Health management blood fat detection method and system based on data recording and analysis

Abstract

The invention discloses a health management blood fat detection method and system based on data recording and analysis, belongs to the technical field of in-vitro detection, and solves the problem that the sensitivity of a blood fat instrument is reduced due to the fact that existing detection equipment cannot identify and detect dirt on the surface of a slot. The method comprises the following steps: acquiring multiple groups of sample detection channel data of the blood fat analyzer based on a photoelectric sensor, performing conversion and normalization processing on the sample detection channel data, and executing a pre-constructed self-learning model to calculate the cleanliness of a detection channel by taking the channel normalization data as input; the detection parameter information serves as input, and a blood fat analysis result is obtained through calculation; data of multiple groups of sample detection channels of the blood fat analyzer are obtained based on the photoelectric sensor, and then the cleanliness of the detection channels is calculated based on the self-learning model, so that the cleanliness of the detection channels can be ensured before blood fat detection analysis, the cleanliness of the detection channels is prevented from being influenced by dirt, and the accuracy of a detection result is improved.

Description

Technical Field

[0001] The present invention belongs to the field of in vitro detection technology, and specifically relates to a health management blood lipid detection method and system based on data record analysis. Background Art

[0002] At present, blood lipid testing is a common test method for evaluating cardiovascular health. Blood lipids mainly include indicators such as cholesterol and triglycerides, which can help assess whether patients are at risk of hyperlipidemia. When testing blood lipids, a blood lipid meter is needed for testing. A blood lipid meter is an instrument that can measure blood lipids. As people pay more attention to their health, many families will be equipped with blood lipid testers. Blood lipid testers are simple to operate and produce results quickly, and are suitable for people of all ages.

[0003] Chinese patent CN115469087B discloses an intelligent blood lipid detection device for health management with data recording and analysis, including an upper shell and a lower shell, the upper shell is snapped into the upper end of the lower shell, a display screen is provided on the left side of the upper end of the upper shell, and an operation key is provided in the middle of the upper end of the upper shell, a slot is provided in the middle of the right side of the upper end of the upper shell, and a detection part is provided on the bottom wall of the slot; a protective mechanism is slidably connected to the inside of the slot, and the protective mechanism includes a slide plate slidably connected to the inside of the slot, and a sealing plate is fixedly connected to the right side of the slide plate, and a storage slot is provided on the upper part of the left side wall of the slot; the existing detection equipment uses a storage slot and a detector slot to only block part of the dirt, but cannot identify and detect the dirt on the surface of the slot, resulting in reduced sensitivity of the blood lipid meter and easy error. In response to the above problems, we propose a health management blood lipid detection method and system based on data recording and analysis. Summary of the invention

[0004] The purpose of the present invention is to address the deficiencies of the prior art and provide a health management blood lipid detection method and system based on data record analysis, which solves the problem that the existing detection equipment uses a storage groove and a detector slot to only block part of the dirt, but cannot identify and detect the dirt on the surface of the slot, resulting in a reduced sensitivity of the blood lipid meter and easy error.

[0005] The present invention is implemented in this way: a health management blood lipid detection method based on data record analysis, the health management blood lipid detection method based on data record analysis includes: Before blood lipid testing, the blood lipid analyzer is pre-started, multiple groups of sample detection channel data of the blood lipid analyzer are obtained based on the photoelectric sensor, and the sample detection channel data are converted and normalized to obtain channel normalized data; Load channel normalization data, use the channel normalization data as input, execute the pre-built self-learning model to calculate the cleanliness of the detection channel, and determine whether the cleanliness of the detection channel is greater than the preset clean threshold. If it is greater than the preset clean threshold, trigger the sample injection instruction; if it is less than the preset clean threshold, trigger the dirt warning instruction; In response to the sample injection instruction, the blood samples enter the plurality of sample detection channels respectively, are mixed with the blood samples by the detection auxiliary agent, and are irradiated and detected by the light source detection module to obtain sample detection signals; Loading sample detection signals and sample spectral absorption data, converting and normalizing the sample detection signals and sample spectral absorption data to obtain detection parameter information, wherein the detection parameter information includes platelet data, reticulocyte data, white blood cell data, total cholesterol data, low-density lipoprotein data, high-density lipoprotein data, and triglyceride data; A blood lipid detection model is pre-built, and the detection parameter information is used as input to execute the blood lipid detection model, calculate the blood lipid analysis results, and determine whether the blood lipid analysis results exceed the preset analysis threshold. If the preset analysis threshold is exceeded, the detection abnormality instruction is triggered. If the preset analysis threshold is not exceeded, the blood lipid analysis results are output.

[0006] In response to the abnormal detection instruction, the density clustering algorithm based on the blood lipid detection model identifies the abnormal clustering points of the sample, calculates the abnormal clustering coefficient of the sample, and determines whether the abnormal clustering coefficient of the sample exceeds the preset clustering threshold. If it exceeds the preset clustering threshold, it is determined that abnormal clustering exists in the blood sample.

[0007] Preferably, the method for converting and normalizing the sample detection channel data specifically includes: Acquire multiple groups of sample detection channel data of the blood lipid analyzer, convert the detection channel data into a format, and obtain the sample detection channel data electrical signal data; Establishing a raw data feature table based on the collected sample detection channel data, and marking the features of the sample detection channel data in the raw data feature table, using a rectangular annotation tool to mark the feature coordinates and labels of the raw data feature table, and saving the raw data feature table, feature coordinates and labels; Retrieve the standard data set of the sample detection channel, compare the standard data set with the original data feature table, identify the local outliers in the original data feature table, traverse the original data feature table, obtain the feature coordinates and labels corresponding to the local outliers, and integrate them to obtain the channel normalized data.

