WO2022064731A1 - 自動分析装置、推奨アクション通知システムおよび推奨アクション通知方法 - Google Patents
自動分析装置、推奨アクション通知システムおよび推奨アクション通知方法 Download PDFInfo
- Publication number
- WO2022064731A1 WO2022064731A1 PCT/JP2021/005819 JP2021005819W WO2022064731A1 WO 2022064731 A1 WO2022064731 A1 WO 2022064731A1 JP 2021005819 W JP2021005819 W JP 2021005819W WO 2022064731 A1 WO2022064731 A1 WO 2022064731A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- action
- predetermined
- learning
- recommended
- abnormality
- Prior art date
Links
- 230000009471 action Effects 0.000 title claims abstract description 194
- 238000000034 method Methods 0.000 title claims abstract description 18
- 230000005856 abnormality Effects 0.000 claims abstract description 63
- 238000012423 maintenance Methods 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims description 23
- 238000011156 evaluation Methods 0.000 claims description 18
- 238000001514 detection method Methods 0.000 claims description 13
- 238000007689 inspection Methods 0.000 claims description 4
- 230000004044 response Effects 0.000 claims 3
- 239000000284 extract Substances 0.000 claims 1
- 239000000523 sample Substances 0.000 description 81
- 238000006243 chemical reaction Methods 0.000 description 56
- 230000007246 mechanism Effects 0.000 description 33
- 239000003153 chemical reaction reagent Substances 0.000 description 23
- 238000004140 cleaning Methods 0.000 description 23
- 238000005259 measurement Methods 0.000 description 18
- 238000003756 stirring Methods 0.000 description 13
- 230000000694 effects Effects 0.000 description 10
- 230000002159 abnormal effect Effects 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 230000010485 coping Effects 0.000 description 5
- 238000002835 absorbance Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000007423 decrease Effects 0.000 description 3
- 230000010365 information processing Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000003908 quality control method Methods 0.000 description 3
- 238000012951 Remeasurement Methods 0.000 description 2
- 238000011109 contamination Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 210000002966 serum Anatomy 0.000 description 2
- 238000005406 washing Methods 0.000 description 2
- 210000004369 blood Anatomy 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 239000012898 sample dilution Substances 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 239000000243 solution Substances 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0633—Workflow analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N35/00—Automatic analysis not limited to methods or materials provided for in any single one of groups G01N1/00 - G01N33/00; Handling materials therefor
- G01N35/00584—Control arrangements for automatic analysers
- G01N35/00594—Quality control, including calibration or testing of components of the analyser
- G01N35/00613—Quality control
- G01N35/00623—Quality control of instruments
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N35/00—Automatic analysis not limited to methods or materials provided for in any single one of groups G01N1/00 - G01N33/00; Handling materials therefor
- G01N35/00584—Control arrangements for automatic analysers
- G01N35/00594—Quality control, including calibration or testing of components of the analyser
- G01N35/00613—Quality control
- G01N35/00623—Quality control of instruments
- G01N2035/00643—Quality control of instruments detecting malfunctions in conveying systems
Definitions
- the present invention relates to an automated analyzer, a recommended action notification system for notifying an operator of recommended actions for an automated analyzer, and a recommended action notification method.
- Patent Document 1 in order to facilitate the investigation of the cause of poor quality control, information on the measurement data acquired by the sample analyzer and information on the operation of the sample analyzer are displayed in association with each other in a time-series graph or the like. By doing so, it is described that the operator can generate data that can grasp the state of the sample analyzer at a glance.
- the operator guesses the cause from the knowledge such as what kind of device data is necessary to identify the cause and the collected device data. Sufficient experience is required to determine the appropriate coping method for the device in consideration of the operating conditions, and it may be difficult for an inexperienced operator to select the appropriate coping method.
- Patent Document 1 information such as quality control of a standard sample, which was conventionally displayed only on a screen or a paper medium, is displayed in association with information related to the operation of the sample analyzer. This makes it easier for an operator with sufficient experience to intuitively understand from the displayed contents whether or not the cause of the quality control failure is due to the device side. However, it may still be difficult for an inexperienced operator to guess the cause from the displayed information and deal with it.
- the automated analyzer monitors the measurement results and maintenance results of the sample, and if it exceeds a preset threshold value, it detects it as an abnormality and notifies the operator.
- a preset threshold value there may be various causes for the measurement result to exceed the threshold value, such as an abnormality in the sample, an abnormality in the reagent, or an abnormality in each mechanism of the device. Is difficult to determine.
- the conventional countermeasures notified to the operator when an abnormality is detected all countermeasures corresponding to possible causes are presented.
- the present invention provides a system and a method for recommending a coping method considered appropriate to the operator from the stage of predicting an abnormality.
- a recommended action notification system includes a plurality of automated analyzers including a first automated analyzer, and a learning device connected to the plurality of automated analyzers via a network, and the operator is informed of the first. It is a recommended action notification system that recommends the action to be executed for 1 automated analyzer.
- the related device data including the sample analysis result data or maintenance result data is input to the training model, and the execution of a predetermined action output by the training model.
- a processing unit that recommends the operator to execute a predetermined action to the first automated analyzer, and a processing unit. It is provided with an update unit that updates the learning model based on the training data sets from a plurality of automatic analyzers.
