item selection transmitted from the selection reception unit 152 and the information of the selected item. The type of inference target data acquired by the inference data acquisition unit 153 and the predetermined interval for acquisition are determined according to the selected item information based on the inference target data acquisition rules defined separately. The inference target data acquisition unit 153 finally transmits the inference target data acquired from the time series data to the inference unit 155 and proceeds to step S304.

[0028] In step S304, the inference unit 155 selects an inference model corresponding to the information of the selected items received from the selection receiving unit 152 in step S302. The inference unit 155 selects and acquires from the storage unit 13 a first inference model (keyword inference model) that performs inference using the first inference method and is used by the first inference unit 156, and a second inference model (input candidate inference model) that performs inference using the second inference unit 157. The inference model to acquire is determined according to the information of the selected items based on acquisition rules defined separately. Note that the selection of an inference model is not mandatory; a pre-configured inference model may be used, or a highly general-purpose inference model may be used. If model selection is not required, the model selection is skipped as appropriate.

[0029] The inference model is a model based on machine learning or deep learning. This model is generated by applying a learning process using training data in a known manner. The training data consists of, for example, time series data as training data and ground truth labels, which are labels that represent the correct answers. The training data may be time series data converted into vector information using various morphological analysis processes, or time series data may be used as pseudo-image data. The ground truth labels are information about the class that the model infers. For example, when performing inference about prescriptions among the input items, the model is generated by training the model using training data with ground truth labels that include drug names, prescription amounts, durations, etc.

[0030] When the inference unit 155 selects an inference model, it applies the selected inference model to the data to be inferred transmitted from the data acquisition unit 153 and infers input candidates for the selected items. Specifically, the first inference unit 156, which constitutes the inference unit 155, uses the first inference model (keyword inference model) to infer keywords related to the input items included in the data to be inferred. Furthermore, the second inference unit 157, which constitutes the inference unit 155, uses the second inference model (input candidate inference model) to infer input candidates for the selected items from the keywords inferred by the first inference unit 156. The inference model outputs the input candidates and the likelihood for each input candidate as the inference result. The inference unit 155 transmits the inference result from the inference model to the display control unit 158 ​​and proceeds to step S305.

[0031] In step S305, the display control unit 158 ​​causes the display unit 14 to display input candidates based on the inference results. The display control unit 158 ​​displays, for example, the input candidates with the highest likelihood based on the likelihood of the inference results. The display control unit may also obtain the size of the display area on which the input candidates are displayed and may vary the number of input candidates and the font size depending on the size of the display area. The selection reception unit 152 can then receive a selection from the user for the input candidates displayed by the display control unit 158, and the selected content can be reflected in the electronic medical record.

[0032] By adopting this configuration, the electronic medical record creation support device 10 can perform inferences from a vast amount of time-series data, focusing on the information necessary for inferring input candidates for selected items. By providing users with input candidates with high accuracy, it enables support for creating electronic medical records as intended by the user.

[0033] (modified version) Up to this point, we have described a configuration in which the display control unit 158 ​​displays input candidates for the selected item based on the inference results. Here, using Figure 4, in addition to the above-described flow, the inference unit 155 further determines whether the input candidates obtained as inference results satisfy predetermined conditions. The predetermined conditions are those that define whether the number of input candidates included in the inference results exceeds a threshold set for at least one of the likelihood of the input candidates.

[0034] The processes from steps S301 to S304 in Figure 4 are the same as those of the corresponding steps explained using Figure 3, so their explanation is omitted. However, in this modified example, the process proceeds to step S307 after completing step S304. Also, the process proceeds to step S303 after completing step S309.

[0035] In step S307, the inference unit 155 determines whether a sufficient number of input candidates have been obtained from the inference result of the inference model. The inference unit 155 may make this determination by checking whether there is a predetermined number of input candidates with a likelihood of a predetermined value or higher. If the inference result satisfies the predetermined conditions, the inference unit 155 proceeds to step S310; otherwise, it proceeds to step S308.

[0036] In step S308, the inference data acquisition unit 153 determines whether there are any unanalyzed intervals in the time series data acquired from the data acquisition unit 151. If there are intervals that have not been analyzed, the process proceeds to step S309.

