CN114118201A - Medical equipment performance index detection method and device based on active learning - Google Patents
Medical equipment performance index detection method and device based on active learning Download PDFInfo
- Publication number
- CN114118201A CN114118201A CN202111138481.9A CN202111138481A CN114118201A CN 114118201 A CN114118201 A CN 114118201A CN 202111138481 A CN202111138481 A CN 202111138481A CN 114118201 A CN114118201 A CN 114118201A
- Authority
- CN
- China
- Prior art keywords
- data
- abnormal
- marking
- classifier
- normal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 71
- 230000002159 abnormal effect Effects 0.000 claims abstract description 188
- 238000012549 training Methods 0.000 claims abstract description 58
- 238000000034 method Methods 0.000 claims abstract description 49
- 238000007637 random forest analysis Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 6
- 238000012163 sequencing technique Methods 0.000 claims description 3
- 239000003550 marker Substances 0.000 claims 1
- 238000002372 labelling Methods 0.000 abstract description 29
- 239000000523 sample Substances 0.000 description 21
- 230000008569 process Effects 0.000 description 16
- 230000005856 abnormality Effects 0.000 description 7
- 230000006870 function Effects 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000011895 specific detection Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
- G16H40/40—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the management of medical equipment or devices, e.g. scheduling maintenance or upgrades
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- General Business, Economics & Management (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The application provides a medical equipment performance index detection method and device based on active learning, and the method comprises the following steps: s1: training a first classifier through abnormal labeled data and unlabeled data in a training data set, and labeling normal data from the unlabeled data through the first classifier; s2: training a second classifier through the abnormal marking data and the normal data; s3: predicting each unlabeled data through a second classifier, and determining candidate abnormal data according to the prediction scores; s4: manually judging whether the candidate abnormal data are abnormal data, if so, marking the candidate abnormal data as abnormal marking data; s5: judging whether the quantity of the abnormal annotation data and the normal data reaches a preset value, if not, repeatedly executing S3 and S4 until the quantity of the abnormal annotation data and the normal data reaches the preset value; s6: and detecting Key Performance Indicators (KPI) flow of the medical equipment according to the acquired marking data. The method can avoid the normal sample from being wrongly marked as the abnormal sample, and improves the accuracy of the abnormal detection.
Description
Technical Field
The application relates to the technical field of data detection, in particular to a medical equipment performance index detection method and device based on active learning.
Background
At present, in the medical field, medical equipment often generates a large amount of monitoring data, namely KPI flow, in the working process, and professional staff observes the data in real time to monitor whether the equipment is abnormal, the data is generally a time sequence, and if the data is artificially detected to judge whether the equipment is abnormal, great manpower and cost are consumed, so that some algorithms are usually adopted to assist in abnormality detection of the KPI flow of the medical equipment in the medical field.
However, the auxiliary algorithm in the related art needs to label a large amount of monitoring data to complete detection, but the cost of labeling the monitoring data is high, and the data labeling method in the related art is prone to falsely label a normal sample as an abnormal sample, resulting in low detection accuracy. Therefore, a method for improving the accuracy of labeling samples and detecting abnormalities with less labeling amount is needed.
Disclosure of Invention
The present application is directed to solving, at least to some extent, one of the technical problems in the related art.
Therefore, a first objective of the present application is to provide a medical device performance index detection method based on active learning, which optimizes the labeling process of active learning according to the characteristics of abnormal samples and normal samples of a time sequence, ensures that as few normal samples as possible are labeled as abnormal in the active learning stage, greatly reduces the probability of false alarm, improves the application of abnormal detection in the PU learning medical field, can obtain more labeled samples, and improves the accuracy of abnormal detection on the performance index of the medical device.
The second purpose of the invention is to provide a medical equipment performance index detection device based on active learning.
A third object of the invention is to propose a non-transitory computer-readable storage medium.
