Annotation collection system and annotation collection method

Abstract

To reduce a burden on a medical worker responding to an alert.SOLUTION: The monitor device 11 outputs an alert. Annotation adding device 12 provides input user interface 121 for inputting correspondence information indicating whether the alert is an actionable alarm or a non-actionable alarm, and adds an annotation corresponding to the correspondence information to the alert.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to an annotation collection system and an annotation collection method for inputting response information to an alert output from a medical device. [Background technology]

[0002] Patent Document 1 discloses, as an example of the above medical equipment, a monitor that displays all of the biological information of a subject such as a patient. The monitor is configured to output an alert when an abnormality occurs in the measured values ​​of the biological information. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2001-070257 Summary of the Invention [Problem to be solved by the invention]

[0004] There is a need to reduce the burden on medical professionals who respond to alerts. [Means for solving the problem]

[0005] One example aspect provided by the present disclosure is an annotation collection system, comprising: A medical device that outputs an alert; an annotation assignment device that provides an input user interface for inputting response information indicating whether the alert is an actionable alarm or a non-actionable alarm, and assigns an annotation corresponding to the response information to the alert; It is equipped with:

[0006] One example of an aspect that may be provided by the present disclosure is an annotation collection method, comprising: providing an input user interface for inputting response information indicating whether the alert output from the medical device was an actionable alarm or a non-actionable alarm; An annotation corresponding to the response information is added to the alert.

[0007] For example, even if a medical device outputs an alarm indicating tachycardia for a certain patient, there are cases where the tachycardia actually requires treatment (actionable), and cases where the temporary increase in pulse rate is due to crying or a sensor misdetection and no treatment is required (non-actionable). Since the possibility of an actionable alarm cannot be ruled out, medical professionals are required to confirm the alert status. However, the frequency of non-actionable alarms is higher than that of actionable alarms. As the number of sensors and medical devices monitoring patients increases, the number of times they have to respond to non-actionable alarms, which are relatively less urgent, also tends to increase, leading to an increased burden on medical professionals known as "alarm fatigue."

[0008] Therefore, medical devices are required to be able to appropriately distinguish between actionable and non-actionable alarms and suppress the output of the latter. One idea is to implement an inference engine that has acquired this discrimination ability through machine learning into medical devices, but creating an inference engine with high-precision discrimination capabilities requires a huge amount of training data. Training data is, for example, a combination of the biological information that triggered the output of an alert and an annotation indicating whether the alert was an actionable alarm or a non-actionable alarm.

[0009] According to the configurations of the above-described exemplary embodiments, the medical professional who responded to the alert can annotate the alert through the input user interface, which can promote the efficient collection of annotations that more closely reflect the biometric monitoring environment in the medical setting where the medical professional works. The accumulation of cases, particularly of frequently occurring non-actionable alarms, can contribute to the preparation of high-quality training data. As a result, the burden on medical professionals who respond to alerts can be reduced. [Brief explanation of the drawings]

[0010] [Figure 1] 1 illustrates a configuration of an annotation collection system according to an embodiment. [Figure 2] 2 illustrates an example of the functional configuration of the annotation collection system of FIG. 1. [Figure 3] 3 shows an example of the appearance of the input user interface of FIG. 2. [Figure 4] 3 shows another example of the appearance of the input user interface of FIG. 2. [Figure 5] 3 illustrates an example of a process performed by the annotation device of FIG. 2. [Figure 6] 1 shows an example of the appearance of an auxiliary input user interface. DETAILED DESCRIPTION OF THE INVENTION

[0011] Exemplary embodiments will now be described in detail with reference to the accompanying drawings.

[0012] 1 illustrates an example of the configuration of an annotation collection system 10 (hereinafter abbreviated as collection system 10) according to one embodiment. The collection system 10 includes a monitor device 11 and an annotation assignment device 12 (hereinafter abbreviated as assignment device 12). The monitor device 11 and the assignment device 12 are communicatively connected via a communication network 20.

[0013] The monitor device 11 is a device that acquires biological information of the subject 30 through a sensor (not shown). The sensor may be configured to be worn by the subject 30 to acquire the biological information, or may be configured to acquire the biological information in a non-contact manner. Examples of the biological information include an electrocardiogram, pulse, blood pressure, body temperature, percutaneous arterial oxygen saturation (SpO2), respiratory rate, and end-tidal carbon dioxide (EtCO2). The monitor device 11 is configured to output an alert. The monitor device 11 is an example of a medical device.

[0014] The providing device 12 is a device that provides an input user interface for the medical professional 40 to input corresponding information. The providing device 12 may be configured to display biological information such as waveforms and numerical values ​​acquired by the monitoring device 11. In this example, the providing device 12 is a mobile device that can be carried by the medical professional 40. Examples of the mobile device include a mobile phone and a personal digital assistant. Examples of the personal digital assistant include a general-purpose terminal device such as a smartphone or a tablet terminal, and a dedicated input device.

