CN114202009A

CN114202009A - Medical equipment performance index abnormity detection method and device based on PU learning

Info

Publication number: CN114202009A
Application number: CN202111136341.8A
Authority: CN
Inventors: 李姗姗; 孙新国; 赵晨宇; 张圣林
Original assignee: Nankai University
Current assignee: Nankai University
Priority date: 2021-09-27
Filing date: 2021-09-27
Publication date: 2022-03-18

Abstract

The application provides a medical equipment performance index abnormity detection method and device based on PU learning, and the method comprises the following steps: clustering the training data according to the similarity degree by taking the KPI (Key Performance indicator) flow as the training data to obtain a centroid curve of each cluster, and labeling the centroid curve of each cluster to obtain first abnormal labeling data and first non-labeling data; constructing a binary classifier through positive-case unlabeled PU learning based on the first abnormal labeling data and the first unlabeled data, and acquiring an abnormal label and a normal label of a centroid curve of each cluster by combining active learning; the method comprises the steps of obtaining a label on a centroid curve of a cluster corresponding to the KPI to be detected, training an abnormality detection model corresponding to the KPI to be detected through semi-supervised learning, and detecting the KPI to be detected through the abnormality detection model. The method improves the accuracy of the abnormal detection of the performance index of the medical equipment while reducing the marking workload to the maximum extent.

Description

Medical equipment performance index abnormity detection method and device based on PU learning

Technical Field

The application relates to the technical field of data detection, in particular to a medical equipment performance index abnormity detection method and device based on PU learning.

Background

At present, in order to ensure the reliability of medical equipment, professionals in the medical field need to continuously collect and monitor a large number of Key Performance Indicator (KPI) data streams of the medical equipment, so that an abnormality occurs in the medical equipment, and the equipment can be repaired in time, therefore, KPI abnormality detection is important for medical equipment management.

In the related art, in the medical field, KPI flow anomaly detection technologies generally include a supervised algorithm, a semi-supervised algorithm, and an unsupervised algorithm. Wherein, the supervision algorithm needs to label all samples of the training set manually, and the time sequence abnormity detection is completed by using the machine learning algorithm with the characteristics and labels of the KPI flow as input; the unsupervised algorithm does not need to be labeled, and the KPI flow is processed and then input into a deep learning model, so that an abnormal detection result can be obtained; the semi-supervised algorithm integrates the advantages of the two algorithms, namely only part of labeled data is needed, and an anomaly detection algorithm is designed by using labeled data points and non-labeled data points.

However, the applicant finds that, in the above algorithms, a supervised algorithm needs a large amount of labeled data, but in practical situations, KPIs are large-scale and diverse, and the labeling work needs a large amount of time and effort cost and is often difficult to implement; the unsupervised algorithm has low precision, needs a large amount of training data, and in an actual scene, the mode of the KPI flow is dynamically changed frequently, so the actual applicability of the algorithm is poor; although the semi-supervised algorithm reduces the labeling cost compared with the supervised algorithm and improves the detection accuracy of the model compared with the unsupervised algorithm, the semi-supervised algorithm still needs the staff to accurately label all the abnormalities in a certain time period in a plurality of KPI flows, and repeatedly confirms whether the abnormalities exist in a section of KPI and still brings a large workload. Therefore, the detection method in the related art, no matter the supervised learning method, the semi-supervised learning method or the unsupervised learning method, cannot realize accurate abnormality detection of large-scale, diversified and dynamically-changed KPI flows with little marking work.

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, the first objective of the present application is to provide a medical device performance index abnormality detection method based on PU learning, which is based on the PU learning method, and realizes that accurate KPI abnormality detection can be performed only with a small number of tags, and improves the accuracy of abnormality detection while reducing labeling workload to the maximum extent through integrated clustering, PU learning, and semi-supervised learning.

The second purpose of the invention is to provide a medical equipment performance index abnormality detection device based on PU learning.

A third object of the invention is to propose a non-transitory computer-readable storage medium.