In order to achieve the above object, a first aspect of the present invention provides a method for detecting a performance index of a medical device based on active learning, including the following steps:
s1: acquiring a training data set, training a first classifier through abnormal labeled data and unlabeled data in the training data set, and marking normal labeled data from the unlabeled data through the first classifier;
s2: training a second classifier through the abnormal marking data and the normal marking data;
s3: predicting each unlabeled data through the second classifier, and determining candidate abnormal data according to the prediction scores;
s4: manually judging whether the candidate abnormal data is abnormal data, if the candidate abnormal data is abnormal data, marking the candidate abnormal data as abnormal marking data, and updating the abnormal marking data, the normal marking data and the unmarked data;
s5: judging whether the quantity of the abnormal annotation data and the normal annotation data reaches a preset quantity, if not, repeatedly executing S3 and S4 until the quantity of the abnormal annotation data and the normal annotation data reaches the preset quantity;
s6: and carrying out abnormity detection on key performance indicator KPI flow of the medical equipment according to the acquired marking data.
Optionally, in an embodiment of the present application, the first classifier is a linear model classifier, and the second classifier is a random forest model classifier.
Optionally, in an embodiment of the present application, determining candidate abnormal data according to the prediction scores includes determining the prediction score of each unlabeled data, and sorting the prediction scores of all the unlabeled data according to the magnitude of the score value; and selecting the unmarked data with the highest score as the candidate abnormal data.
Optionally, in an embodiment of the present application, after the marking the candidate abnormal data as abnormal annotation data, the method further includes: and acquiring unmarked data with the lowest prediction score, and automatically marking the unmarked data with the lowest prediction score as normal marked data.
Optionally, in an embodiment of the present application, performing anomaly detection on a key performance indicator KPI flow of a medical device according to acquired marking data includes: and training a supervised anomaly detection model according to the acquired marking data, and carrying out anomaly detection on the KPI to be detected through the supervised anomaly detection model.
In order to achieve the above object, a second aspect of the present application provides an apparatus for detecting a performance index of a medical device based on active learning, including the following modules:
the system comprises a first marking module, a second marking module and a third marking module, wherein the first marking module is used for acquiring a training data set, training a first classifier through abnormal marking data and unmarked data in the training data set, and marking normal marking data from the unmarked data through the first classifier;
the training module is used for training a second classifier through the abnormal marking data and the normal marking data;
the prediction module is used for predicting each unmarked data through the second classifier and determining candidate abnormal data according to the prediction score;
the second marking module is used for manually judging whether the candidate abnormal data are abnormal data or not, marking the candidate abnormal data as abnormal marking data if the candidate abnormal data are abnormal data, and updating the abnormal marking data, the normal marking data and the unmarked data;
the iteration module is used for judging whether the quantity of the abnormal marking data and the normal marking data reaches a preset quantity, and if not, controlling the prediction module and the second marking module to repeatedly run until the quantity of the abnormal marking data and the normal marking data reaches the preset quantity;
and the detection module is used for carrying out abnormity detection on the key performance indicator KPI flow of the medical equipment according to the acquired marking data.
Optionally, in an embodiment of the present application, the first classifier is a linear model classifier, and the second classifier is a random forest model classifier.
Optionally, in an embodiment of the present application, the prediction module is specifically configured to: determining the prediction score of each unmarked data, and sequencing the prediction scores of all the unmarked data according to the score value; and selecting the unmarked data with the highest score as the candidate abnormal data.
Optionally, in an embodiment of the present application, the second marking module is further configured to: and acquiring unmarked data with the lowest prediction score, and automatically marking the unmarked data with the lowest prediction score as normal marked data.
Optionally, in an embodiment of the present application, the detection module is specifically configured to: and training a supervised anomaly detection model according to the acquired marking data, and carrying out anomaly detection on the KPI to be detected through the supervised anomaly detection model.