[0015] When an alert is output from the monitoring device 11, a medical professional checks the content of the alert and takes appropriate action. This action includes a primary action, such as stopping the output of the alert, and a secondary action, such as checking the condition of the subject 30 and the condition of medical devices including the monitoring device 11 and taking appropriate action.

[0016] The response information indicates whether the alert output from the monitor device 11 was an actionable alarm or a non-actionable alarm. An "actionable alarm" is an alarm that notifies of a condition that requires either a therapeutic response or a non-therapeutic response. A "non-actionable alarm" is an alarm that notifies of a condition that does not require either a therapeutic response or a non-therapeutic response.

[0017] Examples of therapeutic measures include head-up, positioning, cooling, suctioning airway secretions, use of inhalers or nebulizers, manual ventilation, changing oxygen flow rates, administering emergency medications and supplementary foods, etc. Examples of non-therapeutic measures include room visits, examinations, vital signs measurement, blood glucose measurement, replacing sensor electrodes, feeding, and putting the patient to sleep.

[0018] The alerts include vital alarms and technical alarms. A vital alarm is an alarm that is generated based on a change in the vital signs of the subject 30. A technical alarm is an alarm that is generated based on a change in the measurement status of the medical device or the subject 30.

[0019] Each of the vital alarms and technical alarms includes true alarms and false alarms. A true alarm corresponds to a case where the event that should be notified by the alarm matches the event that is actually occurring. A false alarm corresponds to a case where the event that should be notified by the alarm does not match the event that is actually occurring. Note that whether the state notified by the alarm is actionable or non-actionable does not necessarily correspond to the true or false state of the alarm.

[0020] An example of an actionable true vital alarm is when a patient with sepsis develops tachycardia and an alarm is triggered for tachycardia. Another example of an actionable true vital alarm is when a patient develops respiratory distress and an alarm is triggered for low SpO2.

[0021] An example of a non-actionable true vital alarm is when a patient experiences a bradycardia alarm due to respiratory fluctuations. Another example of a non-actionable true vital alarm is when a crying pediatric patient experiences a tachycardia alarm.

[0022] An example of a true actionable technical alarm is when an alarm is output for a bedridden patient indicating a sensor electrode has become disconnected. An example of a true non-actionable technical alarm is when an alarm is output for a sensor contact failure due to patient movement.

[0023] An example of an actionable false vital alarm would be an alarm indicating ventricular tachycardia in a patient experiencing a seizure, while an example of a non-actionable false vital alarm would be an alarm indicating cardiac arrest in a patient with a dislodged ECG electrode.

[0024] An example of an actionable false technical alarm is when an alarm indicating a sensor has come off is output when the signal output from the sensor is too weak to measure.An example of a non-actionable false technical alarm is when an alarm requesting status confirmation is output due to temporary sensor mis-attachment.

[0025] 2 illustrates an example of the functional configuration of the providing device 12. The providing device 12 includes an input user interface 121, a reception interface 122, a processor 123, and an output interface .

[0026] 3 shows an example of the input user interface 121. The input user interface 121 includes an actionable button 121a and a non-actionable button 121b. Each of the actionable button 121a and the non-actionable button 121b may be a button switch that can be mechanically operated, or may be a button image displayed on a touch panel.

[0027] Actionable button 121a is a button for inputting, as response information, that the alert output from monitoring device 11 was an actionable alarm. Non-actionable button 121b is a button for inputting, as response information, that the alert output from monitoring device 11 was a non-actionable alarm. After performing an operation to stop the alert output from monitoring device 11, medical worker 40 inputs the response information by operating actionable button 121a or non-actionable button 121b.

[0028] 2, the monitor device 11 is configured to output output information indicating that an alert has been output. The output information includes information such as the type of abnormality in the detected biological information and the date and time when the alarm was output. The output information may be in the form of digital data or analog data.

[0029] The receiving interface 122 is configured as a hardware interface that receives the output information. When the output information is in the form of analog data, the receiving interface 122 includes an appropriate conversion circuit including an A / D converter.

[0030] The processor 123 is configured to enable the input user interface 121 to accept corresponding information when the reception interface 122 accepts output information.

[0031] The input of the response information to the input user interface 121 does not necessarily have to be performed by operating a button switch or button image. In addition to or instead of this, the response information can be input through voice input or gesture input. In this case, words or gestures indicating an actionable alarm and words or gestures indicating a non-actionable alarm are determined in advance.

[0032] 4 shows another example of the input user interface 121 that allows voice input. When the processor 123 allows the reception of corresponding information, for example, an icon 121c requesting voice input is displayed or lit.