- FIG. 1 It is a figure which shows the whole picture of the recommended action notification system. It is a functional block diagram of the recommended action notification system. It is a schematic diagram of an automatic analyzer. It is an update flowchart of a learning model. This is an example of the action evaluation input screen. This is an example of related device data. This is an example of a learning model using a neural network. This is an example of an operation table. It is a flowchart which outputs a recommended action using a learning model. This is an example of the recommended action display screen. This is an example of a learning model for the action "light source lamp replacement". This is an example of time-series absorbance data, which is photometer check data. This is an example of reaction monitor data. This is an example of the recommended action display screen.
- FIG. 1A is an overview of the recommended action notification system 50 of this embodiment.
- the recommended action notification system 50 includes a plurality of automatic analyzers 51a to z and a learning device 52 connected via a network 53.
- the automatic analyzer 51 is installed in a clinical laboratory, a clinical laboratory center, or the like.
- the number of automatic analyzers 51 connected to the learning device 52 is not particularly limited.
- the hardware of the learning device 52 is an information processing device including a storage device such as an HDD (Hard disk drive) and an SSD (Solid State Drive).
- FIG. 1B which will be described later, the functions executed by the automatic analyzer 51 or the learning device 52 are represented by a functional block diagram, and each function reads the software program code into the main memory and the program code read by the processor.
- the function realized by the information processing apparatus by a program may be referred to as a "part".
- the data processed by the learning device 52 is a large-scale data, it is not necessary to store the data itself in the storage device of the learning device 52, for example, it is stored in the object storage on the cloud and the target data is accessed. You may save the data path to do so.
- the learning device 52 may be realized by one server or a distributed processing server, and is not limited to the physical configuration of the hardware.
- FIG. 1B is a functional block diagram for showing an outline of the recommended action notification system 50.
- the data set creation unit 61 determines the effect of the action from the effect information of the action input by the operator of the device from the input unit 11 and the device data stored in the database 62.
- a learning data set is an action taken as a trigger when the abnormality detection unit 60 detects an abnormality from a sample measurement result or a maintenance result and notifies the operator (for example, displays it on the display unit 10). The target of the action to create.
- the learning data set generated by the data set creation unit 61 is transmitted to the learning device 52 via the network 53.
- the updating unit 71 creates a teacher signal based on the learning data set and updates the learning model 72.
- the updated learning model 72 is delivered to the automated analyzer 51 via the network 53.
- the automatic analyzer 51 every time the sample measurement result or the maintenance result is output, the related device data stored in the database 62 is input to the learning model 72 by the processing unit 63, and the recommended action is output. The recommended action is notified to the operator by being displayed on the display unit 10 of the automatic analyzer 51.
- FIG. 2 shows an outline of the automatic analyzer 51.
- the automatic analyzer 51 mainly includes a sample disk 1, a reagent disk 2, a reaction disk 3, a reaction tank 4, a sampling mechanism 5, a pipetting mechanism 6, a stirring mechanism 7, a photometric mechanism 8, a cleaning mechanism 9, a display unit 10, and an input. It is configured to include a unit 11, a storage unit 12, and a control unit 13.
- reagent disk 2 On the reagent disk 2, a plurality of reagent bottles 18 containing reagents for mixing and reacting with a sample are fixedly arranged on the circumference of the circular disk 19, and the surroundings thereof are temperature-controlled. It is a cold storage 20. Further, the circular disk 19 rotates in the circumferential direction so as to be positioned by a drive mechanism composed of a motor (not shown), a rotating shaft, or the like. A plurality of reaction vessel holders 22 holding reaction vessels 21 for storing samples and reagents are attached to the reaction disk 3, and the drive mechanism 23 repeats circumferential rotation and stop in a fixed cycle to react. The container 21 is intermittently transferred.
- the reaction vessel 4 is installed along the movement trajectory of the reaction vessel 21, and in order to promote the chemical reaction between the sample and the reagent, for example, the sample and the reagent in the reaction vessel 21 are mixed with temperature-controlled constant temperature water. It is a constant temperature bath that controls the reaction solution to a constant temperature.
- the reaction vessel 21 moves in the reaction vessel 4.
- the sampling mechanism 5 includes a probe 24, an arm 26 attached to the support shaft 25, and a drive mechanism that allows the support shaft 25 to reciprocate between the sample disk 1 and the reaction disk 3 around the center of rotation. ,
- the sample in the sample container 16 transferred to the fixed position by the rotation of the sample disk 1 is supplied to the reaction container 21 according to a predetermined sequence.
- the pipetting mechanism 6 includes a probe 27, an arm 29 attached to the support shaft 28, and a drive mechanism that allows the support shaft 28 to reciprocate between the reagent disk 2 and the reaction disk 3 around the rotation center.
- the reagent in the reagent bottle 18, which is configured to be prepared and transferred to a fixed position by the rotation of the reagent disk 2, is supplied to the reaction vessel 21 according to a predetermined sequence.
- Each of the sample container 16 and the reagent bottle 18 contains different types of samples and reagents, and a required amount is supplied to the reaction container 21.
- the stirring mechanism 7 is a stirring mechanism that stirs and mixes the samples and reagents in the reaction vessel 21 that have been transferred to the position (stirring position).
