A blood analysis equipment detection performance evaluation method based on big data

Abstract

The application provides a blood analysis equipment detection performance evaluation method based on big data, which comprises the following steps: collecting the background count error, precision and analysis frequency data of multiple parameters of the blood analysis equipment under a selected detection mode; constructing a background count error prediction model based on a BP neural network and a precision prediction model based on a BP neural network; training the BP neural networks of the background count error prediction model and the precision prediction model; collecting real-time data of the blood analysis equipment to be evaluated and inputting the real-time data into the trained background count error prediction model and precision prediction model respectively to evaluate the performance of the blood analysis equipment in a future period. The application can predict and evaluate the performance of the blood analysis equipment in a future period, which can ensure the operation accuracy of the equipment, reduce the time and frequency of equipment calibration, and improve the operation stability of the equipment.

Description

Technical Field

[0001] This invention relates to the field of blood analysis equipment performance evaluation technology, and in particular to a blood analysis equipment performance evaluation method based on big data. Background Technology

[0002] With the development of medical science and technology, blood analysis equipment has been widely used in the fields of medical testing and disease diagnosis. The accuracy of blood tests is related to the precision of disease diagnosis. Therefore, ensuring the accuracy of blood analysis equipment is an important part of improving disease diagnosis. In order to ensure the accuracy of blood analysis equipment, existing blood analysis equipment needs to be evaluated regularly for its testing performance. When the testing accuracy is lower than the preset threshold, the design needs to be readjusted.

[0003] As blood analysis equipment operates for longer periods and performs more tests, it ages, resulting in fewer stable tests after a single calibration. Current performance evaluations of blood analysis equipment rely on experience to adjust evaluation intervals, which cannot be adjusted based on the actual operating conditions of the equipment. Furthermore, calibration of blood analysis equipment is performed only when its detection accuracy exceeds a preset threshold, making it impossible to predict and evaluate the equipment's operating status. Summary of the Invention

[0004] The purpose of this invention is to provide a method for evaluating the detection performance of blood analysis equipment based on big data.

[0005] The objective of this invention is achieved through the following technical solution, the specific steps of which are as follows:

[0006] 1) Data acquisition: Under the selected detection mode, the baseline counting error and precision of multiple parameters of the blood analysis device are collected at time intervals of Δt, as well as the number of analyses performed by the analysis device within the sampling period, to obtain the collected dataset;

[0007] 2) Data preprocessing: The collected dataset is randomly divided into training and test sets according to a set ratio;

[0008] 3) Model Construction: Construct a background counting error prediction model based on BP neural network and a precision prediction model based on BP neural network respectively;

[0009] 4) Training Construction: Input the training set data divided in step 2) into the background count error prediction model and the precision prediction model constructed in step 3) respectively, and train the BP neural networks of the background count error prediction model and the precision prediction model respectively.

[0010] 5) Performance evaluation: Collect real-time data from the blood analysis device to be evaluated and input it into the baseline count error prediction model and precision prediction model trained in step 4) respectively to evaluate the performance of the blood analysis device at a future time period.

[0011] Furthermore, the specific steps for data collection in step 1) are as follows:

[0012] 2-1) Under the selected detection mode, the background counting errors of five items in the blood analyzer were measured at time intervals of Δt: white blood cell count (WBC), red blood cell count (RBC), hemoglobin (HGB), hematocrit (HCT), and platelet count (PLT). Data is collected, where j∈(1~5) represents 5 items respectively, and t is the sampling time;

[0013] 2-2) Under the selected detection mode, the precision CV of five items of the blood analyzer were measured at time intervals of Δt: white blood cell count (WBC), red blood cell count (RBC), hemoglobin (HGB), mean corpuscular volume (MCV), and platelet count (PLT). t j Data is collected, where j∈(1~5) represents 5 items respectively, and t is the sampling time;

[0014] 2-3) Collect the number of blood analysis cycles performed by the blood analysis equipment within each sampling period, N = {n1, n2, ..., n}. i ,...,n m}, where: m is the maximum number of samples, n i The number of analyses performed by the blood analyzer during the i-th sampling period.

