Method for selecting training data set of quality control model and sample analysis system

Abstract

The invention relates to a quality control model training data set selection method and a sample analysis system. The method comprises the steps that historical detection data of a detection device for a specified detection item within preset time is acquired, the historical detection data has a first preset number, and a second preset number of historical detection data is continuously selected from the historical detection data based on a preset observation window, and calculating a detection quality evaluation value for the historical detection data in each observation window, evaluating a trend value and / or inaccuracy of the historical detection data in each observation window according to the detection quality evaluation value, and selecting the historical detection data of the observation windows of which the trend values and / or inaccuracy meet statistical standards to form a training data set. According to the scheme provided by the invention, the training data set can be selected according to the actual condition of the detection device, and the sensitivity and accuracy of PBRTQC can be improved without depending on the hypothesis that the detection device is in the in-control state within the time of the training data set.

Description

Technical Field

[0001] The present application relates to the field of medical diagnosis technology, and in particular to a method for selecting a training data set for a quality control model and a sample analysis system. Background Art

[0002] Internal quality control (IQC) has been established as a central pillar of analytical quality control. However, current IQC procedures present challenges in detecting erroneous results produced by analytical instruments, such as discontinuous quality control processes, poor interchangeability of IQC materials, and unreliable quality control materials. Patient-based real-time quality control (PBRTQC) is considered a valuable supplement to IQC because it can prevent these problems. The key to the application of the PBRTQC model lies in the determination of parameters such as the upper truncation limit (UTL), lower truncation limit (LTL), floating window size (N), upper control limit (UCL) and lower control limit (LCL). Whether the upper and lower control limits are set reasonably will directly affect the sensitivity and accuracy of the PBRTQC method.

[0003] The method for determining upper and lower control limits in related technologies is mainly to first collect patient sample data (called training data set) of continuous equipment testing within a certain period of time, and based on the assumption that the equipment is in a controlled state during this period of time, the PBRTQC calculated values ​​generated by the patient sample data are used to obtain the mean and standard deviation of the calculated values, and the corresponding control limits are calculated based on the normal distribution or other distribution characteristics. However, in actual application, since the quality control of the equipment in the time period of the data in the training data set is not real-time, even if it is assumed that the equipment is in a controlled state during this period of time, there may still be unreasonable selection of the training data set due to factors such as reagent batch replacement, equipment recalibration, and changes in equipment precision, which in turn leads to unreasonable control limit calculation problems. Summary of the invention

[0004] In order to solve or partially solve the problems existing in the related art, the present application provides a method for selecting a training data set for a quality control model and a sample analysis system, which can select a training data set according to the actual situation of the detection device of the sample analysis system, and does not rely on the assumption that the detection device of the sample analysis system is in a controlled state during the time when the training data set is located, thereby improving the sensitivity and accuracy of PBRTQC.

[0005] The first aspect of the present application provides a method for selecting a training data set for a quality control model, comprising:

[0006] Acquire historical test data of a detection device of a sample analysis system for a specified test item within a preset time, wherein the historical test data has a first preset number;

[0007] Continuously selecting a second preset number of historical detection data from the historical detection data based on a preset observation window;

[0008] Calculating a detection quality evaluation value for the historical detection data in each observation window, and evaluating a trend value and / or imprecision of the historical detection data in each observation window according to the detection quality evaluation value;

[0009] The historical detection data of the observation window whose trend value and / or imprecision meet the statistical standard are selected to form the training data set.

[0010] Optionally, the step before continuously selecting a second preset amount of historical detection data from the historical detection data based on a preset observation window includes:

[0011] Truncating the first preset amount of historical detection data;

[0012] The truncated historical detection data are transformed into normal distribution data.

[0013] Optionally, the detection quality assessment value includes a moving average of the second preset number of historical detection data selected in the observation window.

[0014] Optionally, evaluating the trend value and / or imprecision of the historical detection data in each observation window according to the detection quality evaluation value includes:

[0015] Calculating the mean and standard deviation of the moving average;

[0016] The trend value and / or imprecision of the historical detection data in each observation window is evaluated based on the moving average, the average value and the standard deviation.

[0017] Optionally, evaluating the trend value and / or imprecision of the historical detection data in each observation window according to the moving mean, the average value and the standard deviation includes:

[0018] generating a trend value evaluation equation based on the moving mean and the average value, and calculating the slope and intercept of the trend value evaluation equation;

[0019] Evaluate the trend value of the historical detection data in each observation window according to the slope and the intercept; and / or

[0020] The imprecision is evaluated based on the mean value and the standard deviation. Optionally, the imprecision is evaluated based on the mean value and the standard deviation, comprising:

[0021] Calculate the coefficient of variation of the moving mean based on the mean and the standard deviation;

[0022] When the coefficient of variation is not greater than a preset threshold, it is determined that the imprecision of the historical detection data in the observation window meets the statistical standard.