[0037] In step S309, the data acquisition unit 153 acquires data for inference again from the time series data, targeting intervals that are not subject to analysis, in other words, intervals different from the predetermined intervals determined by the selected items and the timing of selection, and proceeds to step S304.

[0038] Step S310 is the same as step S305, so its explanation is omitted.

[0039] With the configuration disclosed in this modified version, even if the predetermined interval determined based on the timing of the input item selection and the selected item does not contain enough data for inference by the inference unit 155, data targeting an interval different from the predetermined interval can be acquired as inference target data, and the inference process can be performed again to provide the user with highly accurate input candidates.

[0040] Here, using Figure 5, an example of an electronic medical record in this embodiment and its modified form will be explained. In the example in Figure 5, the electronic medical record 40 is composed of multiple input items, including basic information of the subject such as name, as well as "symptoms," "findings," "prescriptions," and "treatment plan" obtained during the examination.

[0041] A physician can select input items and input content for the electronic medical record displayed on the display unit 14 of the electronic medical record creation support device 10 using a mouse, keyboard, etc. (not shown) connected to the electronic medical record creation support device 10.

[0042] Furthermore, the user can select the input content to be reflected in the electronic medical record from among the input candidates corresponding to the selection items displayed on the display unit 14. When a doctor opens the electronic medical record for examination, the data acquisition unit 151 of the electronic medical record creation support device 10 starts collecting time-series data from the time-series data generation device 20 installed in the examination room. Figure 4 shows an example in which the input item "prescription" is selected with the mouse cursor 41. When a doctor selects the input item "prescription," the selection reception unit 152 of the electronic medical record creation support device 10 detects the selection of the input item and infers and displays input candidates for the input item according to the processing procedure shown in Figures 3 and 4.

[0043] Figure 6 shows an example of the acquisition rule for inference target data in the first embodiment and its modified form. The acquisition rule stores, in association with the type of inference target data to be acquired and a predetermined interval starting from the timing of the user's selection, according to the selected item. The rule may be stored in a table format or other format depending on the item.

[0044] If the selected item is "prescription," the inference target data acquisition unit 151 uses voice data, which is time-series data collected from the microphone 21 (an example of the time-series data generation device 20), as the inference target data. It also extracts the most recent 2-minute interval, going back in time from the moment the selected item was chosen, as the inference target data. Furthermore, the inference target data acquisition unit 151 targets the voice data of the doctor included in the time-series data. Here, since the target is the voice data of the doctor from the time-series data, the inference unit 155 may use the first inference unit 156 to identify the speaker and infer keywords, and the second inference unit 157 to infer input candidates from the keywords. In other words, the voices of the subject or nurse, or the sounds of operating equipment, etc., are not used as input candidates. Furthermore, if the inference unit 155 cannot obtain a predetermined number (5) of input candidates in the most recent 2 minutes, it may perform the inference process again using time-series data collected from the microphone 21 by going back in time, according to the flow described in the modified example above, as the inference target.

[0045] Figure 7 shows an example of inferring potential inputs for an electronic medical record from audio data.

[0046] Figure 7(a) shows an example of inference target data (voice data) acquired by the inference target data acquisition unit 153. The inference target data acquisition unit 153 extracts the interval (the most recent 2 minutes) defined by the acquisition rules in Figure 6 from the time-series data acquired by the microphone 21 recorded from the start of the examination to the present, and uses it as the inference target data. In this case, several exchanges of conversation between the doctor and the subject are recorded as voice data.

[0047] Figure 7(b) shows an example of keywords inferred by the first inference unit 156. The first inference unit 156 infers keywords related to the input item (prescription) from the data to be inferred in Figure 7(a) using the first inference model, the keyword inference model. The inference rules at this time may be incorporated into the keyword inference model, or conditions may be passed to the keyword inference model as parameters. In this case, the first inference unit 156 extracts the keywords "iron supplement" and "about one month" from the doctor's statement. Thus, the keyword inference model is characterized by having the function of identifying the speaker in the audio data and extracting keywords only from the statement of the speaker to be recognized. The first inference model is an inference model that has been trained using time-series data as training data and training data with keywords as the correct labels. Alternatively, it may be configured to extract predetermined keywords by pattern matching or the like.