In order to achieve the above object, an embodiment of a first aspect of the present invention provides a method for detecting an abnormality of a performance index of a medical device based on PU learning, including the following steps:

clustering training data according to similarity by taking a historical key performance index KPI flow as training data to obtain a centroid curve of each cluster, and labeling the centroid curve of each cluster to obtain first abnormal labeling data and first non-labeling data;

constructing a binary classifier through positive-case unlabeled PU learning based on the first abnormal labeling data and the first unlabeled data, and acquiring an abnormal label and a normal label of the centroid curve of each cluster by combining active learning;

the method comprises the steps of obtaining a label on a centroid curve of a cluster corresponding to the KPI to be detected, training an abnormality detection model corresponding to the KPI to be detected through semi-supervised learning, and detecting the KPI to be detected through the abnormality detection model.

Optionally, in an embodiment of the present application, labeling the centroid curve of each cluster, and acquiring first abnormal labeling data and first non-labeling data, includes: and manually marking an abnormal section on the centroid curve of each cluster, taking the marking data in the abnormal section as the first abnormal marking data, and taking the data on the centroid curve except the abnormal section as the first unmarked data.

Optionally, in an embodiment of the present application, the obtaining the abnormal label and the normal label of the centroid curve of each cluster includes determining first normal labeling data in the first label-free data through the binary classifier; performing iterative marking on the first unmarked data for multiple times by an active learning method, and determining second normal marked data and second abnormal marked data in the first unmarked data; and taking the first abnormal marking data and the second abnormal marking data as the abnormal labels, and taking the first normal marking data and the second normal marking data as the normal labels.

Optionally, in an embodiment of the present application, acquiring a label on a centroid curve of a cluster corresponding to a KPI flow to be detected includes: determining a cluster corresponding to the KPI to be detected by calculating the shape similarity of the KPI to be detected and the centroid curve of each cluster; and reading the abnormal label and the normal label on the centroid curve of the cluster corresponding to the KPI flow to be detected.

Optionally, in an embodiment of the present application, training, by semi-supervised learning, an anomaly detection model corresponding to the KPI flow to be detected includes: selecting a preset amount of data from the KPI stream to be detected as second unmarked data; and training an abnormal detection model corresponding to the KPI to be detected based on the abnormal label, the normal label and the second unmarked data on the centroid curve of the cluster corresponding to the KPI to be detected.

In order to achieve the above object, a second aspect of the present application provides a medical device performance index abnormality detection apparatus based on PU learning, including the following modules:

the clustering module is used for clustering the training data according to the similarity degree by taking the KPI (Key Performance indicator) flow as the training data to obtain the centroid curve of each cluster, labeling the centroid curve of each cluster and obtaining first abnormal labeling data and first non-labeling data;

an obtaining module, configured to construct a binary classifier through regular-case unlabeled PU learning based on the first abnormal labeling data and the first unlabeled data, and obtain an abnormal label and a normal label of the centroid curve of each cluster by combining active learning;

the training module is used for acquiring a label on a centroid curve of a cluster corresponding to the KPI to be detected, training an abnormality detection model corresponding to the KPI to be detected through semi-supervised learning, and detecting the KPI to be detected through the abnormality detection model.

Optionally, in an embodiment of the present application, the clustering module is specifically configured to: and marking an abnormal section on the centroid curve of each cluster based on manual work, taking the marking data in the abnormal section as the first abnormal marking data, and taking the data on the centroid curve except the abnormal section as the first unmarked data.

Optionally, in an embodiment of the present application, the obtaining module is specifically configured to: determining first normal labeled data in the unlabeled data through the binary classifier; performing iterative marking on the first unmarked data for multiple times by an active learning method, and determining second normal marked data and second abnormal marked data in the first unmarked data; and taking the first abnormal marking data and the second abnormal marking data as the abnormal labels, and taking the first normal marking data and the second normal marking data as the normal labels.

