Triglyceride risk prediction method and system based on near infrared face spectrum
Through near-infrared face spectroscopy technology and machine learning model, non-invasive triglyceride risk prediction is achieved, solving the problems of disease infection and poor experience of traditional detection methods, providing fast and accurate detection methods, which is of great clinical significance.
Patent Information
- Application Number
- CN202411184049.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-12-03
AI Technical Summary
Traditional triglyceride detection methods require blood collection, which poses a risk of disease infection and has a poor experience, making it impossible to achieve non-invasive, fast and accurate detection.
A risk prediction method based on near-infrared face spectrum is adopted to collect spectral samples by scanning faces, preprocessing and constructing a risk prediction model, and predicting triglyceride risk levels using support vector machines and particle swarm optimization algorithms.
It has achieved non-invasive, reagent-free, fast and accurate triglyceride content detection and risk warning, reducing the damage and cost of biochemical tests, and is of great clinical significance.
Smart Images

Figure CN119055196A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of blood component content detection, and specifically, to a triglyceride risk prediction method and system based on near-infrared facial spectroscopy. Background Art
[0002] At present, blood lipids are a general term for lipids contained in human plasma, including cholesterol, triglycerides, cholesterol esters, phospholipids, etc. Hyperlipidemia is a common metabolic disease, which is mainly manifested by abnormally high levels of lipids (including cholesterol and triglycerides) in the blood. As one of the important indicators for determining hyperlipidemia, elevated levels of triglycerides have a significant impact on cardiovascular health. According to the American Heart Association's potential relationship between triglyceride levels and the risk of hyperlipidemia, myocardial infarction and other diseases in the human body, risk screening can be completed based on different triglyceride levels to achieve the purpose of early intervention and early treatment.
[0003] Traditional methods for detecting triglycerides can generally be divided into chemical methods, chromatography methods, and enzymatic methods. All of the above methods can accurately detect triglycerides in human blood. However, since they are all invasive detection methods, blood must be collected in advance and chemical reagents must be used. Therefore, the risk of disease infection cannot be avoided, and the test subject experience is poor. Summary of the invention
[0004] The main purpose of this application is to provide a triglyceride risk prediction method and system based on near-infrared facial spectrum, so as to realize non-invasive, rapid and accurate detection of blood triglyceride content and risk warning for human beings.
[0005] In order to achieve the above object, the present application provides a triglyceride risk prediction method based on near-infrared face spectrum, the method comprising:
[0006] Scan the face of the person being tested and collect the face spectrum sample, and detect the triglyceride content in the blood of the person being tested;
[0007] Preprocessing the face spectrum samples of the tested persons, dividing the risk levels according to the triglyceride content in the blood of the tested persons, and constructing a spectrum sample data set according to the face spectrum samples and the risk levels;
[0008] constructing a risk prediction model based on the spectral sample data set;
[0009] Obtain a face spectrum sample to be tested, and input it into the risk prediction model to obtain a corresponding triglyceride risk prediction result.
[0010] Furthermore, the method of scanning the face of the person being tested, collecting the face spectrum sample, and obtaining the triglyceride content in the blood of the person being tested includes:
[0011] Scan the faces of the tested people at least once, use the Dlib model to detect the face target area in the scanned image, and generate a face detection box (x1, x2, x3, x4), where x i is a tuple of pixel i, indicating the row and column pixel value of the pixel;
[0012] Crop the core area of the face according to the face detection frame, average and convert at least one scan data of the core area to obtain a face spectrum sample;
[0013] The triglyceride content in the blood of the tested population is tested by using a blood test method to obtain the triglyceride content;
[0014] Establish the correspondence between the facial spectrum sample data and triglyceride content of the tested population.
[0015] Further, the method of preprocessing the face spectrum samples of the tested person, dividing the risk level according to the triglyceride content in the blood of the tested person, and constructing the spectrum sample data set according to the face spectrum samples and the risk level includes:
[0016] Using a sliding average filter method or a Savitzky-Golay filter method to eliminate noise in the face spectrum sample;
[0017] Correcting the baseline drift of the face spectrum sample by using polynomial baseline correction or least squares baseline correction;
[0018] Normalizing the face spectrum samples and scaling them to the same range to obtain preprocessed face spectrum samples;
[0019] According to the triglyceride content in the blood of the subjects, the subjects are divided into low-risk level, normal critical level, medium-risk level and high-risk level;
[0020] A spectral sample data set is constructed based on the preprocessed facial spectral samples and the risk level of triglyceride content corresponding to the subject.