[0008] Preferably, the self-learning model is iteratively trained based on a self-attention mechanism model, the self-learning model uses channel normalized data as a training set, and the self-learning model includes an input layer, a fully connected layer, a resampler, and an activation layer; The method of executing a pre-built self-learning model to calculate the cleanliness of a detection channel includes: The input layer traverses the channel normalized data, identifies the local outliers in the channel normalized data, and obtains the feature vectors of the local outliers; The fully connected layer performs self-attention weighting on the feature vector of the local outlier based on the self-attention mechanism, and then transforms the weighted feature vector of the local outlier into the feature vector of the local outlier containing the hidden quantity through wavelet transform; Among them, the weighted calculation formula of the self-attention mechanism is expressed as follows: (1) in, The outlier weight value for the feature vector of the local outlier is weighted by self-attention, is the dimension of the feature vector, is the feature vector input representation of the local outlier point, and is the weighting function, is the attention value of the self-attention mechanism of the local outlier point; (2) in, is the activation function, They are the query vector Query, key vector Key and value vector Value after feature vector conversion. is the feature vector input representation of the local outlier point, is the label vector of the local outlier point, is the mixing coefficient of the activation function, Calculated by matrix multiplication respectively; (3) in, is the weighting function, and is the cosine similarity representation of the correlation scoring function between the feature vector and the label vector, is the number of outliers, is the deviation coefficient of the weighting function; The resampler resamples the local outlier points based on the feature vector of the local outlier points containing the hidden quantity to obtain an enhanced feature vector; The activation layer locally enhances the enhanced feature vector through the activation function and introduces the Gini coefficient to calculate the cleanliness of the detection channel; (4) in, To test the cleanliness of the channel, is the weighted value of the outlier point, Enhance the value for the outlier point; (5) in, is the weighted value of the outlier point, is the outlier enhancement value.

[0009] Preferably, the method for constructing the blood lipid detection model specifically includes: Based on the data recording module, a complex data set for training the blood lipid detection model is obtained, and the complex data set is divided into a training set and a test set, where the ratio of the training set to the test set is 3:1; The EfficientNet model backbone network that introduces the density clustering algorithm is used to extract feature information of complex data sets; The Transformer structure is used to improve the MBConv convolution module in the EfficientNet model, set the loss function of the blood lipid detection model, and build a blood lipid detection model; Obtain a training set, and perform alternate training on the blood lipid detection model based on the training set until the target model converges or reaches Nash equilibrium; The test set is used to verify the accuracy of the blood lipid detection model to verify whether the accuracy of the blood lipid detection model reaches the preset accuracy. If it exceeds the preset accuracy, the blood lipid detection model is output.

[0010] Preferably, the blood lipid detection model includes: The input layer is used to identify the parameter detection sequence input of the complex data set input, and the number of channels of the input layer is 64; Encoder, used to encode the parameter detection sequence input into the target detection network of the blood lipid detection model. The number of encoder channels is 128; Among them, the input sequence of the encoder is expressed as: (6) in, is the input sequence, For parameters The input value of For parameters The data dimension, is the length of the input sequence; C2f module, with 256 channels, is used to fuse sample detection signals and sample spectral absorption data to improve the detection accuracy of the blood lipid detection model; Improved MBConv convolution module with 512 channels for re-weighting parameter detection sequence; SE attention mechanism module, with 512 channels, identifies abnormal clustering points of samples based on the attention mechanism.

[0011] Preferably, the method of executing the blood lipid detection model to calculate the blood lipid analysis results specifically includes: The input sequence of detection parameter information is obtained. The blood lipid detection model introduces a sliding window mechanism, and the blood sample is segmented at equal intervals based on machine vision to obtain multiple groups of segmentation windows. The C2f module integrates the sample detection signal and sample spectral absorption data features in the detection parameter information to obtain the number of platelets, reticulocytes, leukocytes, total cholesterol, low-density lipoprotein, high-density lipoprotein and triglycerides in a single group segmentation window; The real-time window sliding strategy of the MBConv convolution module was improved, and the total number of platelets, reticulocytes, leukocytes, total cholesterol, low-density lipoprotein, high-density lipoprotein and triglycerides in blood samples was obtained by combining the reweighted processing of the parameter detection sequence.

[0012] Preferably, the method of identifying abnormal clustering points of samples and calculating abnormal clustering coefficients of samples by using a density clustering algorithm based on a blood lipid detection model specifically includes: Load the blood lipid analysis results. The blood lipid detection model defines the abnormal clustering energy function based on the SimAM attention mechanism; Among them, the abnormal clustering energy function is expressed as: (7) in, is the abnormal clustering energy value, is the particle detection density in the lipid analysis results, is the characteristic output of sample spectral absorption data, Output mean variance of sample spectral absorption data features, is the total number of target cluster points in the blood sample; Based on the abnormal clustering energy function, the target clustering points in the blood sample are found, and the weighted fusion algorithm is used to perform weighted fusion on the target clustering points to obtain the weighted fusion value of the clustering points; (8) in, is the weighted fusion value of the cluster points, is the abnormal clustering energy value; The density clustering algorithm is used to identify sample abnormal clustering points in the clustering points, and the sample abnormal clustering coefficient is calculated; (9) in, is the abnormal clustering coefficient of abnormal clustering points, is the weighted fusion value of the cluster points, is the bias coefficient of the density function in the density clustering algorithm.

[0013] The present invention also provides a health management blood lipid detection system based on data record analysis, and the health management blood lipid detection system based on data record analysis includes: The photoelectric sensing module obtains multiple groups of sample detection channel data of the blood lipid analyzer based on the photoelectric sensor, converts and normalizes the sample detection channel data, and obtains channel normalized data; Channel detection module: The channel detection module is used to load channel normalization data, take the channel normalization data as input, execute the pre-built self-learning model to calculate the cleanliness of the detection channel, and determine whether the cleanliness of the detection channel is greater than the preset clean threshold. If it is greater than the preset clean threshold, the sample injection instruction is triggered; if it is less than the preset clean threshold, the dirt warning instruction is triggered; The light source detection module responds to the sample injection instruction, and the blood samples enter the plurality of sample detection channels respectively, and the blood samples are irradiated and detected by mixing the detection auxiliary agent with the blood samples to obtain the sample detection signal; A parameter information processing module, which is used to load sample detection signals and sample spectral absorption data, convert and normalize the sample detection signals and sample spectral absorption data, and obtain detection parameter information, wherein the detection parameter information includes platelet data, reticulocyte data, leukocyte data, total cholesterol data, low-density lipoprotein data, high-density lipoprotein data, and triglyceride data; The blood lipid analysis module pre-builds a blood lipid detection model, takes the detection parameter information as input, executes the blood lipid detection model, calculates the blood lipid analysis result, and determines whether the blood lipid analysis result exceeds the preset analysis threshold. If it exceeds the preset analysis threshold, the detection abnormality instruction is triggered. If it does not exceed the preset analysis threshold, the blood lipid analysis result is output; A cluster identification module, in response to an abnormality detection instruction, identifies abnormal clustering points of the sample based on a density clustering algorithm of a blood lipid detection model, calculates an abnormal clustering coefficient of the sample, and determines whether the abnormal clustering coefficient of the sample exceeds a preset clustering threshold. If it exceeds the preset clustering threshold, it is determined that abnormal clustering exists in the blood sample.