Optionally, in an embodiment of the present application, the training module is specifically configured to: determining a cluster corresponding to the KPI to be detected by calculating the shape similarity of the KPI to be detected and the centroid curve of each cluster; and reading the abnormal label and the normal label on the centroid curve of the cluster corresponding to the KPI flow to be detected.

Optionally, in an embodiment of the present application, the training module is further configured to: selecting a preset amount of data from the KPI stream to be detected as second unmarked data; and training an abnormal detection model corresponding to the KPI to be detected based on the abnormal label, the normal label and the second unmarked data on the centroid curve of the cluster corresponding to the KPI to be detected.

The application has the following technical effects: the method comprises the steps that a history key performance index KPI flow is used as training data, the training data are clustered according to the similarity degree, a centroid curve of each cluster is obtained, the centroid curve of each cluster is labeled, and first abnormal labeling data and first non-labeling data are obtained; constructing a binary classifier through positive-case unlabeled PU learning based on the first abnormal labeling data and the first unlabeled data, and acquiring an abnormal label and a normal label of the centroid curve of each cluster by combining active learning; the method comprises the steps of obtaining a label on a centroid curve of a cluster corresponding to the KPI to be detected, training an abnormality detection model corresponding to the KPI to be detected through semi-supervised learning, and detecting the KPI to be detected through the abnormality detection model. According to the method based on PU learning, the KPI abnormity detection can be accurately carried out only by a small number of labels, and the abnormity detection accuracy is improved while the marking workload is reduced to the maximum extent through integration of clustering, PU learning and semi-supervised learning.

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 thereon, and when being executed by a processor, the computer program implements the method for detecting abnormality of performance index of medical device based on PU 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 abnormality of performance index of medical equipment based on PU learning according to an embodiment of the present application;

fig. 2 is a schematic flowchart of a specific method for detecting abnormality of performance index of medical device based on PU learning according to an embodiment of the present application;

fig. 3 is a schematic structural diagram of a medical device performance index abnormality detection apparatus based on PU 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. On one hand, if a PU learning model is trained for each KPI flow, the workload of whole labeling is very large, and on the other hand, if a PU learning model is trained for all KPI flows, because different KPI flows have different modes, the most suitable anomaly detectors and parameters of different KPI flows may be significantly different, so that the model will suffer from low precision.

The following describes a medical device performance index abnormality detection method and apparatus based on PU learning according to an embodiment of the present application with reference to the drawings.

Fig. 1 is a flowchart of a method for detecting an abnormality of a performance index of a medical device based on PU learning according to an embodiment of the present application, as shown in fig. 1, the method includes the following steps:

step S101: and clustering the training data according to the similarity degree by taking the KPI (Key Performance indicator) flow as the training data to obtain the centroid curve of each cluster, labeling the centroid curve of each cluster, and obtaining first abnormal labeling data and first non-labeling data.

The Key Performance Indicator (KPI) flow is a data flow that reflects a Key Indicator of the operation Performance of the medical device and 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 usually a time series.

Specifically, in the present application, through different manners, a pre-stored historical KPI flow may be obtained as a training set of the KPI flow abnormality detection model of the present application. Then, clustering the training data according to the similarity degree, dividing the training data into different clusters through clustering, and determining the mass center of each cluster.

It should be noted that, since the present application is directed to clustering KPI streams of a time sequence, that is, KPI streams are represented in a line form, in an embodiment of the present application, a method of clustering discrete curves may be adopted, training data are clustered according to similarity of the curves, and then a centroid curve of each cluster after clustering is obtained.

Further, labeling the centroid curve of each cluster to obtain abnormal data and label-free data in each centroid curve. In an embodiment of the application, a worker can manually mark some abnormal sections on the centroid curve of each cluster in a manual marking mode, the number of the marked abnormal sections can be set according to actual needs, marking data in the abnormal sections are used as first abnormal marking data, and data except the marked abnormal sections on each centroid curve are used as first unmarked data.