The application has the following technical effects: the method comprises the steps of obtaining a training data set, training a first classifier through abnormal marking data and unmarked data in the training data set, and marking normal marking data from the unmarked data through the first classifier; training a second classifier through the abnormal marking data and the normal marking data; predicting each unlabeled data through a second classifier, and determining candidate abnormal data according to the prediction scores; manually judging whether the candidate abnormal data is abnormal data, if the candidate abnormal data is abnormal data, marking the candidate abnormal data as abnormal marking data, and updating the abnormal marking data, the normal marking data and the unmarked data; judging whether the quantity of the abnormal marking data and the normal marking data reaches a preset quantity, if not, repeating the step of iterative marking until the quantity of the abnormal marking data and the normal marking data reaches the preset quantity; and carrying out abnormity detection on key performance indicator KPI flow of the medical equipment according to the acquired marking data. The method optimizes the labeling process of active learning aiming at the characteristics of the abnormal samples and the normal samples of the time sequence, selects the samples which are most probably abnormal in the iteration of data labeling every time and labels the samples as the abnormal labeling data instead of the samples close to the classification boundary, thereby ensuring that the normal samples as few as possible are labeled as the abnormality in the active learning stage, greatly reducing the probability of misinformation, obtaining more labeled samples, perfecting the application of abnormality detection in the field of PU learning medicine, and improving the accuracy of abnormality detection on the performance indexes of the medical equipment.
To achieve the above object, a non-transitory computer-readable storage medium is provided in an embodiment of a third aspect of the present application, and a computer program is stored on the non-transitory computer-readable storage medium, and when executed by a processor, the computer program implements the method for detecting a performance index of a medical device based on active learning according to the embodiment of the first aspect of the present application.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The foregoing and/or additional aspects and advantages of the present application will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a flowchart of a method for detecting a performance index of a medical device based on active learning according to an embodiment of the present application;
FIG. 2 is a flowchart illustrating a method for detecting performance indicators of a medical device based on active learning according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of a medical device performance index detection apparatus based on active learning according to an embodiment of the present application.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
It should be noted that, for a medical device, the number of KPI flows generated during the operation is large and the pattern is diverse, and some algorithms are needed to assist the abnormality detection of the KPI flows of the medical device. One of the methods for detecting an anomaly by using PU learning is generally accomplished by using existing data points with and without an anomaly label and a two-step strategy. The method comprises the following steps that firstly, an existing abnormal sample and an existing unlabeled sample are used for training a classifier, so that a sample which is judged to be normal in the unlabeled sample is labeled; and the second step is to train a second classifier on the current marked data set, continue to mark the unmarked data and repeat the process repeatedly until enough marked samples can finish the final training of the anomaly detection model.
However, in practical applications in the medical field, the main problem of applying the existing PU learning to perform anomaly detection on the time sequence is that the label is insufficient, and the performance of the PU learning method is limited. Active learning can be applied to increase the number of tags. However, the active learning method in the related art generally marks abnormal samples near the classification boundary, which easily causes the normal samples to be misclassified as abnormal, thereby generating many false positives, and if one normal sample is misclassified as an abnormal sample, more and more normal samples will be marked as abnormal samples because the similarity between the normal samples is higher than the similarity between the abnormal sample and the normal sample, thereby causing the accuracy of the abnormality detection to decrease.
Based on this, the present application proposes a detection method using a simple and efficient active learning model, which selects the samples that are most likely to be abnormal in each iteration, rather than those close to the classification boundary, and ensures that as few normal samples as possible are labeled as abnormal in the active learning phase, reducing false positives.
The method and apparatus for detecting performance index of medical device based on active learning according to the embodiments of the present application are described below with reference to the accompanying drawings.
Fig. 1 is a flowchart of a method for detecting a performance index of a medical device based on active learning according to an embodiment of the present application, as shown in fig. 1, the method includes the following steps:
s1: the method comprises the steps of obtaining a training data set, training a first classifier through abnormal marked data and unmarked data in the training data set, and marking normal marked data from the unmarked data through the first classifier.
The training data set is a data set for training an anomaly detection model for detecting a Key Performance Indicator flow of the medical device, the Key Performance Indicator (KPI) flow is a data flow of the Key Indicator reflecting the operation Performance of the medical device, which is obtained when the medical device is monitored, for example, the KPI flow may include a response delay of the medical device or a data flow of a network throughput, and the KPI flow is generally a time sequence. In the embodiment of the application, the pre-stored historical KPI flow can be acquired in different ways as a training set of the medical equipment performance index detection model of the application.