[0033] 2, the input user interface 121 is configured to output the corresponding information input by the medical professional 40. The corresponding information may be in the form of digital data or analog data.

[0034] The receiving interface 122 is configured to receive the corresponding information. If the corresponding information is in the form of analog data, the receiving interface 122 includes an appropriate conversion circuit including an A / D converter.

[0035] The processor 123 is configured to perform annotation processing, as exemplified in Fig. 5, based on the response information received by the reception interface 122. The annotation processing is processing for associating an alert output from the monitor device 11 with a response made by the medical worker 40 in response to the alert.

[0036] First, the processor 123 determines whether the reception interface 122 has received output information from the monitor device 11 (STEP 1). This process is repeated until it is determined that the output information has been received (NO in STEP 1).

[0037] If it is determined that the output information has been received (YES in STEP 1), processor 123 determines whether reception interface 122 has received corresponding information within a predetermined time period since the output information was received (STEP 2).

[0038] If it is determined that the correspondence information has been accepted before the predetermined time has elapsed (YES in STEP 2), processor 123 determines whether the correspondence information indicates an actionable alarm (STEP 3).

[0039] If it is determined that the response information indicates an actionable alarm (YES in STEP 3), processor 123 performs processing to add an annotation indicating an actionable alarm to the output information (STEP 4).

[0040] If it is determined that the response information does not indicate an actionable alarm (NO in STEP 3), the processor 123 performs processing to add an annotation indicating a non-actionable alarm to the output information (STEP 5).

[0041] 2, the processor 123 is configured to output the history information to which the annotations have been added from the output interface 124. The output interface 124 is configured as a hardware interface.

[0042] The history information may be in the form of digital data or analog data. If the history information is in the form of analog data, the output interface 124 includes an appropriate conversion circuit including a D / A converter.

[0043] As mentioned above, even when a monitoring device outputs an alarm indicating tachycardia for a certain subject, there are cases where the tachycardia actually requires treatment (actionable), and cases where the temporary increase in pulse rate is due to crying or a sensor misdetection and no treatment is required (non-actionable). Since the possibility of an actionable alarm cannot be ruled out, medical professionals are required to confirm the alert status. However, the frequency of non-actionable alarms is higher than that of actionable alarms. As the number of sensors and medical devices monitoring subjects increases, the number of times they have to respond to non-actionable alarms, which are relatively less urgent, also tends to increase, leading to an increased burden on medical professionals known as "alarm fatigue."

[0044] Therefore, monitoring devices are required to be able to appropriately distinguish between actionable and non-actionable alarms and suppress the output of the latter. One possible solution is to implement an inference engine that has acquired discriminatory capabilities through machine learning in the monitoring device, but creating an inference engine with highly accurate discriminatory capabilities requires a huge amount of training data. The training data is, for example, a combination of the biometric information that triggered the output of an alert and an annotation indicating whether the alert was an actionable alarm or a non-actionable alarm.

[0045] According to the configuration of this embodiment, the medical professional who responded to the alert can annotate the alert through the input user interface, which can promote the efficient collection of annotations that more closely reflect the biometric monitoring environment in the medical setting where the medical professional works. The system also promotes the accumulation of cases, particularly for frequently occurring non-actionable alarms, which can contribute to the preparation of high-quality training data. As a result, the burden on medical professionals who respond to alerts can be reduced.

[0046] 5, if no corresponding information is received within a predetermined time period after the output information is received (NO in STEP 2), the processor 123 adds an annotation to the output information indicating that there was no reaction to the alert (STEP 6). That is, the alert corresponding to the output information is recorded as an alert for which no reaction was received.

[0047] Examples of situations where there was no response to an alert include when healthcare professionals' sensitivity to alerts has decreased due to excessive alerts, or when healthcare professionals are performing higher-priority tasks. By adding data annotated with no response to an alert to the collection, it will contribute more to the preparation of learning data to suppress the output of alarms with relatively low urgency.

[0048] 1, in addition to or instead of a mobile device carried by a healthcare professional 40, the monitoring device 11 may function as the annotation device 12. That is, the input user interface 121 may be part of the monitoring device 11.

[0049] With this configuration, the medical staff who has performed an operation to cancel the alert on the monitor device 11 can then proceed to provide annotations, thereby reducing the psychological burden of additional work.

[0050] 3 and 4, it is preferable to place the input user interface 121 adjacent to the cancel user interface 111 for canceling the alert provided on the monitor device 11. This configuration can facilitate a smooth transition from the alert cancel operation to the annotation addition operation.

[0051] 6 , the annotation device 12 may provide an auxiliary input user interface 125. The auxiliary input user interface 125 includes a first group 125 a, a second group 125 b, and a third group 125 c. After annotating an alert through the input user interface 121, the healthcare professional 40 can input supplemental information for the alert through the auxiliary input user interface 125.

[0052] The first group 125a includes buttons or button images for selectively inputting whether an alarm is true or false. As described above, true alarms and false alarms each include actionable alarms and non-actionable alarms. Therefore, by adding information indicating whether the annotated alert was a true alarm or a false alarm, more detailed information about the background of the alert output can be provided to machine learning. As a result, it is possible to realize inference of the alert type based on more thorough judgment.

[0053] The second group 125b includes buttons or button images for selectively inputting the healthcare professional's 40 impressions of the alert. In this example, two buttons are shown: one indicating that the alert was helpful and the other indicating that the alert was unnecessary. By adding information indicating the healthcare professional's 40 impressions of the appropriateness of the alert corresponding to the annotation, more detailed information regarding the contribution of the alert can be provided to machine learning. As a result, it is possible to realize inference of the alert type based on more thorough judgment.

[0054] The third group 125c includes buttons or button images for selectively inputting the type of response taken by the healthcare professional 40 in response to the alert. The types of responses include those appropriately selected from the therapeutic and non-therapeutic responses exemplified above. By adding information indicating the type of response required for the actionable or non-actionable alarm corresponding to the annotation, more detailed information regarding the relationship between the alert and the response can be provided to machine learning. As a result, it is possible to realize inference of the alert type based on more thorough judgment.

[0055] In addition, the auxiliary input user interface 125 is configured to allow selective input of supplementary information, thereby reducing the psychological burden on the medical staff 40 of additional work.

[0056] The auxiliary input user interface 125 may include at least one of the first group 125a, the second group 125b, and the third group 125c.

[0057] 1 , the collection system 10 may include an imaging device 131. The imaging device 131 is installed so as to acquire an image that captures at least one of the monitor device 11 and the subject 30. In other words, the imaging device 131 is installed so as to record how the medical staff 40 responds to an alert output from the monitor device 11. The imaging device 131 may be mounted on a specific medical device including the monitor device 11.

[0058] The collection system 10 includes a display device 14 that displays an image acquired by the imaging device 131 at least when an alert is output. The display device 14 may be an independent device or may be part of the monitor device 11 or the providing device 12, as long as it can receive image data from the imaging device 131 via the communication network 20.

[0059] With this configuration, even in a location other than the medical site where the monitor device 11 is installed, an annotation can be added to the alert at any time after the alert is output by referring to the image displayed on the display device 14. This means that spatial and time constraints on the annotation work can be alleviated. In addition, annotations can be added by a third party other than the medical professionals working at the medical site. Examples of such third parties include engineers working for a medical-related company and medical license holders on parental leave.

[0060] To obtain similar advantages, in addition to or instead of the imaging device 131, the collection system 10 may include a sound collection device 132. The sound collection device 132 is installed so as to acquire sounds emitted from at least one of the monitoring device 11 and the subject 30. In other words, the sound collection device 132 is installed so as to be able to record how the medical staff 40 responds to an alert output from the monitoring device 11.

[0061] Even with this configuration, it is possible to refer to the sound acquired by the sound collection device 132 at any timing after the alert is output and to add annotations to the alert at a location different from the medical site where the monitor device 11 is installed. In other words, it is possible to alleviate the spatial and temporal constraints on the annotation work.

[0062] As illustrated in FIG. 1 , the collection system 10 includes an annotation collection device 15 (hereinafter, abbreviated as the collection device 15). The collection device 15 can communicate with the monitoring device 11 and the assignment device 12 via a communication network 20. The collection device 15 may be a general-purpose or dedicated device installed within the medical facility MF, or may be a general-purpose or dedicated device installed outside the medical facility MF. The display device 14 described above may be a part of the collection device 15. The collection device 15 may be a part of the monitoring device 11 or the assignment device 12.

[0063] As illustrated in FIG. 2, the collection device 15 includes a reception interface 151 , a processor 152 , a storage 153 , and an output interface 154 .

[0064] The receiving interface 151 is configured as a hardware interface that receives history information from at least one providing device 12. When the history information is in the form of analog data, the receiving interface 151 is provided with an appropriate conversion circuit including an A / D converter. This description also applies to other information and data that can be received by the receiving interface 151, which will be described later.

[0065] The processor 152 is configured to store the history information in the storage 153. The storage 153 can be realized by a semiconductor memory, a hard disk device, a magnetic tape device, or the like.

[0066] The reception interface 151 is configured to be able to receive biometric information acquired by the monitoring device 11. The reception interface 151 may receive the biometric information directly from the monitoring device 11, or may receive biometric information copied by a mirroring hub device installed on the communication network 20 and transmitted from the monitoring device 11 to another device. The expression "biometric information acquired by a monitoring device" used in this specification includes the latter case.

[0067] Examples of biological information include information indicating the value of a specific biological parameter, information indicating the change over time (waveform or trend) of the value, etc. The biological information is configured to include information indicating the time point when it was acquired by the monitor device 11.

[0068] The processor 152 can be configured to acquire at least the biological information at the time when the monitor device 11 outputs an alert, and store the acquired information in the storage 153 in association with history information generated based on the alert.

[0069] For example, all of the biological information acquired by the monitoring device 11 may be temporarily stored in a predetermined storage. The storage may be the storage 153, or may be a storage provided in a device on the communication network 20 that is capable of communicating with the collecting device 15. In this case, the processor 152 may refer to the history information to identify the time when an alarm was output, and acquire from the storage the biological information for a predetermined length of time including that time.

[0070] This configuration automates the process of associating an annotated alert with the biometric information that triggered the alert, thereby streamlining the process of preparing data to be used for analyzing historical information (described later) and for machine learning by the inference engine 112 implemented in the monitor device 11.

[0071] In addition to the above-mentioned biological information, the reception interface 151 can receive at least the conditions for acquiring biological information by the monitor device 11 when an alert is output (such as the sensor type), the conditions for outputting an alert, and the like.

[0072] 2, the collection system 10 may include a database 16. The database 16 is a database for storing and managing electronic medical records. Although not shown, the database 16 can communicate with the collection device 15 via a communication network 20. The database 16 may be installed outside the medical facility MF.

[0073] 2, the processor 152 of the collection device 15 may be configured to acquire personal information of the subject 30 from the database 16 through the reception interface 151 when an alert is output from the monitoring device 11. The personal information may include at least one of the subject 30's gender, height, weight, medical history, current medical condition, etc.

[0074] In this case, the processor 152 may be configured to associate the history information generated based on the annotation added to the alert through the adding device 12 with the personal information and store them in the storage 153 .

[0075] The personal characteristics of the subject 30 may be influencing the output of an alert. With the above configuration, it is possible to take such influences into account in the history information. This makes it possible to improve the quality of data used in the analysis of the history information (described later) and the machine learning of the inference engine 112 implemented in the monitor device 11.

[0076] In addition, if it is possible to identify the occurrence of a specific serious event (sudden death, ischemic attack, myocardial infarction, vascular rupture, etc.) in the electronic medical record stored in database 16 and it is possible to obtain biological information at the time of the occurrence of the serious event, an annotation indicating an actionable alarm may be added to the biological information at any time. In other words, adding an annotation indicating an actionable alarm does not necessarily mean that an actual response to the alert has been taken. The personal information of the subject who experienced the serious event and the history information generated based on the annotation are associated and stored in storage 153.

[0077] According to this configuration, it is possible to give appropriate annotations to serious events that occur infrequently, and it is possible to improve the accuracy (specificity) of actionable alarm output.

[0078] The processor 152 of the collection device 15 may accept personal information of the medical professional 40 who inputs the correspondence information through the reception interface 151. The personal information may be, for example, information that identifies attributes of the medical professional 40. Examples of attributes include doctor, resident, nurse, and work history. The personal information may be input, for example, through the input user interface 121 or auxiliary input user interface 125 of the assignment device 12.

[0079] In this case, the processor 152 may be configured to associate the history information generated based on the annotation added to the alert through the adding device 12 with the personal information of the medical worker 40 and store it in the storage 153.

[0080] The quality of annotations added by the annotation device 12 may be affected by the attributes of the medical professional 40 who inputs the correspondence information. With the above configuration, such influences can be taken into account in the history information. Therefore, the quality of data used for analysis of the history information (described later) and machine learning by the inference engine 112 implemented in the monitor device 11 can be improved.

[0081] As illustrated in FIG. 2, the collection system 10 may include an output device 17. The output device 17 is a device that visualizes desired information. Examples of the output device 17 include a general-purpose device with a display, a printer, a data output device, etc. The output device 17 may be part of the providing device 12 or the collecting device 15. As illustrated in FIG. 1, the output device 17 can communicate with the collecting device 15 via a communication network 20. Although not illustrated, the output device 17 may be installed outside the medical facility MF.

[0082] The processor 152 of the collection device 15 may be configured to generate statistical information about the annotations based on the historical information. For example, the statistical information may include the ratio of the number of annotations indicating actionable alarms, annotations indicating non-actionable alarms, and annotations indicating no response to an alert to the total number of annotations.

[0083] The processor 152 may be configured to output an output control signal from the output interface 154, which causes the statistical information to be output to the output device 17. The output interface 154 is configured as a hardware interface. The output control signal may be a digital signal or an analog signal. If the output control signal is an analog signal, the output interface 154 includes an appropriate conversion circuit including a D / A converter.