[0015] Furthermore, the specific steps for constructing the model in step 3) are as follows:

[0016] 3-1) Construct background counting error prediction models for white blood cell count (WBC), red blood cell count (RBC), hemoglobin (HGB), hematocrit (HCT), and platelet count (PLT), respectively:

[0017] The background counting error at the three sampling times t-3△t, t-2△t, and t-△t And the number of analyses (n) performed by the blood analyzer within the sampling period corresponding to the three sampling times. t-3△t n t-2△t n t-△t As input, the baseline counting error at time t. To produce the output, a background counting error prediction model based on a BP neural network is constructed.

[0018] 3-2) Construct precision prediction models for white blood cell count (WBC), red blood cell count (RBC), hemoglobin (HGB), mean corpuscular volume (MCV), and platelet count (PLT), respectively:

[0019] Precision at three sampling times: t-3△t, t-2△t, and t-△t And the number of analyses (n) performed by the blood analyzer within the sampling period corresponding to the three sampling times. t-3△t n t-2△t n t-△t As input, the precision CV at time t t j To produce the output, a precision prediction model based on a BP neural network is constructed.

[0020] 3-3) The BP neural networks of the five background count error prediction models and the five precision prediction models all have 6 input layer variables, 1 output layer variable, and 12 hidden layer neurons.

[0021] Initialize the parameters of the BP neural network, including the weights between layers, the threshold of the hidden layer, the learning accuracy, the maximum number of iterations, and the learning rate.

[0022] Furthermore, the specific steps for constructing the model in step 4) are as follows:

[0023] The BP neural networks of the five baseline count error prediction models and the five precision prediction models are trained using the training set data divided in step 2). After each iteration of training, the error between the actual output and the expected output in the two training processes is calculated.

[0024] If the current error is greater than the learning accuracy, then update the weights and threshold parameters of the BP neural network;

[0025] If the current error is less than or equal to the learning accuracy or the current iteration count is greater than or equal to the maximum iteration count, then output the weights and threshold parameters of the current BP neural network as the trained BP neural network.

[0026] Furthermore, the specific method for performance evaluation in step 5) is as follows:

[0027] 5-1) Obtain the time t to be evaluated f The first 3 sampling periods at time t f -3△t、t f -2△t、t f -△t's baseline recording error and precision And the number of analyses performed by the blood analyzer within the sampling period corresponding to the three sampling times. Obtain real-time data;

[0028] 5-2) Input the real-time data from step 5-1) into the 10 BP neural network models trained in step 4), respectively, at the time t to be evaluated. f Background counting error and precision Make predictions;

[0029] 5-3) If the predicted background count error in step 5-2) and precision If all values ​​are less than the preset threshold, the blood analysis equipment will continue to be used; if the baseline counting error is... and precision If any of the preset thresholds for rainfall is set, then in t f The blood analysis equipment is calibrated at time -Δt.

[0030] 5-4) Calculate t f -2△t、t f - Rate of change of background counting error at time Δt and precision change rate Obtain the mean rate of change of the baseline counting error and the mean of the rate of change of precision

[0031]

[0032]

[0033] Based on the average rate of change of the baseline counting error and the mean of the rate of change of precision Blood analysis equipment is divided into different lifecycle stages, and the sampling and evaluation time intervals are adjusted according to these stages.

[0034] like The blood analysis equipment is in the initial stage of its lifespan, during which the equipment operates very stably and the sampling evaluation time interval is equal to the initial sampling interval Δt.

[0035] like The blood analysis equipment is in the early to middle stage of its lifespan. During this stage, the equipment operates relatively stably, and the time interval for sampling and evaluation is equal to 0.8Δt of the initial sampling interval.

[0036] like The blood analysis equipment is in the middle of its lifespan. During this stage, the equipment is operating stably, and the time interval for sampling and evaluation is equal to 0.6Δt of the initial sampling interval.

[0037] like The blood analysis equipment is in the middle to late stage of its lifespan. During this stage, the equipment's operating status is relatively unstable. The time interval for sampling and evaluation is equal to 0.5△t of the initial sampling interval.

[0038] like The blood analysis equipment is at the end of its lifespan. During this stage, the equipment's operating status is extremely unstable. The time interval for sampling and evaluation is equal to the initial sampling interval of 0.4Δt.

[0039] in: All of these are preset change rate thresholds based on experience.

[0040] Furthermore, the detection modes in step 1) include three detection modes: automated-whole blood, closed-whole blood, and closed-predilution.