In the initial stage of training, the obtained training set includes a small amount of abnormal labeled data and a large amount of unlabeled data, that is, the training set is composed of several samples with abnormal labels and a large amount of unlabeled samples. The sample of the abnormal annotation may be obtained by manually marking some abnormal data on the KPI flow by a worker in advance, and the like, which is not limited herein.
In the embodiment of the application, the first classifier is trained through the abnormal labeled data and the unlabeled data in the training data set, and some normal labeled data are marked from the unlabeled data through the first classifier, as an example, in order to mark more reliable normal labeled data in the pre-training stage more accurately, a linear model classifier can be used as the first classifier, that is, the linear model classifier is trained through the abnormal labeled data and the unlabeled data in the training data set, and then the normal labeled data are marked from the unlabeled data through the linear model classifier.
S2: and training a second classifier through the abnormal labeling data and the normal labeling data.
S3: and predicting each unlabeled data through a second classifier, and determining candidate abnormal data according to the prediction scores.
In an embodiment of the present application, the second classifier may be a random forest model classifier, that is, the present application trains a random forest model classifier according to the abnormal annotation data and the normal annotation data obtained in step S1, predicts each unlabeled data in the training set through the random forest model classifier, and outputs a prediction score of each unlabeled data, thereby determining candidate abnormal data that may be abnormal data.
In specific implementation, as a possible implementation manner, the prediction score of each unlabeled data can be determined according to the output of the random forest model classifier, the prediction scores of all the unlabeled data are sorted according to the score value, and then the unlabeled data with the highest score is selected as the candidate abnormal data.
Of course, in some other embodiments of the present application, other ways of determining candidate abnormal data may also be determined according to actual needs, for example, unlabeled data with a prediction score of the first three may be used as candidate abnormal data.
S4: and manually judging whether the candidate abnormal data is abnormal data, if so, marking the candidate abnormal data as abnormal marking data, and updating the abnormal marking data, the normal marking data and the unmarked data.
Specifically, after the candidate abnormal data is determined, the worker can manually judge whether the candidate abnormal data is abnormal data, that is, the worker can manually judge whether the candidate abnormal data is abnormal data by synthesizing various factors, and if the candidate abnormal data is abnormal data, the candidate abnormal data is manually marked as abnormal marking data, so that the sample marked as abnormal is ensured to be a real abnormal sample, and the normal sample is prevented from being mistakenly marked as an abnormal sample.
In an embodiment of the present application, after the candidate abnormal data is marked as the abnormal labeled data, the unlabeled data with the lowest score may be determined according to the obtained prediction score of each unlabeled data, and the unlabeled data with the lowest prediction score is automatically marked as the normal labeled data. Therefore, abnormal marking data and normal marking data with higher authenticity and reliability are marked in the self-training process of active learning. It should be noted that, the step of automatically labeling the normal labeling data may be performed after, before, or simultaneously with the step of labeling the candidate abnormal data as the abnormal labeling data, and is not limited herein.
Furthermore, after the data is marked, the abnormal marking data, the normal marking data and the unmarked data are updated, that is, the currently marked abnormal marking data, the normal marking data and the unmarked data which are not marked yet are determined again, and the data which is not marked in the previous round is transferred to the corresponding abnormal marking data set or the normal marking data set. In an embodiment of the present application, the method further includes updating respective numbers of the marked abnormal annotation data, the marked normal annotation data, and the marked unmarked data.
S5: and judging whether the quantity of the abnormal annotation data and the normal annotation data reaches the preset quantity, if not, repeatedly executing S3 and S4 until the quantity of the abnormal annotation data and the normal annotation data reaches the preset quantity.
The preset number is the number of the marking samples which are determined to be required for the on-line detection according to the current detection mode of the performance indexes of the medical equipment.