[0084] The above statistical information can be used to analyze the operational status of the vital signs monitoring system in the medical site where the monitor device 11 is installed. For example, a situation in which the ratio of annotations indicating actionable alarms is high suggests an operational status in which abnormalities in subjects that require a relatively high level of response are appropriately reported by alerts.

[0085] On the other hand, a high ratio of annotations indicating non-actionable alarms suggests that the algorithm of the inference engine 112 implemented in the monitoring device 11 may not be aligned with the actual situation at the medical site. A high ratio of annotations indicating no reaction to the alert suggests that there may be an operational problem with the vital sign monitoring system.

[0086] Therefore, users (healthcare professionals, consultants, etc.) who come into contact with the statistical information can distinguish between unnecessary alert responses caused by algorithm deficiencies and those caused by deficiencies in the operating environment, and then implement measures to improve the algorithm and measures to improve the operating environment.

[0087] Improvements to the algorithm and operating environment will lead to improved quality of care provided to patients and reduced workloads for healthcare professionals. Examples of improvements to the algorithm include changing the alert output conditions in the monitoring device 11 and relearning the inference engine 112. Examples of improvements to the operating environment include changing the alert output conditions in the monitoring device 11, performing maintenance and inspections of sensors and probes attached to patients 30, and changing the allocation of healthcare professionals.

[0088] The processor 152 of the collection device 15 may be configured to determine whether or not it is necessary to change the alert output conditions based on the historical information stored in the storage 153. For example, if the number or ratio of unresponsive alerts exceeds a predetermined threshold, it is determined that the alert output conditions need to be changed. Additionally or alternatively, if the number or ratio of alarms determined to be unnecessary exceeds a predetermined threshold, it is determined that the alert output conditions need to be changed. Specifically, it is determined that it is necessary to change the conditions (such as adjusting the threshold) so that the output of alerts is suppressed.

[0089] If it is determined that the alert output conditions need to be changed, the processor 152 outputs an output control signal to the output device 17 from the output interface 154 to notify the output device 17 of the determination. The output device 17 notifies the user that the alert output conditions need to be changed through at least one of a visual notification, an auditory notification, and a tactile notification. The user who receives the notification changes the settings of the monitor device 11 as necessary.

[0090] As mentioned above, setting appropriate alert output conditions contributes to both improving the algorithm and the operational environment, thereby improving the quality of care provided to patients and reducing the burden on medical professionals.

[0091] As described above, the monitoring device 11 includes the inference engine 112. The inference engine 112 is an algorithm generated through machine learning using a neural network. The inference engine 112 is configured to receive input of biometric information acquired from the subject 30 and output the probability that a specific abnormality has occurred in the biometric information. If the probability exceeds a threshold, the monitoring device 11 outputs an alert corresponding to the abnormality.

[0092] The inference engine 112 may be mounted on a device capable of data communication with the monitor device 11. The data communication may be performed via the communication network 20 illustrated in FIG. 1 or via short-range wireless communication.

[0093] The processor 152 of the collection device 15 may be configured to determine whether retraining of the inference engine 112 is necessary based on the historical information stored in the storage 153. For example, a determination that retraining of the inference engine 112 is necessary may be made if the number or proportion of unresponsive alerts exceeds a predetermined threshold. Additionally or alternatively, a determination that retraining of the inference engine 112 is necessary may be made if the number or proportion of alarms determined to be unnecessary exceeds a predetermined threshold.

[0094] If it is determined that re-learning of the inference engine 112 is necessary, the processor 152 outputs an output control signal from the output interface 154 to the output device 17 to notify the determination. The output device 17 notifies the user that re-learning of the inference engine 112 is necessary through at least one of a visual notification, an auditory notification, and a tactile notification. The user who receives the notification re-learns the inference engine 112 as necessary. When re-learning, the history information associated with the biometric information and stored in the storage 133 is used as learning data.

[0095] The algorithm of the inference engine 112 generated through machine learning intended for overall optimization does not necessarily fit the operating environment of a particular medical site. On the other hand, the history information stored in the storage 153 of the collection device 15 is highly likely to reflect the actual conditions of the medical site where the monitoring device 11 equipped with the inference engine 112 is installed. Therefore, by re-learning the inference engine 112 using the history information as learning data, it is possible to obtain alert output characteristics of the monitoring device 11 that are more suited to the operating environment of the medical site.

[0096] It should be noted that the inference engine 112 does not necessarily have to be generated through machine learning using a neural network. The inference engine 112 can be generated through other machine learning algorithms. Examples of other machine learning algorithms include decision trees, random forests, and support vector machines.

[0097] The range of medical sites from which the collection device 15 collects historical information can be determined as appropriate. It may be determined by hospital room, ward, or medical department in a specific medical facility, or collection may be performed from multiple medical facilities that can communicate via the communication network 20. When collection is performed from multiple medical facilities, it is preferable to define the targets so that the medical departments match.