[0048] Figure 7(c) shows an example of input candidates inferred by the second inference unit 157. The second inference unit 157 uses the second inference model, the input candidate inference model, to infer input candidates for selection items (prescriptions) from the keywords in Figure 7(b). Here, from the information "iron supplements," it infers the input candidates "AAA tablets 105mg" and "1 tablet / day." It also infers the input candidate "35 days' worth" from the information "about one month." This second inference model is an inference model that has been trained using keywords as training data and at least one of the following as the correct label: drug name, prescription amount, or duration.

[0049] Furthermore, since the inference so far has only generated three input candidates, "AAA tablets 105mg", "1 tablet / day", and "35 days' supply", the inference unit 155 analyzes the time-series data of microphone 21 by going further back in time. At this time, the next analysis target may be the 2 minutes from 2 minutes ago to 4 minutes ago, or the next target may be the 2 minutes from 1.5 minutes ago to 3.5 minutes ago to correct for analysis errors at the boundaries of the intervals.

[0050] Furthermore, while Figure 7(c) only infers a maximum of one "drug name," "daily dosage," and "duration of medication," it is also acceptable to infer multiple input candidates, as shown in Figure 7(c'). For example, from "iron supplement," two drug names with the same active ingredient, "AAA Tablets 105mg" and "BBB Tablets 100mg," may be inferred. Alternatively, two "daily dosages," "1 tablet / day" and "2 tablets / day," may be inferred from "iron supplement," allowing for adjustment of the dosage according to symptoms, weight, etc.

[0051] Finally, Figure 8 shows an example in which the display control unit 158 ​​displays the input candidates inferred by the electronic medical record creation support device 10 in the prescription field of the electronic medical record. After the doctor clicks the "Prescription" input item in Figure 5, the electronic medical record creation support device 10 displays the input candidates 42 according to the processing procedure shown in Figure 3 and / or Figure 4. The input candidates 42 display the input candidates related to "Drug Name" ("AAA Tablets 105mg" and "BBB Tablets 100mg") from Figure 7(c') as a pull-down menu in the currently focused "Name" field. The doctor may select one of these input candidates, or they may ignore the input candidates and enter the drug name as usual using the keyboard, etc. When the doctor selects an input candidate, the selected drug name is reflected in the currently focused "Name" field. Similarly, when the focus shifts to "Quantity" or "Days," input candidates such as "1 tablet / day," "2 tablets / day," and "35 days' supply" and "30 days' supply" are displayed as pull-down menus. Furthermore, the number and combination of inference models used by the inference unit 155 are not limited to those described in this embodiment and may be replaced by known natural language processing methods.

[0052] This configuration allows for the acquisition of data to be inferred from time-series data, and then infers and displays input candidates from that data, thereby providing users with highly accurate input candidates. Furthermore, if the input candidates match the candidates expected by the physician, the physician can create electronic medical record information by selecting the desired candidate from the input candidates, allowing the physician to communicate with the patient in a way that is eye-to-eye. Even if the candidates do not meet expectations, the physician can still enter the drug name using the same procedure as before. This configuration also has the effect of preventing the physician from having to perform any extra operations even if an inference error occurs in the electronic medical record creation support device (inferring an incorrect candidate or no candidates).

[0053] In this example, the input options are displayed in a dropdown menu, but it is also acceptable to automatically fill in the input fields. The doctor can then complete the electronic medical record entry by making corrections to the automatically entered information if necessary. If no corrections are needed, the doctor can immediately return their attention to the patient.

[0054] (Modification 2) The embodiments and modifications described above illustrate an example of acquiring data for a predetermined interval from time-series data based on selected items and timing. In this modification, further rules for acquiring data to be inferred are defined, including an inference model, data type, and interval, corresponding to the selected items, and inference of input candidates is performed in accordance with these rules. That is, the data acquisition unit determines the type of data to be acquired in accordance with the data acquisition rules, and the inference unit performs inference using an inference method (inference model) corresponding to the determined type.