- the stirring mechanism 7 is a stirring mechanism based on a non-contact stirring method, and drives a fixing portion 31 for fixing a piezoelectric element (not shown) at a position where sound waves can be irradiated from the side surface of the reaction vessel 21 at the stirring position, and a piezoelectric element.
- the piezoelectric element driver 14 and the stirring mechanism controller 15 are included.
- the control unit 13 sends a control signal to the sample disk 1, the reagent disk 2, the reaction disk 3, the reaction tank 4, the sampling mechanism 5, the pipetting mechanism 6, the stirring mechanism 7, the photometric mechanism 8, and the cleaning mechanism 9, and each operation thereof is performed.
- the function of the recommended action notification system of this embodiment specifically, the functions of the abnormality detection unit 60, the data set creation unit 61, and the processing unit 63 shown in FIG. 1B are executed.
- the display unit 10 displays various screens such as analysis items and analysis results, and the input unit 11 inputs various information such as analysis items.
- the storage unit 12 stores various information such as a predetermined sequence (program) and analysis items for controlling each mechanism.
- the updated learning model 72 delivered from the database 62 and the learning device 52 shown in FIG. 1B is also stored in the storage unit 12.
- FIG. 3 is an update flowchart of the learning model 72. Since the learning model 72 has different input / output depending on the action to be learned, the recommended action notification system 50 individually creates a learning model according to the action. The action to be learned is selected in the system, but it is assumed that the operator operation to be performed when the device is abnormal, such as maintenance execution, reagent exchange, and sample dilution remeasurement, is selected.
- the learning model is initially learned using the device abnormality data collected in advance by the recommended action notification system 50, the corresponding recommended action, and the cause of the abnormality as a teacher signal.
- the abnormality detection unit 60 of the automatic analyzer 51 detects some abnormality from the sample inspection result or maintenance result performed by the apparatus, it notifies the operator of an alarm (S10).
- the operator executes an action for dealing with the abnormality (S01).
- the control unit 13 of the automatic analyzer records the operator's action as the operation log of the device (S11), and when it is determined from the operation log that the operator has performed the action to be learned, the data set creation unit 61 of the automatic analyzer After the action is executed, the input screen for inputting the validity evaluation of the operator's action is displayed on the display unit 10 (S12).
- this flowchart is applied when the abnormality detection unit 60 detects the occurrence of some abnormality in the device and the operator executes an action triggered by that, and the action in periodic maintenance is applied. , Not subject to collection of training datasets. This is because learning can be optimized by collecting as a learning data set only when the presence or absence of the effect of the action is clear.
- FIG. 4 shows an example of the action evaluation input screen 80.
- the action executed by the operator is displayed on the execution content display unit 81 (the specific action name performed by the operator is displayed in [Action name] in the figure).
- the operator inputs the validity evaluation of the action by selecting any of the effect buttons 82 (S02). Further, the cause of the abnormality found by the action of the operator is input to the abnormality cause selection unit 83 (S02). In order to reduce the number of input steps of the operator and suppress the variation in expression, the input is made by the selection method.
- the action evaluation and the cause of the abnormality input by the operator are recorded in the storage unit 12 of the automatic analyzer 51 (S13).
- the data set creation unit 61 creates a learning data set including the evaluation, the cause of abnormality, and the related device data input by the operator for each action (S14).
- the created learning data set is transmitted to the learning device 52 (S15).
- the related device data is device data collected for a preset period based on the operator's action execution date and time in order to update the learning model.
- An example of related device data is shown in FIG.
- device data it is not limited to the mechanism and hardware of the device, but broadly includes information on the mechanism related to the operation of the device, hardware, reagents, samples, analysis result data, and the like. Since the device data that shows the sign differs depending on the abnormality (abnormal cause) for which the operator's action execution is valid, the related device data is set for each action.
- FIG. 5 the types of related device data set for each action are collectively illustrated for each data set group.
- a dataset group is a grouping of actions according to their content.
- the three dataset groups of "device”, "reagent”, and “sample” are groups of actions related to the device, reagent, and sample, respectively, and have high commonality in the types of related device data to be collected.
- the learning data set include time information regarding the action.
- the time information includes the execution date and time of the action and the error occurrence date and time for the abnormality (error) for which the execution of the action detected by the abnormality detection unit 60 was effective.
- the update unit 71 of the learning device 52 receives such a learning data set (S21), and updates the learning model 72 with a teacher signal based on the received learning data set (S22).
- the updated learning model is delivered to the automatic analyzer 51 at a predetermined timing.
- FIG. 6 shows an example of a learning model using a neural network.
- the learning model 90 propagates the information input to the input layer 91 to the intermediate layer 92, further propagates to the output layer 93 in order, and outputs an inference result based on the information input from the output layer 93 to the input layer 91. do.
- the input layer 91, the intermediate layer 92, and the output layer 93 each have a plurality of input units, intermediate units, and output units indicated by circles.
- the intermediate layer of the neural network is usually a plurality of layers, it is represented by the intermediate layer 92 in the figure.
- the information input to each input unit of the input layer 91 is weighted by the coupling coefficient between the input unit and the intermediate unit, and is input to each intermediate unit.
- the value of the intermediate unit of the intermediate layer 92 is calculated by adding the value from the input unit.
- the output from each intermediate unit of the intermediate layer 92 is weighted by the coupling coefficient between the intermediate unit and the output unit, and is input to each output unit.