It should be further noted that, after clustering, each cluster corresponds to one centroid curve, and the number of clusters is much smaller than that of the time series curves in the training set, so that the workload of manual labeling by workers in the application can be minimized.

Step S102: based on the first abnormal labeling data and the first unmarked data, a binary classifier is constructed through normal-case unmarked PU learning, and an abnormal label and a normal label of a centroid curve of each cluster are obtained through active learning.

The binary classifier is used for marking abnormal data and normal data in the data without labels, and the abnormal labels and the normal labels are abnormal labeled samples and normal labeled samples with higher reliability.

In an embodiment of the application, first abnormal labeling data and first non-labeling data are obtained based on manual labeling, a binary classifier is constructed through normal non-labeling PU learning, and first normal labeling data are determined in the first non-labeling data through the binary classifier. And then, continuously marking normal data and abnormal data in the unmarked data in an active learning mode, repeatedly executing the marking step based on the result of the previous marking, and carrying out iterative marking on the unmarked data for multiple times, so that samples which are possibly abnormal are selected from the unmarked data in the active learning mode for marking, and second normal marked data and second abnormal marked data which are more in quantity and higher in reliability are determined. And finally, taking the first abnormal labeling data and the second abnormal labeling data corresponding to each centroid curve as the abnormal label of the centroid curve, and taking the first normal labeling data and the second normal labeling data of each centroid curve as the normal label of the centroid curve.

Thus, the abnormal label and the normal label of the centroid curve of each cluster are obtained.

Step S103: the method comprises the steps of obtaining a label on a centroid curve of a cluster corresponding to the KPI to be detected, training an abnormality detection model corresponding to the KPI to be detected through semi-supervised learning, and detecting the KPI to be detected through the abnormality detection model.

In an embodiment of the present application, for any newly acquired KPI flow that needs to be subjected to anomaly detection, a cluster corresponding to the KPI flow to be detected may be determined by calculating shape similarity between the KPI flow to be detected and a centroid curve of each cluster determined after classification, that is, the KPI flow to be detected may be assigned to an existing cluster. Then, since the abnormal label and the normal label of the centroid curve of each cluster have been determined in step S102, the abnormal label and the normal label on the centroid curve of the cluster (i.e. the cluster to which the KPI flow to be detected belongs) corresponding to the KPI flow to be detected can be directly read.

Further, based on the obtained label on the centroid curve of the cluster corresponding to the KPI flow and the data of the KPI flow, an anomaly detection model corresponding to the KPI flow to be detected is trained through semi-supervised learning. As an example, training the anomaly detection model corresponding to the KPI flow to be detected through semi-supervised learning includes selecting a preset amount of data from the KPI flow to be detected as second unmarked data, where the selected data may be set according to actual needs, for example, the first 20% of data is selected from the KPI flow to be detected. Namely, a preset amount of data is selected from KPI flows to be detected as non-labeled data to participate in model training. And then, training an abnormal detection model corresponding to the KPI to be detected through a semi-supervised algorithm in the correlation technique based on the abnormal label, the normal label and the second unmarked data on the centroid curve of the cluster corresponding to the KPI to be detected.

Furthermore, the abnormal detection model is trained to detect the KPI flows to be detected. In the embodiment of the present application, the anomaly detection model is completed through training corresponding to the KPI flow to be detected, the remaining data in the KPI flow to be detected is detected, and referring to the above example, if the first 20% of the data selected from the KPI flow to be detected is unmarked data for model training, the anomaly detection model can be completed through training to perform anomaly detection on 80% or 60% of the data after the KPI flow to be detected.

It should be noted that, in other embodiments of the present application, a preset amount of second non-labeled data selected from the KPI flow to be detected may also be returned to the PU learning, that is, the first non-labeled data and the second non-labeled data are combined to be used as non-labeled data, after the PU learning is performed to construct the binary classifier for active learning, the abnormal label and the normal label of the centroid curve of each cluster are obtained, that is, the preset amount of data selected from the KPI flow to be detected is used as non-labeled data to participate in the PU learning and the active learning, so as to determine the abnormal label and the normal label of the centroid curve of the cluster corresponding to the KPI flow to be detected, and then the determined abnormal label, the normal label and the second non-labeled data are combined to train the abnormal detection model corresponding to the KPI flow to be detected.