In the embodiment of the application, in each round of marking process, after the abnormal marking data and the normal marking data are marked in the manner of the above embodiment, the updated quantities of the abnormal marking data and the normal marking data are compared with the preset quantity, whether the quantities of the abnormal marking data and the normal marking data reach the preset quantity is judged, if the updated quantities of the abnormal marking data and the normal marking data are less than the preset quantity, S3 and S4 are repeatedly executed, that is, unmarked samples are iteratively marked, so that more marked samples are obtained through iterative marking, until the quantities of the abnormal marking data and the normal marking data reach the preset quantity, and enough unmarked samples are marked to perform subsequent abnormal detection.
S6: and carrying out abnormity detection on key performance indicator KPI flow of the medical equipment according to the acquired marking data.
Specifically, after enough marking data are acquired, the KPI flow of the medical device to be detected is subjected to anomaly detection according to the acquired marking data, and a specific detection process can be set according to actual needs, for example, a supervised anomaly detection model is trained according to the acquired marking data, the features and the labels of the KPI flow are used as inputs, and a machine learning algorithm is used for completing time series anomaly detection, that is, the KPI flow to be detected is subjected to anomaly detection through the supervised anomaly detection model.
Therefore, the medical equipment performance index detection method based on active learning ensures that as few as possible normal samples are marked as abnormal in the active learning stage, greatly reduces the probability of misinformation, and obtains more marked samples through iterative marking.
In summary, in the method for detecting performance indexes of medical equipment based on active learning according to the embodiment of the present application, a training data set is obtained, a first classifier is trained through abnormal labeled data and unlabeled data in the training data set, and normal labeled data is labeled from the unlabeled data through the first classifier; training a second classifier through the abnormal marking data and the normal marking data; predicting each unlabeled data through a second classifier, and determining candidate abnormal data according to the prediction scores; manually judging whether the candidate abnormal data is abnormal data, if the candidate abnormal data is abnormal data, marking the candidate abnormal data as abnormal marking data, and updating the abnormal marking data, the normal marking data and the unmarked data; judging whether the quantity of the abnormal marking data and the normal marking data reaches a preset quantity, if not, repeating the step of iterative marking until the quantity of the abnormal marking data and the normal marking data reaches the preset quantity; and carrying out abnormity detection on key performance indicator KPI flow of the medical equipment according to the acquired marking data. The method optimizes the labeling process of active learning aiming at the characteristics of abnormal samples and normal samples of a time sequence, and selects the samples which are most likely to be abnormal in each iteration of data labeling to be labeled as abnormal labeling data instead of the samples close to classification boundaries, so that the normal samples are ensured to be labeled as abnormal as little as possible in the active learning stage, the probability of false alarm is greatly reduced, more labeled samples can be obtained, the application of abnormal detection in the PU learning medical field is perfected, and the accuracy of abnormal detection on the performance indexes of the medical equipment is improved.
In order to more clearly illustrate the method for detecting a performance index of a medical device based on active learning of the present application, a specific example is described below with reference to fig. 2.
As shown in fig. 2, the method for detecting performance index of medical device based on active learning includes a pre-training process and a self-training process based on active learning, and in specific implementation, the pre-training process may adopt a first step of conventional PU learning, and a linear classifier is used to mark a sample that is most likely to be normal. In the initial stage, the training set consists of several samples with abnormal labels (i.e. Ω (P) in fig. 2) and a large number of unlabeled samples (i.e. Ω (U) in fig. 2). To more carefully find reliable normal samples from Ω (U), a linear model may be employed as a classifier.
Further, the self-training process based on active learning is to iteratively label more unlabeled samples. First, a random forest model classifier is trained on Ω (P) and the reliable normal samples obtained in the previous step (i.e., Ω (N) in fig. 2) to obtain the prediction score of Ω (U) in each iteration. Then, some unlabeled samples are labeled iteratively, Ω (U) is ranked according to the prediction scores, and the unlabeled sample with the highest score is set as a candidate abnormal sample. Next, the staff manually marks these candidate samples to ensure that the samples marked as abnormal are true abnormal samples, i.e., no normal samples are mis-marked as abnormal. For the unlabeled sample with the lowest score, it can be directly labeled as a normal sample. Thereafter, Ω (P), Ω (N), and Ω (U) are updated, and then the next iteration begins until enough unlabeled samples are labeled.
