CN114944208A

CN114944208A - Quality control method, quality control device, electronic device, and storage medium

Info

Publication number: CN114944208A
Application number: CN202210575556.8A
Authority: CN
Inventors: 周睿; 陈超; 宋彪; 王哲
Original assignee: Inner Mongolia Weishu Data Technology Co ltd
Current assignee: Inner Mongolia Weishu Data Technology Co ltd
Priority date: 2021-05-27
Filing date: 2022-05-25
Publication date: 2022-08-26
Anticipated expiration: 2042-05-25
Also published as: CN113539397A; CN114944208B

Abstract

The application provides a quality control method, a quality control device, electronic equipment and a storage medium, and relates to the technical field of inspection medicine. The method comprises the following steps: acquiring sample data of a detection item; determining a quality control model corresponding to the detection item according to the variation coefficient of the detection item and an error scene in which the detection item is positioned, wherein the quality control model comprises a floating mean algorithm and a control limit, and the quality control models corresponding to the same detection item under different error scenes are different; determining whether the sample data is a monitoring result of out-of-control data or not through the quality control model; the quality control model is used for calculating the sample data by using the floating mean algorithm to obtain a floating mean value, and determining whether the sample data is a monitoring result of out-of-control data by judging whether the floating mean value exceeds the control limit. The method improves the precision of the sample data inspection.

Description

Quality control method, quality control device, electronic device, and storage medium

Technical Field

The present application relates to the field of medical testing technologies, and in particular, to a quality control method and apparatus, an electronic device, and a storage medium.

Background

Clinical examination is an important auxiliary reference for doctors to make scientific diagnosis for patients, but the examination results often have errors due to some objective or subjective reasons, so that the doctors make wrong diagnosis, further delay treatment, and even cause serious harm to the patients. Errors may occur in every stage of the assay process, such as sample collection, transport or handling, instrumentation, personnel manipulation, result interpretation, etc. In general, errors generated during the inspection process are mostly related to instruments and equipment, and laboratories generally use internal quality control programs to monitor, guarantee and manage the detection quality of the inspection overall process.

In order to ensure the reliability of the detection result, the method of performing indoor quality control by using a quality control substance is one of the most important means for monitoring the inspection performance, which is introduced in medical laboratories and is commonly used at present. Meanwhile, quality control is also an instructive evaluation index in grade evaluation of medical institutions, and specifically, the national health industry standard "clinical examination quantitative determination indoor quality control" clearly stipulates how to design a traditional quality control scheme of indoor quality control in the medical detection process, and is effectively implemented.

However, the conventional quality control schemes are greatly affected by quality control materials (i.e., the above-mentioned quality control materials) in clinical applications, such as: the concentration level of the quality control product is limited, the cost is higher, the detection frequency is lower, and the like. Research has shown that quality Control methods such as real-time quality Control (PBRTQC) based on Patient data can sensitively detect the change of the deviation of the test result, and can detect an error event before the next quality Control action, thereby reducing the number of unreliable Patient test results in clinical reports to the maximum extent, being not influenced by quality Control quality matrix effect, and monitoring the whole analysis process of the test. Once set up is successful in a medical laboratory, no additional inspection and maintenance costs are required. In 2020, the international association of clinical chemistry and laboratory medicine consortium (IFCC) analysis quality working group issued an expert consensus on the establishment and testing of the PBRTQC quality control method. However, the quality control method of PBRTQC still has the following defects: aiming at the same detection item, when different error scenes are dealt with, the technical defect of low precision of specimen data detection caused by monitoring by adopting the same quality control model is overcome.

Disclosure of Invention

The application provides a quality control method, a quality control device, electronic equipment and a storage medium, which are used for monitoring by adopting different quality control models when different error scenes are dealt with, so that high accuracy of standard data inspection is ensured.

According to a first aspect of the present application, there is provided a quality control method comprising:

obtaining sample data of a detection item;

determining a quality control model corresponding to the detection item according to a variation coefficient of the detection item and an error scene in which the detection item is positioned, wherein the quality control model comprises a floating mean algorithm and a control limit, the variation coefficient is a numerical value made by the detection item according to biological variation and is used for reflecting the dispersion degree of the detection item, and the corresponding quality control models of the same detection item under different error scenes are different;

determining whether the sample data is a monitoring result of out-of-control data or not through the quality control model; the quality control model is used for calculating the sample data by using the floating mean algorithm to obtain a floating mean value, and determining whether the sample data is a monitoring result of out-of-control data by judging whether the floating mean value exceeds the control limit.