Therefore, the PU learning-based medical equipment performance index abnormity detection method reduces the labeling workload, improves the abnormity detection accuracy, and improves the practicability in practical application.

In summary, in the method for detecting the performance index abnormality of the medical device based on the PU learning according to the embodiment of the present application, the historical key performance index KPI flow is used as the training data, the training data are clustered according to the similarity degree, the centroid curve of each cluster is obtained, the centroid curve of each cluster is labeled, and the first abnormal labeling data and the first non-labeling data are obtained; constructing a binary classifier through positive-case unlabeled PU learning based on the first abnormal labeling data and the first unlabeled data, and acquiring an abnormal label and a normal label of the centroid curve of each cluster by combining active learning; the method comprises the steps of obtaining a label on a centroid curve of a cluster corresponding to the KPI to be detected, training an abnormality detection model corresponding to the KPI to be detected through semi-supervised learning, and detecting the KPI to be detected through the abnormality detection model. The method is based on the PU learning method, the KPI abnormity detection can be accurately carried out only by a small number of labels, and the abnormity detection accuracy is improved while the marking workload is reduced to the maximum extent through integrating clustering, PU learning and semi-supervised learning.

In order to more clearly describe the method for detecting abnormality of performance index of medical device based on PU learning of the present application, a specific example is described below with reference to fig. 2.

As shown in fig. 2, the anomaly detection framework PUAD based on PU learning includes an offline training process and an online detection process, and specifically includes three steps, i.e., first, clustering is performed: the time series of the training set are clustered according to the similarity degree, and for each cluster, a worker can manually mark some abnormal segments for the centroid curve, and because the number of clusters is much smaller than that of the time series curves, the marking work of the worker can be minimized. Furthermore, a newly emerging KPI flow can be assigned to an existing cluster by calculating its shape similarity to each cluster centroid. Secondly, PU learning is carried out: for each cluster centroid, PUAD applies PU learning to construct a binary classifier using existing anomalous labeled and unlabeled samples. And then, obtaining a reliable abnormal sample in multiple iterations by adopting an active learning method. This results in a cluster centroid that contains reliable outlier and normal labels. Thirdly, performing semi-supervised learning: for each KPI flow, an anomaly detection model is trained according to the label on the clustering centroid. It should be noted that, as can be seen from fig. 2, before the second step of PU learning, feature extraction may also be performed on the centroid curve and the KPI flow, so as to perform anomaly detection on the extracted features.

In order to achieve the above object, as shown in fig. 3, a second aspect of the present application provides a PU learning-based medical device performance index abnormality detection apparatus, including: clustering module 100, acquisition module 200, and training module 300.

The clustering module 100 is configured to cluster the training data according to the similarity degree by using the KPI stream as the training data, obtain a centroid curve of each cluster, label the centroid curve of each cluster, and obtain first abnormal labeling data and first non-labeling data.

An obtaining module 200, configured to construct a binary classifier through regular non-labeled PU learning based on the first abnormal labeled data and the first non-labeled data, and obtain the abnormal label and the normal label of the centroid curve of each cluster by combining active learning.

The training module 300 is configured to acquire a label on a centroid curve of a cluster corresponding to a KPI flow to be detected, train an abnormality detection model corresponding to the KPI flow to be detected through semi-supervised learning, and detect the KPI flow to be detected through the abnormality detection model.

Optionally, in an embodiment of the present application, the clustering module 100 is specifically configured to mark an abnormal segment on the centroid curve of each cluster based on manual work, and use the labeled data in the abnormal segment as first abnormal labeled data, and use the data on the centroid curve except the abnormal segment as first unlabeled data.