[0055] The electronic medical record creation support device 10 can perform highly accurate inferences for each input item by inferring based on an inference model, time-series data, and intervals corresponding to the selected items.

[0056] Figure 6 shows the rules for acquiring the data to be inferred for each selected item. Note that the inference model is automatically determined based on the selected items and the data to be inferred, so this is omitted from the explanations from Figure 5 onwards.

[0057] The acquisition rules for this modified example will be explained in accordance with Figure 6. If the user selects "prescription," then, as previously explained, audio data from the "microphone" is used as time-series data, and the doctor's statements are further used as the data to be inferred as input candidates.

[0058] When the selection reception unit 152 receives a selection for the input item corresponding to the "prescription" field, the inference target data acquisition unit 153 extracts a predetermined interval defined by the inference target data acquisition rules from the time series data acquired from the "microphone" and generates the inference target data.

[0059] Subsequently, in the inference unit 155, the first inference unit 156 infers keywords from the data to be inferred, and the second inference unit 157 infers input candidates for the electronic medical record from the keywords.

[0060] Here, since prescription information is very likely to be explained by a doctor to the subject, the inference unit 155 can increase the likelihood of appropriately inferring prescription information by focusing its inference on the doctor's statements. Furthermore, by excluding statements from the subject, who is unlikely to speak about prescriptions, from the inference target, the likelihood of presenting incorrect input candidates is reduced.

[0061] Furthermore, the data acquisition unit 152 for inference limits the data to be acquired from time-series data to the most recent two minutes prior to the time when the input item was selected. Since prescription information to be entered into the electronic medical record is very likely to be explained immediately before the start of electronic medical record input, inferring from data from the most recent time increases the likelihood of appropriately inferring prescription information. Moreover, as explained in Modification Example 1, if input candidates cannot be inferred, the data acquisition unit for inference 152 performs inference on the data to be inferred in the preceding time interval. Also, regarding prescriptions, it is unlikely that a prescription different from the one last explained by the doctor to the subject will be given, and it is also unlikely that the doctor will finalize the prescription without explaining it to the subject. Therefore, by establishing the data acquisition rules for inference as shown in Figure 6, it is possible to provide the user with highly accurate input candidates.

[0062] For the reasons mentioned above, actively inferring input candidates even from older data increases the likelihood of appropriately inferring prescription information. Furthermore, if a sufficient number of input candidates cannot be inferred, the data acquisition unit 152 acquires data by going further back in time, and the inference unit 155 performs the inference.

[0063] Since doctors may be describing multiple types of prescriptions, the inference unit 155 is more likely to be able to infer prescription information completely by inferring multiple input candidates. Actively inferring input candidates from older data also increases the risk of presenting incorrect input candidates. However, since prescription information is entered using names, quantities, etc., rather than free text, doctors can relatively easily decide whether to accept or reject a reasonable number of candidates. For these reasons, the inference unit 155 is more likely to be able to infer prescription information completely while avoiding problems that would hinder inference electronic medical record input by inferring a reasonable number of input candidates in order from newest to oldest.

[0064] On the other hand, if the input items are "disease name" and "treatment," for example, time-series data obtained from "microphone" and "camera" will be adopted as the type of data to be inferred, and the data from "microphone" will be used to infer the doctor's statements.

[0065] When the selection reception unit 152 detects the start of input into the "Disease Name" or "Treatment" field, the inference target data reception unit 153 generates inference target data by extracting data from a predetermined interval defined by the inference target data acquisition rules from the time-series data acquired from the "microphone" and "camera".

[0066] Subsequently, the inference unit 155 has a first inference unit 156 that infers keywords from the data to be inferred, and a second inference unit 157 that infers input candidates for the electronic medical record from the keywords. At this time, the data acquisition unit 153 for inference may detect specific events from the time-series data of "camera," such as a doctor palpating a subject or using medical tools, and acquire audio data from around the time the event occurred as data to be inferred. The inference unit 155 may also focus on inferring keywords from the audio data from around the time the event occurred.