According to a second aspect of the present application, there is provided a quality control apparatus comprising:

the data acquisition module is used for acquiring sample data of the detection items;

the data analysis module is used for determining a quality control model corresponding to the detection item according to a variation coefficient of the detection item and an error scene where the detection item is located, wherein the quality control model comprises a floating mean algorithm and a control limit, the variation coefficient is a numerical value made by the detection item according to biological variation and is used for reflecting the discrete degree of the detection item, and the corresponding quality control models of the same detection item under different error scenes are different;

the abnormal recognition module is used for determining whether the sample data is a monitoring result of out-of-control data or not through the quality control model; the quality control model is used for calculating the sample data by using the floating mean algorithm to obtain a floating mean value, and determining whether the sample data is a monitoring result of out-of-control data by judging whether the floating mean value exceeds the control limit.

According to a third aspect of the present application, there is provided an electronic device comprising: at least one processor and memory;

the memory stores computer execution instructions;

the at least one processor executing the computer-executable instructions stored by the memory causes the at least one processor to perform the quality control method as described above in the first aspect.

According to a fourth aspect of the present application, there is provided a computer-readable storage medium having stored therein computer-executable instructions for implementing the quality control method of the first aspect as described above when executed by a processor.

The application provides a quality control method, device, electronic equipment and storage medium, not only regard the discrete degree of patient data (being standard data) as the model selection foundation when the quality control model is confirmed, the error scene that the detection project is in has still been considered, the effective influence to accurate modeling through the error scene can make the monitoring result of quality control model more accurate, when being directed at same detection project under different error scenes, the floating mean algorithm in the quality control model is no longer single, and then can improve the accuracy and the timeliness of standard data inspection.

It should be understood that the statements in this section are not intended to identify key or critical features of the embodiments of the present application, nor are they intended to limit the scope of the present application. Other features of the present application will become apparent from the following description.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and together with the description, serve to explain the principles of the application.

Fig. 1 is a schematic diagram of an application scenario related to an embodiment of the present application;

fig. 2 is a schematic flowchart of a quality control method according to an embodiment of the present application;

fig. 3 is a schematic flow chart of another quality control method provided in the embodiments of the present application;

FIG. 4 is a schematic diagram of a test of the floating mean algorithm;

FIG. 5 is a MNped graph of the floating mean algorithm of 5 detection items under the proportional error;

FIG. 6 is a diagram showing the effects of quality control in various inspection items;

fig. 7 is a schematic structural diagram of a quality control apparatus according to an embodiment of the present disclosure;

fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

With the above figures, there are shown specific embodiments of the present application, which will be described in more detail below. These drawings and written description are not intended to limit the scope of the inventive concepts in any manner, but rather to illustrate the inventive concepts to those skilled in the art by reference to specific embodiments.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application.

For ease of understanding, an application scenario of the embodiment of the present application is first described.

Fig. 1 is a schematic diagram of an application scenario related to an embodiment of the present application. As shown in fig. 1, an application scenario of the embodiment relates to an instrument device and a monitoring device, wherein the instrument device is configured to provide detection data of a certain patient, the monitoring device monitors the detection data through a PBRTQC quality control method, and when the detection data is out of control, an alarm is given.

The existing PBRTQC quality control method still has the defects that: 1) specifically, the existing PBRTQC technology is a quality control method based on a floating mean value, that is, a mean value, a median value, or other quantities of a sample are monitored in a rolling manner through a sliding window with a fixed length, and if the mean value, the median value, or other quantities exceed a set range (control limit), the PBRTQC technology is regarded as being out of control. The control limit is statistically selected based on the floating amount of the normal sample. The method is extremely sensitive to data performance, and different methods have different performances under different projects, so that the establishment of a quality control system is lacked to make up for the defect of a single algorithm; 2) in the early stage of setting, the characteristics of a laboratory patient group and an analysis method used by a detection project need to be known, that is, the performance of the current different PBRTQC quality control methods is different under the data with different characteristics, and the characteristics of the data are closely related to the patient group and the types of instruments, so that the characteristics of the data need to be summarized according to the characteristics of the different patient groups and the working principles of the instruments, and the medicine is taken according to the symptoms and finally matched with a special quality control method; 3) the detection accuracy and the detection speed of the method can not completely meet the clinical diagnosis and treatment requirements, in practical application, different quality control methods correspond to different monitored quantities, the quality control methods are sensitive to data quality, and the selection process of the method and the setting modes of parameters such as window length, control limit and the like lack guiding standards. Once the unreasonable result is given, only the larger deviation can be effectively identified, and finally the identification range is reduced and the accuracy is low. However, an excessively long window results in hundreds of consecutive runaway data being required to be identified, and the slowness of the alarm affects the detection speed. Especially for the project requirement with important clinical value, the detection efficiency for small error is still to be improved, such as: prostate Specific Antigen (PSA) detection is crucial for prostate cancer diagnosis, monitoring and treatment. In a continuous PSA assay, a change of 0.2. mu.g/L may be considered by the system as an indication of a poorer prognosis. If the small deviation with clinical significance can be accurately identified, the occurrence of subsequent misdiagnosis can be effectively reduced. 4) The functional requirements for laboratory information systems are high. Therefore, it is necessary to provide quality control models including different floating mean algorithm types for the same detection item under different error scenarios to form a quality control system capable of coping with different error scenarios.