Optionally, in an embodiment of the present application, the obtaining module 200 is specifically configured to: determining first normal labeled data in the unlabeled data through a binary classifier; performing iterative marking on the first unmarked data for multiple times by an active learning method, and determining second normal marked data and second abnormal marked data in the first unmarked data; and taking the first abnormal labeling data and the second abnormal labeling data as abnormal labels, and taking the first normal labeling data and the second normal labeling data as normal labels.

Optionally, in an embodiment of the present application, the training module 300 is specifically configured to: determining a cluster corresponding to the KPI to be detected by calculating the shape similarity of the KPI to be detected and a centroid curve of each cluster; and reading the abnormal label and the normal label on the centroid curve of the cluster corresponding to the KPI flow to be detected.

Optionally, in an embodiment of the present application, the training module 300 is further configured to: selecting a preset amount of data from KPI flows to be detected as second unmarked data; and training an abnormity detection model corresponding to the KPI to be detected based on the abnormity label, the normal label and the second non-labeled data on the centroid curve of the cluster corresponding to the KPI to be detected.

It should be noted that the foregoing description of the embodiment of the method for detecting abnormality of performance index of medical device based on PU learning is also applicable to the embodiment of the apparatus, and the implementation principle is the same, which is not described herein again.

To sum up, the medical equipment performance index anomaly detection device based on PU learning of the embodiment of the application is based on the PU learning method, realizes that only a small number of labels are needed to carry out accurate KPI anomaly detection, and improves the accuracy of anomaly detection while reducing marking workload to the maximum extent through integrating clustering, PU learning and semi-supervised learning.

In order to implement the foregoing 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 a method for detecting performance index abnormality of a medical device based on PU learning according to the embodiments 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

1. A medical equipment performance index abnormity detection method based on PU learning is characterized by comprising the following steps:

2. The method according to claim 1, wherein the labeling the centroid curve of each cluster, and acquiring first abnormal labeling data and first label-free data comprises:

and manually marking an abnormal section on the centroid curve of each cluster, taking the marking data in the abnormal section as the first abnormal marking data, and taking the data on the centroid curve except the abnormal section as the first unmarked data.

3. The method of claim 1 or 2, wherein the obtaining of the outlier label and the normal label of the centroid curve for each cluster comprises:

determining first normal labeling data in the first label-free data through the binary classifier;

performing iterative marking on the first unmarked data for multiple times by an active learning method, and determining second normal marked data and second abnormal marked data in the first unmarked data;

and taking the first abnormal marking data and the second abnormal marking data as the abnormal labels, and taking the first normal marking data and the second normal marking data as the normal labels.

4. The method according to claim 1, wherein the obtaining of the label on the centroid curve of the cluster corresponding to the KPI flow to be detected comprises:

determining a cluster corresponding to the KPI to be detected by calculating the shape similarity of the KPI to be detected and the centroid curve of each cluster;

and reading the abnormal label and the normal label on the centroid curve of the cluster corresponding to the KPI flow to be detected.

5. The method according to claim 4, wherein the training of the anomaly detection model corresponding to the KPI flows to be detected through semi-supervised learning comprises:

selecting a preset amount of data from the KPI stream to be detected as second unmarked data;

and training an abnormal detection model corresponding to the KPI to be detected based on the abnormal label, the normal label and the second unmarked data on the centroid curve of the cluster corresponding to the KPI to be detected.

6. A medical equipment performance index anomaly detection device based on PU learning is characterized by comprising:

7. The detection apparatus according to claim 6, wherein the clustering module is specifically configured to: and marking an abnormal section on the centroid curve of each cluster based on manual work, taking the marking data in the abnormal section as the first abnormal marking data, and taking the data on the centroid curve except the abnormal section as the first unmarked data.

8. The detection apparatus according to claim 6 or 7, wherein the acquisition module is specifically configured to:

determining first normal labeled data in the unlabeled data through the binary classifier;

9. The detection apparatus according to claim 6, wherein the training module is specifically configured to:

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 PU learning based medical device performance indicator anomaly detection method according to any one of claims 1-5.