[0021] Furthermore, the method for constructing a risk prediction model based on the spectral sample data set includes:
[0022] Support vector machine SVM is selected as the benchmark model;
[0023] Dividing the spectral sample data set into a training set and a test set;
[0024] Construct multiple binary sample classifiers;
[0025] Training the training set by using a plurality of the binary classifiers and pre-generating a risk prediction model;
[0026] The pre-generated risk prediction model is tested and optimized through the test set, and finally the constructed risk prediction model is output.
[0027] Further, the method of training the training set by using a plurality of the binary classifiers and pre-generating a risk prediction model includes:
[0028] Establishing the optimal classification line between the triglyceride content data at different risk levels, specifically including:
[0029] Get M risk-to-predict face spectrum samples, and the characteristic vector of the wavelength of each face spectrum sample is h i (i=1,2,…,M); where the four risk levels corresponding to the M face spectrum samples are y j ∈{0,1,2,3}, respectively low risk level, normal critical level, medium risk level and high risk level;
[0030] Construct a sample set (h) based on the face spectrum samples to be tested and the risk level i ,y j ), where i = 1, 2, ..., M, j = 0, 1, 2, 3;
[0031] When the risk level of face spectrum samples is classified into linear inseparable categories, the kernel function k(h m ,h n ) Input vector h m and h n Mapped to a high-dimensional feature space, where the input vector h m and h n are two arbitrary data samples without label y, h m ∈h i 、h n ∈h i ;
[0032] After nonlinear mapping, h m and h n Perform dot product operation<φ(h m )·φ(h n )>Construct the optimal classification line.
[0033] Further, the method of testing and optimizing the pre-generated risk prediction model through a test set and finally outputting the constructed risk prediction model includes:
[0034] The PSO particle swarm optimization algorithm is used to optimize the parameters of the support vector machine SVM to complete the optimization of the risk prediction model.
[0035] Furthermore, the triglyceride risk prediction results include low-risk level, normal critical level, medium-risk level and high-risk level.
[0036] This embodiment further discloses a triglyceride risk prediction system based on near-infrared facial spectrum, including:
[0037] A face scanning module is used to scan the face of the person being tested and collect face spectrum samples, and detect the content of triglycerides in the blood of the person being tested;
[0038] A data preprocessing module is used to preprocess the face spectrum samples of the tested person, divide the risk level according to the triglyceride content in the blood of the tested person, and construct a spectrum sample data set based on the face spectrum samples and risk levels;
[0039] A model building module, used to build a risk prediction model based on the spectral sample data set;
[0040] The risk prediction module is used to obtain the face spectrum sample to be tested and input it into the risk prediction model to obtain the corresponding triglyceride risk prediction result.
[0041] The present invention further discloses a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program implements the above method when executed by a processor.
[0042] The present invention further discloses an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the program.
[0043] Compared with the prior art, the triglyceride risk prediction method and system based on near-infrared facial spectrum provided by the present invention have the following beneficial effects: by constructing a risk prediction model, a close connection can be achieved between the spectral sample of the face and the triglyceride content in the human blood; after obtaining the spectral sample to be tested by scanning the face, the spectral sample can be input into the risk prediction model to obtain the triglyceride content in the blood, thereby ultimately achieving non-invasive, reagent-free, rapid and accurate detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] The drawings constituting a part of this application are used to provide a further understanding of this application, so that other features, purposes and advantages of this application become more obvious. The schematic embodiment drawings and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application. In the drawings:
[0045] Figure 1 This is a schematic diagram of the process of the triglyceride risk prediction method based on the near-infrared face spectrum of this embodiment;
[0046] Figure 2This is a schematic diagram of the structure of the triglyceride risk prediction system based on near-infrared facial spectrum in this embodiment;
[0047] Figure 3 It is a schematic diagram of the structure of the electronic device disclosed in this embodiment. DETAILED DESCRIPTION
[0048] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.