- the values of the output units of the output layer 93 are calculated.
- the processing in the intermediate layer 92 corresponds to nonlinearly converting the value of the input data input to the input layer 91 and outputting it as the output data of the output layer 93.
- the learning model 90 shown in FIG. 6 is a learning model related to the action “reaction cell cleaning” and is shown together with an example of input / output data.
- Related device data (or its processing data, hereinafter referred to as related device data without distinction) is input to the input layer 91, and the calculation result corresponding to the input is output from the output layer 93.
- the related device data input to the input layer 91 includes time-series measurement results 94 such as analysis results and maintenance data defined for each learning model, and analysis information 95 such as reagent information and sample information that affect actions. Is done.
- the output data from the output layer 93 is an inference result related to the action “reaction cell cleaning”, and includes the action-related information 97 and the abnormality cause 98.
- the action-related information 97 includes whether or not an action (here, “reaction cell cleaning”) is recommended, an average remaining time until an error occurs, and an average remaining time until an action is executed.
- the average remaining time until an error occurs refers to the average remaining time required for the abnormality detection unit 60 to detect an abnormality (error) corresponding to the action, and the average remaining time until the action is executed is the action by the operator. Refers to the average remaining time required to execute.
- the average time remaining before an error occurs is a good indicator of when the need for action will occur in the future, even if the action is not recommended.
- the remaining time until the action is executed is further taken into account the operation at each facility and the experience value of the operator, and is generally a guide for knowing when the action is being executed in the future.
- the output values of the output unit "Recommended reaction cell cleaning (action)” and “Abnormal cause (dirt of cells)” are probability values, and actions are taken depending on whether the output probability value is above or below a predetermined threshold. It is recommended as, or it is determined whether it is the cause of the abnormality. For example, the closer the output unit "recommended reaction cell cleaning” is to 1, the more recommended it is, and the closer it is to 0, the less recommended. Similarly, the closer the output unit “cause of abnormality (dirt of cell)” is to 1, the more certain it is, and the closer it is to 0, the more suspicious it is. On the other hand, the output from the output unit “average remaining time until the error occurs” and the output unit "average remaining time until the action is executed” is the value of the remaining time itself.
- the update unit 71 creates a teacher signal corresponding to the input / output data of the learning model 90 based on the learning data set.
- the input value of the teacher signal is the related device data.
- the output value of the teacher signal is as follows in this example. If the effectiveness evaluation of the action is effective, the output value of "recommended reaction cell cleaning” is set to 1, and if it is not effective, the output value of "recommended reaction cell cleaning” is set to 0. If the cause of the abnormality is "dirt in the cell", the output value of "cause of abnormality (dirt in the cell)” is set to 1, otherwise the output value of "cause of abnormality (dirt in the cell)” is set to 0. ..
- the elapsed time from the predetermined reference time of the error occurrence date and time and the action execution date and time is calculated and used as the output value of the teacher signal.
- the predetermined reference time point may be, for example, the execution date and time of the previous action.
- the update unit 71 inputs the input value (related device data) of the created teacher signal to the input layer 91, and the coupling coefficient between the units is such that the value output from the output layer 93 becomes the output value of the teacher signal. Adjust the value.
- FIG. 7 is an example of the operation table 100.
- the learning device 52 collects information on the operation status of the automatic analyzer for each facility and manages it in the operation table 100.
- a typical operation example is defined as a group (here, group A is shown), and it is determined whether or not to be a target of the group based on the closeness between the operation status defined by the group and the operation status of each facility. ..
- the operation status information column 101 the operation status information defined for the group and the operation status information of each facility are registered.
- the correlation value between the operation status defined for the group and the operation status of each facility is registered in the approximation column 102, and the group (group A) determined based on the correlation value is registered in the target determination column (group A) 103. Whether or not it is the target of is registered.
- facility A is determined to be a target facility of group A, and other facilities are excluded.
- it is not a condition of the group that all the items of the operation status are close to each other, but a condition of the group that one or more items of the operation status are close to each other depending on the cause of the abnormality. So good.
- FIG. 8 shows a flowchart in which the automated analyzer 51 outputs a recommended action using the learning model 72. It is desirable to execute the diagnosis using the learning model in real time every time the automatic analyzer 51 outputs the sample analysis result, the maintenance result, or the like (S31). This makes it possible to detect an abnormality in the device at an early stage and notify the operator.
- the processing unit 63 selects a learning model for executing the diagnosis (S32).
- a learning model in which the output result of the automatic analyzer 51 is set as the related device data may be extracted, and a plurality of learning models may be extracted.
- the related device data including the output result corresponding to the learning model is input (S33).
- the inference result such as the recommended user action output from the learning model is acquired (S34) and displayed on the screen (S35).
- FIG. 9 shows an example of a display screen of the recommended action displayed on the display unit 10.
- the recommended action display screen 110 includes a recommended action display unit 111, a recommended implementation time display unit 112, and a reference information display unit 113.
- whether or not the action that is the output value from the learning model is recommended is determined by whether the probability value output from the learning model is above or below a certain threshold value, but the recommended action is not limited to one, and a plurality of actions are recommended. Actions may be recommended. In this way, when a plurality of actions are recommended, the degree of recommendation can be determined by the probability value, so the recommended action having a large difference between the threshold value and the probability value is displayed on the screen as the recommended action with high priority.