[0023] Optionally, the method further comprises:

[0024] When the trend value and / or imprecision do not meet the statistical standards, the historical test data is postponed by a second preset number of data in chronological order, and the step of obtaining the historical test data of the specified test item within the preset time of the detection device of the sample analysis system is returned.

[0025] A second aspect of the present application provides a sample analysis system, comprising a detection device for detecting a blood sample, the sample analysis system comprising a quality control model for detection quality, and the sample analysis system further comprising:

[0026] An acquisition module, used to acquire test samples;

[0027] An analysis module is used to analyze the test sample through the quality control model; wherein the quality control model is obtained by training a training data set; the training data set is formed by: obtaining historical test data of a detection device of a sample analysis system for a specified test item within a preset time, the historical test data having a first preset number; continuously selecting a second preset number of historical test data from the historical test data based on a preset observation window; calculating a test quality evaluation value for the historical test data in each observation window, and evaluating a trend value and / or imprecision of the historical test data in each observation window based on the test quality evaluation value; and selecting historical test data of an observation window whose trend value and / or imprecision meets statistical standards;

[0028] A generation module is used to generate analysis results corresponding to the detection sample.

[0029] A third aspect of the present application provides an electronic device, including:

[0030] Processor; and

[0031] The memory stores executable codes thereon, and when the executable codes are executed by the processor, the processor is caused to execute the method as described above.

[0032] A fourth aspect of the present application provides a computer-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor is caused to execute the method as described above.

[0033] The technical solution provided by this application may have the following beneficial effects:

[0034] Acquire historical detection data of a specified detection item within a preset time from a detection device of a sample analysis system, the historical detection data having a first preset number, continuously select historical detection data of a second preset number from the historical detection data based on a preset observation window, calculate a detection quality assessment value for the historical detection data in each observation window, evaluate the trend value and / or imprecision of the historical detection data in each observation window based on the detection quality assessment value, select the historical detection data of the observation window whose trend value and / or imprecision meets the statistical standard to form a training data set, thereby selecting the training data set according to the actual situation of the detection device of the sample analysis system, and not relying on the assumption that the detection device of the sample analysis system is in a controlled state during the time when the training data set is located, thereby improving the sensitivity and accuracy of PBRTQC.

[0035] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The above and other objects, features and advantages of the present application will become more apparent by describing in more detail the exemplary embodiments of the present application in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the present application.

[0037] Figure 1 It is a flowchart of a method for selecting a training data set for a quality control model shown in an embodiment of the present application;

[0038] Figure 2 is another flow chart of a method for selecting a training data set for a quality control model shown in an embodiment of the present application;

[0039] Figure 3 It is a schematic diagram of test results shown in an embodiment of the present application without making a trendless judgment on historical test data;

[0040] Figure 4 It is a schematic diagram of the test results after the historical test data is judged to be trendless according to the embodiment of the present application;

[0041] Figure 5 It is a schematic diagram of the test results shown in the embodiment of the present application without making an imprecision judgment on the historical test data;

[0042] Figure 6 It is a schematic diagram of the test results after the imprecision judgment is made on the historical test data shown in the embodiment of the present application;

[0043] Figure 7is a structural schematic diagram of a sample analysis system shown in an embodiment of the present application;

[0044] Figure 8 It is a schematic diagram of the structure of an electronic device shown in an embodiment of the present application. DETAILED DESCRIPTION

[0045] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0046] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0047] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0048] In the related art, since the quality control process of the equipment in the time period of the data in the training data set is not carried out in real time, unreasonable selection of the training data set may occur, which in turn leads to unreasonable calculation of the control limit.

[0049] In response to the above problems, an embodiment of the present application provides a method for selecting a training data set for a quality control model, which can make the setting of upper and lower control limits more reasonable and improve the sensitivity and accuracy of PBRTQC.

[0050] The technical solution of the embodiments of the present application is described in detail below with reference to the accompanying drawings.

[0051] Figure 1 It is a flowchart of a method for selecting a training data set for a quality control model shown in an embodiment of the present application.

[0052] See also Figure 1 , the training data set is used to train the quality control model of the sample analysis system, the method comprising:

[0053] Step 101, obtaining historical test data of a detection device of a sample analysis system for a specified test item within a preset time, where the historical test data has a first preset number;

[0054] PBRTQC is a quality control method that uses the test results of patient clinical specimens to monitor the analytical performance of the test process in real time and continuously. In related technologies, it is necessary to obtain the mean and standard deviation of the calculated values ​​based on the assumption that the test device is in a controlled state within a preset time period, and to calculate the corresponding control limits based on the normal distribution or other distribution characteristics. However, in actual application, since the quality control of the test device in the time period of the data in the training data set is not real-time, even if it is assumed that the test device is in a controlled state within this time period, there may still be unreasonable selection of the training data set due to factors such as replacement of reagent batches, recalibration of the test device, and changes in the precision of the test device, which may lead to unreasonable calculation of control limits.