[0041] Because of the adoption of the above technical solution, the present invention has the following advantages:

[0042] 1. This application predicts and evaluates the performance of blood analysis equipment at a future time, which not only ensures the operating accuracy of the equipment, but also reduces the time and frequency of equipment calibration and improves the operating stability of the equipment.

[0043] 2. The operating error of blood analysis equipment is related to the operating time and the number of tests. As the operating time and the number of tests increase, the operating accuracy of the equipment will decrease and the error will increase. This application constructs a BP neural network with the background counting error, the number of tests, and the precision and the number of tests as inputs at the first three time points to predict the background counting error and precision of the equipment, thereby improving the prediction accuracy of the background counting error and precision and providing a theoretical basis for the prediction and evaluation of equipment performance.

[0044] 3. This application divides the blood analysis equipment into different life cycles by calculating the background count error change rate and precision change rate. This reduces the number of times the equipment performance needs to be evaluated during the initial stage of the equipment life cycle, while ensuring the operating accuracy of the equipment at the end of the equipment life cycle.

[0045] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0046] The accompanying drawings of this invention are described below.

[0047] Figure 1 This is a flowchart of the blood analysis device detection performance evaluation method based on big data according to the present invention. Detailed Implementation

[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0049] Example 1:

[0050] like Figure 1 The method for evaluating the detection performance of a blood analysis device based on big data, as shown, includes the following steps:

[0051] 1) Data Acquisition: Under the selected detection mode, the baseline counting error and precision of multiple parameters of the blood analysis device are collected at time intervals of Δt, as well as the number of analyses performed by the analysis device within the sampling period, to obtain the collected dataset. The specific steps are as follows:

[0052] 1-1) Under the selected detection mode, the background counting errors of five items in the blood analyzer were measured at time intervals of Δt: white blood cell count (WBC), red blood cell count (RBC), hemoglobin (HGB), hematocrit (HCT), and platelet count (PLT). Data is collected, where j∈(1~5) represents 5 items respectively, and t is the sampling time;

[0053] 1-2) Under the selected detection mode, the precision CV of five items of the blood analyzer were measured at time intervals of Δt: white blood cell count (WBC), red blood cell count (RBC), hemoglobin (HGB), mean corpuscular volume (MCV), and platelet count (PLT). t j Data is collected, where j∈(1~5) represents 5 items respectively, and t is the sampling time;

[0054] 1-3) Collect the number of blood analysis cycles performed by the blood analysis equipment within each sampling period, N = {n1, n2, ..., n} i ,...,n m}, where: m is the maximum number of samples, n i The number of analyses performed by the blood analyzer during the i-th sampling period.

[0055] In this invention example, a blood analyzer of a certain brand and model BC5380 is used as the sampling object. This blood analyzer includes three detection modes: automatic-whole blood, closed-whole blood, and closed-predilution. With a sampling interval of 10 days, the background count error and precision of multiple parameters of the three detection modes of the equipment are measured before each morning shift. The measurement steps are as follows:

[0056] a) Pre-use confirmation: Determine whether the environmental parameters of the blood analyzer, the supporting reagents, and the equipment tubing meet the equipment requirements. If they do, proceed with the measurement.

[0057] b) Measurement of background counting error: Taking closed-cell whole blood as an example, select the device and click the "Count" function, select the measurement mode of "Closed-cell whole blood" + "CBD + DIFF", and set the sample number. Take a new anticoagulant-free vacuum blood collection tube, fill it with 2ml of clean diluent, open the sample cap and place it in the sample chamber, close the chamber door, click the device's "Count", and perform 3 consecutive measurements. Take the largest result of the 3 measurements as the output data. After the measurement is completed, check the measurement results. If the background counting error meets the parameter threshold in Table 1, then the device meets the background counting requirements.

[0058] Table 1. Baseline Count Error Thresholds under Closed-Whole Blood Mode

[0059]

[0060] c) Precision measurement: Select one fresh venous anticoagulated blood sample with normal concentration as required. The venous anticoagulated blood sample must meet the parameter requirements in Table 2. Mix the test tube, open the cap, put it into the sample chamber, close the chamber door, click "Count" on the device, and repeat the test 10 times. Click "CV" on the device to collect precision data. If the precision meets the parameter threshold in Table 3, then the device meets the precision requirements.