- the recommended degree is displayed by the bar 114 for each recommended action, and the contents of the recommended action are displayed in order from the highest priority.
- the action “reaction cell cleaning” having a higher priority is displayed on the preceding recommended action display screen 110a, and the action “reaction cell exchange” is displayed on the subsequent recommended action display screen 110b.
- the recommended action display unit 111 displays the name of the recommended action and the degree of recommendation thereof. The larger the hatched portion of the bar 114, the greater the degree of recommendation.
- the recommended implementation time display unit 112 also displays the implementation history of the action in the automated analyzer and the date and time of those actions if there is a schedule.
- the shaded date in the calendar indicates today, and the black circle indicates that the action "reaction cell cleaning" is recommended.
- the error occurrence date and time and the action execution date and time in a general automatic analyzer are estimated from the average remaining time output from the learning model and displayed as reference information. It is possible for the operator to actually perform the recommended action based on the timing in general and average cases.
- the reference information display unit 113 can display the implementation history of device data that is highly important for determining the necessity of a predetermined recommended action, for example, reaction cell blank measurement data in the case of reaction cell cleaning. This makes it possible for the operator to confirm whether the recommended action is appropriate in the light of the operator's experience, without the operator collecting device data again.
- the light source lamp is generally replaced at regular intervals or after a certain period of time, whichever comes first, or when the photometer check value exceeds the threshold value. Even if the exchange threshold value of the photometer check value is set, the lamp does not always turn off after the threshold value is exceeded, so it is recommended to exchange it regularly or after using it for a certain period of time or longer. However, even though regular replacement is recommended, the appropriate period will vary depending on the start-up time of the equipment at each facility. For example, in a facility that uses 5 hours a day and a facility that uses 10 hours a day, a facility that uses 10 hours a day needs to be replaced more frequently. Therefore, in the following description, it is assumed that the learning data set used for learning collects data from the automated analyzer of the facility whose device operating time is close.
- time-series measurement data related to the light source lamp is used as related device data, timing when an abnormality occurs in the measurement data due to the light source lamp as time information related to the action, and learning data including the operator's light source lamp replacement timing. It collects sets and recommends effective light source lamp replacement timing.
- Time-series photometer check data of the light source lamp, time-series main wavelength, and reaction monitor data for each sub-wavelength, which are related device data are input to the input layer 121 of the learning model 120, and the light source lamp is replaced from the output layer 122. Whether it is recommended or not, as reference information, the average remaining time until an abnormality starts to appear in the reaction monitor data in the automatic analyzer of the facility where the equipment operation is similar, and the average remaining time until the light source lamp replacement is executed are output.
- FIG. 11 shows time-series absorbance data which is an example of photometer check data used for learning or inference. As time passes after the light source lamp is replaced with a new one, the amount of light decreases and the absorbance tends to gradually increase.
- FIG. 12 shows an example of reaction monitor data used for learning or inference. When the life of the light source lamp is shortened, the amount of light becomes unstable, spike noise 131 in which the absorbance jumps is generated in the reaction process, and linearity abnormality occurs. Spike noise 131 tends to increase over time.
- the learning model 120 infers whether or not to recommend the replacement of the light source lamp by grasping such a feature amount from the time series data.
- FIG. 13 shows a recommended action display screen 140.
- the recommended action display screen 140 includes a recommended action display unit 141, a recommended implementation time display unit 142, and a reference information display unit 143. Since each display unit is the same as the display unit shown in FIG. 9, overlapping description will be omitted.
- the operator can recommend the light source lamp based on the time series data of the related device data, and the light source at an effective timing based on the average error occurrence timing and the operator replacement timing in the automatic analyzer performing the same operation. The lamp can be replaced.
- a learning example of the action "cleaning the outer wall of the sample probe” will be described with reference to FIGS. 14 to 17.
- the outer wall of the sample probe should be cleaned daily after the measurement is complete. Failure to clean may reduce sample dispensing accuracy. If the sample dispensing accuracy decreases, the sample measurement data will be affected, but it is not clear from the measurement data alone whether the sample is originally an abnormal value or the sample dispensing accuracy has decreased.
- the user is asked to input whether or not the measurement data is abnormal and the effect is obtained when the outer wall of the sample probe is cleaned. That is, if there is an effect, it can be presumed that the cause was a decrease in sample dispensing accuracy.
- the learning data set used for learning collects data from an automatic analyzer of a facility whose operation status is close to that which affects the degree of contamination of the sample probe.
- the time-series reaction monitor data of the main wavelength and the sub-wavelength are used as the related device data, the timing when the reaction monitor data is abnormal due to the dirt on the outer wall of the sample probe as the time information related to the action, and the operator sample. It collects training data sets including probe exterior wall cleaning timing and recommends effective sample probe exterior wall cleaning timing.
- Reaction monitor data for each of the main wavelength and sub-wavelength of the time series, which are related device data is input to the input layer 151 of the learning model 150.
- the average remaining time until an abnormality starts to appear in the reaction monitor data and the average remaining time until the sample probe outer wall cleaning is executed in the automatic analyzer of the facility that is similar to are output.