[0055] In the embodiment of the present application, it is possible to not rely on the assumption that the detection device is in a controlled state during the time period of the training data set, and to obtain a first preset amount of historical detection data generated by the detection device through specified detection items within a preset time as the training data set.

[0056] In this application, historical test data can be understood as the test data of the sample analysis system for the blood sample. "Blood sample" or "test sample" should not be understood as a quality control product or calibrator with a test standard prepared through a specific preparation procedure, but refers to a blood sample that needs to be tested and the concentration of the substance to be tested is unknown before the test.

[0057] In one example, it may be specified that within a preset time, when a historical test is selected for the first time, the first 1000 test results of the current test device arranged in order of test time are used as the historical test.

[0058] Step 102, continuously selecting a second preset amount of historical detection data from the historical detection data based on a preset observation window;

[0059] In an embodiment of the present application, a second preset number of historical detection data can be continuously selected by setting an observation window, and the number of historical detection data selected each time can be controlled by setting the size of the observation window. In one example, the number of data selected by the observation window can be set to 50, that is, 50 historical detection data are selected from the first preset number of historical detection data each time in the order of test time, and the data selected for the first time can be the historical detection data arranged in the order of test time from the 1st to the 50th, and the data selected for the second time can be the historical detection data arranged in the order of test time from the 2nd to the 51st, and so on, until the last data in the historical detection data is selected. It can be understood by those skilled in the art that the number of data selected by the observation window in the embodiment of the present application is only an example, and those skilled in the art can set the number of data selected by the observation window according to actual needs, and the present application does not limit this.

[0060] Step 103, calculating a detection quality evaluation value for the historical detection data in each observation window, and evaluating the trend value and / or imprecision of the historical detection data in each observation window according to the detection quality evaluation value;

[0061] For the second preset number of historical detection data selected each time in the observation window, a corresponding detection quality evaluation value can be calculated, and the trend value, or imprecision, or the trend value and imprecision of the second preset number of historical detection data selected each time in the observation window can be evaluated according to the detection quality evaluation value. Among them, the trend value can reflect the data distribution trend of the historical detection data, and the imprecision can reflect the fluctuation of the historical detection data.

[0062] Step 104 , selecting historical detection data of the observation window whose trend value and / or imprecision meet the statistical standard to form a training data set.

[0063] When at least one of the trend value and the imprecision meets the statistical standard, the historical data of the observation window is used to generate a training data set. If the trend value meets the statistical standard, it means that the detection quality evaluation value of the first preset number of historical detection data obtained has no trend, and the deviation value of the detection quality evaluation value is very small. If the imprecision meets the statistical standard, it means that the detection quality evaluation value of the first preset number of historical detection data obtained has a small volatility.

[0064] An embodiment of the present application discloses a method for selecting a training data set for a quality control model, by obtaining historical detection data of a specified detection item within a preset time of a detection device, the historical detection data having a first preset number, continuously selecting a second preset number of historical detection data from the historical detection data based on a preset observation window, calculating a detection quality evaluation value for the historical detection data in each observation window, evaluating the trend value and / or imprecision of the historical detection data in each observation window based on the detection quality evaluation value, selecting the historical detection data of the observation window whose trend value and / or imprecision meets statistical standards to form a training data set, thereby selecting a training data set according to the actual situation of the detection device, and not relying on the assumption that the detection device is in a controlled state during the time of the training data set, thereby improving the sensitivity and accuracy of PBRTQC.

[0065] Figure 2 It is another flowchart of a method for selecting a training data set for a quality control model shown in an embodiment of the present application.

[0066] See also Figure 2 , the training data set is used to train the quality control model of the sample analysis system, the method comprising:

[0067] Step 201, obtaining historical detection data of a detection device for a specified detection item within a preset time, where the historical detection data has a first preset number;

[0068] In the embodiment of the present application, it is possible to obtain a first preset amount of historical detection data generated by the detection device through specified detection items within a preset time without relying on the assumption that the detection device is in a controlled state during the time period of the training data set.

[0069] In one example, it may be specified that within a preset time, when historical detection data is selected for the first time, the first 1000 detection results of the current detection device arranged in order of test time are used as historical detection data.

[0070] Step 202, truncating a first preset amount of historical detection data;

[0071] After obtaining the first preset amount of historical detection data, the historical detection data can be truncated to discard the data that is particularly high or low in the historical detection data. In general, data that is too high or too low can be considered as abnormal data, which will affect subsequent processing. Therefore, the abnormal data needs to be truncated first and removed from the historical detection data.