Further, when there are enough labeled samples, a subsequent anomaly detection process can be performed, such as training a supervised model for detecting anomalies.
In order to achieve the above object, as shown in fig. 3, a second aspect of the present application provides an active learning-based medical device performance index detection apparatus, including: a first labeling module 100, a training module 200, and a prediction module 300, a second labeling module 400, an iteration module 500, and a detection module 600.
The first labeling module 100 is configured to obtain a training data set, train a first classifier through abnormal labeling data and unlabeled data in the training data set, and label normal labeling data from the unlabeled data through the first classifier.
And a training module 200, configured to train the second classifier according to the abnormal labeling data and the normal labeling data.
The prediction module 300 is configured to predict each unlabeled data through the second classifier, and determine candidate abnormal data according to the prediction score;
the second marking module 400 is configured to manually determine whether the candidate abnormal data is abnormal data, mark the candidate abnormal data as abnormal annotation data if the candidate abnormal data is abnormal data, and update the abnormal annotation data, the normal annotation data, and the unlabeled data.
And the iteration module 500 is configured to determine whether the number of the abnormal annotation data and the normal annotation data reaches a preset number, and if not, control the prediction module and the second marking module to repeatedly run until the number of the abnormal annotation data and the normal annotation data reaches the preset number.
The detection module 600 is configured to perform anomaly detection on a key performance indicator KPI flow of the medical device according to the acquired label data.
Optionally, in an embodiment of the present application, the first classifier is a linear model classifier, and the second classifier is a random forest model classifier.
Optionally, in an embodiment of the present application, the prediction module is specifically configured to determine a prediction score of each unlabeled data, and sort the prediction scores of all the unlabeled data according to a size of the score value; and selecting the unmarked data with the highest score as candidate abnormal data.
Optionally, in an embodiment of the present application, the second labeling module is further configured to obtain unlabeled data with the lowest prediction score, and automatically label the unlabeled data with the lowest prediction score as normal labeled data.
Optionally, in an embodiment of the present application, the detection module is specifically configured to train a supervised anomaly detection model according to the obtained tag data, and perform anomaly detection on the KPI flow to be detected through the supervised anomaly detection model.
It should be noted that the foregoing description of the embodiment of the method for detecting a performance index of a medical device based on active learning is also applicable to the embodiment of the apparatus, and the implementation principle is the same, and is not repeated here.
To sum up, the medical equipment performance index detection device based on active learning of the embodiment of the application optimizes the labeling process of active learning aiming at the characteristics of abnormal samples and normal samples of a time sequence, selects the most possible abnormal samples to be labeled as abnormal labeling data in the iteration of data labeling every time, but not the samples close to classification boundaries, thereby ensuring that the normal samples as few as possible are labeled as abnormal in the active learning stage, greatly reducing the probability of misinformation, obtaining more labeled samples, perfecting the application of abnormal detection in the PU learning medical field, and improving the accuracy of abnormal detection on the performance index of the medical equipment.
In order to achieve the above embodiments, the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements an active learning-based medical device performance index detection method according to an embodiment of the first aspect of the present application.
Although the present application has been disclosed in detail with reference to the accompanying drawings, it is to be understood that such description is merely illustrative and not restrictive of the application of the present application. The scope of the present application is defined by the appended claims and may include various modifications, adaptations, and equivalents of the invention without departing from the scope and spirit of the application.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing steps of a custom logic function or process, and alternate implementations are included within the scope of the preferred embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present invention.
The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
It will be understood by those skilled in the art that all or part of the steps carried by the method for implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and when the program is executed, the program includes one or a combination of the steps of the method embodiments.
In addition, functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium.