- FIG. 15 shows the measurement results when the outer wall of the sample probe is neglected
- FIGS. 16A and 16B show the reaction monitor data when the outer wall of the sample probe is neglected
- FIG. 16A is the monitor data of the ALB reaction process
- FIG. 16B is the monitor data of the CRE reaction process.
- FIG. 17 shows a recommended action display screen 160.
- the recommended action display screen 160 includes a recommended action display unit 161 and a recommended implementation time display unit 162. Since each display unit is the same as the display unit shown in FIG. 9, overlapping description will be omitted.
- the operator can recommend cleaning the outer wall of the sample probe based on the time-series data of the related equipment data, and the effective timing based on the average error occurrence timing and operator replacement timing in the automatic analyzer performing the same operation. Allows cleaning of the outer wall of the sample probe.
- the recommended action display screen 160 may be provided with a reference information display unit to display time-series data as shown in FIG. 16A.
- the present invention has been described above based on the examples and modifications, the present invention is not limited to the above-mentioned examples and modifications, and various modifications can be made without changing the gist of the invention.
- the recommended action notification system is configured by a plurality of automatic analyzers and a learning device, it can also be configured by a single automated analyzer.
- a learning data set from a plurality of automated analyzers for learning it is possible to collect information from a large number of automated analyzers at an early stage and promote the learning effect.
- learning data from an automated analyzer other than the self it is possible to obtain inference results in the average automated analyzer as described above, and it is possible to determine whether or not to actually execute the recommended action. Can be used for.
- the processing unit 63 may be provided in the learning device or other information processing device to perform the inference.
- the recommended action notification system may be provided with a configuration for notifying the operator with a mobile terminal such as a smartphone. This allows you to be notified when a recommended action occurs, without limiting the operator's location. Further, in addition to the screen input for the effect input of the operator's action, another input method such as voice input that reduces the input load of the operator may be adopted.
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Automatic Analysis And Handling Materials Therefor (AREA)
Abstract
Description
第1の自動分析装置からの検体分析結果データまたはメンテナンス結果データを受けて、検体分析結果データまたはメンテナンス結果データを含む関連装置データを学習モデルに入力し、学習モデルが出力する所定のアクションの実行を推奨する確率値が所定の閾値以上である場合にはオペレータに第1の自動分析装置に対して所定のアクションの実行を推奨する処理部と、
複数の自動分析装置からの学習用データセットに基づき学習モデルの更新を行う更新部とを備える。
Claims (14)
- 第1の自動分析装置を含む複数の自動分析装置と、前記複数の自動分析装置とネットワークで接続される学習装置とを含み、オペレータに前記第1の自動分析装置に対して実行するアクションを推奨する推奨アクション通知システムであって、
前記第1の自動分析装置からの検体分析結果データまたはメンテナンス結果データを受けて、前記検体分析結果データまたは前記メンテナンス結果データを含む関連装置データを学習モデルに入力し、前記学習モデルが出力する所定のアクションの実行を推奨する確率値が所定の閾値以上である場合にはオペレータに前記第1の自動分析装置に対して前記所定のアクションの実行を推奨する処理部と、
前記複数の自動分析装置からの学習用データセットに基づき前記学習モデルの更新を行う更新部とを備える推奨アクション通知システム。 - 請求項1において、
前記第1の自動分析装置は前記処理部を備え、前記学習装置は前記更新部を備え、
前記更新部によって更新された前記学習モデルが、前記第1の自動分析装置に配信される推奨アクション通知システム。 - 請求項1において、
前記学習モデルは、前記所定のアクションに応じて設定された関連装置データが入力される入力層と、前記入力層への前記関連装置データの入力を受けて前記所定のアクションに関係する推論結果を出力する出力層とを備え、