[0072] In an embodiment of the present application, a preset quantile may be set, and the historical detection data may be truncated according to the preset quantile. The preset quantile may be selected according to the previous experience of those skilled in the art. In one example, the 5th percentile and the 95th percentile of the historical detection data may be selected, and the historical detection data less than the 5th percentile and greater than the 95th percentile may be truncated, thereby eliminating abnormal data that is particularly high or low from the historical detection data.

[0073] Step 203, converting the truncated historical detection data into normal distribution data.

[0074] After the historical detection data is truncated, the truncated historical detection data can be normally transformed, thereby converting the truncated historical detection data into normally distributed data, so that the data as a whole presents a state of approximate normal distribution, which is convenient for subsequent data processing.

[0075] In one example, the Box-Cox normal transformation can be used. The Box-Cox normal transformation is a data transformation commonly used in statistical modeling, which is used when the continuous response variable does not satisfy the normal distribution. After the Box-Cox normal transformation, the correlation between the unobservable error and the predictor variable can be reduced to a certain extent. The main operation is to transform the dependent variable so that the transformed dependent variable has a linear dependence on the regression independent variable, the error also obeys the normal distribution, and the error components are equal variance and independent of each other. Using the Box-Cox normal transformation can make the dependent variable obtain some properties, such as stationarity in time series analysis, or make the dependent variable distribution normally distributed. It can be understood that the present application does not limit the method of normal transformation, and those skilled in the art can choose a suitable transformation method according to the actual situation, such as logarithmic transformation, inverse transformation, etc.

[0076] Step 204, continuously selecting a second preset amount of historical detection data from the historical detection data based on a preset observation window;

[0077] In an embodiment of the present application, a second preset number of historical detection data can be continuously selected by setting an observation window, and the number of historical detection data selected each time can be controlled by setting the size of the observation window. In one example, the observation window can be set to 50, that is, 50 sample data are selected from the normally distributed data each time in the order of the test time. The data selected for the first time can be the sample data arranged in the order of the test time from the 1st to the 50th, and the data selected for the second time can be the sample data arranged in the order of the test time from the 2nd to the 51st, and so on, until the last sample data in the training data set is selected. It can be understood by those skilled in the art that the number of data selected by the observation window in the embodiment of the present application is only an example, and those skilled in the art can set the number of data selected by the observation window according to actual needs, and the present application does not limit this.

[0078] Step 205, calculating a detection quality evaluation value for the historical detection data in each observation window, and evaluating the trend value and / or imprecision of the historical detection data in each observation window according to the detection quality evaluation value;

[0079] For the second preset number of historical detection data selected each time in the observation window, a corresponding detection quality evaluation value can be calculated, and the trend value, or imprecision, or the trend value and imprecision of the second preset number of historical detection data selected each time in the observation window can be evaluated according to the detection quality evaluation value. Among them, the trend value can reflect the data distribution trend of the historical detection data, and the imprecision can reflect the fluctuation of the historical detection data.

[0080] In an optional embodiment of the present application, the detection quality evaluation value includes a moving average of a second preset number of historical detection data selected by the observation window.

[0081] In an embodiment of the present application, the detection quality assessment value includes the moving average MA of the second preset number of historical detection data selected each time by the observation window. The moving average is the moving average value of the time series data in a specified time period. In an embodiment of the present application, after each normal distribution data is selected through the observation window, the moving average of the selected data can be calculated.

[0082] In an optional embodiment of the present application, step 205 includes:

[0083] Sub-step S11, calculating the average value and standard deviation of the moving average;

[0084] The mean is a quantity that represents the trend of a set of data. It refers to the sum of all the data in a set of data divided by the number of data in this set. The mean is an indicator that reflects the trend of a data set. The standard deviation is the arithmetic square root of the arithmetic mean of the squares of the deviations from the mean (i.e., the variance). The standard deviation is also called the standard deviation, or the experimental standard deviation, and is most commonly used in probability statistics as a measure of the degree of statistical distribution. In statistical work, the mean and standard deviation are the two most important measurements that describe the central tendency and dispersion of data.

[0085] After calculating the moving average of the second preset number of historical detection data selected in each observation window, the average value and standard deviation of all moving averages may be calculated.

[0086] Sub-step S12, evaluating the trend value and / or imprecision of the historical detection data in each observation window according to the moving mean, the average value and the standard deviation.

[0087] The trend value, imprecision, or trend value and imprecision of the second preset number of historical detection data selected each time in the observation window can be evaluated based on the moving average, the mean value and the standard deviation.

[0088] In an optional embodiment of the present application, sub-step S12 includes:

[0089] Sub-step S121, generating a trend value evaluation equation based on the moving mean and the average value, and calculating the slope and intercept of the trend value evaluation equation;

[0090] The trend value evaluation equation can be generated based on the moving mean and the average value, and the slope and intercept of the trend value evaluation equation can be calculated at the same time. The trend value evaluation equation can be used to evaluate the trend value.