[0014] Preferably, the photoelectric sensor module comprises: A format conversion unit, the format conversion unit is used to obtain multiple groups of sample detection channel data of the blood lipid analyzer, perform format conversion on the detection channel data, and obtain sample detection channel data electrical signal data; A feature table construction unit, which establishes a raw data feature table based on the collected sample detection channel data, and marks the features of the sample detection channel data in the raw data feature table, uses a rectangular annotation tool to mark the feature coordinates and labels of the raw data feature table, and saves the raw data feature table, feature coordinates and labels; Data normalization unit, the data normalization unit is used to retrieve the standard data set of the sample detection channel, compare the standard data set with the original data feature table, identify the local abnormal points in the original data feature table, traverse the original data feature table, obtain the feature coordinates and labels corresponding to the local abnormal points, and integrate them to obtain channel normalized data.

[0015] Compared with the prior art, the embodiments of the present application have the following beneficial effects: In an embodiment of the present invention, multiple groups of sample detection channel data of a blood lipid analyzer are obtained based on a photoelectric sensor, and then the cleanliness of the detection channel is calculated based on a self-learning model to determine whether the cleanliness of the detection channel is greater than a preset cleanliness threshold. This ensures the cleanliness of the detection channel before blood lipid detection and analysis, avoids the cleanliness of the detection channel being affected by dirt, and further interferes with the blood lipid detection and analysis results, thereby improving the accuracy of the detection results.

[0016] In an embodiment of the present invention, a health management blood lipid detection method based on data record analysis is provided. When detecting blood lipids, the sample detection signal and the sample spectral absorption data are combined to characterize the particles and cells in the blood lipids, and then the blood lipid detection model is executed to calculate the blood lipid analysis results, thereby ensuring the accuracy of blood lipid detection and significantly reducing detection errors.

[0017] In the embodiment of the present invention, a blood lipid detection model is constructed by improving the EfficientNet model, and the blood lipid detection model introduces the EfficientNet model backbone network of the density clustering algorithm to extract the characteristic information of complex data sets, thereby realizing the effective fusion of photoelectric vision and spectral absorption, which can not only ensure the accuracy of blood lipid analysis and detection, but also identify abnormal aggregation points of blood lipid particles, which is aimed at the limitations of the current biochemical analyzer that the blood lipid test time is too long and the abnormal particle aggregation cannot be identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart for implementing the health management blood lipid detection method based on data record analysis provided by the present invention.

[0019] Figure 2 It is a flow chart of the implementation of the method for converting and normalizing sample detection channel data provided by the present invention.

[0020] Figure 3 It is a flow chart of the implementation of the method for calculating the cleanliness of a detection channel by executing a pre-built self-learning model provided by the present invention.

[0021] Figure 4 It is a flow chart of the implementation of the method for constructing a blood lipid detection model provided by the present invention.

[0022] Figure 5It is a flow chart of the implementation of the method for executing the blood lipid detection model and calculating the blood lipid analysis results provided by the present invention.

[0023] Figure 6 The present invention provides a flow chart for implementing a method for identifying abnormal clustering points of samples and calculating abnormal clustering coefficients of samples using a density clustering algorithm based on a blood lipid detection model.

[0024] Figure 7 It is a structural schematic diagram of the health management blood lipid detection system based on data record analysis provided by the present invention. DETAILED DESCRIPTION

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of this application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0026] The existing detection equipment uses storage slots and detector slots to only block some dirt, but cannot identify and detect dirt on the surface of the slot, resulting in reduced sensitivity of the blood lipid analyzer and prone to errors. In response to the above problems, we proposed a health management blood lipid detection method and system based on data record analysis. In short, the method includes obtaining multiple groups of sample detection channel data of the blood lipid analyzer based on a photoelectric sensor, converting and normalizing the sample detection channel data, using the channel normalization data as input, executing a pre-built self-learning model to calculate the cleanliness of the detection channel, and judging whether the cleanliness of the detection channel is greater than the preset clean threshold. After determining that the detection channel is clean, the blood samples enter the multiple groups of sample detection channels respectively, are mixed with the blood samples through the detection auxiliary agent, and are irradiated and detected based on the light source detection module to obtain the sample detection signal, and then the blood lipid detection model is executed with the detection parameter information as input to calculate the blood lipid analysis results. In an embodiment of the present invention, multiple groups of sample detection channel data of a blood lipid analyzer are obtained based on a photoelectric sensor, and then the cleanliness of the detection channel is calculated based on a self-learning model to determine whether the cleanliness of the detection channel is greater than a preset cleanliness threshold. This ensures the cleanliness of the detection channel before blood lipid detection and analysis, avoids the cleanliness of the detection channel being affected by dirt, and further interferes with the blood lipid detection and analysis results, thereby improving the accuracy of the detection results.

[0027] The embodiment of the present invention provides a health management blood lipid detection method based on data record analysis. Figure 1The implementation flow chart of the health management blood lipid detection method based on data record analysis is shown. The health management blood lipid detection method based on data record analysis specifically includes: Step S10, before blood lipid detection, pre-start the blood lipid analyzer, obtain multiple groups of sample detection channel data of the blood lipid analyzer based on the photoelectric sensor, convert and normalize the sample detection channel data, and obtain channel normalized data; It should be noted that the photoelectric sensor is built into the blood lipid analyzer. The photoelectric sensor includes but is not limited to a through-type sensor, a reflective sensor, and a slot sensor. The photoelectric sensor uses a light source such as an LED or a laser to illuminate the target object (sample detection channel). In this embodiment, the sample detection channel is used for sample matching, detection and analysis. At least one group of sample detection channels is provided, and the sample detection channel can be a circular slot, a square slot, or a rectangular slot structure.

[0028] Step S20, loading the channel normalization data, taking the channel normalization data as input, executing the pre-built self-learning model to calculate the cleanliness of the detection channel; After the cleanliness of the detection channel is calculated by the self-learning model, considering that blood testing and blood lipid analysis require high accuracy, a preset cleanliness threshold is required to measure the cleanliness of the detection channel. The embodiment of the present invention further includes step S30, determining whether the cleanliness of the detection channel is greater than the preset cleanliness threshold. If it is greater than the preset cleanliness threshold, a sample injection instruction is triggered, wherein the cleanliness threshold can be set to 0.98-1; Step S40: If the value is less than the preset clean threshold, a dirt warning instruction is triggered. The blood lipid analyzer responds to the dirt warning instruction, starts the channel self-cleaning mode, and issues an audible and visual alarm.

[0029] Step S50, in response to the sample injection instruction, the blood samples enter the plurality of sample detection channels respectively, the blood samples are mixed with the detection auxiliary agent, and the blood samples are irradiated and detected based on the light source detection module to obtain the sample detection signal; Among them, detection auxiliary agents include but are not limited to fluorescent dyes, hemolytic agents, cholesterol oxidase, triglyceride esterase, and phospholipase A2. By acting on blood samples with detection auxiliary agents, blood / blood lipid particles / cells can be calibrated, thereby facilitating the light source detection module to receive scattered light and fluorescence under light irradiation.

[0030] Step S60, loading the sample detection signal and the sample spectral absorption data, converting and normalizing the sample detection signal and the sample spectral absorption data to obtain detection parameter information, wherein the detection parameter information includes platelet data, reticulocyte data, white blood cell data, total cholesterol data, low-density lipoprotein data, high-density lipoprotein data and triglyceride data; Step S70, pre-constructing a blood lipid detection model, taking the detection parameter information as input, executing the blood lipid detection model, and calculating the blood lipid analysis result; Step S80, determine whether the blood lipid analysis result exceeds the preset analysis threshold. In this embodiment, the preset analysis threshold is divided into multiple groups, and the preset analysis threshold is set based on platelet data, reticulocyte data, white blood cell data, total cholesterol data, low-density lipoprotein data, high-density lipoprotein data and triglyceride data.

[0031] Step S90, if it exceeds the preset analysis threshold, trigger the abnormal detection instruction; Step S110: if the blood lipid analysis result is not exceeded, output the blood lipid analysis result.

[0032] Step S120, in response to the abnormality detection instruction, identifying abnormal clustering points of the sample based on the density clustering algorithm of the blood lipid detection model, and calculating the abnormal clustering coefficient of the sample; Step S130, determining whether the sample abnormal clustering coefficient exceeds a preset clustering threshold; Step S140: If the value exceeds the preset clustering threshold, it is determined that the blood sample has abnormal clustering. In this embodiment, the preset clustering threshold is set to 0.6-0.8.

[0033] Step S150: If the number of clusters does not exceed the preset clustering threshold, it is determined that there is no abnormal clustering in the blood sample.

[0034] In an embodiment of the present invention, a health management blood lipid detection method based on data record analysis is provided. When detecting blood lipids, the sample detection signal and the sample spectral absorption data are combined to characterize the particles and cells in the blood lipids, and then the blood lipid detection model is executed to calculate the blood lipid analysis results, thereby ensuring the accuracy of blood lipid detection and significantly reducing detection errors.

[0035] In an embodiment of the present invention, multiple groups of sample detection channel data of a blood lipid analyzer are obtained based on a photoelectric sensor, and then the cleanliness of the detection channel is calculated based on a self-learning model to determine whether the cleanliness of the detection channel is greater than a preset cleanliness threshold. This ensures the cleanliness of the detection channel before blood lipid detection and analysis, avoids the cleanliness of the detection channel being affected by dirt, and further interferes with the blood lipid detection and analysis results, thereby improving the accuracy of the detection results.

[0036] The embodiment of the present invention provides a method for converting and normalizing sample detection channel data. Figure 2 The flowchart of the method for converting and normalizing the sample detection channel data is shown. The method for converting and normalizing the sample detection channel data specifically includes: Step S101, obtaining multiple groups of sample detection channel data of a blood lipid analyzer, performing format conversion on the detection channel data, and obtaining sample detection channel data electrical signal data; When the detection channel data is formatted, the photodiode, phototransistor or other type of photosensitive device converts the optical signal into an electrical signal, which is then sent to the processing circuit for amplification, filtering and digitization. This step ensures that the signal output from the photoelectric sensor can be accurately parsed and used.

[0037] Step S102, establishing a raw data feature table based on the collected sample detection channel data, and marking the features of the sample detection channel data in the raw data feature table, using a rectangular annotation tool to mark the feature coordinates and labels of the raw data feature table, and saving the raw data feature table, feature coordinates and labels; Step S103, retrieve the standard data set of the sample detection channel, compare the standard data set with the original data feature table, identify local abnormal points in the original data feature table, traverse the original data feature table, obtain the feature coordinates and labels corresponding to the local abnormal points, and integrate them to obtain channel normalized data.

[0038] In an embodiment of the present invention, by converting and normalizing the sample detection channel data, local abnormal points in the detection channel can be accurately identified, and the characteristic coordinates and labels corresponding to the local abnormal points can also be collected, thereby facilitating the self-learning model to identify and judge the surface cleanliness of the detection channel.

[0039] It should be noted that, in an embodiment of the present invention, the self-learning model is iteratively trained based on a self-attention mechanism model. The self-learning model uses channel normalized data as a training set. The self-learning model includes an input layer, a fully connected layer, a resampler, and an activation layer. The input layer is connected to the fully connected layer, the fully connected layer is connected to the resampler, and the resampler is connected to the activation layer.

[0040] The embodiment of the present invention provides a method for executing a pre-built self-learning model to calculate the cleanliness of a detection channel. Figure 3 The implementation flow chart of the method for calculating the cleanliness of the detection channel by executing the pre-built self-learning model is shown. The method for calculating the cleanliness of the detection channel by executing the pre-built self-learning model specifically includes: Step S201, the input layer traverses the channel normalized data, identifies local outliers in the channel normalized data, and obtains feature vectors of the local outliers; Among them, the abnormality of the channel normalized data is measured by the feature vector of the local abnormal point, so as to obtain a more accurate detection of channel defect points.

[0041] Step S202, the fully connected layer performs self-attention weighting on the feature vector of the local outlier based on the self-attention mechanism, and then converts the weighted feature vector of the local outlier into a feature vector of the local outlier containing the hidden amount through wavelet transformation; Among them, converting the feature vector of the weighted local outlier point into the feature vector of the local outlier point containing the hidden amount after wavelet transform can correct the photoelectric detection error, thereby realizing the identification of the hidden outlier point, and fully connecting the hidden outlier point with the local outlier point to obtain the feature vector of the corrected local outlier point.

[0042] Among them, the weighted calculation formula of the self-attention mechanism is expressed as follows: (1) in, The outlier weight value for the feature vector of the local outlier is weighted by self-attention, is the dimension of the feature vector, is the feature vector input representation of the local outlier point, and is the weighting function, is the attention value of the self-attention mechanism of the local outlier point; The dimension of the feature vector can be 2-5, and the weighting function can be a hierarchical analysis method weighting function. By building a hierarchical model and comparing the importance of various factors, it is suitable for decision analysis of complex problems.

[0043] (2) in, is the activation function, They are the query vector Query, key vector Key and value vector Value after feature vector conversion. is the feature vector input representation of the local outlier point, is the label vector of the local outlier point, is the mixing coefficient of the activation function. In this embodiment, the mixing coefficient of the activation function is 0.1 or 0.15. Calculated by matrix multiplication respectively; (3) in, is the weighting function, and is the cosine similarity representation of the correlation scoring function between the feature vector and the label vector, is the number of outliers, is the deviation coefficient of the weighting function. In this embodiment, the deviation coefficient of the weighting function may be an integer between 1 and 5.

[0044] Step S203, the resampler resamples the local outlier points based on the feature vector of the local outlier points containing the hidden quantity to obtain an enhanced feature vector; Step S204, the activation layer locally enhances the enhanced feature vector through an activation function, and introduces the Gini coefficient to calculate the cleanliness of the detection channel; (4) in, To test the cleanliness of the channel, is the weighted value of the outlier point, is the outlier enhancement value, and is the gradient parameter of the activation function. In this embodiment, the activation function may be a softmax activation function, and the gradient parameter of the activation function is 0.5-0.6; (5) in, is the weighted value of the outlier point, is the outlier enhancement value.

[0045] The embodiment of the present invention provides a method for constructing a blood lipid detection model. Figure 4 The flowchart of the method for constructing the blood lipid detection model is shown. The method for constructing the blood lipid detection model specifically includes: Step S301, obtaining a complex data set for training a blood lipid detection model based on a data recording module, and dividing the complex data set into a training set and a test set, wherein the ratio of the training set to the test set is 3:1; It should be noted that the complex data set for training the blood lipid detection model is obtained from the OCTA image data set and BCCD data set released by Professor Chen Qiang and his team from Nanjing University of Science and Technology.

[0046] Step S302, extracting feature information of complex data sets using the EfficientNet model backbone network that introduces a density clustering algorithm; Step S303, using the Transformer structure to improve the MBConv convolution module in the EfficientNet model, setting the loss function of the blood lipid detection model, and building the blood lipid detection model; Step S304, obtaining a training set, and performing alternating training on the blood lipid detection model based on the training set until the target model converges or reaches Nash equilibrium; Step S305, using the test set to verify the accuracy of the blood lipid detection model, verifying whether the accuracy of the blood lipid detection model reaches the preset accuracy, if it exceeds the preset accuracy, output the blood lipid detection model.

[0047] In this embodiment, the preset accuracy is set to 0.85-0.8, and the number of times of alternating training of the blood lipid detection model based on the training set can be 200-2000 times, and the learning rate of the blood lipid detection model is 0.002.

[0048] In the embodiment of the present invention, a blood lipid detection model is constructed by improving the EfficientNet model, and the blood lipid detection model introduces the EfficientNet model backbone network of the density clustering algorithm to extract the characteristic information of complex data sets, thereby realizing the effective fusion of photoelectric vision and spectral absorption, which can not only ensure the accuracy of blood lipid analysis and detection, but also identify abnormal aggregation points of blood lipid particles, which is aimed at the limitations of the current biochemical analyzer that the blood lipid test time is too long and the abnormal particle aggregation cannot be identified.

[0049] Among them, in the embodiment of the present invention, the blood lipid detection model includes: An input layer is used to identify the parameter detection sequence input of a complex data set. The number of channels of the input layer is 64, and the input layer is connected to the encoder. Encoder, used to encode the parameter detection sequence input into the target detection network of the blood lipid detection model. The number of encoder channels is 128; Among them, the input sequence of the encoder is expressed as: (6) in, is the input sequence, For parameters The input value of For parameters In this embodiment, the parameter The data dimension can be 2-10. is the length of the input sequence; C2f module, with 256 channels, is used to fuse sample detection signals and sample spectral absorption data to improve the detection accuracy of the blood lipid detection model; It should be noted that the C2f module has three layers of convolution, and the convolution kernel size is 3*3. The C2f module is an efficient network structure unit and is widely used in deep learning and computer vision. It optimizes the information flow and feature extraction process through its unique design.

[0050] Improved MBConv convolution module with 512 channels for re-weighting parameter detection sequence; SE attention mechanism module, with 512 channels, identifies abnormal clustering points of samples based on the attention mechanism.

[0051] The embodiment of the present invention provides a method for executing a blood lipid detection model and calculating a blood lipid analysis result. Figure 5 The flowchart of the method for executing the blood lipid detection model and calculating the blood lipid analysis result is shown. The method for executing the blood lipid detection model and calculating the blood lipid analysis result specifically includes: Step S401, obtaining an input sequence of detection parameter information, introducing a sliding window mechanism into the blood lipid detection model, performing equal-interval segmentation on the blood sample based on machine vision, and obtaining multiple groups of segmentation windows; Step S402, the C2f module fuses the sample detection signal and the sample spectral absorption data characteristics in the detection parameter information to obtain the number of platelets, reticulocytes, white blood cells, total cholesterol, low-density lipoprotein, high-density lipoprotein and triglycerides in a single group segmentation window; Step S403, improving the real-time window sliding strategy of the MBConv convolution module, combined with reweighting the parameter detection sequence, to obtain the total number of platelets, reticulocytes, leukocytes, total cholesterol, low-density lipoprotein, high-density lipoprotein and triglycerides in the blood sample.

[0052] In this embodiment, the provided method for calculating the blood lipid analysis results can integrate the sample detection signal and the sample spectral absorption data characteristics in the detection parameter information, thereby obtaining the total number of platelets, reticulocytes, white blood cells, total cholesterol, low-density lipoprotein, high-density lipoprotein and triglycerides in the blood sample, providing data support for blood and blood lipid analysis.