前記推論結果として、前記所定のアクションの実行を推奨する確率値と前記所定のアクションの実行が有効な所定の異常が生じている確率値とを含む推奨アクション通知システム。 - 請求項3において、
前記自動分析装置からの異常発生検知を受けて、オペレータが前記所定のアクションを実行したとき、
前記自動分析装置のデータセット作成部は、前記自動分析装置の表示部にアクション評価入力画面を表示し、前記所定のアクションの実行日時を基準にあらかじめ設定された期間について前記所定のアクションに応じて設定された関連装置データを収集し、収集した前記関連装置データと、前記アクション評価入力画面より入力された前記所定のアクションの有効性評価及び異常原因とを含む学習用データセットを作成する推奨アクション通知システム。 - 請求項4において、
前記推論結果として、前記複数の自動分析装置において前記所定の異常が発生するまでの平均的な残り時間と、前記複数の自動分析装置において前記所定のアクションが実行されるまでの平均的な残り時間とを含み、
前記自動分析装置のデータセット作成部は、前記学習用データセットに、さらに前記自動分析装置において所定の基準時点から前記所定の異常が発生するまでの経過時間及び前記自動分析装置において所定の基準時点から前記所定のアクションを実行するまでの経過時間を含ませる推奨アクション通知システム。 - 請求項3において、
前記処理部は、前記検体分析結果データまたは前記メンテナンス結果データが、その入力層に入力される関連装置データとして設定されている複数の前記学習モデルを抽出し、
前記処理部は、前記検体分析結果データまたは前記メンテナンス結果データを含む関連装置データを抽出された前記複数の学習モデルに入力し、当該学習モデルが出力する所定のアクションの実行を推奨する確率値が所定の閾値以上である学習モデルが複数ある場合には、前記確率値が所定の閾値以上であった複数の学習モデルに対応する複数の所定のアクションをオペレータに推奨する推奨アクション通知システム。 - 請求項5において、
前記処理部は、前記第1の自動分析装置の表示部に推奨アクション表示画面を表示し、
前記推奨アクション表示画面には、前記学習モデルによる、前記所定のアクションの実行を推奨する確率値、前記複数の自動分析装置において前記所定の異常が発生するまでの平均的な残り時間、及び前記複数の自動分析装置において前記所定のアクションが実行されるまでの平均的な残り時間についての推論結果に基づき、オペレータに実行を推奨する前記所定のアクションの名称、前記複数の自動分析装置において前記所定の異常が発生すると推定される日時、及び前記複数の自動分析装置において前記所定のアクションが実行されると推定される日時が表示される推奨アクション通知システム。 - 請求項1において、
前記更新部は、前記複数の自動分析装置のうち、運用及び検査内容を含む運用状況の近似性に基づき前記複数の自動分析装置をグルーピングし、運用状況が近似するとしてグループ化された自動分析装置からの前記学習用データセットに基づき前記学習モデルの更新を行う推奨アクション通知システム。 - 検体分析またはメンテナンスを行う自動分析装置であって、
オペレータが自動分析装置に対して実行する所定のアクションに応じて設定された関連装置データが入力される入力層と、前記入力層への前記関連装置データの入力を受けて前記所定のアクションに関係する推論結果を出力する出力層とを備える学習モデルを記憶する記憶部と、
前記記憶部から前記学習モデルを呼び出し、前記検体分析の結果データまたは前記メンテナンスの結果データを含む関連装置データを前記学習モデルに入力し、前記学習モデルが出力する前記所定のアクションの実行を推奨する確率値が所定の閾値以上である場合にはオペレータに前記所定のアクションの実行を推奨する処理部と、
前記処理部により推奨された前記所定のアクションの名称を表示する表示部とを有する自動分析装置。 - 請求項9において、
異常を検知する異常検知部と、
前記異常検知部からの異常発生検知を受けて、オペレータが前記所定のアクションを実行したとき、前記表示部にアクション評価入力画面を表示し、前記所定のアクションの実行日時を基準にあらかじめ設定された期間について前記所定のアクションに応じて設定された関連装置データを収集し、収集した前記関連装置データと、前記アクション評価入力画面より入力された前記所定のアクションの有効性評価及び異常原因とを含む学習用データセットを作成するデータセット作成部とを有する自動分析装置。 - 請求項10において、
前記学習用データセットは、学習装置に送信され、前記学習モデルの更新に使用される自動分析装置。 - 第1の自動分析装置を含む複数の自動分析装置と、前記複数の自動分析装置とネットワークで接続される学習装置とを含む推奨アクション通知システムにより、オペレータに前記第1の自動分析装置に対して実行するアクションを推奨する推奨アクション通知方法であって、
前記複数の自動分析装置は、学習用データセットを前記学習装置に送信し、
前記学習装置は、前記複数の自動分析装置からの前記学習用データセットに基づき学習モデルを更新し、
前記学習装置は、更新された前記学習モデルを前記第1の自動分析装置に配信し、
前記第1の自動分析装置は、検体分析またはメンテナンスを行い、
前記第1の自動分析装置は、前記検体分析の結果データまたは前記メンテナンスの結果データを含む関連装置データを前記学習モデルに入力し、前記関連装置データの入力を受けて前記学習モデルが出力する所定のアクションの実行を推奨する確率値が所定の閾値以上である場合には、オペレータに前記所定のアクションの実行を推奨する推奨アクション通知方法。 - 請求項12において、
前記複数の自動分析装置のいずれかの自動分析装置が異常の発生を検知したとき、
前記異常の発生を検知した自動分析装置は、オペレータに前記異常の検知を通知し、
前記異常の発生を検知した自動分析装置は、オペレータが前記異常の検知の通知を受けて前記所定のアクションを実行したとき、当該自動分析装置の表示部にアクション評価入力画面を表示し、
前記異常の発生を検知した自動分析装置は、前記所定のアクションの実行日時を基準にあらかじめ設定された期間について前記所定のアクションに応じて設定された関連装置データを収集し、収集した前記関連装置データと、前記アクション評価入力画面より入力された前記所定のアクションの有効性評価及び異常原因とを含む学習用データセットを作成する推奨アクション通知方法。 - 請求項12において、
前記学習装置は、前記複数の自動分析装置のうち、運用及び検査内容を含む運用状況の近似性に基づき前記複数の自動分析装置をグルーピングし、運用状況が近似するとしてグループ化された自動分析装置からの学習用データセットに基づき前記学習モデルの更新を行う推奨アクション通知方法。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202180050948.6A CN115956205A (zh) | 2020-09-28 | 2021-02-17 | 自动分析装置、推荐动作通知***及推荐动作通知方法 |
EP21871868.2A EP4220502A1 (en) | 2020-09-28 | 2021-02-17 | Automatic analysis device, recommended action notification system, and recommended action notification method |
US18/025,117 US20230375580A1 (en) | 2020-09-28 | 2021-02-17 | Automatic Analyzer, Recommended Action Notification System, and Recommended Action Notification Method |
JP2022551127A JP7490793B2 (ja) | 2020-09-28 | 2021-02-17 | 自動分析装置、推奨アクション通知システムおよび推奨アクション通知方法 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2020-162044 | 2020-09-28 | ||