[0091] In one example, the trend value evaluation equation may be a linear regression equation. Linear regression is a statistical analysis method that uses regression analysis in mathematical statistics to determine the quantitative relationship of mutual dependence between two or more variables.

[0092] When performing linear regression on the moving mean, first calculate the percentage deviation between the moving mean and the average value. Secondly, determine the rank corresponding to the moving mean, which is the order in which the second preset number of historical detection data is selected for the observation window corresponding to the calculated moving mean. In one example, the rank of the moving mean of the 1st to 50th sample data selected for the first time is 1, the rank of the moving mean of the 2nd to 51st sample data selected for the second time is 2, and so on. Perform linear regression with rank as the independent variable X and percentage deviation as the dependent variable Y to obtain the linear regression equation.

[0093] Sub-step S122, evaluating the trend value of the historical detection data in each observation window according to the slope and the intercept; and / or

[0094] Imprecision was estimated based on the mean and standard deviation.

[0095] The trend value of the historical detection data in the observation window can be evaluated according to the slope and intercept of the trend value evaluation equation. The imprecision of the historical detection data in the observation window can be evaluated according to the mean value and standard deviation.

[0096] In one example, the slope can be evaluated by a t-test P value and a 95% confidence interval for the slope, and the intercept can be evaluated by a t-test P value and a 95% confidence interval for the intercept.

[0097] If the historical detection data has no trend, the linear regression equation is a straight line parallel to the x-axis, and the slope is 0. In an embodiment of the present application, when the 95% confidence interval of the slope contains 0, the slope can be regarded as 0. The intercept can be regarded as the average of the percentage deviation number. If the historical detection data has no trend, the percentage deviation number fluctuates around 0. When the 95% confidence interval of the intercept contains 0, it means that the percentage deviation fluctuates around 0. If the t-test p-values ​​of the slope and the intercept are both greater than the test threshold, and the 95% confidence intervals of the slope and the intercept both contain 0, it means that the trend value of the first preset number of historical detection data meets the statistical standard. In one example, the test threshold can be set to 0.05. This application does not limit the value of the test threshold, and those skilled in the art can set it according to actual needs.

[0098] Reference Figure 3 , this figure does not evaluate the trend value of historical detection data. Figure 3 It can be clearly seen that the data in the previous short period of time within this period were abnormally high, but the quality control chart showed that the detection device was under control during the entire period. This was due to the unreasonable selection of historical detection data and the trend of the data set, which led to the control limit being set too wide. The linear regression results are shown in Table 1.

[0099]

[0100] Table 1

[0101] Reference Figure 4 , the scheme in the embodiment of the present application is adopted, and the historical detection data of the graph is evaluated by trend value, and the problem of abnormally high data in the previous short period of time in the period can be quickly found. The linear regression results are shown in Table 2.

[0102]

[0103]

[0104] Table 2

[0105] Depend on Figure 3 and Figure 4 From the comparison, it can be seen that when there is an unrecognized trend in the historical test data, it means that the test quality of the sample analysis system at this time may be within the quality control range, but it is not the optimal and most stable test quality, and there may be a risk of out-of-control. If this trend is not identified, the data of the test quality deviation will affect the control limit selection of the PBRTQC model, which will directly lead to inaccurate identification of out-of-control test quality.

[0106] Therefore, with the training data set selection method of the present application, even if the quality control display of the sample analysis system is always in a controlled state, it is possible to well determine whether there is a trend of out-of-control detection quality in the historical data, thereby selecting historical detection data with more stable detection quality to form a training data set. Therefore, with the method of the present application, a better quality training data set can be selected for training the PBRTQC model, and then more accurate detection quality out-of-control identification can be achieved when the sample analysis system performs PBRTQC quality control.

[0107] In an optional embodiment of the present application, sub-step S122 includes:

[0108] Sub-step S1221, calculating the coefficient of variation of the moving mean based on the mean value and the standard deviation;

[0109] When you need to compare the degree of dispersion of two sets of data, if the measurement scales of the two sets of data are too different, or the data dimensions are different, it is not appropriate to use the standard deviation directly for comparison. At this time, the influence of the measurement scale and dimension should be eliminated, and the coefficient of variation (CV) can do this. It is the ratio of the standard deviation of the original data to the mean of the original data. Generally speaking, the higher the average level of the variable value, the larger the measurement value of its dispersion, and vice versa.

[0110] The coefficient of variation can be calculated from the mean and standard deviation. The coefficient of variation can be calculated by dividing the standard deviation by the mean.

[0111] Sub-step S1222, when the coefficient of variation is not greater than a preset threshold, determining that the imprecision of the historical detection data in the observation window meets the statistical standard.