[0098] By increasing the number of sources of historical information, it will not only be possible to compensate for the lack of alert response cases at each medical facility, but it will also reduce the burden on each medical facility in accumulating alert response cases.

[0099] Similarly, the inference engine 112 of the monitoring device 11 that has been retrained based on the determinations made by the collection device 15 does not necessarily need to be applied only to the medical site where the monitoring device 11 is installed. The algorithm of the inference engine 112 that has been retrained based on the operational environment of one medical site can be applied to other medical sites with similar operational environments. Only the data used for retraining may be shared.

[0100] This configuration can efficiently improve the compatibility of the algorithms of the inference engine 112 with the operating environment of a particular medical site, and also reduce the burden on the medical site associated with accumulating alert response cases to improve compatibility.

[0101] When history information is provided by multiple entities, the processor 152 of the collection device 15 may be configured to acquire the contribution of each source of the history information. The contribution may be defined based on various criteria. For example, the contribution may be determined based on the number of times the history information is provided, the contribution to determining whether or not to change the alert output conditions, the contribution to determining whether or not to retrain the inference engine 112, etc. The processor 152 may output an output control signal from the output interface 154 to cause the output device 17 to output the determined contribution.

[0102] This configuration makes it possible to build a system that provides rewards to providers of history information according to their contribution. Under such a system, incentives act on inputting corresponding information, and a virtuous cycle can be expected that leads to the collection of more history information.

[0103] The processor 123 of the providing device 12 and the processor 152 of the collecting device 15, each having the various functions described above, may be realized by a general-purpose microprocessor operating in cooperation with general-purpose memory. Examples of general-purpose microprocessors include a CPU, an MPU, and a GPU. Examples of general-purpose memory include a ROM and a RAM. In this case, a computer program for executing the above-described processes may be stored in the ROM. The ROM is an example of a non-transitory computer-readable medium for storing a computer program. The general-purpose microprocessor specifies at least a portion of the program stored in the ROM, expands it in the RAM, and executes the above-described processes in cooperation with the RAM. The computer program may be pre-installed in the general-purpose memory or may be downloaded from an external server via the communication network 20 and then installed in the general-purpose memory. In this case, the external server is an example of a non-transitory computer-readable medium for storing a computer program.

[0104] The processor 123 of the providing device 12 and the processor 152 of the collecting device 15, each having the various functions described above, may be realized by a dedicated integrated circuit capable of executing the computer program, such as a microcontroller, an ASIC, or an FPGA. In this case, the computer program is pre-installed in a memory element included in the dedicated integrated circuit. The memory element is an example of a computer-readable medium storing a computer program.

[0105] Each of the processor 123 of the application device 12 and the processor 152 of the collection device 15, which have the various functions described above, may also be realized by a combination of a general-purpose microprocessor and a dedicated integrated circuit.

[0106] The various configurations described above are merely examples for facilitating understanding of the present disclosure. Each configuration example can be appropriately modified or combined with other configurations within the scope of the present disclosure.

[0107] The medical device that outputs the alert is not limited to the monitor device 11 placed near the subject 30. A central monitor installed in a nurse's center or an appropriate testing device can also be an example of the medical device, as long as it can acquire biological information of the subject 30 and output an alert based on abnormalities.

[0108] The timing when the medical professional annotates the alert through the input user interface is arbitrary. For example, the medical professional may annotate the alert when responding to the alert, or when no alert has been output. This is because a random visit to the room (when no alert has been output) may be a more suitable time for the medical professional to annotate than when they are busy responding to an alert.

[0109] When a medical professional adds an annotation to the alert through the input user interface, the medical professional may add the annotation while referring to biological information such as waveforms and numerical values ​​displayed on the annotation device 12.