JP2020162044 | 2020-09-28 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022064731A1 true WO2022064731A1 (ja) | 2022-03-31 |
Family
ID=80845127
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2021/005819 WO2022064731A1 (ja) | 2020-09-28 | 2021-02-17 | 自動分析装置、推奨アクション通知システムおよび推奨アクション通知方法 |
Country Status (5)
Country | Link |
---|---|
US (1) | US20230375580A1 (ja) |
EP (1) | EP4220502A1 (ja) |
JP (1) | JP7490793B2 (ja) |
CN (1) | CN115956205A (ja) |
WO (1) | WO2022064731A1 (ja) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2019174424A (ja) | 2018-03-29 | 2019-10-10 | シスメックス株式会社 | 検体分析装置のモニタリングデータの生成装置、検体分析装置、検体分析装置のモニタリングデータ生成システム、前記システムの構築方法、検体分析装置のモニタリングデータの生成方法及び検体分析装置のモニタリング方法 |
JP2019536049A (ja) * | 2016-12-02 | 2019-12-12 | エフ.ホフマン−ラ ロシュ アーゲーF. Hoffmann−La Roche Aktiengesellschaft | 生物学的試料を分析する自動分析装置のための故障状態予測 |
JP2020135007A (ja) * | 2019-02-13 | 2020-08-31 | セイコーエプソン株式会社 | 情報処理装置、学習装置及び学習済モデル |
-
2021
- 2021-02-17 JP JP2022551127A patent/JP7490793B2/ja active Active
- 2021-02-17 US US18/025,117 patent/US20230375580A1/en active Pending
- 2021-02-17 CN CN202180050948.6A patent/CN115956205A/zh active Pending
- 2021-02-17 EP EP21871868.2A patent/EP4220502A1/en active Pending
- 2021-02-17 WO PCT/JP2021/005819 patent/WO2022064731A1/ja active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2019536049A (ja) * | 2016-12-02 | 2019-12-12 | エフ.ホフマン−ラ ロシュ アーゲーF. Hoffmann−La Roche Aktiengesellschaft | 生物学的試料を分析する自動分析装置のための故障状態予測 |
JP2019174424A (ja) | 2018-03-29 | 2019-10-10 | シスメックス株式会社 | 検体分析装置のモニタリングデータの生成装置、検体分析装置、検体分析装置のモニタリングデータ生成システム、前記システムの構築方法、検体分析装置のモニタリングデータの生成方法及び検体分析装置のモニタリング方法 |
JP2020135007A (ja) * | 2019-02-13 | 2020-08-31 | セイコーエプソン株式会社 | 情報処理装置、学習装置及び学習済モデル |
Also Published As
Publication number | Publication date |
---|---|
US20230375580A1 (en) | 2023-11-23 |
JPWO2022064731A1 (ja) | 2022-03-31 |
EP4220502A1 (en) | 2023-08-02 |
CN115956205A (zh) | 2023-04-11 |
JP7490793B2 (ja) | 2024-05-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP5562421B2 (ja) | 自動分析装置、分析方法及び情報処理装置 | |
EP0871034B1 (en) | Automatic analyzer and support system therefor | |
JP5425487B2 (ja) | 自動分析装置 | |
EP2775307B1 (en) | Automated analysis system | |
JP5183457B2 (ja) | 自動分析装置、その支援システム | |
JP2010261822A (ja) | 自動分析装置、及び分析方法 | |
JP4933198B2 (ja) | 自動分析装置 | |
US11537968B2 (en) | Predictive inventory control including scheduling and performing bio-fluid tests out of order based on reagent inventory expiration | |
JP2008209329A (ja) | 自動分析装置 | |
WO2015171581A1 (en) | Intelligent service assistant - instrument side software client | |
US11971424B2 (en) | Automatic analyzer | |
JP6629832B2 (ja) | 自動分析装置 | |
JP5147906B2 (ja) | 精度管理システム | |
JP2018017676A (ja) | 自動分析装置及びプログラム | |
CN104508490A (zh) | 数据处理装置以及使用该数据处理装置的自动分析装置 | |
WO2022064731A1 (ja) | 自動分析装置、推奨アクション通知システムおよび推奨アクション通知方法 | |
JP6991363B2 (ja) | 自動分析装置 | |
JP7174066B2 (ja) | 自動分析装置 | |
JP2006292698A (ja) | 臨床検査用自動分析装置の精度管理方法、及び自動分析装置 | |
JP2009300340A (ja) | 自動分析装置管理システム | |
JP6219757B2 (ja) | 自動分析装置及び異常判定方法 | |
JP2014021034A (ja) | 自動分析装置および保守サポートシステム | |
JP2014016218A (ja) | 自動分析装置 | |
CN113574390A (zh) | 数据解析方法、数据解析***以及计算机 | |
JP7357557B2 (ja) | 自動分析装置及び反応異常判定方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21871868 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2022551127 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2021871868 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2021871868 Country of ref document: EP Effective date: 20230428 |