[0112] The preset threshold of the coefficient of variation can be determined based on the experience of those skilled in the art. In one example, 20% can be selected. When the coefficient of variation is not greater than 20%, it is determined that the imprecision of the historical detection data in the observation window meets the statistical standard.

[0113] Reference Figure 5The figure does not make any imprecision judgment on the historical test data. It can be clearly seen from the figure that there are periods of large MA fluctuations in the historical test data, indicating that the historical test data has abnormal imprecision. After calculation, the coefficient of variation of the MA of the historical test data is 24.56%.

[0114] Reference Figure 6 According to the embodiment of the present application, the preset threshold of the MA coefficient of variation of the historical detection data is set. As can be seen from the figure, the imprecision anomaly within the period can be discovered more quickly, thereby improving the sensitivity of the PBRTQC method.

[0115] Step 206 , selecting historical detection data of the observation window whose trend value and / or imprecision meet the statistical standard to form a training data set.

[0116] When at least one of the trend value and the imprecision meets the statistical standard, the historical data of the observation window is used to generate a training data set. If the trend value meets the statistical standard, it means that the detection quality evaluation value of the first preset number of historical detection data obtained has no trend, and the deviation value of the detection quality evaluation value is very small. If the imprecision meets the statistical standard, it means that the detection quality evaluation value of the first preset number of historical detection data obtained has a small volatility.

[0117] Compared with trend, imprecision describes the volatility in test quality. It reflects the change in test quality between different periods of time. Therefore, it is also of great significance to the test quality. When selecting a training data set, test data with as little volatility as possible should be selected so that the selection of control limits can be more accurate.

[0118] In an optional embodiment of the present application, the method further includes:

[0119] When the trend value and / or imprecision do not meet the statistical standard, the historical detection data is postponed by a second preset number of data in chronological order, and the step of obtaining the historical detection data of the specified detection item within the preset time of the detection device is returned.

[0120] When at least one of the trend value and the imprecision does not meet the statistical standard, it means that the selection of historical test data is unreasonable, and the historical test data for the specified test items within the preset time of the detection device can be re-acquired. At this time, the second preset number of data can be postponed in chronological order. In one example, if the historical test data of the 1st to 1000th cases are selected for the first time, when at least one of the trend value and the imprecision does not meet the statistical standard, 50 samples can be shifted backward and return to step 201, that is, the test results of the 51st to 1050th cases in chronological order are used as historical test data, and the above process is repeated until the acquired historical test data meets the requirements.

[0121] In an optional embodiment of the present application, after step 206, the following steps are further included:

[0122] The control limits of the quality control model are calculated based on the mean value and standard deviation. The control limits are used by the quality control model to test samples for specified test items.

[0123] Control limit refers to the control range specified when implementing quality control procedures for analytical tests, including upper control limit and lower control limit. The reasonable selection of upper and lower control limits is of great significance to the test results of the controlled equipment. If the upper and lower control limits are set too wide, the test results may fall within the range of the upper and lower control limits, and abnormal results cannot be detected, resulting in missed alarms; if the upper and lower control limits are set too narrow, normal results may also exceed the upper and lower control limits, causing false alarms and increasing the false alarm rate.

[0124] After generating the training data set, the control limits of the quality control model can be calculated based on the mean value and the standard deviation. In one example, the upper and lower control limits of the control limits of the quality control model are calculated by adding or subtracting N times the standard deviation from the mean value; wherein N is a positive integer.

[0125] Those skilled in the art will appreciate that the value of N in the embodiments of the present application can be set according to actual needs, and the present application does not impose any limitation on this.

[0126] By improving the selection rules of the training data set, we can ensure that the calculated control limits are more reasonable and more applicable to the test items corresponding to the training data set. When the quality control model detects the patient's sample data in real time, it can detect abnormal data more sensitively and accurately.

[0127] An embodiment of the present application discloses a method for selecting a training data set for a quality control model, by obtaining historical detection data of a specified detection item within a preset time of a detection device, the historical detection data having a first preset number, truncating the historical detection data of the first preset number, converting the truncated historical detection data into normally distributed data, continuously selecting a second preset number of historical detection data from the historical detection data based on a preset observation window, calculating a detection quality assessment value for the historical detection data in each observation window, evaluating the trend value and / or imprecision of the historical detection data in each observation window based on the detection quality assessment value, selecting the historical detection data of the observation window whose trend value and / or imprecision meets statistical standards to form a training data set, thereby selecting a training data set according to the actual situation of the detection device, and not relying on the assumption that the detection device is in a controlled state during the time of the training data set, thereby improving the sensitivity and accuracy of PBRTQC.

[0128] Corresponding to the aforementioned application function implementation method embodiment, the present application also provides a sample analysis system, an electronic device and corresponding embodiments.

[0129] Figure 7 It is a structural schematic diagram of a sample analysis system shown in an embodiment of the present application.

[0130] See also Figure 7 The sample analysis system includes a detection device for detecting a blood sample, and also includes a quality control model for the detection quality, the system comprising:

[0131] An acquisition module 701 is used to acquire a test sample;

[0132] In the present application, the collection time of the first preset amount of historical test data may be completely unrelated to the collection time of the test sample of the acquisition module 701. For example, the test sample is the patient sample data collected in real time by the sample analysis system. The historical data is the historical test data of the sample analysis system for 90 days starting from yesterday.

[0133] The analysis module 702 is used to analyze the test sample through the quality control model; wherein the quality control model is obtained by training the training data set; the training data set is obtained by obtaining the historical test data of the test device for the specified test item within a preset time, and the historical test data has a first preset number; based on the preset observation window, a second preset number of historical test data is continuously selected from the historical test data; the test quality evaluation value is calculated for the historical test data in each observation window, and the trend value and / or imprecision of the historical test data in each observation window is evaluated according to the test quality evaluation value; the historical test data of the observation window whose trend value and / or imprecision meets the statistical standard is selected to form;

[0134] After obtaining the test sample, the test sample can be analyzed by the quality control model in the sample analysis system, wherein the quality control model is trained by a training data set, and the training data set is formed in the following manner: obtaining historical test data for a specified test item within a preset time of the detection device, the historical test data having a first preset number; continuously selecting a second preset number of historical test data from the historical test data based on a preset observation window; calculating a test quality evaluation value for the historical test data in each observation window, and evaluating the trend value and / or imprecision of the historical test data in each observation window based on the test quality evaluation value; and selecting the historical test data of the observation window whose trend value and / or imprecision meets statistical standards.

[0135] A generating module 703 is used to generate an analysis result corresponding to the test sample;

[0136] After the test sample is analyzed through the quality control model, an analysis result corresponding to the test sample can be generated.

[0137] In an embodiment of the present application, the sample analysis system is used to detect the sampled object's body fluid (such as blood) sample obtained in real time, mainly to detect specific markers in the blood. The sample analysis system can be applied to analytical instruments such as biochemical analyzers, chemiluminescent immunoassay analyzers, fluorescent immunoassay analyzers, immunoturbidimetric analyzers, biochemical immunoassay machines, and gene sequencers. The sample analyzer may include a detection device for detecting a blood sample, a sample compartment for placing a blood sample, a reagent compartment for placing a detection reagent, and other necessary components.

[0138] The sample analysis system can also be composed of multiple sample analyzers connected. Multiple sample analyzers can be managed by an information management system to uniformly manage sample detection related processes. In addition, in addition to the sample analyzer, the sample analysis system can also include a remote analysis platform, such as a cloud platform. When connected to a cloud platform, the sample analyzer in the sample analysis system transmits the detection data to the cloud platform and can receive various instructions from the cloud platform, such as maintenance, quality control, and calibration instructions. Therefore, the cloud platform can also perform detection quality control and then execute the method of the present application.

[0139] A database is provided in the sample analysis system, and the database can store historical test data that has been tested and analyzed. If historical test data is needed in the subsequent data processing process, it can be retrieved from the database.

[0140] The sample analysis system also includes a quality control model, which is mainly used to analyze the test sample, thereby performing quality inspection on the test sample. A processor is provided in the analysis instrument, and the specific operation of each module in the sample analysis system can be executed by the processor.

[0141] The embodiment of the present application discloses a sample analysis system, which obtains test samples and analyzes the test samples through a quality control model; wherein the quality control model is obtained by training a training data set; the training data set is obtained by obtaining historical test data of a specified test item within a preset time of a test device, and the historical test data has a first preset number; a second preset number of historical test data is continuously selected from the historical test data based on a preset observation window; a test quality assessment value is calculated for the historical test data in each observation window, and the trend value and / or imprecision of the historical test data in each observation window is evaluated based on the test quality assessment value; historical test data of the observation window whose trend value and / or imprecision meets the statistical standard are selected to form an analysis result corresponding to the test sample, so that the quality control model of the sample analysis system can be trained based on the training data set selected according to the actual situation of the test device, without relying on the assumption that the test device is in a controlled state during the time period of the training data set, thereby improving the sensitivity and accuracy of PBRTQC.

[0142] Regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated again here.

[0143] Figure 8 It is a schematic diagram of the structure of an electronic device shown in an embodiment of the present application.

[0144] See also Figure 8 , the electronic device 800 includes a memory 810 and a processor 820 .

[0145] The processor 820 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0146] The memory 810 may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage devices. Among them, ROM can store static data or instructions required by the processor 1020 or other modules of the computer. The permanent storage device may be a readable and writable storage device. The permanent storage device may be a non-volatile storage device that does not lose the stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (such as a magnetic or optical disk, flash memory) as a permanent storage device. In some other embodiments, the permanent storage device may be a removable storage device (such as a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory may store some or all instructions and data required by the processor at run time. In addition, the memory 810 may include any combination of computer-readable storage media, including various types of semiconductor storage chips (such as DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks may also be used. In some embodiments, the memory 810 may include a readable and / or writable removable storage device, such as a laser disc (CD), a read-only digital versatile disc (e.g., DVD-ROM, double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (e.g., SD card, mini SD card, Micro-SD card, etc.), a magnetic floppy disk, etc. The computer-readable storage medium does not include carrier waves and transient electronic signals transmitted wirelessly or wired.

[0147] The memory 810 stores executable codes, and when the executable codes are processed by the processor 820 , the processor 820 can execute part or all of the methods described above.

[0148] In addition, the method according to the present application may also be implemented as a computer program or a computer program product, which includes computer program code instructions for executing some or all of the steps in the above method of the present application.

[0149] Alternatively, the present application can also be implemented as a computer-readable storage medium (or non-transitory machine-readable storage medium or machine-readable storage medium) on which executable code (or computer program or computer instruction code) is stored. When the executable code (or computer program or computer instruction code) is executed by a processor of an electronic device (or server, etc.), the processor executes part or all of the steps of the above-mentioned method according to the present application.

[0150] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for selecting a training data set for a quality control model, It is characterized in that The training data set is used to train a quality control model of a sample analysis system, and the method comprises: Acquire historical test data of a detection device of a sample analysis system for a specified test item within a preset time, wherein the historical test data has a first preset number; Continuously selecting a second preset number of historical detection data from the historical detection data based on a preset observation window; Calculating a detection quality evaluation value for the historical detection data in each observation window, and evaluating a trend value and / or imprecision of the historical detection data in each observation window according to the detection quality evaluation value; The historical detection data of the observation window whose trend value and / or imprecision meet the statistical standard are selected to form the training data set.

2. The method according to claim 1, It is characterized in that The step before continuously selecting a second preset amount of historical detection data from the historical detection data based on the preset observation window includes: Truncating the first preset amount of historical detection data; The truncated historical detection data are transformed into normal distribution data.

3. The method according to claim 1, It is characterized in that The detection quality evaluation value includes a moving average of the second preset number of historical detection data selected by the observation window.

4. The method according to claim 3, It is characterized in that The step of evaluating the trend value and / or imprecision of the historical detection data in each observation window according to the detection quality evaluation value includes: Calculating the mean and standard deviation of the moving average; The trend value and / or imprecision of the historical detection data in each observation window is evaluated based on the moving average, the average value and the standard deviation.

5. The method according to claim 4, It is characterized in that The step of evaluating the trend value and / or imprecision of the historical detection data in each observation window based on the moving mean, the average value and the standard deviation includes: generating a trend value evaluation equation based on the moving mean and the average value, and calculating the slope and intercept of the trend value evaluation equation; Evaluate the trend value of the historical detection data in each observation window according to the slope and the intercept; and / or The imprecision is estimated based on the mean and the standard deviation.

6. The method according to claim 5, It is characterized in that The step of evaluating the imprecision based on the mean value and the standard deviation comprises: Calculate the coefficient of variation of the moving mean based on the mean and the standard deviation; When the coefficient of variation is not greater than a preset threshold, it is determined that the imprecision of the historical detection data in the observation window meets the statistical standard.

7. The method according to claim 1, It is characterized in that The method further comprises: When the trend value and / or imprecision do not meet the statistical standards, the historical test data is postponed by a second preset number of data in chronological order, and the step of obtaining the historical test data of the specified test item within the preset time of the detection device of the sample analysis system is returned.

8. A sample analysis system, comprising a detection device for detecting a blood sample, It is characterized in that The sample analysis system includes a quality control model for detection quality, and the sample analysis system also includes: An acquisition module, used to acquire test samples; An analysis module is used to analyze the detection quality of the detection sample by the sample analysis system through the quality control model; wherein the quality control model is obtained by training a training data set; the training data set is formed by the following method: obtaining historical detection data of the detection device of the sample analysis system for a specified detection item within a preset time, the historical detection data having a first preset number; continuously selecting a second preset number of historical detection data from the historical detection data based on a preset observation window; calculating a detection quality evaluation value for the historical detection data in each observation window, and evaluating the trend value and / or imprecision of the historical detection data in each observation window according to the detection quality evaluation value; and selecting the historical detection data of the observation window whose trend value and / or imprecision meets the statistical standard; A generation module is used to generate analysis results corresponding to the detection sample.

9. An electronic device, It is characterized in that include: processor; as well as A memory having executable codes stored thereon, which, when executed by the processor, causes the processor to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having executable codes stored thereon, which, when executed by a processor of an electronic device, causes the processor to execute the method according to any one of claims 1 to 7.

Images