The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc. Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (10)
1. A medical equipment performance index detection method based on active learning is characterized by comprising the following steps:
s1: acquiring a training data set, training a first classifier through abnormal labeled data and unlabeled data in the training data set, and marking normal labeled data from the unlabeled data through the first classifier;
s2: training a second classifier through the abnormal marking data and the normal marking data;
s3: predicting each unlabeled data through the second classifier, and determining candidate abnormal data according to the prediction scores;
s4: manually judging whether the candidate abnormal data is abnormal data, if the candidate abnormal data is abnormal data, marking the candidate abnormal data as abnormal marking data, and updating the abnormal marking data, the normal marking data and the unmarked data;
s5: judging whether the quantity of the abnormal annotation data and the normal annotation data reaches a preset quantity, if not, repeatedly executing S3 and S4 until the quantity of the abnormal annotation data and the normal annotation data reaches the preset quantity;
s6: and carrying out abnormity detection on key performance indicator KPI flow of the medical equipment according to the acquired marking data.
2. A method as claimed in claim 1, wherein the first classifier is a linear model classifier and the second classifier is a random forest model classifier.
3. The method of claim 1 or 2, wherein determining candidate anomaly data based on the prediction scores comprises:
determining the prediction score of each unmarked data, and sequencing the prediction scores of all the unmarked data according to the score value;
and selecting the unmarked data with the highest score as the candidate abnormal data.
4. The method of claim 3, wherein after said marking said candidate anomaly data as anomaly marked data, further comprising:
and acquiring unmarked data with the lowest prediction score, and automatically marking the unmarked data with the lowest prediction score as normal marked data.
5. The method according to claim 1, wherein said anomaly detection of a key performance indicator, KPI, flow of a medical device from acquired marker data comprises:
and training a supervised anomaly detection model according to the acquired marking data, and carrying out anomaly detection on the KPI to be detected through the supervised anomaly detection model.
6. A medical equipment performance index detection device based on active learning is characterized by comprising:
the system comprises a first marking module, a second marking module and a third marking module, wherein the first marking module is used for acquiring a training data set, training a first classifier through abnormal marking data and unmarked data in the training data set, and marking normal marking data from the unmarked data through the first classifier;
the training module is used for training a second classifier through the abnormal marking data and the normal marking data;
the prediction module is used for predicting each unmarked data through the second classifier and determining candidate abnormal data according to the prediction score;
the second marking module is used for manually judging whether the candidate abnormal data are abnormal data or not, marking the candidate abnormal data as abnormal marking data if the candidate abnormal data are abnormal data, and updating the abnormal marking data, the normal marking data and the unmarked data;
the iteration module is used for judging whether the quantity of the abnormal marking data and the normal marking data reaches a preset quantity, and if not, controlling the prediction module and the second marking module to repeatedly run until the quantity of the abnormal marking data and the normal marking data reaches the preset quantity;
and the detection module is used for carrying out abnormity detection on the key performance indicator KPI flow of the medical equipment according to the acquired marking data.
7. The apparatus of claim 6, wherein the first classifier is a linear model classifier and the second classifier is a random forest model classifier.
8. The apparatus according to claim 6 or 7, wherein the prediction module is specifically configured to:
determining the prediction score of each unmarked data, and sequencing the prediction scores of all the unmarked data according to the score value;
and selecting the unmarked data with the highest score as the candidate abnormal data.