[0061] Table 2 Range of parameters for fresh venous anticoagulated blood specimens

[0062]

[0063] Table 3 Precision thresholds in closed-whole blood mode

[0064]

[0065] d) Analysis frequency statistics: Record the total number of measurements taken by the device within 7 days.

[0066] e) Repeat steps a)-d) to obtain m sets of sampled data.

[0067] 2) Data preprocessing: The collected dataset is randomly divided into training set and test set in a 9:1 ratio.

[0068] 3) Model Construction: Construct a background count error prediction model based on a BP neural network and a precision prediction model based on a BP neural network, respectively. The specific steps are as follows:

[0069] 3-1) Construct background counting error prediction models for white blood cell count (WBC), red blood cell count (RBC), hemoglobin (HGB), hematocrit (HCT), and platelet count (PLT), respectively:

[0070] The background counting error at the three sampling times t-3△t, t-2△t, and t-△t And the number of analyses (n) performed by the blood analyzer within the sampling period corresponding to the three sampling times. t-3△t n t-2△t n t-△t As input, the baseline counting error at time t. To produce the output, a background counting error prediction model based on a BP neural network is constructed.

[0071] 3-2) Construct precision prediction models for white blood cell count (WBC), red blood cell count (RBC), hemoglobin (HGB), mean corpuscular volume (MCV), and platelet count (PLT), respectively:

[0072] Precision at three sampling times: t-3△t, t-2△t, and t-△t And the number of analyses (n) performed by the blood analyzer within the sampling period corresponding to the three sampling times. t-3△t n t-2△t n t-△t As input, the precision CV at time t t j To produce the output, a precision prediction model based on a BP neural network is constructed.

[0073] 3-3) The BP neural networks of the five background count error prediction models and the five precision prediction models all have 6 input layer variables, 1 output layer variable, and 12 hidden layer neurons.

[0074] Initialize the parameters of the BP neural network, including the weights between layers, the threshold of the hidden layer, the learning accuracy, the maximum number of iterations, and the learning rate.

[0075] In this invention, the operating error of the blood analysis device is related to the device's operating time and the number of tests. As the device's operating time and the number of tests increase, the device's operating accuracy decreases and the error increases. This application constructs a BP neural network using the baseline counting error, the number of tests, and the precision and the number of tests as inputs at the first three time points to predict the device's baseline counting error and precision, thereby improving the prediction accuracy of the baseline counting error and precision and providing a theoretical basis for the prediction and evaluation of device performance.

[0076] In this embodiment of the invention, the initial weights of the BP neural network are random numbers between [-1, 1] that follow a normal distribution, the initial biases are all 0, the learning rate is 0.005, and the maximum number of iterations is 200.

[0077] 4) Training Construction: Input the training set data divided in step 2) into the baseline count error prediction model and the precision prediction model constructed in step 3), respectively, and train the BP neural networks of the baseline count error prediction model and the precision prediction model. The specific method is as follows:

[0078] The BP neural networks of the five baseline count error prediction models and the five precision prediction models are trained using the training set data divided in step 2). After each iteration of training, the error between the actual output and the expected output in the two training processes is calculated.

[0079] If the current error is greater than the learning accuracy, then update the weights and threshold parameters of the BP neural network;

[0080] If the current error is less than or equal to the learning accuracy or the current iteration count is greater than or equal to the maximum iteration count, then output the weights and threshold parameters of the current BP neural network as the trained BP neural network.

[0081] In this invention, the gradient descent method is used to update the weights and biases of the BP neural network.

[0082] 5) Performance Evaluation: Real-time data from the blood analysis device to be evaluated is collected and input into the baseline count error prediction model and precision prediction model trained in step 4), respectively, to evaluate the performance of the blood analysis device at a future time period. The specific steps are as follows:

[0083] 5-1) Obtain the time t to be evaluated f The first 3 sampling periods at time t f -3△t、t f -2△t、t f -△t's baseline recording error and precision And the number of analyses performed by the blood analyzer within the sampling period corresponding to the three sampling times. Obtain real-time data;

[0084] 5-2) Input the real-time data from step 5-1) into the 10 BP neural network models trained in step 4), respectively, at the time t to be evaluated. f Background counting error and precision Make predictions;

[0085] 5-3) If the predicted background count error in step 5-2) and precision If all values ​​are less than the preset threshold, the blood analysis equipment will continue to be used; if the baseline counting error is... and precision If any of the preset thresholds for rainfall is set, then in t f The blood analysis equipment is calibrated at time -Δt.

[0086] 5-4) Calculate t f -2△t、t f - Rate of change of background counting error at time Δt and precision change rate

[0087]

[0088]

[0089]

[0090]

[0091] Obtain the mean rate of change of the baseline counting error and the mean of the rate of change of precision

[0092]

[0093]

[0094] Based on the average rate of change of the baseline counting error and the mean of the rate of change of precision Blood analysis equipment is divided into different lifecycle stages, and the sampling and evaluation time intervals are adjusted according to these stages.

[0095] like The blood analysis equipment is in the initial stage of its lifespan, during which the equipment operates very stably and the sampling evaluation time interval is equal to the initial sampling interval Δt.

[0096] like The blood analysis equipment is in the early to middle stage of its lifespan. During this stage, the equipment operates relatively stably, and the time interval for sampling and evaluation is equal to 0.8Δt of the initial sampling interval.

[0097] like The blood analysis equipment is in the middle of its lifespan. During this stage, the equipment is operating stably, and the time interval for sampling and evaluation is equal to 0.6Δt of the initial sampling interval.

[0098] like The blood analysis equipment is in the middle to late stage of its lifespan. During this stage, the equipment's operating status is relatively unstable. The time interval for sampling and evaluation is equal to 0.5△t of the initial sampling interval.

[0099] like The blood analysis equipment is at the end of its lifespan. During this stage, the equipment's operating status is extremely unstable. The time interval for sampling and evaluation is equal to the initial sampling interval of 0.4Δt.

[0100] in: All of these are preset change rate thresholds based on experience.

[0101] In an example of the present invention, The values ​​are 1%, 2%, 3%, and 5%. This application divides the blood analysis equipment into different life cycles by calculating the background count error change rate and precision change rate. In the early stage of the equipment's life cycle, the evaluation sampling cycle is longer, which reduces the frequency of sampling evaluation during the initial stage of the equipment's life cycle and reduces the workload of operators. In the late stage of the equipment's life cycle, the evaluation sampling cycle is shorter, which ensures the accuracy of the equipment and facilitates timely calibration of the equipment's status.

[0102] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0103] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0104] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0105] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the function specified in one or more boxes.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for evaluating performance of a blood analysis device based on big data, the method comprising: The specific steps are: 1) Data collection: in the selected detection mode, collect the background count error, precision of multiple parameters of the blood analysis equipment with a time interval of Δt, and the analysis times of the analysis equipment within the sampling period to obtain the collected data set; 2) Data preprocessing: the collected data set is randomly divided into training set and test set according to the proportion; 3) Model construction: constructing a background count error prediction model based on BP neural network and a precision prediction model based on BP neural network; 4) Model training: input the training set data divided in step 2) into the background count error prediction model and the precision prediction model constructed in step 3) respectively, and train the BP neural network of the background count error prediction model and the precision prediction model respectively; 5) Performance evaluation: collect real-time data of the blood analysis equipment to be evaluated, and input the data into the background count error prediction model and the precision prediction model trained in step 4) respectively to evaluate the performance of the blood analysis equipment in a certain period in the future; The specific steps of data collection in step 1) are: 2-1) In the selected detection mode, the background counting error of the blood analysis equipment for 5 items of white blood cell count WBC, red blood cell count RBC, hemoglobin determination HGB, hematocrit determination HCT, and platelet determination PLT is counted respectively with a time interval of Δt Collecting, wherein: j∈(1~5) respectively represents 5 items, and t is the sampling time; 2-2) Precision CV of 5 items of blood analysis equipment, including white blood cell count WBC, red blood cell count RBC, hemoglobin determination HGB, mean corpuscular volume MCV, and platelet determination PLT, respectively, in a selected detection mode with a time interval of Δt t j Collection is performed, wherein: j ∈ (1 ~ 5) respectively represent 5 items, and t is the sampling time; 2-3) Collect the number of blood analysis cycles performed by the blood analysis equipment within each sampling period, N = {n1, n2, ..., n}. i ,...,n m }, where: m is the maximum number of samples, n i The number of analyses performed by the blood analyzer during the i-th sampling period; The specific method of performance evaluation in step 5) is: 5-1) Acquisition of the instant t to be evaluated f First three sampling instants t f - 3Δt, t f - 2Δt, t f - Background counting error at -Δt And precision And the number of analyses of the blood analysis device in the sampling period corresponding to the three sampling instants Obtaining real-time data; 5-2) input the real-time data in step 5-1) into the 10 BP neural network models trained in step 4) respectively, to predict the background count error and precision at the time t to be evaluated f and precision ​​ 5-3) if the background count error and the precision predicted in step 5-2) are both less than a preset threshold, the blood analysis device continues to be used; if either the background count error and the precision is greater than the preset threshold, a calibration operation is performed on the blood analysis device at time t f - Δt. 5-4) Calculate t f -2Δt, t f - Rate of change of background count error at time Δt and rate of change of precision Obtain mean rate of change of background count error and mean rate of change of precision According to the mean rate of change of background count error and the mean rate of change of precision The blood analysis device is divided into different life cycle stages, and the time interval of sampling evaluation is adjusted according to different life cycle stages: If then the blood analysis device is in its initial phase of life, in which the device state runs very stable and the time interval of the sampling evaluation equals the initial sampling interval Δt; If then the blood analysis device is in the middle of its life, where the device state is more stable, and the time interval for sampling evaluation is equal to the initial sampling interval 0.8Δt; If then the blood analysis device is in the middle of its life, with a stable device state operation, and the time interval of its sampling evaluation is equal to the initial sampling interval 0.6Δt; If then the blood analysis device is in the middle-late stage of its life, which is a less stable stage of operation, and the time interval of the sampling evaluation is equal to the initial sampling interval 0.5Δt; If then the blood analysis device is at the end of its life, in a stage where the device operating conditions are extremely unstable, and the time interval of its sampling evaluation is equal to the initial sampling interval 0.4Δt; wherein: are both empirically predetermined change rate thresholds.

2. The method of claim 1, wherein the method is based on a big data-based blood analysis device detection performance evaluation method, characterized by, The specific steps of model construction in step 3) are: 3-1) Construct a background count error prediction model for white blood cell count WBC, red blood cell count RBC, hemoglobin determination HGB, hematocrit determination HCT, and platelet determination PLT respectively; The background count error at the three sampling time points t-3Δt, t-2Δt and t-Δt And the analysis number n of the blood analysis equipment in the sampling period corresponding to the three sampling time points t-3Δt , n t-2Δt , n t-Δt The background count error at the time point t is taken as input The background count error prediction model based on the BP neural network is constructed 3-2) Construct a precision prediction model for white blood cell count WBC, red blood cell count RBC, hemoglobin determination HGB, mean corpuscular volume MCV, and platelet determination PLT respectively; The precision at the three sampling time points t-3Δt, t-2Δt and t-Δt And the number of analyses n of the blood analysis equipment in the sampling period corresponding to the three sampling time points t-3Δt , n t-2Δt , n t-Δt As input, the precision CV t j At the time t is output, a precision prediction model based on a BP neural network is constructed; 3-3) The input layer variables of the BP neural network of the five background count error prediction models and the five precision prediction models are all six, the output layer variable is one, and the hidden layer neurons are all twelve; Initialize the parameters of the BP neural network, including the weights between layers, the threshold value of the hidden layer, the learning precision, the maximum number of iterations, and the learning rate.

3. The method of claim 1, wherein the method is based on a big data-based blood analysis device detection performance evaluation method, characterized by, The specific steps of model construction in step 4) are: Use the training set data divided in step 2) to train the BP neural network of the five background count error prediction models and the five precision prediction models respectively, and calculate the error between the actual output and the expected output after each iteration training; If the current error is greater than the learning precision, update the weight and threshold value parameters of the BP neural network; If the current error is less than or equal to the learning precision or the current iteration number is greater than or equal to the maximum number of iterations, output the weight and threshold value parameters of the current BP neural network as the trained BP neural network.

4. The method of claim 1, wherein the method is a method of evaluating the performance of a blood analysis device based on big data, characterized by, The detection modes in step 1) include automatic-whole blood, closed-whole blood, and closed-pre-dilution three detection modes.

Images