[0110] The configurations listed below also form part of this disclosure. Item 1: A medical device that outputs an alert; an annotation providing device that provides an input user interface for inputting response information indicating whether the alert is an actionable alarm or a non-actionable alarm, and provides an annotation corresponding to the response information; Equipped with Annotation collection system. Item 2: the annotation providing device provides an annotation indicating that there was no reaction to the alert when the response information is not input within a predetermined time period after the alert is output; Item 1. An annotation collection system according to item 1. Item 3: the input user interface is part of the medical device; Item 3. The annotation collection system according to item 1 or 2. Item 4: the input user interface is arranged adjacent to a cancellation user interface for canceling the alert; Item 3. The annotation collection system according to item 3. Item 5: the annotation device provides an auxiliary input user interface that allows selective input of at least one of the truth or falsity of the alert, the type of response made by the medical professional in response to the alert, and the medical professional's opinion on the validity of the alert; 5. An annotation collection system according to any one of items 1 to 4. Item 6: an imaging device that captures an image in which at least one of the subject and the medical device is captured; a display device that displays the image at least when the alert is output; and It is equipped with the input user interface is configured to input the corresponding information while referring to the image. 6. An annotation collection system according to any one of items 1 to 5. Item 7: a sound collection device for collecting sounds emitted from at least one of the subject and the medical device; the input user interface is configured to input the corresponding information while referring to the voice. 7. An annotation collection system according to any one of items 1 to 6. Item 8: an annotation collection device that acquires and stores history information of annotations from at least one of the annotation assigning devices; 8. An annotation collection system according to any one of items 1 to 7. Item 9: The annotation collection device is configured to acquire biometric information of the subject at least at the time of outputting the alert and associate the biometric information with the history information. Item 9. The annotation collection system according to item 8. Item 10: The annotation collection device is configured to acquire personal information of at least one of the subject and the person who inputs the correspondence information, and associate the acquired personal information with the history information. 10. The annotation collection system according to item 8 or 9. Item 11: the annotation collection device is configured to cause an output device to output statistical information related to the annotations generated based on the history information. 11. An annotation collection system according to any one of items 8 to 10. Item 12: The annotation collection device is configured to determine whether or not it is necessary to change an output condition of the alert based on the history information. 12. An annotation collection system according to any one of items 8 to 11. Item 13: the medical device is configured to determine whether an abnormality has occurred in the biological information by using an inference engine that has been machine-learned using a combination of the biological information of the subject and the annotation as learning data; The annotation collection device is configured to determine whether or not re-learning of the inference engine is necessary based on the history information. 13. An annotation collection system according to any one of items 8 to 12. Item 14: The annotation collection device is configured to acquire a contribution degree of a source of the history information. 14. An annotation collection system according to any one of items 8 to 13. [Explanation of symbols]

[0111] 10: annotation collection system, 11: monitor device, 111: release user interface, 112: inference engine, 12: annotation assignment device, 121: input user interface, 125: auxiliary input user interface, 131: imaging device, 132: sound collection device, 14: display device, 15: annotation collection device, 17: output device, 30: subject

Claims

1. A medical device that outputs an alert; an annotation providing device that provides an input user interface for inputting response information indicating whether the alert is an actionable alarm or a non-actionable alarm, and provides an annotation corresponding to the response information; Equipped with Annotation collection system.

2. the annotation providing device provides an annotation indicating that there was no reaction to the alert when the response information is not input within a predetermined time period after the alert is output; The annotation collection system according to claim 1 .

3. the input user interface is part of the medical device; The annotation collection system according to claim 1 .

4. the input user interface is arranged adjacent to a cancellation user interface for canceling the alert; The annotation collection system according to claim 3 .

5. the annotation device provides an auxiliary input user interface that allows selective input of at least one of the truth or falsity of the alert, the type of response made by the medical professional in response to the alert, and the medical professional's opinion on the validity of the alert; The annotation collection system according to claim 1 .

6. an imaging device that captures an image in which at least one of the subject and the medical device is captured; a display device that displays the image at least when the alert is output; and It is equipped with the input user interface is configured to input the corresponding information while referring to the image. The annotation collection system according to claim 1 .

7. a sound collection device for collecting sounds emitted from at least one of the subject and the medical device; the input user interface is configured to input the corresponding information while referring to the voice. The annotation collection system according to claim 1 .

8. an annotation collection device that acquires and stores history information of annotations from at least one of the annotation assigning devices; The annotation collection system according to claim 1 .

9. The annotation collection device is configured to acquire biometric information of the subject at least at the time of outputting the alert and associate the biometric information with the history information. The annotation collection system according to claim 8 .

10. The annotation collection device is configured to acquire personal information of at least one of the subject and the person who inputs the correspondence information, and associate the acquired personal information with the history information. The annotation collection system according to claim 8 .

11. the annotation collection device is configured to cause an output device to output statistical information related to the annotations generated based on the history information. The annotation collection system according to claim 8 .

12. The annotation collection device is configured to determine whether or not it is necessary to change an output condition of the alert based on the history information. The annotation collection system according to claim 8 .

13. the medical device is configured to determine whether an abnormality has occurred in the biological information by using an inference engine that has been machine-learned using a combination of the biological information of the subject and the annotation as learning data; The annotation collection device is configured to determine whether or not re-learning of the inference engine is necessary based on the history information. The annotation collection system according to claim 8 .

14. The annotation collection device is configured to acquire a contribution degree of a source of the history information. The annotation collection system according to claim 8 .

15. providing an input user interface for inputting response information indicating whether the alert output from the medical device was an actionable alarm or a non-actionable alarm; adding an annotation corresponding to the corresponding information to the alert; Annotation collection methods.

Images