[0067] While it is highly likely that doctors will explain the diagnosis and treatment to the subject, and because these can be written in free text, making inference difficult, highly accurate inference becomes possible by acquiring event information such as the areas touched by the doctor on the subject or the measuring instruments used during the examination, and then obtaining the data to be inferred based on this event information. Event information includes, for example, information on the timing when the subject was measured with a measuring instrument. Furthermore, if the input items are "diagnosis" or "treatment," the interval for acquiring the data to be inferred from the time-series data is set to the most recent two minutes. If nothing can be inferred from the time-series data of the most recent two minutes, the entire time-series data may be input to the first inference unit as the data to be inferred without setting an interval.

[0068] The inference target data acquisition unit 153 acquires inference target data from the camera's inference target data using the timing when the physician palpates or measures the subject, or when the physician uses a medical tool. The inference unit 155 then performs inference processing on this inference target data and infers keywords from the time before and after the most recent time. Note that the order of switching between "the last 2 minutes" and "the time before and after the time of palpation, etc." may be reversed, or only one of them may be performed. A physician's judgment may change over time, and input candidates for free-form text such as "disease name" and "treatment" can be cumbersome for the physician to review. For these reasons, by having the inference target data acquisition unit 153 use the most recent statement or statements around the time of palpation / tool ​​use as the inference target data for disease name / treatment information, the physician can reduce the effort required to review input candidates that are unlikely to be reflected in the electronic medical record.

[0069] Although the data acquisition unit 153 for inference narrows down the interval to be inferred, sometimes no input candidates are presented at all. However, doctors can still fill in the "Disease Name" and "Treatment" fields themselves as before, thus maintaining work efficiency.

[0070] Furthermore, if the input item is "symptoms," "microphone" and "camera" data are used as time-series data, and the "microphone" data consists of statements made by the doctor and the subject. When the selection reception unit 152 detects a selection for an item corresponding to the "symptoms" field, the inference target data reception unit 153 extracts a predetermined interval defined by the acquisition rules from the time-series data of "microphone" and "camera" to generate inference target data. At this time, the inference target data acquisition unit 153 may further detect the subject's speech events from the inference target data and use conversations within one minute of the subject's speech as the inference target data.

[0071] The inference unit 155 then infers input candidates for the electronic medical record based on the data to be inferred. Since much of the information about symptoms can only be obtained from the subject's "chief complaint," inferring from the subject's statements increases the likelihood of appropriately inferring symptom information. Since information about the case is likely to appear in the subject's own statements and the subsequent statements of the doctor, the inference unit 155 increases the likelihood of appropriately inferring symptom information by focusing its inference on the subject's statements. Since input candidates for "symptoms," which are free-form text, are likely to be noise, the data acquisition unit 152 for inference targets avoids inference from time-series data from times far removed from the subject's statements, thereby reducing the risk of displaying noise as input candidates.

[0072] Furthermore, similar to "disease name" and "treatment," restricting the keyword inference and target time from the camera's inference data (for example, recognizing that the subject is pointing to their abdomen and inferring the keyword "abdomen") increases the likelihood of improving the inference accuracy of the second inference unit 157. Regarding the subject's speech, recognition processing may be performed without limiting the interval based on the selection of input items (i.e., all audio data as inference data). Since the timing of the subject's speech may not be controllable, this reduces the chance of missing the main complaint.

[0073] Note that the interval conditions do not always have to be constant. When inputting the same input item again, only the range since the previous inference may be considered for inference. This prevents unnecessary input candidates from being presented by having the system infer ranges that have already been inferred.

[0074] Furthermore, even if the input items are different, the interval conditions do not always have to be constant. When inputting data sequentially for different items, only the range from the time of inference for the previously entered item onward may be included in the inference target. In other words, the inference target data acquisition unit 153 acquires data as an inference target for inferring input candidates to be entered into the second input item, based on the timing of the user's selection of the first input item.