9. The apparatus of claim 8, wherein the second marking module is further configured to:
and acquiring unmarked data with the lowest prediction score, and automatically marking the unmarked data with the lowest prediction score as normal marked data.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the active learning based medical device performance indicator detection method of any of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111138481.9A CN114118201A (en) | 2021-09-27 | 2021-09-27 | Medical equipment performance index detection method and device based on active learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111138481.9A CN114118201A (en) | 2021-09-27 | 2021-09-27 | Medical equipment performance index detection method and device based on active learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114118201A true CN114118201A (en) | 2022-03-01 |
Family
ID=80441453
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111138481.9A Pending CN114118201A (en) | 2021-09-27 | 2021-09-27 | Medical equipment performance index detection method and device based on active learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114118201A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109299162A (en) * | 2018-11-08 | 2019-02-01 | 南京航空航天大学 | A kind of Active Learning Method classified for positive class and data untagged |
| CN109492026A (en) * | 2018-11-02 | 2019-03-19 | 国家计算机网络与信息安全管理中心 | A kind of Telecoms Fraud classification and Detection method based on improved active learning techniques |
| CN111277459A (en) * | 2020-01-16 | 2020-06-12 | 新华三信息安全技术有限公司 | Equipment anomaly detection method and device and machine-readable storage medium |
| CN112905412A (en) * | 2021-01-29 | 2021-06-04 | 清华大学 | Method and device for detecting abnormity of key performance index data |
| CN113114529A (en) * | 2021-03-25 | 2021-07-13 | 清华大学 | KPI (Key Performance indicator) abnormity detection method and device based on condition variation automatic encoder |
-
2021
- 2021-09-27 CN CN202111138481.9A patent/CN114118201A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109492026A (en) * | 2018-11-02 | 2019-03-19 | 国家计算机网络与信息安全管理中心 | A kind of Telecoms Fraud classification and Detection method based on improved active learning techniques |
| CN109299162A (en) * | 2018-11-08 | 2019-02-01 | 南京航空航天大学 | A kind of Active Learning Method classified for positive class and data untagged |
| CN111277459A (en) * | 2020-01-16 | 2020-06-12 | 新华三信息安全技术有限公司 | Equipment anomaly detection method and device and machine-readable storage medium |
| CN112905412A (en) * | 2021-01-29 | 2021-06-04 | 清华大学 | Method and device for detecting abnormity of key performance index data |
| CN113114529A (en) * | 2021-03-25 | 2021-07-13 | 清华大学 | KPI (Key Performance indicator) abnormity detection method and device based on condition variation automatic encoder |
Non-Patent Citations (1)
| Title |
|---|
| JIAQI ZHANG ET AL.: "Positive and Unlabeled Learning for Anomaly Detection with Multi-features", 《PROCEEDINGS OF THE 2017 ACM MULTIMEDIA CONFERENCE(MM\'17)》, 19 October 2017 (2017-10-19), pages 1 - 7 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114065613B (en) | Multi-working-condition process industrial fault detection and diagnosis method based on deep migration learning | |
| CN110852983B (en) | Method for detecting defect in semiconductor device | |
| EP2905665B1 (en) | Information processing apparatus, diagnosis method, and program | |
| JP6815480B2 (en) | Methods and systems for discovering precursor substrings in time series | |
| US10989908B2 (en) | Automated workflows based on an identification of calibration samples | |
| CN111666993A (en) | Medical image sample screening method and device, computer equipment and storage medium | |
| CN112528975A (en) | Industrial quality inspection method, device and computer readable storage medium | |
| CN112905371B (en) | Software change checking method and device based on heterogeneous multi-source data anomaly detection | |
| Zhang et al. | Few-shot bearing anomaly detection via model-agnostic meta-learning | |
| CN108268689A (en) | Method for diagnosing and evaluating state of heating element and application thereof | |
| CN112613617A (en) | Uncertainty estimation method and device based on regression model | |
| CN113051452B (en) | Operation and maintenance data feature selection method and device | |
| CN117193088B (en) | Industrial equipment monitoring method and device and server | |
| CN112363465B (en) | Expert rule set training method, trainer and industrial equipment early warning system | |
| CN114118201A (en) | Medical equipment performance index detection method and device based on active learning | |
| US7814034B2 (en) | Method and system for automatically developing a fault classification system by segregation of kernels in time series data | |
| CN114202009A (en) | Medical equipment performance index abnormity detection method and device based on PU learning | |
| CN114510994A (en) | Training method and device for urban underground gas leakage recognition model | |
| CN114357185A (en) | Manufacturing equipment fault monitoring model training method based on knowledge graph | |
| CN113076993A (en) | Information processing method and model training method for chest X-ray film recognition | |
| TWI771098B (en) | Fault diagnosis system and method for state of radar system of roadside units | |
| CN114626545A (en) | Data preparation for artificial intelligence models | |
| Liu et al. | Hybrid KANN-DBSCAN with KNN for Anomaly Detection of Power Plug-in Board | |
| Bonilla-Rivas et al. | Adapting YOLO to recognition of real numbers in 7-segment digits | |
| CN116776961A (en) | Equipment fault diagnosis method and device, storage medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |