Data processing method and system applied to flow cytometer

Abstract

The invention discloses a data processing method and system applied to a flow cytometer, and relates to the technical field of data processing. The system comprises an arrangement module, an acquisition module, a screening module, a screening module and a statistics module, the arrangement module is used for obtaining photoacoustic signal data of cells of types to be acquired from a database, processing the photoacoustic signal data, calculating a standard threshold interval and obtaining characteristic screening types; according to the technical key points, the characteristics of to-be-collected type cells are analyzed and processed, a multi-stage screening scheme aiming at the characteristics is formulated, the cells are recognized and judged in a multi-stage screening mode, most of data can be screened out based on characteristic screening, the data size of subsequent data processing is greatly reduced, and the efficiency is improved. Therefore, the equipment cost required by data processing is effectively reduced, the efficiency is greatly improved, the use effect is good, and the method has a good use prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a data processing method and system applied to a flow cytometer. Background Art

[0002] Flow cytometer is a device for automatic analysis and sorting of cells. It can rapidly measure, store and display a series of important biophysical and biochemical characteristic parameters of dispersed cells suspended in liquid, and can sort out designated cell subpopulations according to a preselected parameter range.

[0003] In recent years, cell detection methods based on fluorescence effects have developed rapidly. People have invented methods for detecting a certain cell population using cell markers. The specific method is to label cells with fluorescent molecules in vitro and then inject the cells into the body for detection. The markers on the target cells can achieve optical absorption detection. This method is suitable for early diagnosis of major diseases such as malignant tumors in the laboratory. However, since this detection method requires various dyes or markers, and the injection after labeling will be toxic to the normal internal environment of the animal, the reliability of the measurement results is reduced.

[0004] At present, a Chinese patent with the existing patent document number "CN116698709B" discloses a data processing method for a flow cytometer and a flow cytometer, which records pre-setting imaging test parameters of different lasers of the flow cytometer, and constructing a marking template set under different test parameters; wherein the marking template is used to mark the cell data detected by the flow cytometer; the image signal data output by the flow cytometer is obtained, and the image signal data is imported into the marking template according to the laser type, a plurality of image data sets are generated, and the image deviation is calculated; wherein each image data set is provided with a particle time code of the cell; through the image data set, the cell images cut under different lasers are determined; according to the cell code image and the particle time code, different cells are counted, and the counting deviation between the particle counting data and the mean data under different lasers is determined; the cell detection data corresponding to the closest image deviation and counting deviation is judged and used as the final detection data.

[0005] However, during the implementation of the above technical solution, it was found that there were at least the following technical problems: First, it uses an image-based flow cytometer, which is suitable for early diagnosis of major diseases such as malignant tumors in the laboratory. However, since this detection method requires various dyes or markers, and the injection after labeling will be toxic to the normal internal environment of the animal, the reliability of the measurement results will be reduced. Similarly, since it requires pre-processing, it is not applicable in many scenarios, so its applicability is low; Secondly, for photoacoustic flow cytometers, the scheme used when identifying cells is to process the collected data and input it into a neural network model for comprehensive comparison with all the data of this type of cells, and then output the identification result. Since the photoacoustic flow cytometer collects data based on the time domain, the photoacoustic data of a cell is very complicated. This identification method needs to process a large amount of data and a lot of data to be compared, which makes the efficiency of data processing very low, and the requirements for related equipment are very high, which does not meet the needs of technicians in this field. To this end, a data processing method and system for flow cytometers are provided. Summary of the invention

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0007] A data processing system for flow cytometer, the system comprises a sorting module, a collection module, a screening module, a sieving module and a statistics module: Arrangement module: obtain the photoacoustic signal data of the type of cells to be collected from the database, process the photoacoustic signal data, calculate the standard threshold interval, and obtain the characteristic screening category; Acquisition module: pre-processing the photoacoustic signals collected by the flow cytometer to obtain a photoacoustic signal data set; Screening module: Screen the preprocessed photoacoustic signal data set with the help of the standard threshold interval, remove the photoacoustic signals that do not meet the standard threshold interval, and obtain the remaining photoacoustic signal set; Screening module: obtain the number of identification collection points and the corresponding data interval corresponding to the collection type cell characteristic screening category, obtain the data corresponding to the number of identification collection points of the photoacoustic signal in the remaining photoacoustic signal set, and compare it with the data interval, statistically compare the results, and calculate the degree of conformity; Statistical module: Compare the conformity with the preset interval, and use the classification algorithm to divide the photoacoustic signals in the photoacoustic signal set into a standard set, a suspected set, and a rejection set. The classification algorithm here adopts the common threshold comparison classification method, and performs a comprehensive analysis on the photoacoustic signals in the suspected set for further division. The number of photoacoustic signals in the standard set is counted, and the data and conformity are marked to generate a detection data report.

[0008] Furthermore, the photoacoustic signal data of the type of cells to be collected is a certified real photoacoustic signal set containing the type of cells to be collected, the number of real photoacoustic signals in the real photoacoustic signal set is M groups, and the photoacoustic signal data includes but is not limited to average signal intensity, signal integral value, pulse height, pulse width and pulse area.

[0009] Furthermore, the steps of processing the photoacoustic signal data and calculating the standard threshold interval are as follows: Obtain the type of collected cells, obtain the preset category ranking of the cell type based on cell biological characteristics and the correlation between photoacoustic signals and cell characteristics from the database, and select the first K screening categories, 3≤K≤5; A feature extraction algorithm is used to extract M groups of photoacoustic signal data according to the screening categories to obtain K data sets; Use statistical methods to process the data in K data sets one by one to generate K standard intervals; The K standard intervals are aggregated to obtain the standard threshold interval.

[0010] Furthermore, the steps of using statistical methods to process the data in the K data sets one by one and generate K standard intervals are as follows: Calculate the average value of K data sets , where is the mean value of the ith data set, for The result is rounded to 3 decimal places. is the jth group of data in the i-th data set; Determine the maximum and minimum values ​​of the data in the K data sets, and calculate the difference ratios between the maximum and minimum values ​​and the average value, compare the difference ratios, obtain the maximum difference ratio Q, and calculate the adjusted ratio Qt. , where To adjust the ratio, 3%≤ ≤6%; Calculate K standard intervals based on Qt .

[0011] Furthermore, the preprocessing of the photoacoustic signal collected in real time by the flow cytometer is to perform wavelet analysis denoising on the collected photoacoustic signal, and the wavelet analysis denoising adopts a default threshold denoising.

[0012] Furthermore, the steps of screening the preprocessed photoacoustic signal data set are as follows: Acquire photoacoustic signal data corresponding to the collected photoacoustic signal, compare K groups of data in the photoacoustic signal data with the K standard intervals recorded in the standard threshold interval one by one, calculate the Euclidean distance between the feature vector and the boundary of the standard interval through the boundary recognition model, and determine whether the K groups of data are located in the corresponding K standard intervals; If it is outside the standard range, the photoacoustic signal is deleted; If it is within the standard interval, the photoacoustic signal is retained, and C groups of pulse height data are randomly extracted from the retained photoacoustic signal and M groups of real photoacoustic signal data. The C groups of data extracted from the retained photoacoustic signal are subjected to similarity analysis one by one with the C groups of data randomly extracted from the M groups of real photoacoustic signals to obtain M groups of similarity data. The formula for calculating the similarity is as follows; In the formula, is the similarity between the detected photoacoustic signal and the i-th group of real photoacoustic signals, is the photoacoustic signal data of the jth type in the collected photoacoustic signal, is the j-th type of photoacoustic signal data extracted from the i-th group of real photoacoustic signals, is the adjustment factor; Obtaining the maximum value among the M groups of similarity data, that is, the maximum similarity, and comparing the maximum similarity with the standard similarity, if the maximum similarity is less than the standard similarity, deleting the photoacoustic signal; If the maximum similarity is greater than or equal to the standard similarity, the photoacoustic signal is retained; The photoacoustic signals retained after the deletion are aggregated to form a set of remaining photoacoustic signals.

[0013] Furthermore, the characteristic screening categories include but are not limited to instantaneous intensity, instantaneous frequency and instantaneous phase, the collection type cell characteristic screening category is one of them, the value range of the number of identification collection points L is 30 to 50, and the data interval corresponding to the identification collection point is , The characteristic screening category data of the i-th identification collection point in the photoacoustic signal corresponding to the maximum similarity, B is the fluctuation coefficient, 2%≤B≤5%, and the statistical comparison results are compared. The number of data that do not meet the data interval is P group, and the calculated compliance .

[0014] Furthermore, the preset interval includes a standard data interval , suspected data interval and remove data intervals , ; The conformity is compared with the preset interval, and the classification algorithm is used to divide the photoacoustic signals in the photoacoustic signal set into a standard set, a suspected set and a rejection set.

[0015] Further, the steps of comprehensive analysis of the suspected concentrated photoacoustic signal are as follows: The photoacoustic signal in the suspected concentration and its corresponding photoacoustic signal with the maximum similarity are equally divided into E segments by using signal processing technology; Compare the pulse heights in segment E to obtain the half-maximum full width and frequency bandwidth of the photoacoustic signal in the segment with the highest pulse height; The half-maximum width and frequency bandwidth of the photoacoustic signal in the suspected concentration are compared with the half-maximum width and frequency bandwidth of the photoacoustic signal corresponding to the maximum similarity, and the cell similarity index is calculated; The specific calculation formula is as follows: In the formula, is the cell similarity index, is the half-maximum full width of the suspected concentrated photoacoustic signal, The maximum similarity corresponds to the half-maximum full width of the photoacoustic signal. is the weight ratio corresponding to the difference ratio of the half-height full width value, is the frequency bandwidth of the suspected concentrated photoacoustic signal, is the frequency bandwidth of the photoacoustic signal corresponding to the maximum similarity, is the weight ratio corresponding to the frequency bandwidth difference ratio; If the calculated cell similarity index is less than the set value, the photoacoustic signal is classified into the rejection set; If the calculated cell similarity index is greater than or equal to the set value, the photoacoustic signal is classified into the standard set.

[0016] Furthermore, a data processing method applied to a flow cytometer comprises the following steps: Acquire photoacoustic signal data of the type of cells to be collected from the database, process the photoacoustic signal data, calculate the standard threshold interval, and obtain the characteristic screening category; Preprocessing the photoacoustic signals collected by the flow cytometer to obtain a photoacoustic signal data set; The preprocessed photoacoustic signal data set is screened by using the standard threshold interval, and the photoacoustic signals that do not meet the standard threshold interval are eliminated to obtain the remaining photoacoustic signal set; Obtain the number of identification collection points and the corresponding data interval corresponding to the collection type cell characteristic screening category, obtain the data corresponding to the number of identification collection points of the photoacoustic signal in the remaining photoacoustic signal set, and compare it with the data interval, statistically compare the results, and calculate the degree of conformity; The conformity is compared with the preset interval, and the classification algorithm is used to divide the photoacoustic signals in the photoacoustic signal set into a standard set, a suspected set and a rejection set. The photoacoustic signals in the suspected set are comprehensively analyzed and divided again. The number of photoacoustic signals in the standard set is counted, and the data and conformity are marked to generate a detection data report.

[0017] The present invention provides a data processing method and system for flow cytometer, which has the following beneficial effects: (1) The present invention provides a data processing method and system for flow cytometer, which analyzes and processes the characteristics of the type of cells to be collected, formulates a multi-level screening scheme based on the characteristics, and uses the multi-level screening method to realize cell identification and judgment. Based on the characteristic screening, most of the data can be screened out, so that the amount of data for subsequent data processing is greatly reduced, thereby effectively reducing the equipment cost required for data processing, and greatly improving the efficiency. The use effect is good, it has a good use prospect, and solves the problems raised in the background technology.

[0018] (2) The present invention provides a data processing method and system for flow cytometers. In the process of data screening, the method does not adopt a complete data comparison method. The real-time monitored photoacoustic data is processed based on the commonality of characteristic data. This method greatly reduces the amount of data comparison. Then, a random data sampling comparison method is used to obtain the closest recognized photoacoustic signal. Finally, further analysis and judgment are performed. After this multi-level screening method, the data processing volume can be reduced by 65%. The use effect is good and has a good application prospect.

[0019] (3) The present invention provides a data processing method and system for flow cytometer, which sets a suspected set, adds an analysis and identification scheme for the suspected set, and makes a separate judgment on the photoacoustic signal in the suspected set. In addition, it uses the half-maximum full width value and frequency bandwidth of the most obvious characteristic time period for screening, which can realize cell identification more accurately and quickly. In addition, by adopting this method, the amount of data calculation increases less, but the accuracy of cell identification is greatly improved, basically reaching more than 96%, and the use effect is good. Therefore, this scheme has good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is the overall flow chart of the present invention; Figure 2 A diagram showing the steps for calculating the standard threshold interval in the present invention; Figure 3 A diagram showing the steps of screening the preprocessed photoacoustic signal data set in the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] With the development of society, cell detection methods based on fluorescence effects have developed rapidly. People have invented methods of detecting a certain cell population with cell markers. The specific method is to label cells with fluorescent molecules in vitro, and then inject the cells into the body for detection. The markers on the target cells can realize optical absorption detection. This method is suitable for early diagnosis of major diseases such as malignant tumors in the laboratory. However, since this detection method requires various dyes or markers, and the injection after labeling will be toxic to the normal internal environment of the animal, the reliability of the measurement results will be reduced.

[0023] In the photoacoustic detection of biological tissues, researchers use a pulsed laser of a certain wavelength and, through a series of optical path modulation, allow the laser spot to shine on the tissues of living animals. Due to the photoacoustic effect in biological tissues, the target cells in the animal body will produce photoacoustic signals. Different target cells have different absorption of lasers, so the ultrasonic signals they emit will also differ in some aspects. The in vivo photoacoustic detection method can avoid the various defects of in vitro labeling and maintain the stability of the internal environment of the experimental individual.

[0024] In short, research and development of data processing methods for in vivo photoacoustic flow cytometry has the advantages of real-time cell monitoring, non-invasive cell diagnosis, and no labeling, which also avoids the inaccuracy caused by in vitro detection. In addition, it combines the advantages of ultrasonic signals such as non-cytotoxicity and deep penetration in living tissues, and can be especially used for the identification of special cells such as cancer cells.

[0025] like Figure 1 As shown, the specific implementation plan is as follows: Instrument selection: The focus of this program is the innovation of data processing methods for cell photoacoustic signals. It is committed to proposing data processing methods based on solving the current problems of photoacoustic signal analysis and processing. The photoacoustic signals generated in the experiment are derived from in vivo photoacoustic flow cytometers, and relevant data are directly obtained through IoT data.

[0026] When different cells are irradiated by pulsed lasers, the cells absorb light differently due to their different internal structures. After absorbing light, the temperature inside the cells will change, resulting in changes in regional structure or volume. Therefore, when we use pulsed light sources to irradiate cells, the continuous change in temperature will cause the cell volume to expand and contract, allowing the cells to emit sound waves outward. When cells flow through the focal plane of the focused laser, since the excitation light of the cells in this band has a high absorption spectrum different from the background, they will absorb electromagnetic wave energy, increase tissue temperature, and expand the local volume, thereby generating vibrations and emitting ultrasonic waves. Different cells will radiate ultrasonic signals of different frequencies due to their different characteristics, so quantitative detection of cells can be achieved.

[0027] Data reception: First, data organization: The data collection in the present invention is to use a server to connect to a database through the Internet of Things, and can directly collect historically collected and certified standard cell databases. There are many types of data stored in the standard cell database, and they are classified, that is, the photoacoustic data of the same cells collected under the same collection parameters of the in vivo photoacoustic flow cytometer.

[0028] When the standard cell database was constructed, relevant analysis was performed to mark the obvious photoacoustic data characteristics of different cell types. These data were summarized and sorted to form a category sorting based on cell biological characteristics and the correlation between photoacoustic signals and cell characteristics, and the characteristic screening categories were marked.

[0029] Before use, it is necessary to understand the type of cells that need to be identified, and then understand the acquisition parameters set for the in vivo photoacoustic flow cytometer, and then find the classified data set corresponding to its acquisition parameters. For example: the number of points collected per unit time is the same. If they are different, they cannot be compared, so preprocessing is required. This process is based on the sorting module of this solution.

[0030] The sorting module obtains the photoacoustic signal data of the type of cells to be collected from the database, processes the photoacoustic signal data, calculates the standard threshold interval, and obtains the characteristic screening category.

[0031] Photoacoustic signals are the basic data collected by in vivo photoacoustic flow cytometer, while photoacoustic signal data are the data obtained after photoacoustic signal processing.

[0032] The steps for processing the photoacoustic signal data and calculating the standard threshold range are as follows: Obtain the type of collected cells, obtain the preset category ranking of the cell type based on cell biological characteristics and the correlation between photoacoustic signals and cell characteristics from the database, and select the first K screening categories, 3≤K≤5; for example, for tumor cells, the first 3 categories with a high correlation with tumor characteristics can be selected.

[0033] Different types of cells exhibit different situations in the photoacoustic signals, that is, there are some obvious characteristics. For example, there are many types of cell sizes, so there are many cells with similar sizes. At this time, the number of cells with similar pulse areas is relatively large. Therefore, the effect of screening by pulse area is not good. Therefore, the database makes corresponding evaluations when it is built, and lists the categories based on cell biological characteristics and the correlation between photoacoustic signals and cell characteristics to facilitate subsequent comparison, so it is more convenient to use later.

[0034] The selection of the top K screening categories can be further manually participated. For example, the external environment during the batch detection process is different, and the category sorting based on cell biological characteristics and the correlation between photoacoustic signals and cell characteristics is not completely applicable to a certain extent. Therefore, the external environment and the cells to be collected can be input into the computer in advance, and the category sorting based on cell biological characteristics and the correlation between photoacoustic signals and cell characteristics in this case can be retrieved from the database. The category sorting based on cell biological characteristics and the correlation between photoacoustic signals and cell characteristics is not completely consistent. It is generated during historical detection, and the purpose is to facilitate subsequent detection.

[0035] like Figure 2 As shown, a feature extraction algorithm is used to extract M groups of photoacoustic signal data according to the screening categories to obtain K data sets. This step classifies the data, which is more convenient for comparison. Each set corresponds to a screening category, reflecting cell-specific information; such as principal component analysis, linear discriminant analysis, etc., to screen and reduce the dimensions of features, remove redundant or irrelevant features, and retain the most representative and discriminative features for subsequent classification and recognition tasks; Use statistical methods to process the data in K data sets one by one to generate K standard intervals. The specific steps are as follows: Calculate the average value of K data sets , where is the mean value of the ith data set, for The result is rounded to 3 decimal places. For the jth group of data in the ith data set, the effect of designing decimals is mainly to facilitate calculation and reduce the amount of data processing; Determine the maximum and minimum values ​​in the K data sets, and calculate the difference ratios between the maximum and minimum values ​​and the average value respectively. Compare the difference ratios to get the maximum difference ratio Q. For example, if the maximum value is 25, the average value is 20, and the minimum value is 16, then the difference ratio between the maximum value and the average value is (25-20) / 20×100%=25%, and the difference ratio between the maximum value and the average value is (20-16) / 20×100%=20%, then the maximum difference ratio Q is 25%; And calculate the adjusted ratio Qt, , where To adjust the ratio, 3%≤ ≤6%; The K standard intervals are aggregated to obtain the standard threshold interval.

[0036] Calculate K standard intervals based on Qt , each interval is used for a separate comparison. This step is set in advance and can be directly called during subsequent comparisons. There is no need to perform on-site calculations during data comparison, thereby improving the efficiency of data comparison.

[0037] Secondly, data collection: the data collection in the present invention is to use the server to connect with the in vivo photoacoustic flow cytometer through the Internet of Things to obtain the acoustic and optical data collected by the in vivo photoacoustic flow cytometer, and this process is based on the existing collection module.

[0038] Acquisition module: pre-processing the photoacoustic signals collected by the flow cytometer to obtain a photoacoustic signal data set; The photoacoustic signal data of the type of cells to be collected is a certified real photoacoustic signal set containing the type of cells to be collected. The number of real photoacoustic signals in the real photoacoustic signal set is M groups. The photoacoustic signal data includes but is not limited to average signal intensity, signal integral value, pulse height, pulse width and pulse area.

[0039] The preprocessing of the photoacoustic signal collected in real time by the flow cytometer is to perform wavelet analysis denoising on the collected photoacoustic signal, and the wavelet analysis denoising adopts the default threshold denoising.

[0040] The collected photoacoustic signals are subjected to wavelet analysis and denoising to ensure the purity and accuracy of the signal data and improve the reliability of subsequent analysis. The signal is decomposed by wavelet transform, the noise component is eliminated, and the effective signal characteristics are retained, and finally a high-quality photoacoustic signal is obtained, laying a solid foundation for the subsequent accurate identification and analysis of cells. High-quality data ensures the accuracy of the analysis and improves the reliability and scientificity of the experimental results.

[0041] Wavelet denoising can characterize and represent the local characteristic information of the signal in both the time domain and the frequency domain. The theoretical basis of wavelet analysis is to apply wavelet functions to various analyses. After the signal is decomposed by wavelet, its wavelet coefficient amplitude will be higher than the wavelet coefficient amplitude of the noise. When denoising, the wavelet coefficient of the signal is preserved and the wavelet coefficient of the noise is removed. The denoised signal is obtained by inverse transforming the retained wavelet coefficients.

[0042] The process of wavelet analysis denoising can generally be divided into the following three steps: wavelet decomposition of the signal, threshold quantization of the high-frequency coefficients obtained by wavelet decomposition, and wavelet reconstruction of the coefficients. The key is how to choose the threshold and perform threshold quantization. To some extent, the choice of threshold is related to the quality of wavelet denoising.

[0043] The default threshold denoising process selects the wdencmp function which performs well in compression or noise reduction of one-dimensional and two-dimensional signals.

[0044] Regarding the above-mentioned wavelet analysis denoising scheme, this application does not improve its method, but is only a related application of the existing technology. Moreover, those skilled in the art also have schemes for photoacoustic processing using wavelet analysis denoising. Therefore, no further description will be given, and those skilled in the art are clear and understand how to implement the related technology.

[0045] Secondly, data screening: after collecting the complete photoacoustic signal, the collected photoacoustic signal is further processed to remove part of the interference signal to obtain the photoacoustic signal of the cell. For further analysis, the obtained photoacoustic signal of the cell needs to be further screened to obtain a more standard photoacoustic signal. The screening of the photoacoustic signal of the cell depends on the screening module.

[0046] Screening module: Screen the preprocessed photoacoustic signal data set with the help of the standard threshold interval, remove the photoacoustic signals that do not meet the standard threshold interval, and obtain the remaining photoacoustic signal set; like Figure 3 As shown, the steps for screening the preprocessed photoacoustic signal data set are as follows: Acquire photoacoustic signal data corresponding to the collected photoacoustic signal, compare K groups of data in the photoacoustic signal data with the K standard intervals recorded in the standard threshold interval one by one, calculate the Euclidean distance between the feature vector and the boundary of the standard interval through the boundary recognition model, and determine whether the K groups of data are located in the corresponding K standard intervals; The boundary recognition model uses the K standard intervals recorded in the standard threshold interval as prior knowledge. The converted feature vector is input into the model, and the model uses internal calculation logic to quickly calculate the Euclidean distance between the feature vector and the boundary of the standard interval, and then accurately determines whether the K groups of data are within the corresponding K standard intervals.

[0047] If it is outside the standard interval, it means that the data is very different from the required data, so it is not the required data, so the photoacoustic signal is deleted. Because the standard interval is formulated based on all the data, its range is very wide. If the photoacoustic signal does not meet the requirements in this case, it means that the photoacoustic signal definitely does not belong to the photoacoustic signal of the collected cell, so the photoacoustic signal of the cell is deleted, thereby reducing the amount of subsequent data processing, which is more convenient to use and has better results.

[0048] If it is within the standard range, the photoacoustic signal is retained. In this case, it indicates that the signal may meet the requirements, but it is not certain whether it is. Therefore, secondary verification is still required in the future, so subsequent steps are required to process the photoacoustic data; Randomly extract C types of photoacoustic signal data from the retained photoacoustic signal and M groups of real photoacoustic signal data, and perform similarity analysis on the C groups of data extracted from the retained photoacoustic signal and the C groups of data randomly extracted from the M groups of real photoacoustic signals one by one; when extracting, extract C types, and then retrieve the photoacoustic signal data corresponding to the C types of photoacoustic signals and the real photoacoustic signals, so as to obtain C groups of data; The random data extraction method and the local comparison method can greatly reduce the amount of data comparison, mainly because the amount of photoacoustic signal data of normal cells is very large. The full comparison or waveform comparison method requires a large amount of data to be compared. M groups of similarity data are obtained, and the formula for calculating the similarity is as follows; In the formula, is the similarity between the detected photoacoustic signal and the i-th group of real photoacoustic signals, is the photoacoustic signal data of the jth type in the collected photoacoustic signal, is the j-th type of photoacoustic signal data extracted from the i-th group of real photoacoustic signals, is the adjustment coefficient, C>K; The above similarity calculation formula is mainly obtained by improving the existing Pearson correlation coefficient. The Pearson correlation coefficient is mainly used to calculate the correlation of each corrected data. The higher the correlation, the closer the two data are, and therefore, the higher the similarity.

[0049] The calculation of similarity is mainly to determine which identified photoacoustic data the photoacoustic data of the cell is closer to, so as to facilitate further comparison and judgment. It is equivalent to a basic identification process, and its purpose is also to screen.

[0050] The scheme for calculating similarity in this paper adds corresponding adjustment coefficients, and the comparison effect is better.

[0051] The adjustment factor is to correct the data. The adjustment coefficient is not weight data. The adjustment coefficient is the characteristic setting of the cell photoacoustic signal data type. Different types of data are different. In addition, the same type also has large differences at different times. Therefore, adjustment is required; For example, the pulse height data in the cell photoacoustic signal is basically in the thousands, while some data are in the hundreds or single digits. If the similarity is calculated directly, the difference is very large. Therefore, it is necessary to make separate corrections to ensure the accuracy of the calculation. At different times, the same type of data also has large differences. For example, the pulse height data of the cell photoacoustic signal is completely different at different stages. It is mainly based on the current detection result setting. For example, there are two sets of data in the pulse height of a single cell photoacoustic signal, which are 6000 and 2000 respectively. Since the same ratio is used, the adjustment coefficient will make the 6000 data have a greater impact. Therefore, when designing, the corresponding 6000 smaller, while 2000 corresponds to Larger, thereby eliminating the impact of data fluctuations at different stages and improving the accuracy of similarity analysis. For the convenience of calculation, when the pulse height is selected, the maximum pulse height is generally used for calculation.

[0052] Therefore, the photoacoustic signal data must be corrected, and the adjustment coefficient is determined when the database is constructed, without the need for subsequent calculations.

[0053] Obtain the maximum value of the M groups of similarity data, that is, the maximum similarity, and compare the maximum similarity with the standard similarity. If the maximum similarity is less than the standard similarity, delete the photoacoustic signal. This situation is mainly because the data similarity is relatively low, indicating that it is not similar to the data, so it is removed; For example, if M is 5, the calculated similarity data are 0.65, 0.68, 0.72, 0.61, and 0.68, respectively. The maximum value of 0.62 is taken. If the standard similarity is 0.70, the data is retained. If the standard similarity is 0.75, the data is removed. This method effectively screens high-quality data and ensures the accuracy of subsequent analysis. By continuously optimizing the adjustment coefficient, the recognition accuracy is improved, laying a solid foundation for subsequent in-depth research.

[0054] If the maximum similarity is greater than or equal to the standard similarity, the photoacoustic signal is retained, indicating that it is relatively similar and there is a certain possibility that it is the photoacoustic signal that needs to be collected; The photoacoustic signals retained after the deletion are aggregated to form a set of remaining photoacoustic signals.

[0055] The present invention provides a data processing method and system applied to a flow cytometer, which analyzes and processes the characteristics of cells of a to-be-collected type, formulates a multi-level screening scheme for the characteristics, and uses the multi-level screening method to realize cell identification and judgment. Based on the characteristic screening, most of the data can be screened out, so that the amount of data for subsequent data processing is greatly reduced, thereby effectively reducing the equipment cost required for processing data, and also greatly improving the efficiency, with good use effect and good use prospects, solving the problems raised in the background technology.

[0056] After that, data screening: after collecting the complete photoacoustic signal, the collected photoacoustic signal is further processed, and after removing a part of the interference signal, the photoacoustic signal of the cell is obtained. After further analysis, the obtained photoacoustic signal of the cell is further screened and processed to obtain the remaining photoacoustic signal set. Then, further judgment is made to determine whether the photoacoustic signals in the remaining photoacoustic signal set meet the collection requirements. Therefore, the photoacoustic signals in the remaining photoacoustic signal set need to be further screened, and this process is based on the screening module.

[0057] Screening module: obtain the number of identification collection points and the corresponding data interval corresponding to the collection type cell characteristic screening category, obtain the data corresponding to the number of identification collection points of the photoacoustic signal in the remaining photoacoustic signal set, and compare it with the data interval, statistically compare the results, and calculate the degree of compliance.

[0058] The characteristic screening categories include but are not limited to instantaneous intensity, instantaneous frequency and instantaneous phase. The characteristic screening categories are mainly for real-time data. Instead of global comparison of the overall data, they focus on the characteristic changes at a specific moment, so as to more accurately capture the dynamic characteristics of cells. Through this meticulous screening, it is ensured that the collected photoacoustic signals are both real and representative, providing reliable data support for subsequent in-depth analysis and research.

[0059] The cell characteristic screening category of the acquisition type is one of them. The value range of the number of acquisition points L is 30 to 50. The higher the number of acquisition points L, the higher the accuracy, but the data processing volume will also increase, and the data processing time will also increase accordingly, thus affecting the overall efficiency. The existing scheme uses a convolutional neural network for one-time identification. Usually, 512 or more point data will be taken from the photoacoustic signal, and then they will be compared one by one with the photoacoustic data identified in the database. There are a lot of data to be compared, and when there is a local data discrepancy, multiple data need to be compared, that is, repeated comparisons of more than 500 groups of data are repeated, which is cumbersome and requires a lot of data to be processed. Therefore, the corresponding hardware requirements are relatively high; The data interval corresponding to the identified collection point is , The characteristic screening category data of the i-th identification collection point in the photoacoustic signal corresponding to the maximum similarity, B is the fluctuation coefficient, 2%≤B≤5%, and the statistical comparison results are compared. The number of data that do not meet the data interval is P group, and the calculated compliance , or you can collect it directly.

[0060] The higher the degree of conformity, the closer the collected photoacoustic signal is to the identified photoacoustic signal, indicating that the similarity is higher, and the greater the possibility that the required type of cells are collected.

[0061] The present invention provides a data processing method and system for flow cytometer, which does not adopt a complete data comparison method in the process of data screening, but processes the photoacoustic data monitored in real time based on the commonality of characteristic data, which greatly reduces the amount of data comparison, and then adopts a random data sampling comparison method to obtain the closest recognized photoacoustic signal, and finally further analyzes and judges, and through the multi-level screening method, can reduce 65% of the data processing volume, has good use effect, and has a good application prospect.

[0062] Finally, data statistics: After analyzing the conformity of the photoacoustic signal, further comparison and judgment based on the conformity is required to identify the photoacoustic signal, and this process is based on the screening module.

[0063] Statistical module: Compare the conformity with the preset interval, and use the classification algorithm to divide the photoacoustic signals in the photoacoustic signal set into a standard set, a suspected set, and a rejection set. The classification algorithm here adopts the common threshold comparison classification method, and performs a comprehensive analysis on the photoacoustic signals in the suspected set for further division. The number of photoacoustic signals in the standard set is counted, and the data and conformity are marked to generate a detection data report.

[0064] The conformity is compared with the preset interval, and the classification algorithm is used to divide the photoacoustic signals in the photoacoustic signal set into a standard set, a suspected set and a rejection set.

[0065] The preset intervals include standard data intervals , suspected data interval and remove data intervals , ; The photoacoustic signals within the standard data interval are regarded as high-confidence samples, those within the suspected data interval need further verification, and those within the excluded data interval are excluded. Through this process, the accuracy and reliability of the final photoacoustic signal identification are ensured.

[0066] The steps for comprehensive analysis of the suspected concentrated photoacoustic signal are as follows: The suspected concentrated photoacoustic signal and its corresponding photoacoustic signal with the maximum similarity are equally divided into E segments by using signal processing technology; for example, in this process, a Gaussian model can be used to describe the distribution characteristics of the signal to ensure the scientificity and accuracy of signal segmentation; Compare the pulse heights in segment E to obtain the half-maximum full width and frequency bandwidth of the photoacoustic signal in the segment with the highest pulse height; the half-maximum full width can directly reflect the time width of the signal pulse at half the intensity, while the frequency bandwidth reflects the frequency range covered by the signal; The half-maximum width and frequency bandwidth of the photoacoustic signal in the suspected concentration are compared with the half-maximum width and frequency bandwidth of the photoacoustic signal corresponding to the maximum similarity, and the cell similarity index is calculated; The specific calculation formula is as follows: In the formula, is the cell similarity index, is the half-maximum full width of the suspected concentrated photoacoustic signal, The maximum similarity corresponds to the half-maximum full width of the photoacoustic signal. is the weight ratio corresponding to the difference ratio of the half-height full width value, is the frequency bandwidth of the suspected concentrated photoacoustic signal, is the frequency bandwidth of the photoacoustic signal corresponding to the maximum similarity, is the weight ratio corresponding to the frequency bandwidth difference ratio, < , + =1; If the calculated cell similarity index is less than the set value, the photoacoustic signal is classified into the rejection set; If the calculated cell similarity index is greater than or equal to the set value, the photoacoustic signal is classified into the standard set.

[0067] The data and compliance are marked, and the distribution uniformity data can be marked when generating the test data report, which is conducive to further testing.

[0068] The detection data report also includes the final proportion data, that is, the proportion of the number of cells of this type detected in the total number of cells; the data after screening the pre-processed photoacoustic signal data set through the standard threshold interval is the photoacoustic signal of the cell, and the number of photoacoustic signals can be counted.

[0069] If the existing convolutional neural network model wants to achieve this aspect, it has to increase the output structure and also increase the calculation layer corresponding to cell recognition, which makes the data processing volume even larger, greatly reduces the efficiency and greatly increases the cost. Therefore, the existing convolutional neural network model does not involve data output in this aspect, so this aspect is difficult to achieve with the existing convolutional neural network model.

[0070] In actual use, since the number of suspected concentrated photoacoustic signals is less than the original data, which is basically less than 10%, a convolutional neural network processing model can also be constructed to perform photoacoustic signal recognition and authentication. Through the convolutional neural network model, the photoacoustic signal features are further extracted to improve the recognition accuracy.

[0071] For example, an MP neuron model is constructed. That is, when a neuron receives input signals from multiple neurons in front of it after the transmission path weight addition calculation, it will calculate the total input value received by the current neuron and compare it with the excitation threshold of this neuron. The activation function is used to determine whether this neuron is activated, that is, whether the neuron is excited, and it is passed on in sequence according to this rule. Similar to logistic regression, the sigmoid function is used as the activation function in the neural network model.

[0072] In this solution, the built-in neural network function package of MATLAB is used to build and train the feedforward neural network. The feedforward neural network is constructed by calling the newff() function, which includes three parameters: normalized training samples, normalized training sample outputs, and the number of hidden layer neurons.

[0073] The training set is a mixed data set of 20-30% authenticated photoacoustic data and 70-80% authenticated other cell photoacoustic data.

[0074] During the training process, the input matrix is ​​randomly divided into training set, validation set and test set. After training the training set, the error calculation is performed on the validation set. If the training requirements are met, the training stops. Otherwise, the training continues and the parameters are optimized. This process is repeated until the training results are obtained and the recognition accuracy reaches more than 95%.

[0075] The present invention provides a data processing method and system for flow cytometer, which sets a suspected set, adds an analysis and identification scheme for the suspected set, and makes a separate judgment on the photoacoustic signal in the suspected set. Moreover, the half-maximum full width value and frequency bandwidth of the most obvious characteristic time period are used for screening, so that the cell recognition can be realized more accurately and quickly. Moreover, by adopting this method, the amount of data calculation increases less, but the accuracy of cell recognition is greatly improved, basically reaching more than 96%, and the use effect is good. Therefore, the scheme has good application prospects.

[0076] The weight coefficient is determined by the coefficient of variation method, which is a method of weighting each indicator according to the degree of variation between the current value of each evaluation indicator and the target value. If the numerical difference of an indicator is large, it can clearly distinguish the evaluated objects, indicating that the indicator has rich discrimination information, and thus the indicator should be given a larger weight. On the contrary, if the numerical difference of each evaluated object on a certain indicator is small, then the ability of this indicator to distinguish the evaluation objects is weak, and thus the indicator should be given a smaller weight. This method directly uses the information contained in each indicator and obtains the weight of the indicator through calculation, so it is objective.

[0077] A data processing method applied to a flow cytometer comprises the following steps: Acquire photoacoustic signal data of the type of cells to be collected from the database, process the photoacoustic signal data, calculate the standard threshold interval, and obtain the characteristic screening category; Preprocessing the photoacoustic signals collected by the flow cytometer to obtain a photoacoustic signal data set; The preprocessed photoacoustic signal data set is screened by using the standard threshold interval, and the photoacoustic signals that do not meet the standard threshold interval are eliminated to obtain the remaining photoacoustic signal set; Obtain the number of identification collection points and the corresponding data interval corresponding to the collection type cell characteristic screening category, obtain the data corresponding to the number of identification collection points of the photoacoustic signal in the remaining photoacoustic signal set, and compare it with the data interval, statistically compare the results, and calculate the degree of conformity; The conformity is compared with the preset interval, and the classification algorithm is used to divide the photoacoustic signals in the photoacoustic signal set into a standard set, a suspected set and a rejection set. The photoacoustic signals in the suspected set are comprehensively analyzed and divided again. The number of photoacoustic signals in the standard set is counted, and the data and conformity are marked to generate a detection data report.

[0078] In the application, the several formulas involved are all calculated by taking their numerical values ​​after removing the dimensions, and the formulas are established by collecting a large amount of data and performing software simulation to obtain a formula for the most recent actual situation. Some coefficients or weights in the formulas are set by technical personnel in this field according to actual conditions, so they will not be elaborated here.

[0079] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0080] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0081] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A data processing system for a flow cytometer, characterized in that: The system includes: Arrangement module: obtain the photoacoustic signal data of the type of cells to be collected from the database, process the photoacoustic signal data, calculate the standard threshold interval, and obtain the characteristic screening category; Acquisition module: pre-processing the photoacoustic signals collected by the flow cytometer to obtain a photoacoustic signal data set; Screening module: Screen the preprocessed photoacoustic signal data set with the help of the standard threshold interval, remove the photoacoustic signals that do not meet the standard threshold interval, and obtain the remaining photoacoustic signal set; Screening module: obtain the number of identification collection points and the corresponding data interval corresponding to the collection type cell characteristic screening category, obtain the data corresponding to the number of identification collection points of the photoacoustic signal in the remaining photoacoustic signal set, and compare it with the data interval, statistically compare the results, and calculate the degree of conformity; Statistical module: Compare the conformity with the preset interval, use the classification algorithm to divide the photoacoustic signals in the photoacoustic signal set into a standard set, a suspected set and a rejection set, and conduct a comprehensive analysis of the photoacoustic signals in the suspected set for further division, count the number of photoacoustic signals in the standard set, mark the data and conformity, and generate a test data report.

2. A data processing system for flow cytometer according to claim 1, characterized in that: The photoacoustic signal data of the type of cells to be collected is a certified real photoacoustic signal set containing the type of cells to be collected. The number of real photoacoustic signals in the real photoacoustic signal set is M groups. The photoacoustic signal data includes average signal intensity, signal integral value, pulse height, pulse width and pulse area.