[0075] For example, when the "chief complaint" is entered first, and then the "treatment" is entered after the medical interview, the inference for the "treatment" can be made from a range after the time the "chief complaint" was inferred. Alternatively, the interval conditions for the later entered items can be limited based on the input candidates used for the previously entered items. For example, if the "chief complaint" is selected from the "input candidates" first, then when entering the "treatment," the inference can be made from a range after the time the keywords that form the basis of the "selection candidate" selected for the "chief complaint" were extracted. Generally, when there is a sequence of events for each input item, the range where related keywords are unlikely to exist is inferred again, preventing unnecessary input candidates from being presented. In other words, the inference target data acquisition unit 153 acquires inference target data for inferring input candidates to be entered in the second input item, using data from the timing of the selection of the first input item as a predetermined interval.

[0076] Since some input items generally have no contextual relationship, it may be possible to decide whether or not to impose such inference interval restrictions on each interval, or to allow for dynamic setting. For example, one could consider not imposing an inference interval restriction between "disease name" and "treatment."

[0077] Note that the time-series data used for inference may be of a type other than those shown in Figure 5. Furthermore, multiple time-series data generation devices may be used to collect time-series data from each. Each will be described in the acquisition rules as a separate time-series data type. The inference target data acquisition unit 152 acquires time-series data from the stethoscope 23, allowing the first inference unit 156 to identify the "time the stethoscope was used" and the "time heart and respiratory sounds were confirmed." By acquiring the "collected heart and respiratory sounds" from the stethoscope 23, keywords representing the characteristics of the heart and respiratory sounds (such as extra heart sounds) can be inferred. Additionally, by acquiring time-series data from the IT-enabled tongue depressor container 24, the first inference unit 156 can identify the "time the tongue depressor was removed" and infer keywords related to the "number of tongue depressors used." By providing multiple microphones 21 and individually recording the voices of the subject and the doctor, the first inference unit can also improve the accuracy of speaker recognition. In this way, the first inference unit can optimize the "time period" for keyword extraction and infer "new keywords" that are not present in audio or video by combining various time-series generating data devices 20. It is not necessary to use audio data as time-series data. It can also infer "symptoms" and "diagnoses" from "heart sounds and respiratory sounds" collected from a stethoscope 23.

[0078] Furthermore, the data used for inference is not limited to time-series data. Unchanging data such as the subject's gender and age, data obtained from pre-examination interviews, and drug databases may also be used as input to the first inference unit 156 and the second inference unit 157 for inference. This makes it possible to perform inferences that show different trends depending on gender and age, or different trends depending on the patient's physical condition and whether or not they are accompanied by someone on the day.

[0079] Furthermore, the timing of inference is not limited to after the selection of input items. The data acquisition unit 151 may extract keywords in real time from time-series data from the microphone or camera. In this case, the time-series data becomes the "keywords," and the first inference unit 156 may select and generate keywords from the keywords in the time-series data to pass to the second inference unit 157. This shortens the processing time for keyword extraction and reduces the time from when the doctor selects an input item until the display control unit 158 ​​displays input candidates.

[0080] As described above, the electronic medical record creation support device 10 according to this embodiment can present input candidates that are narrowed down to those that are easy for a physician to adopt, by combining information from various time-series data generation devices and limiting the inference target to a range with high accuracy. This can support the creation of electronic medical records as intended by the physician and has the effect of streamlining the editing work of electronic medical records without interfering with the operation.

[0081] (Other embodiments) The present invention can also be realized by supplying a program that implements one or more of the functions of the above-described embodiments to a system or device via a network or storage medium, and by having one or more processors in the computer of that system or device read and execute the program. It can also be realized by a circuit (e.g., an ASIC) that implements one or more functions.