[0053] The embodiment of the present invention provides a method for identifying abnormal clustering points of samples and calculating abnormal clustering coefficients of samples based on a density clustering algorithm of a blood lipid detection model. Figure 6 The flowchart of the method for identifying abnormal clustering points of samples and calculating abnormal clustering coefficients of samples by using the density clustering algorithm based on the blood lipid detection model is shown. The method for identifying abnormal clustering points of samples and calculating abnormal clustering coefficients of samples by using the density clustering algorithm based on the blood lipid detection model specifically includes: Step S501, loading blood lipid analysis results, the blood lipid detection model defines an abnormal clustering energy function based on the SimAM attention mechanism; Among them, the abnormal clustering energy function is expressed as: (7) in, It is the abnormal clustering energy value. It should be noted that the abnormal clustering energy value is a powerful tool for judging the abnormal clustering of particles. It can help people better understand and analyze the clustering structure in the data set. Through further research and experimental verification, in cluster analysis, the energy value can be understood as an abstract concept used to describe the degree of compactness or separation of clusters. At the same time, once the cluster center is determined, the clustering of data points is carried out according to the nearest neighbor rule, that is, each data point is assigned to the cluster represented by the nearest cluster center. is the particle detection density in the lipid analysis results, is the characteristic output of sample spectral absorption data, Output mean variance of sample spectral absorption data features, is the total number of target cluster points in the blood sample; Step S502, searching for target cluster points in the blood sample based on the abnormal clustering energy function, and using a weighted fusion algorithm to perform weighted fusion on the target cluster points to obtain weighted fusion values ​​of the cluster points; (8) in, is the weighted fusion value of the cluster points, is the abnormal clustering energy value; Step S503, using a density clustering algorithm to identify sample abnormal clustering points among the clustering points, and calculating the sample abnormal clustering coefficient; (9) in, is the abnormal clustering coefficient of abnormal clustering points, is the weighted fusion value of the cluster points, is the bias coefficient of the density function in the density clustering algorithm. In this embodiment, the bias coefficient of the density function in the density clustering algorithm may be 1-3.

[0054] In this embodiment, the sample abnormal clustering coefficient is used to characterize the degree of discreteness of clustering of different blood and blood lipid particles. The present invention adopts a density clustering algorithm to identify sample abnormal clustering points in the clustering points to suppress clustering points with smaller weights, and improves the recognition accuracy of the blood lipid detection model through the density clustering algorithm.

[0055] The embodiment of the present invention also provides a health management blood lipid detection system based on data record analysis. Figure 7 The overall block diagram of the health management blood lipid detection system based on data record analysis is shown, and the health management blood lipid detection system based on data record analysis specifically includes: The photoelectric sensor module 100 obtains multiple groups of sample detection channel data of the blood lipid analyzer based on the photoelectric sensor, converts and normalizes the sample detection channel data, and obtains channel normalized data; The channel detection module 200 is used to load the channel normalization data, take the channel normalization data as input, execute the pre-built self-learning model to calculate the cleanliness of the detection channel, and determine whether the cleanliness of the detection channel is greater than the preset clean threshold. If it is greater than the preset clean threshold, a sample injection instruction is triggered; if it is less than the preset clean threshold, a dirt warning instruction is triggered; The light source detection module 300, in response to the sample injection instruction, causes the blood samples to enter the plurality of sample detection channels respectively, mixes the detection auxiliary agent with the blood samples, performs irradiation detection on the blood samples, and obtains the sample detection signal; The parameter information processing module 400 is used to load the sample detection signal and the sample spectral absorption data, convert and normalize the sample detection signal and the sample spectral absorption data, and obtain the detection parameter information, wherein the detection parameter information includes platelet data, reticulocyte data, leukocyte data, total cholesterol data, low-density lipoprotein data, high-density lipoprotein data and triglyceride data; It should be noted that the sample detection signal is obtained through the photoelectric sensor of the light source detection module 300, and the sample spectral absorption data is obtained based on ultraviolet-visible spectrum, infrared spectrum, and Raman spectrum counting.

[0056] The blood lipid analysis module 500 pre-builds a blood lipid detection model, takes the detection parameter information as input, executes the blood lipid detection model, calculates the blood lipid analysis result, determines whether the blood lipid analysis result exceeds the preset analysis threshold, and triggers the abnormal detection instruction if it exceeds the preset analysis threshold. If it does not exceed the preset analysis threshold, the blood lipid analysis result is output; The cluster identification module 600 responds to the abnormality detection instruction, identifies the abnormal clustering points of the sample based on the density clustering algorithm of the blood lipid detection model, calculates the abnormal clustering coefficient of the sample, and determines whether the abnormal clustering coefficient of the sample exceeds the preset clustering threshold. If it exceeds the preset clustering threshold, it is determined that abnormal clustering exists in the blood sample.

[0057] The embodiment of the present invention further provides a photoelectric sensor module 100, Figure 7 The control structure diagram of the photoelectric sensor module 100 is shown. The photoelectric sensor module specifically includes: The format conversion unit 110 is used to obtain multiple groups of sample detection channel data of the blood lipid analyzer, perform format conversion on the detection channel data, and obtain sample detection channel data electrical signal data; A feature table construction unit 120, which establishes a raw data feature table based on the collected sample detection channel data, and marks the features of the sample detection channel data in the raw data feature table, uses a rectangular annotation tool to mark the feature coordinates and labels of the raw data feature table, and saves the raw data feature table, feature coordinates and labels; The data normalization unit 130 is used to retrieve the standard data set of the sample detection channel, compare the standard data set with the original data feature table, identify local abnormal points in the original data feature table, traverse the original data feature table, obtain the feature coordinates and labels corresponding to the local abnormal points, and integrate them to obtain channel normalized data.

[0058] On the other hand, the present invention further provides a computer-readable storage medium, which stores computer program instructions, and the computer program instructions can be executed by a processor. When the computer program instructions are executed, the method of any one of the above embodiments is implemented.

[0059] Another aspect of the embodiments of the present invention further provides a computer device, which includes a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the method of any one of the above embodiments is implemented.

[0060] As a non-volatile computer-readable storage medium, the memory can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as the program instructions / modules corresponding to the health management blood lipid detection method based on data record analysis in the embodiment of the present application. The memory may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application required for at least one function; the storage data area may store data created by the use of the health management blood lipid detection method based on data record analysis, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the local module via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0061] Finally, it should be noted that the computer-readable storage medium (e.g., memory) herein may be a volatile memory or a non-volatile memory, or may include both a volatile memory and a non-volatile memory. As an example and not by way of limitation, a non-volatile memory may include a read-only memory (ROM), a programmable ROM (PROM), an electrically programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), or a flash memory. A volatile memory may include a random access memory (RAM), which may act as an external cache memory. As an example and not by way of limitation, RAM may be obtained in a variety of forms, such as synchronous RAM (DRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct Rambus RAM (DRRAM). The storage devices of the disclosed aspects are intended to include, but are not limited to, these and other suitable types of memory.

[0062] In summary, the present invention provides a health management blood lipid detection method and system based on data record analysis. In an embodiment of the present invention, multiple groups of sample detection channel data of a blood lipid analyzer are obtained based on a photoelectric sensor, and then the cleanliness of the detection channel is calculated based on a self-learning model to determine whether the cleanliness of the detection channel is greater than a preset cleanliness threshold. This ensures the cleanliness of the detection channel before blood lipid detection and analysis, avoids the cleanliness of the detection channel being affected by dirt, and then interferes with the blood lipid detection and analysis results, thereby improving the accuracy of the detection results.

[0063] It should be noted that, for the above-mentioned embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described order of actions, because according to the present invention, some steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0064] In the several embodiments provided in the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the above-mentioned units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be in the form of telecommunication or other forms.

[0065] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they 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.

[0066] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on these embodiments, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in this field can still combine, add, delete or make other adjustments to the features in the various embodiments of the present invention according to the circumstances without conflict, without making creative work, so as to obtain different other technical solutions that do not deviate from the concept of the present invention in essence, and these technical solutions also belong to the scope of protection of the present invention.

Claims

1. A health management blood lipid detection method based on data record analysis, characterized in that: The health management blood lipid detection method based on data record analysis includes: Before blood lipid testing, the blood lipid analyzer is pre-started, multiple groups of sample detection channel data of the blood lipid analyzer are obtained based on the photoelectric sensor, and the sample detection channel data are converted and normalized to obtain channel normalized data; Load channel normalization data, use the channel normalization data as input, execute the pre-built self-learning model to calculate the cleanliness of the detection channel, and determine whether the cleanliness of the detection channel is greater than the preset clean threshold. If it is greater than the preset clean threshold, trigger the sample injection instruction; if it is less than the preset clean threshold, trigger the dirt warning instruction; In response to the sample injection instruction, the blood samples enter the plurality of sample detection channels respectively, are mixed with the blood samples by the detection auxiliary agent, and are irradiated and detected by the light source detection module to obtain sample detection signals; Loading sample detection signals and sample spectral absorption data, converting and normalizing the sample detection signals and sample spectral absorption data to obtain detection parameter information, wherein the detection parameter information includes platelet data, reticulocyte data, white blood cell data, total cholesterol data, low-density lipoprotein data, high-density lipoprotein data, and triglyceride data; A blood lipid detection model is pre-built, and the detection parameter information is used as input to execute the blood lipid detection model, calculate the blood lipid analysis results, and determine whether the blood lipid analysis results exceed the preset analysis threshold. If the preset analysis threshold is exceeded, the detection abnormality instruction is triggered. If the preset analysis threshold is not exceeded, the blood lipid analysis results are output.

2. The method for detecting blood lipids for health management based on data record analysis according to claim 1, characterized in that: The method further comprises: In response to the abnormal detection instruction, the density clustering algorithm based on the blood lipid detection model identifies the abnormal clustering points of the sample, calculates the abnormal clustering coefficient of the sample, and determines whether the abnormal clustering coefficient of the sample exceeds the preset clustering threshold. If it exceeds the preset clustering threshold, it is determined that abnormal clustering exists in the blood sample.

3. The method for detecting blood lipids for health management based on data record analysis according to claim 1, characterized in that: The method for converting and normalizing the sample detection channel data specifically includes: Acquire multiple groups of sample detection channel data of the blood lipid analyzer, convert the detection channel data into a format, and obtain the sample detection channel data electrical signal data; Establishing a raw data feature table based on the collected sample detection channel data, and marking the features of the sample detection channel data in the raw data feature table, using a rectangular annotation tool to mark the feature coordinates and labels of the raw data feature table, and saving the raw data feature table, feature coordinates and labels; Retrieve the standard data set of the sample detection channel, compare the standard data set with the original data feature table, identify the local outliers in the original data feature table, traverse the original data feature table, obtain the feature coordinates and labels corresponding to the local outliers, and integrate them to obtain the channel normalized data.

4. The method for detecting blood lipids for health management based on data record analysis according to claim 3, characterized in that: The self-learning model is iteratively trained based on the self-attention mechanism model. The self-learning model uses channel normalized data as a training set. The self-learning model includes an input layer, a fully connected layer, a resampler, and an activation layer. The method of executing a pre-built self-learning model to calculate the cleanliness of a detection channel includes: The input layer traverses the channel normalized data, identifies the local outliers in the channel normalized data, and obtains the feature vectors of the local outliers; The fully connected layer performs self-attention weighting on the feature vector of the local outlier based on the self-attention mechanism, and then transforms the weighted feature vector of the local outlier into the feature vector of the local outlier containing the hidden quantity through wavelet transform; Among them, the weighted calculation formula of the self-attention mechanism is expressed as follows: （1） in, The outlier weight value for the feature vector of the local outlier is weighted by self-attention, is the dimension of the feature vector, is the feature vector input representation of the local outlier point, and is the weighting function, is the attention value of the self-attention mechanism of the local outlier point; （2） in, is the activation function, They are the query vector Query, key vector Key and value vector Value after feature vector conversion. is the feature vector input representation of the local outlier point, is the label vector of the local outlier point, is the mixing coefficient of the activation function, Calculated by matrix multiplication respectively; （3） in, is the weighting function, and is the cosine similarity representation of the correlation scoring function between the feature vector and the label vector, is the number of outliers, is the deviation coefficient of the weighting function; The resampler resamples the local outlier points based on the feature vector of the local outlier points containing the hidden quantity to obtain an enhanced feature vector; The activation layer locally enhances the enhanced feature vector through the activation function and introduces the Gini coefficient to calculate the cleanliness of the detection channel; （4） in, To test the cleanliness of the channel, is the weighted value of the outlier point, Enhance the value for the outlier point; （5） in, is the weighted value of the outlier point, is the outlier enhancement value.

5. The method for detecting blood lipids for health management based on data record analysis according to claim 2, characterized in that: The method for constructing the blood lipid detection model specifically includes: Based on the data recording module, a complex data set for training the blood lipid detection model is obtained, and the complex data set is divided into a training set and a test set, where the ratio of the training set to the test set is 3:1; The EfficientNet model backbone network that introduces the density clustering algorithm is used to extract feature information of complex data sets; The Transformer structure is used to improve the MBConv convolution module in the EfficientNet model, set the loss function of the blood lipid detection model, and build a blood lipid detection model; Obtain a training set, and perform alternate training on the blood lipid detection model based on the training set until the target model converges or reaches Nash equilibrium; The test set is used to verify the accuracy of the blood lipid detection model to verify whether the accuracy of the blood lipid detection model reaches the preset accuracy. If it exceeds the preset accuracy, the blood lipid detection model is output.

6. The method for detecting blood lipids for health management based on data record analysis according to claim 5, characterized in that: The blood lipid detection model includes: The input layer is used to identify the parameter detection sequence input of the complex data set input, and the number of channels of the input layer is 64; Encoder, used to encode the parameter detection sequence input into the target detection network of the blood lipid detection model. The number of encoder channels is 128; Among them, the input sequence of the encoder is expressed as: （6） in, is the input sequence, For parameters The input value of For parameters The data dimension, is the length of the input sequence; C2f module, with 256 channels, is used to fuse sample detection signals and sample spectral absorption data to improve the detection accuracy of the blood lipid detection model; Improved MBConv convolution module with 512 channels for re-weighting parameter detection sequence; SE attention mechanism module, with 512 channels, identifies abnormal clustering points of samples based on the attention mechanism.

7. The method for detecting blood lipids for health management based on data record analysis according to claim 6, characterized in that: The method of executing the blood lipid detection model to calculate the blood lipid analysis results specifically includes: The input sequence of detection parameter information is obtained. The blood lipid detection model introduces a sliding window mechanism, and the blood sample is segmented at equal intervals based on machine vision to obtain multiple groups of segmentation windows. The C2f module integrates the sample detection signal and sample spectral absorption data features in the detection parameter information to obtain the number of platelets, reticulocytes, leukocytes, total cholesterol, low-density lipoprotein, high-density lipoprotein and triglycerides in a single group segmentation window; The real-time window sliding strategy of the MBConv convolution module was improved, and the total number of platelets, reticulocytes, leukocytes, total cholesterol, low-density lipoprotein, high-density lipoprotein and triglycerides in blood samples was obtained by combining the reweighted processing of the parameter detection sequence.

8. The method for detecting blood lipids for health management based on data record analysis according to claim 7, characterized in that: The method of identifying abnormal clustering points of samples and calculating abnormal clustering coefficients of samples by using a density clustering algorithm based on a blood lipid detection model specifically includes: Load the blood lipid analysis results. The blood lipid detection model defines the abnormal clustering energy function based on the SimAM attention mechanism; Among them, the abnormal clustering energy function is expressed as: （7） in, is the abnormal clustering energy value, is the particle detection density in the lipid analysis results, is the characteristic output of sample spectral absorption data, Output mean variance of sample spectral absorption data features, is the total number of target cluster points in the blood sample; Based on the abnormal clustering energy function, the target clustering points in the blood sample are found, and the weighted fusion algorithm is used to perform weighted fusion on the target clustering points to obtain the weighted fusion value of the clustering points; （8） in, is the weighted fusion value of the cluster points, is the abnormal clustering energy value; The density clustering algorithm is used to identify sample abnormal clustering points in the clustering points, and the sample abnormal clustering coefficient is calculated; （9） in, is the abnormal clustering coefficient of abnormal clustering points, is the weighted fusion value of the cluster points, is the bias coefficient of the density function in the density clustering algorithm.

9. A health management blood lipid detection system based on data record analysis, implemented by a health management blood lipid detection method based on data record analysis as claimed in any one of claims 1 to 8, characterized in that: The health management blood lipid detection system based on data record analysis includes: The photoelectric sensing module obtains multiple groups of sample detection channel data of the blood lipid analyzer based on the photoelectric sensor, converts and normalizes the sample detection channel data, and obtains channel normalized data; Channel detection module: The channel detection module is used to load channel normalization data, take the channel normalization data as input, execute the pre-built self-learning model to calculate the cleanliness of the detection channel, and determine whether the cleanliness of the detection channel is greater than the preset clean threshold. If it is greater than the preset clean threshold, the sample injection instruction is triggered; if it is less than the preset clean threshold, the dirt warning instruction is triggered; The light source detection module responds to the sample injection instruction, and the blood samples enter the plurality of sample detection channels respectively, and the blood samples are irradiated and detected by mixing the detection auxiliary agent with the blood samples to obtain the sample detection signal; A parameter information processing module, which is used to load sample detection signals and sample spectral absorption data, convert and normalize the sample detection signals and sample spectral absorption data, and obtain detection parameter information, wherein the detection parameter information includes platelet data, reticulocyte data, leukocyte data, total cholesterol data, low-density lipoprotein data, high-density lipoprotein data, and triglyceride data; The blood lipid analysis module pre-builds a blood lipid detection model, takes the detection parameter information as input, executes the blood lipid detection model, calculates the blood lipid analysis result, and determines whether the blood lipid analysis result exceeds the preset analysis threshold. If it exceeds the preset analysis threshold, the detection abnormality instruction is triggered. If it does not exceed the preset analysis threshold, the blood lipid analysis result is output; A cluster identification module, in response to an abnormality detection instruction, identifies abnormal clustering points of the sample based on a density clustering algorithm of a blood lipid detection model, calculates an abnormal clustering coefficient of the sample, and determines whether the abnormal clustering coefficient of the sample exceeds a preset clustering threshold. If it exceeds the preset clustering threshold, it is determined that abnormal clustering exists in the blood sample.

10. The health management blood lipid detection system based on data record analysis as claimed in claim 9, characterized in that: The photoelectric sensor module comprises: A format conversion unit, the format conversion unit is used to obtain multiple groups of sample detection channel data of the blood lipid analyzer, perform format conversion on the detection channel data, and obtain sample detection channel data electrical signal data; A feature table construction unit, which establishes a raw data feature table based on the collected sample detection channel data, and marks the features of the sample detection channel data in the raw data feature table, uses a rectangular annotation tool to mark the feature coordinates and labels of the raw data feature table, and saves the raw data feature table, feature coordinates and labels; Data normalization unit, the data normalization unit is used to retrieve the standard data set of the sample detection channel, compare the standard data set with the original data feature table, identify the local abnormal points in the original data feature table, traverse the original data feature table, obtain the feature coordinates and labels corresponding to the local abnormal points, and integrate them to obtain channel normalized data.

Images