3. A data processing system for flow cytometer according to claim 2, characterized in that: The steps for processing the photoacoustic signal data and calculating the standard threshold range are as follows: Obtain the type of collected cells, obtain the preset category ranking of the cell type based on cell biological characteristics and the correlation between photoacoustic signals and cell characteristics from the database, and select the first K screening categories, 3≤K≤5; A feature extraction algorithm is used to extract M groups of photoacoustic signal data according to the screening categories to obtain K data sets; Use statistical methods to process the data in K data sets one by one to generate K standard intervals; The K standard intervals are aggregated to obtain the standard threshold interval.

4. A data processing system for flow cytometer according to claim 3, characterized in that: The steps of using statistical methods to process the data in K data sets one by one and generate K standard intervals are as follows: Calculate the average value of K data sets ; Determine the maximum and minimum values ​​of the data in the K data sets, and calculate the difference ratios between the maximum and minimum values ​​and the average value, compare the difference ratios, obtain the maximum difference ratio Q, and calculate the adjusted ratio Qt. , where To adjust the ratio, 3%≤ ≤6%; Based on Qt, we can get K standard intervals .

5. A data processing system for flow cytometer according to claim 4, characterized in that: The preprocessing of the photoacoustic signal collected in real time by the flow cytometer is to perform wavelet analysis denoising on the collected photoacoustic signal, and the wavelet analysis denoising adopts the default threshold denoising.

6. A data processing system for flow cytometer according to claim 5, characterized in that: The steps for screening the preprocessed photoacoustic signal data set are as follows: Acquire photoacoustic signal data corresponding to the collected photoacoustic signal, compare K groups of data in the photoacoustic signal data with the K standard intervals recorded in the standard threshold interval one by one, calculate the Euclidean distance between the feature vector and the boundary of the standard interval through the boundary recognition model, and determine whether the K groups of data are located in the corresponding K standard intervals; If it is outside the standard interval, the photoacoustic signal is deleted; If it is within the standard interval, the photoacoustic signal is retained, and the retained photoacoustic signal is analyzed for similarity with the M groups of real photoacoustic signals one by one to obtain M groups of similarity data. The formula for calculating the similarity is as follows; In the formula, is the similarity between the detected photoacoustic signal and the i-th group of real photoacoustic signals, C is the number of photoacoustic signal data types for calculating similarity comparison, is the photoacoustic signal data of the jth type in the collected photoacoustic signal, is the j-th type of photoacoustic signal data extracted from the i-th group of real photoacoustic signals, is the adjustment factor; Obtaining the maximum value among the M groups of similarity data, that is, the maximum similarity, and comparing the maximum similarity with the standard similarity, if the maximum similarity is less than the standard similarity, deleting the photoacoustic signal; If the maximum similarity is greater than or equal to the standard similarity, the photoacoustic signal is retained; The photoacoustic signals retained after the deletion are aggregated to form a set of remaining photoacoustic signals.

7. A data processing system for flow cytometer according to claim 6, characterized in that: The characteristic screening categories include instantaneous intensity, instantaneous frequency and instantaneous phase. The acquisition type cell characteristic screening category is one of them. The value range of the number of identification acquisition points L is 30 to 50. The data interval corresponding to the identification acquisition point is , The characteristic screening category data of the i-th identification collection point in the photoacoustic signal corresponding to the maximum similarity, B is the fluctuation coefficient, 2%≤B≤5%, and the statistical comparison results are compared. The number of data that do not meet the data interval is P group, and the calculated compliance .

8. A data processing system for flow cytometer according to claim 7, characterized in that: The preset intervals include standard data intervals , suspected data interval and remove data intervals , ; The conformity is compared with the preset interval, and the classification algorithm is used to divide the photoacoustic signals in the photoacoustic signal set into a standard set, a suspected set and a rejection set.

9. A data processing system for flow cytometer according to claim 8, characterized in that: The steps for comprehensive analysis of the suspected concentrated photoacoustic signal are as follows: The photoacoustic signal in the suspected concentration and its corresponding photoacoustic signal with the maximum similarity are equally divided into E segments by using signal processing technology; Compare the pulse heights in segment E to obtain the half-maximum full width and frequency bandwidth of the photoacoustic signal in the segment with the highest pulse height; The half-width value and frequency bandwidth of the suspected concentrated photoacoustic signal are compared with the half-width value and frequency bandwidth of the photoacoustic signal corresponding to the maximum similarity, and the cell similarity index is calculated. ; The specific calculation formula is as follows: In the formula, is the half-maximum full width of the suspected concentrated photoacoustic signal, The maximum similarity corresponds to the half-maximum full width of the photoacoustic signal. and are weight ratios, is the frequency bandwidth of the suspected concentrated photoacoustic signal, is the frequency bandwidth of the photoacoustic signal corresponding to the maximum similarity; If the cell similarity index is less than the set value, the photoacoustic signal is divided into a rejection set; If the cell similarity index is greater than or equal to the set value, the photoacoustic signal is classified into the standard set.

10. A data processing method for flow cytometer, using any system of claims 1 to 9, characterized in that: The steps include: Acquire photoacoustic signal data of the type of cells to be collected from the database, process the photoacoustic signal data, calculate the standard threshold interval, and obtain the characteristic screening category; Preprocessing the photoacoustic signals collected by the flow cytometer to obtain a photoacoustic signal data set; The preprocessed photoacoustic signal data set is screened by using the standard threshold interval, and the photoacoustic signals that do not meet the standard threshold interval are eliminated to obtain the remaining photoacoustic signal set; Obtain the number of identification collection points and the corresponding data interval corresponding to the collection type cell characteristic screening category, obtain the data corresponding to the number of identification collection points of the photoacoustic signal in the remaining photoacoustic signal set, and compare it with the data interval, statistically compare the results, and calculate the degree of conformity; The conformity is compared with the preset interval, and the classification algorithm is used to divide the photoacoustic signals in the photoacoustic signal set into a standard set, a suspected set and a rejection set. The photoacoustic signals in the suspected set are comprehensively analyzed and divided again. The number of photoacoustic signals in the standard set is counted, and the data and conformity are marked to generate a detection data report.

Images