[0082] Although the present invention has been described above with reference to embodiments, the present invention is not limited to the embodiments described above. Inventions modified insofar as not to contradict the spirit of the present invention, and inventions equivalent to the present invention, are also included in the present invention. Furthermore, the above embodiments and modifications can be combined as appropriate insofar as not to contradict the spirit of the present invention. [Explanation of Symbols]

[0083] 151 Data Acquisition Unit 152 Selection Reception Section 153 Data acquisition unit for inference 155 Reasoning part 156 First Inference Section 157 Second inference section 158 Display Control Unit

Claims

1. An electronic medical record creation support device that assists in the creation of an electronic medical record consisting of multiple input items, A data acquisition method for acquiring time-series data collected during the examination of a subject, A selection receiving unit that receives user instructions to select one of the aforementioned multiple input items as a selection item, An inference target data acquisition unit extracts data for a predetermined interval determined according to the selected item, including the time-series data prior to the timing when the selection reception unit received the user's instruction, from the time-series data as inference target data; An inference unit that performs inference processing on the aforementioned data to be inferred and infers input candidates for the aforementioned selected items, A display control unit that displays input candidates on a display unit based on the inference results from the inference unit, An electronic medical record creation support device characterized by having the following features.

2. The electronic medical record creation support device according to claim 1, characterized in that the inference unit infers the input candidates using a first inference method for inferring keywords from the data to be inferred, and a second inference method for inferring the input candidates corresponding to the selected items from the keywords.

3. The electronic medical record creation support device according to claim 1 or 2, characterized in that the inference target data acquisition unit acquires data for a predetermined interval determined by the selected item, starting from the timing, as inference target data.

4. The aforementioned data acquisition unit for inference further acquires event information in the examination, The electronic medical record creation support device according to any one of claims 1 to 3, characterized in that it acquires data to be inferred using the event information.

5. The electronic medical record creation support device according to claim 4, characterized in that the event information includes information on the timing when the subject was measured by the measuring instrument.

6. The inference target data acquisition unit determines the type of the inference target data, The electronic medical record creation support device according to any one of claims 1 to 5, characterized in that the inference unit performs inference using an inference method corresponding to the determined type.

7. If the input candidate inferred by the inference unit does not satisfy the predetermined conditions, The electronic medical record creation support device according to any one of claims 1 to 6, characterized in that the inference target data acquisition unit acquires the inference target data again from an interval different from the predetermined interval for the time series data.

8. The electronic medical record creation support device according to claim 7, characterized in that the predetermined condition is a condition that defines whether or not the number of input candidates or the likelihood of an input candidate exceeds a threshold set for at least one of the two.

9. The electronic medical record creation support device according to claim 7 or 8, characterized in that the interval different from the predetermined interval is an interval in which the inference unit has not performed analysis on the time series data.

10. The selection receiving unit receives the selection of multiple items by the user, The electronic medical record creation support device according to any one of claims 1 to 9, characterized in that the inference target data acquisition unit acquires a plurality of inference target data based on the timing of receiving a selection for each of the plurality of items and the selected item.

11. The electronic medical record creation support device according to any one of claims 1 to 10, characterized in that the inference target data acquisition unit acquires the inference target data for inferring input candidates to be entered into the second input item based on the timing of the user's selection of the first input item.

12. The electronic medical record creation support device according to claim 11, characterized in that the inference target data acquisition unit acquires the time-series data from the timing of the selection of the first input item onward as inference target data for inferring input candidates to be entered into the second input item.

13. The electronic medical record creation support device according to any one of claims 1 to 12, characterized in that the display control unit displays the inferred input candidates in the display area.

14. The electronic medical record creation support device according to claim 13, characterized in that the selection receiving unit receives a user's selection from the input candidates displayed in the display area and reflects the selected content in the electronic medical record.

15. The electronic medical record creation support device according to any one of claims 1 to 14, characterized in that the aforementioned time-series data includes data based on audio data.

16. An electronic medical record creation support method performed by an electronic medical record creation support device that assists in the creation of an electronic medical record consisting of multiple input items, A data acquisition step to obtain time-series data collected during the examination of the subject, A selection acceptance step that receives user instructions to select one of the aforementioned multiple input items as a selection item, An inference target data acquisition step, which extracts data for a predetermined interval determined according to the selected item, including the time series data prior to the time when the user's instructions were received in the selection acceptance step, from the time series data as inference target data; An inference step in which inference processing is performed on the data to be inferred and input candidates for the selected items are inferred, A display control step which displays input candidates on the display unit based on the inference results from the inference step, A method for supporting the creation of electronic medical records, characterized by having the following features.

17. A program for executing the electronic medical record creation support method described in claim 16 using a computer.

Images