In order to solve at least one of the above technical problems, embodiments of the present application provide a quality control method, an apparatus, an electronic device, and a storage medium, which are intended to introduce the PBRTQC technology into the field of medical examination, and establish a novel real-time dynamic quality control framework based on patient examination results (or referred to as detection results and sample data) that can cover various error types (i.e., error scenes) and sizes. The frame is established and clinically applied, so that the accuracy and timeliness of a test result can be further improved, clinical requirements and industrial and regulatory requirements are better met, medical resources are further saved, the medical cost is reduced, the medical safety is guaranteed, and both doctors and patients are benefited. In order to achieve the above purpose, the present application first provides a real-time quality control system based on patient test results, and the overall framework is divided into the following four layers: the system comprises a data acquisition layer, a data analysis layer, an anomaly identification layer and a reinforcement learning layer. On the basis of the real-time quality control system, the application provides a quality control method.

The following describes the technical solution of the present application and how to solve the above technical problems in detail by specific embodiments. The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present application will be described below with reference to the accompanying drawings.

Fig. 2 is a schematic flow chart of a quality control method according to an embodiment of the present disclosure. As shown in fig. 2, the method of the present embodiment includes:

s201: obtaining sample data of a detection item;

in the embodiment of the present application, the inspection item is also referred to as an inspection item. Conventional test items include, but are not limited to: blood routine items and biochemical items, wherein the blood routine items are serum or protein, and the biochemical items include but are not limited to: glucose (GLU), Total Protein (TP), alanine Aminotransferase (ALT), Aspartase (AST), Albumin (ALB), and the like. The sample data may be understood as patient samples, patient test data, patient test results or patient test results. The sample data may be original sample data or data obtained by processing the original sample data, and details of the sample data are described in S301 to S303 below. Further, S201 may be understood as a specific operation of the data acquisition layer. When the sample data is acquired, the patient information corresponding to the sample data can be recorded, including but not limited to: age, sex, category, time of collection, season, etc. It should be noted that only clinically reported results are included in the specimen data acquisition process.

S202: and determining a quality control model corresponding to the detection item according to the variation coefficient of the detection item and the error scene in which the detection item is positioned, wherein the quality control model comprises a floating mean algorithm and a control limit. The variation coefficient is a numerical value formulated by the detection item according to biological variation and used for reflecting the discrete degree of the detection item, and the corresponding quality control models of the same detection item under different error scenes are different.

In this embodiment, the coefficient of variation is an index characteristic of the detection item, and the coefficient of variation may be determined according to the index characteristic of the detection itemA suitable quality control model is provided that recommends a corresponding floating mean algorithm. S202 may be understood as the operation of the data analysis layer. In the data analysis layer, the embodiment of the present application can determine, for different detection items, whether the detection item is suitable for the floating mean algorithm according to a biological variation coefficient (referred to as a variation coefficient for short) of the detection item. The coefficient of variation includes the coefficient of variation CV within an individual _i And coefficient of variation between individuals CV _g . This embodiment may be based on CV _i And CV _g The degree of dispersion of the detection items is judged, for example: if CV is _i And CV _g All are greater than 60%, which indicates that the possibility of instability of the detection data is very high, and then the detection item is not suitable for the floating mean algorithm.

S203: determining whether the sample data is a monitoring result of out-of-control data or not through the quality control model; the quality control model is used for calculating the sample data by using the floating mean algorithm to obtain a floating mean value, and determining whether the sample data is a monitoring result of out-of-control data by judging whether the floating mean value exceeds the control limit.

In this embodiment, S203 may be understood as an operation of the anomaly recognition layer. And in the abnormal identification layer, monitoring the new patient samples in the same batch in real time through the control limit calculated by the historical patient samples.

Because the existing quality control technology is based on a quality control product, the technical problems of high cost and limited inspection exist, the embodiment of the application is based on a patient sample, the cost problem of the quality control product is avoided, the accuracy and the timeliness of a detection result are further improved, and the clinical requirement and the industrial regulation requirement can be better met.

In one possible implementation manner, since the quality control models in the present application are pre-established before being applied, before determining the quality control model corresponding to the detection item according to the variation coefficient of the detection item and the error scenario in which the detection item is located, the quality control method further includes the following steps S204 to S205, where:

s204: and obtaining the variation coefficient of the detection item, and determining a plurality of alternative floating mean algorithms according to the variation coefficient of the detection item.

The step of determining a plurality of alternative floating mean algorithms is a step of implementing algorithm model selection. One or more of an expanded Weighted Average (EWMA), a Moving Average (MA), a Moving Median (MM), a Moving statistical Median Hd50, a Moving Standard deviation (MovSD), and a Moving partition number MovSO may be provided as an alternative floating Average algorithm.

According to the statistical result of each floating mean algorithm under a normal sample, 6 formulas for calculating the floating mean algorithm are as follows:

MA：

average of data within a fixed window.

EWMA：EWMA _t ＝(1-λ)×EWMA _t-1 +λ×x _t The moving weighted average is based on the MA algorithm.

MM：MM _t ＝Median(x _t-N ，x _t-N+1 ，...，x _t ) The median of the data within the fixed window.

Hd50：Hd50＝(x _t-N ，x _t-N+1 ，...，x _t ) Statistical median of data within a fixed window.

MovSD：MovSD _t ＝SD(x _t-N ，x _t-N+1 ，...，x _t ) Statistical standard deviation of the data within a fixed window.

MovSO：

The number of data in the fixed window that exceeds the clinical decision limit.

In each of the above floating mean algorithms, the step size may be according to [10, 30, 50, 90, 110, 130, 150 ]]The search is performed in the manner of (1). It should be noted that different algorithms have different error differences for the tested test itemsResolution, and the magnitude of the error resolution is related to the numerical distribution of the detected items. The self mean value and the discrete degree of the detection item influence the identification of errors, and the selection of the alternative floating mean value algorithm suitable for the detection item can effectively improve the unbalanced relation between sensitivity and specificity, so that the embodiment can be used for effectively improving the unbalanced relation between sensitivity and specificity according to the inter-individual variation coefficient CV of the detection item _g And judging the discrete degree, and selecting a proper moving average algorithm as an alternative floating average algorithm.

S205: and carrying out a grid search experiment based on the multiple candidate floating mean algorithms and the error scene where the detection item is positioned, and establishing a quality control model corresponding to the detection item based on the experiment result.

In the embodiment of the present application, S204 to S205 may be understood as a statistical process including algorithm type selection, a grid search experiment, and result optimization, where the grid search experiment in S205 includes a series of operations including data filtering, data partitioning, data transition (i.e., normal transition), window length setting, control limit calculation method selection, simulation test, and statistical result selection. The specific analysis of data filtering, data partitioning and data state transition is described in the following S301 to S303. Setting the window length means setting the data displayed in the movable window to be a fixed length. For example: the window length is 50, and the embodiment of the present application monitors the mean, median or other quantity of the samples by rolling the sliding window in units of 50, and the number of samples in the window is kept at 50. The control limit calculation method is a method adopted in the process of setting the left control limit and the right control limit after normal samples are learned, and comprises a symmetry method, an integral position division method and a daily position division method. After the control limit calculation method is selected, the calculation process of the control limit may be implemented. And the details of the simulation test and the statistical result are shown in the following S1-S3.

In a possible implementation manner, the step S205: performing a grid search experiment based on the multiple candidate floating mean algorithms and the error scene where the detection item is located, and establishing a quality control model corresponding to the detection item based on an experiment result, including: S1-S3, wherein:

s1: under the error scene of the detection item, providing a plurality of control limit calculation methods and a plurality of window lengths, and forming a plurality of permutation and combination based on the plurality of alternative floating mean value algorithms, the plurality of control limit calculation methods and the plurality of window lengths.

In the embodiment of the present application, S1 is a step of a simulation test, and according to a clinical actual error scenario, the embodiment of the present application may introduce errors according to three scenarios, namely a proportional error, a constant error, and a random error, and the error magnitude is changed according to an error factor. For example, the embodiment of the application can set multiple simulation days, introduce 1000 continuous error data as a test object every day, and use the statistically affected patient sample number (Nped) as an important index for evaluating the recognition rate.

S2: and counting the index analysis results of the permutation and combination under each preset evaluation index aiming at each permutation and combination.

In an embodiment of the present application, the preset evaluation index includes at least one of: true Positive Rate (TPR), False Positive Rate (FPR), True Negative Rate (TNR), False Negative Rate (FNR), cumulative ANPed (i.e., mean of Nped), cumulative 95NPed (i.e., 95% quantile of Nped), and MNPed (i.e., median of Nped). In practical application, a True Positive Rate (TPR), a False Positive Rate (FPR), a True Negative Rate (TNR) and a False Negative Rate (FNR) are common evaluation criteria under classification statistics, and since the quality control process substantially judges whether the control is in progress or out of control by identifying unbiased and biased samples, the scene of the application substantially belongs to the classification process. If the part of indexes are deleted, the realization of the whole scheme is not influenced. During the whole verification process, the embodiment of the application needs to divide the data set into a plurality of simulation days, and each day contains 1000 continuous deviation data. The number of samples passed by the detection data from the insertion point of the offset until the offset is identified is called NPed. At each deviation, the mean, median and 95 quantiles (i.e., ANPed, MNPed and 95NPed) of NPed for all simulation days can be counted, which is a necessary indicator because it is a clinically recognized indicator.

S2 is a step of statistical results, and is used to summarize index analysis results corresponding to all permutation and combination related to the grid search experiment into a table.

S3: and selecting a target permutation combination from the permutation combinations as a quality control model corresponding to the detection item according to the priority order of all the preset evaluation indexes on the basis of the index analysis results of the permutation combinations under all the preset evaluation indexes.

In the step of optimizing the result, the embodiment of the present application synthesizes each index of each permutation and combination in the statistical result table, and selects an optimal permutation and combination as an optimal solution of the grid search experiment according to a certain priority order.

Before S1, sample data used for modeling may be processed, and the data processing procedure is consistent with the following steps S301 to S303, and will not be described herein again.

In a possible implementation manner, after determining whether the sample data is a monitoring result of the runaway data by determining whether the floating mean value exceeds the control limit, the quality control method further includes S206:

s206: when the sample data are accumulated to a preset number, updating a quality control model corresponding to the detection item; wherein the updated quality control model has re-optimized control limits.

It should be understood that S206 is a specific operation of the reinforcement learning layer, and when the patient samples are accumulated to a certain extent, the model can be updated again according to the statistical process, and the control limit is preferred to improve the accuracy. The updating process is similar to the model establishing process, and is not described herein again. That is, when the model is used for the first time, the control limit is preset through the step of reinforcement learning. The preset number can be set by a user in a self-defined mode, and the calculation method of the control limit or the value is changed when the model is updated.

On the basis of the above embodiments, the technical solutions of the present application are described in more detail below with reference to specific embodiments.

Fig. 3 is a schematic flowchart of another quality control method according to an embodiment of the present application. On the basis of the embodiment shown in fig. 2, the present embodiment focuses on refining S201 in fig. 2. As shown in fig. 3, the method of this embodiment includes:

s301: original sample data are obtained, and the original sample data are filtered to obtain filtered sample data. The acquisition in the embodiment of the present application may be understood as collection. The model can be used as sample data to train the model in the process of establishing the model, and can be used as input data of the model to test in the process of clinical application.

In one possible implementation, S301: the filtering the original sample data to obtain filtered sample data includes: s401, when the filtering processing is intercepting processing, setting an upper cutoff value and a lower cutoff value; s402, according to the upper cutoff value and the lower cutoff value, intercepting abnormal data in the original sample data to obtain intercepted sample data.

In the embodiment of the present application, abnormal data is intercepted, and an interception limit (i.e., a truncation limit) is defined as a quantile t corresponding to a threshold boundary of normal data, where t is [0, 0.01, 0.05, 0.2, 0.4 ═ t []The upper limit of the interception limit (i.e., the upper cutoff value) is: UTL ═ quantile (1-n/2), the lower limit of interception (lower cut-off value) is: LTL is equal to quantile (n/2). The interception processing is to intercept the abnormal point, and the mode of intercepting the abnormal point is divided into two modes of replacement and removal, wherein the formula of the replacement mode is as follows:

the formula of the removal mode is as follows:

wherein NAN is a non-number.

In this embodiment of the application, the filtering process in S301 is to clean the original sample data, and is capable of removing outliers (which may be in a certain proportion, for example, values smaller than the overall quantile of 25% or larger than the overall quantile of 75% are removed as outliers) and nonsensical values (for example, empty values, character strings, negative numbers, etc.); or, the data with the structure not conforming to the conventional data and the sample data in the non-digital format are removed, so that the adverse effect of the abnormal point on the control limit calculation process in the follow-up process is avoided. After the data interception process, a transition may be made.

S302: performing data transformation on the filtered sample data to obtain transformed sample data;

in one possible implementation, S302: the data transformation is performed on the filtered sample data to obtain transformed sample data, and the method comprises the following steps: and performing data transformation on the filtered sample data by using a Box-Cox transformation mode or a prototype transformation mode to obtain transformed sample data.

The prototype transformation mode, also called prototype state conversion, namely the neat type state conversion, refers to processing according to the original level, namely, the original data is directly used without any processing. The Box-Cox transformation mode is called Box-Cox transformation, and the Box-Cox transformation is a generalized power transformation method proposed in 1964, is a data transformation commonly used in statistical modeling, and is used for the condition that continuous response variables do not meet normal distribution. After Box-Cox transformation, the correlation of the unobservable errors and the predictor variables can be reduced to some extent. The Box-Cox transformation can obviously improve the properties of normality, symmetry and variance equality of data. The Box-Cox transformation formula is

Wherein, λ is a transformation parameter carried by a box-cox transformation, which can affect the normal performance after transformation, and the parameter can directly obtain an optimal value by using a programming tool. The data transformation in S302 thus achieves a transition. For the state conversion, the skewness of the data distribution of different types of detection items is different, and if the sample data is converted into the normal state, the accuracy of the experimental result can be improved to a certain extent.

S303: determining the converted sample data as the sample data of the detection item;

in this embodiment, the foregoing S301 to S303 clean and process the original sample data, and can provide a data basis for improving the inspection accuracy of the sample data. In addition, the data processing may include data division in addition to the filtering processing and the data conversion, the data division may count the number of sample points on both sides by using a clinically specified level decision limit as a boundary point of the data division only for the movSD algorithm, and the division results corresponding to the decision limits at different positions are different.

In this embodiment, the data processing can also be classified by controlling age, department and category, and select the control time period of laboratory data quality control. It should be noted that the original sample data collected in the present application is the data of the quality control in the control period.

S304: and determining a quality control model corresponding to the detection item according to the variation coefficient of the detection item and the error scene in which the detection item is positioned, wherein the quality control model comprises a floating mean algorithm and a control limit. The variation coefficient is a numerical value which is made by the detection item according to biological variation and is used for reflecting the discrete degree of the detection item, and the corresponding quality control models of the same detection item under different error scenes are different.

It should be understood that the specific implementation manner of S304 is similar to S202 in fig. 2, and is not described herein again.

S305: determining whether the sample data is a monitoring result of out-of-control data or not through the quality control model; the quality control model is used for calculating the sample data by using the floating mean algorithm to obtain a floating mean value, and determining whether the sample data is a monitoring result of out-of-control data by judging whether the floating mean value exceeds the control limit.

It should be understood that the specific implementation manner of S305 is similar to S203 in fig. 2, and is not described herein again.

With the above specific step analysis, an example analysis of the model building process is given below in an embodiment:

step 11: alanine Aminotransferase (ALT), Aspartase (AST), Glucose (GLU), Total Protein (TP) and Albumin (ALB) were selected as test items by a laboratory information system in a hospital laboratory. The reason why these analytes were selected as the test items is that these 5 analytes represent different data distribution segments in the clinical laboratory and there is no significant gender difference, since the fluctuation of gender distribution would cause unpredictable gender difference in the average value except the analysis performance, so the present embodiment can avoid the above gender difference. For each detection item, data processing such as interception and state transition is performed, and sample data (hereinafter referred to as a sample) is obtained.

Step 12: the samples were divided into 200 days, 1150 samples per day, the first 150 samples were normal samples, and the last 1000 samples were taken as abnormal samples. The process of simulating 1000 error data was analyzed as follows: the errors are defined as three types of proportional error, constant error and random error, wherein the formula of proportional error is x' ═ x (1+ n × TEa), wherein n × TEa [ -50%, -48%, -46%,. -, 46%, 48%, 50%](ii) a The constant error is formulated as

Wherein n [ -3, -2.5, -2, -1.5, -1, -0.5,0.5,1,1.5,2,2.5,3]The TEA is the industry quality standard of each clinical detection item, the corresponding numerical value of each clinical detection item is different, and the TEA is updated by the association every year; the random error is formulated as

Wherein n is [0.5,1,1.5,2,2.5,3 ]]In the above three formulas, x' represents patient data that introduces modeling error, x represents raw patient data (i.e., sample data),

represents the mean of the data set, σ represents the standard deviation of the data set, and n is a factor used to control the magnitude of the introduced error. The above mentioned TEa means the total allowable error in clinic, is the sum of random error (BE) and Systematic Error (SE), and is used to reflect the difference between the measured result and the true value, and the total error of the selected detection method must BE within the range of the clinically acceptable level.

Step 13: and (3) calculating the control limit based on the normal sample in the step (12), firstly obtaining mean value data of the normal sample, and then calculating the control limit according to different methods, wherein the calculation methods of the control limit are specifically divided into a symmetrical method, an integral position dividing method and a daily position dividing method. The symmetric method is characterized in that the average value is taken as a symmetric axis according to the distribution condition of the average value sequence of the normal sample, and numerical values of standard deviation widths are selected from left to right to serve as control limits; the integral quantile method is as follows: taking the 0.5% quantile of the mean value number series data of all the normal samples as the lower limit of the control limit, and taking the 99.5% quantile as the upper limit of the control limit; the daily quantile method is: and (3) respectively and independently counting the minimum value and the maximum value of the mean value sequence of the normal samples every day, wherein 0.5% quantiles serve as the lower limit of the control limit in the minimum value set, and 99.5% quantiles serve as the upper limit of the control limit in the maximum value set.

Step 14: for each control limit provided in step 13, 1150 samples per day are tested according to the control limit, the test mode is as shown in fig. 4, 1000 deviation data (i.e. the above error data) are introduced after the maximum step length 150, when deviation data exists in the window, the floating average value under each sliding average algorithm is calculated according to step length sliding until the control limit is exceeded for alarming, and the number of samples experienced by the window is recorded as Nped. If no deviation data is identified, statistics are taken as 1100. Because the number of tests is large, the tests are not listed one by one, and part of test parameters are shown as follows:

TABLE 1 test parameters

Algorithm	Filtration mode	Limit of filtration	Normal mode of transformation	Step size	Control limit calculation method
						EWMA	2	5	2	7	3
MA	2	5	2	7	3
						MM	2	5	-	7	3
Hd50	-	-	2	7	3
						MovSD	2	5	2	7	3
MovSO	-	-	-	7	3

As can be seen from table 1, in the present embodiment, under a proportional error scenario, 4 floating mean algorithms, such as EWMA, MA, MM, and MovSD, are considered as candidate floating mean algorithms, a filtering manner indicates that a removal manner is adopted, a positive state conversion manner is Box-Cox transformation, a control limit calculation method is a daily quantile method, a window length is 50, and the preset evaluation index may include: false positive rate, true positive rate, MNped, ANped, and 95 NPed. In this example, the embodiment of the present application performs a corresponding test on the four permutations and combinations of the above example to obtain a corresponding index analysis result. An MNped curve formed by the 4 permutation and combination of the 5 detection items is shown in fig. 5, the shape of the curve is used for reflecting the recognition capability of the adopted floating average algorithm under different types and different sizes of deviations, and the performance of each algorithm can be visually judged. The ideal shape is: two sides are low, the middle is narrow, and the two sides are symmetrical. Therefore, the algorithms suitable for different detection items are different.

The priority order of the preset evaluation indexes may be: 1) sorting the false positive columns under each permutation and combination, and reserving the permutation and combination with the false positive rate of less than 5%; 2) under the condition of 1), retaining the permutation and combination with the true positive rate more than or equal to 90 percent; 3) and sorting according to the accumulated MNped, selecting the permutation combination corresponding to the minimum value, and if the screening result is not unique, further comparing ANped with 95NPed, and selecting the permutation combination corresponding to the minimum value. It should be noted that all the 3 calculation methods of the calculation mode of the floating mean, the setting of the window length, and the control limit need to be enumerated once in a certain range, and an optimal permutation and combination is selected through the optimization and judgment rule (i.e., the preset evaluation index) under the final summary table. In the result optimizing process, the embodiment may sort the permutation and combination according to the priority order of the preset evaluation index, and use the best permutation and combination as the optimal solution of the grid search experiment. By performing quality control on various detection items in the above manner, an effect diagram of quality control under various detection items in fig. 6 can be obtained.

Fig. 7 is a schematic structural diagram of a quality control apparatus according to an embodiment of the present disclosure. The apparatus of the present embodiment may be in the form of software and/or hardware. As shown in fig. 7, the quality control device provided in this embodiment includes: a data acquisition module 71, a data analysis module 72 and an anomaly identification module 73. Wherein the content of the first and second substances,

a data acquisition module 71, configured to acquire sample data of the detection item;

the data analysis module 72 is configured to determine a quality control model corresponding to the detection item according to a variation coefficient of the detection item and an error scene where the detection item is located, where the quality control model includes a floating mean algorithm and a control limit, the variation coefficient is a numerical value formulated by the detection item according to biological variation and is used to reflect a discrete degree of the detection item, and the corresponding quality control models of the same detection item are different in different error scenes;

an anomaly identification module 73, configured to determine, through the quality control model, whether the sample data is a monitoring result of uncontrolled data; the quality control model is used for calculating the sample data by using the floating mean algorithm to obtain a floating mean value, and determining whether the sample data is a monitoring result of out-of-control data by judging whether the floating mean value exceeds the control limit.

In one possible implementation, the quality control apparatus further includes: the system comprises an acquisition determining module and a grid search experiment module, wherein:

the acquisition determining module is used for acquiring the variation coefficient of the detection item and determining a plurality of alternative floating mean algorithms according to the variation coefficient of the detection item;

and the grid search experiment module is used for carrying out a grid search experiment based on the multiple candidate floating mean algorithms and the error scene where the detection item is positioned, and establishing a quality control model corresponding to the detection item based on the experiment result.

In one possible implementation, the grid search experiment module includes: a combination unit, a statistical unit and a selection unit, wherein:

a combination unit, configured to provide multiple control limit calculation methods and multiple window lengths in an error scenario in which the detection item is located, and form multiple permutation and combination based on the multiple candidate floating mean algorithms, the multiple control limit calculation methods, and the multiple window lengths;

the statistical unit is used for counting index analysis results of the permutation and combination under each preset evaluation index aiming at each permutation and combination;

and the selecting unit is used for selecting a target permutation combination from the permutation combinations as the quality control model corresponding to the detection item according to the priority order of all the preset evaluation indexes on the basis of the index analysis results of the permutation combinations under all the preset evaluation indexes.

In one possible implementation, the quality control apparatus further includes:

a reinforcement learning module 74, configured to update a quality control model corresponding to the detection item when the sample data is accumulated to a preset amount; wherein the updated quality control model has re-optimized control limits.

In one possible implementation, the data obtaining module 71 includes: a filtering unit, a transformation unit and a determination unit, wherein:

the filtering unit is used for acquiring original sample data and filtering the original sample data to obtain filtered sample data;

the transformation unit is used for carrying out data transformation on the filtered sample data to obtain transformed sample data;

and the determining unit is used for determining the converted sample data as the sample data of the detection item.

In one possible implementation, the filtering unit includes a setting subunit and an intercepting subunit, where:

a setting subunit, configured to set an upper cutoff value and a lower cutoff value when the filtering processing is interception processing;

and the intercepting subunit is used for intercepting the abnormal data in the original sample data according to the upper cutoff value and the lower cutoff value to obtain intercepted sample data.

In a possible implementation manner, the transformation unit is further configured to: and performing data transformation on the filtered sample data by using a Box-Cox transformation mode or a prototype transformation mode to obtain transformed sample data.

The quality control apparatus provided in this embodiment may be configured to execute the quality control method provided in any of the above method embodiments, and the implementation principle and the technical effect are similar, which are not described herein again.

In the technical scheme of the application, the collection, storage, use, processing, transmission, provision, disclosure and other processing of the personal information of the related user are all in accordance with the regulations of related laws and regulations and do not violate the good custom of the public order.

According to an embodiment of the present application, an electronic device and a readable storage medium are also provided.

Fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present application. The electronic device comprises a receiver 80, a transmitter 81, a processor 82 and a memory 83, and the electronic device formed by the above components can be used to implement the above specific embodiments of the present application, and will not be described herein again.

The embodiments of the present application further provide a computer-readable storage medium, in which computer instructions are stored, and when a processor executes the computer instructions, the steps in the method in the foregoing embodiments are implemented.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present disclosure may be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, and the present disclosure is not limited herein.

The above-described embodiments should not be construed as limiting the scope of the present application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A quality control method, comprising:

obtaining sample data of a detection item;

2. The quality control method according to claim 1, wherein before the determining a quality control model corresponding to the detection item according to the variation coefficient of the detection item and the error scenario in which the detection item is located, the method further comprises:

obtaining the variation coefficient of the detection item, and determining a plurality of alternative floating mean algorithms according to the variation coefficient of the detection item;

and carrying out a grid search experiment based on the multiple candidate floating mean algorithms and the error scene where the detection item is positioned, and establishing a quality control model corresponding to the detection item based on the experiment result.

3. The quality control method according to claim 2, wherein performing a grid search experiment based on the multiple candidate floating mean algorithms and an error scenario in which the detection item is located, and establishing a quality control model corresponding to the detection item based on an experiment result comprises:

under the error scene of the detection item, providing a plurality of control limit calculation methods and a plurality of window lengths, and forming a plurality of permutation and combination based on the plurality of alternative floating mean value algorithms, the plurality of control limit calculation methods and the plurality of window lengths;

for each permutation combination, counting index analysis results of the permutation combination under each preset evaluation index;

and selecting a target permutation combination from the permutation combinations as a quality control model corresponding to the detection item according to the priority order of all the preset evaluation indexes on the basis of the index analysis results of the permutation combinations under all the preset evaluation indexes.

4. The quality control method according to claim 2, wherein after the determining whether the sample data is a monitoring result of runaway data by determining whether the floating mean value exceeds the control limit, the method further comprises:

when the sample data are accumulated to a preset number, updating a quality control model corresponding to the detection item; wherein the updated quality control model has re-optimized control limits.

5. The quality control method according to claim 1, wherein the acquiring sample data of the detection item includes:

acquiring original sample data, and filtering the original sample data to obtain filtered sample data;

performing data transformation on the filtered sample data to obtain transformed sample data;

and determining the converted sample data as the sample data of the detection item.

6. The quality control method according to claim 5, wherein the filtering the original sample data to obtain filtered sample data comprises:

when the filtering processing is interception processing, setting an upper cutoff value and a lower cutoff value;

and intercepting abnormal data in the original sample data according to the upper cutoff value and the lower cutoff value to obtain intercepted sample data.

7. The quality control method according to claim 5, wherein the performing data transformation on the filtered sample data to obtain transformed sample data comprises:

and performing data transformation on the filtered sample data by using a Box-Cox transformation mode or a prototype transformation mode to obtain transformed sample data.

8. A quality control apparatus, comprising:

9. An electronic device, comprising: at least one processor and a memory;

the memory stores computer-executable instructions;

the at least one processor executing the computer-executable instructions stored by the memory causes the at least one processor to perform the quality control method of any one of claims 1 to 7.

10. A computer-readable storage medium having computer-executable instructions stored thereon, which when executed by a processor, are configured to implement the quality control method of any one of claims 1 to 7.