[0049] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0050] In the present application, the terms "upper", "lower", "left", "right", "front", "back", "top", "bottom", "inner", "outer", "middle", "vertical", "horizontal", "lateral", "longitudinal" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings. These terms are mainly used to better describe the present application and its embodiments, and are not used to limit the indicated devices, elements or components to have a specific orientation, or to be constructed and operated in a specific orientation.
[0051] In addition, some of the above terms may be used to express other meanings in addition to indicating orientation or positional relationship. For example, the term "on" may also be used to express a certain dependency or connection relationship in some cases. For those of ordinary skill in the art, the specific meanings of these terms in this application can be understood according to specific circumstances.
[0052] In addition, the term "plurality" shall mean two or more.
[0053] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0054] Traditional detection methods for triglycerides can generally be divided into chemical methods, chromatography and enzymatic methods, etc. All of the above methods can achieve accurate detection of triglycerides in human blood. However, since they are all invasive detection methods, blood must be collected in advance and chemical reagents must be used, so the risk of disease infection cannot be avoided, and the subject experience is poor. In order to solve the above problems, the present application has innovated in the relevant technologies of near-infrared spectroscopy, and finally realized a non-invasive, reagent-free, rapid and accurate detection method, which not only reduces the harm and cost of biochemical tests, but also can quickly realize risk screening, which has very important clinical significance.
[0055] refer to Figure 1 , is a schematic diagram of the process of triglyceride risk prediction method under near infrared face spectrum of this application, and the method is as follows:
[0056] S100: Scan the face of the person being tested and collect the face spectrum sample, and detect the triglyceride content in the blood of the person being tested.
[0057] Since the blood flow of the human face is rich and sampling is convenient, this embodiment uses a portable near-infrared spectrometer as a collection device. Other collection devices with both spectral and imaging functions can also be used, and this embodiment does not limit it. Taking a portable near-infrared spectrometer as an example, before collecting faces, the spectrometer is calibrated, the faces of the tested people are scanned uniformly at least once, and the Dlib model is used to detect the target face area in the facial image, and a face detection box box(x1, x2, x3, x4) is generated. i is a tuple of pixel point i, indicating the pixel value of the row and column where the pixel point is located. The face core area is cropped according to the face detection frame, and at least one scan data of the core area is averaged and converted to obtain a face spectrum sample. For example, when scanning once, the scan data is a face spectrum sample. When scanning five times, the face spectrum sample is retained according to the average value of five times. It is also necessary to use a blood test method to detect the triglyceride content in the blood of the current test population (in a fasting state), obtain the triglyceride content, and establish a corresponding relationship between the face spectrum sample data of the test population and the triglyceride content. In this way, a complete sample data detection record is generated.
[0058] S200: pre-processing the face spectrum samples of the tested person, dividing the risk level according to the triglyceride content in the blood of the tested person, and constructing a spectrum sample data set according to the face spectrum samples and the risk level.
[0059] The preprocessing is to preprocess the face spectrum samples and the risk level of triglyceride content. The preprocessing includes noise processing, limit drift correction, standardization, normalization and other operations. Specifically:
[0060] S210: Eliminate noise in the face spectrum sample using a sliding average filtering method or a Savitzky-Golay filtering method.
[0061] S220: Correcting the baseline drift of the face spectrum sample by using polynomial baseline correction or least squares baseline correction, thereby suppressing the negative impact caused by the instability of the spectral instrument or the inhomogeneity of the sample.
[0062] S230: normalizing the face spectrum samples and scaling them to the same range to obtain pre-processed face spectrum samples, thereby ensuring that the differences between face spectrum samples are eliminated and making the spectrum data comparable.
[0063] In order to prepare for further model training and the need for risk identification and reasoning calculations, it is necessary to complete the risk level classification of triglyceride test values, specifically:
[0064] S240: According to the triglyceride content in the blood of the tested person, the tested person is divided into a low-risk level, a normal critical level, a medium-risk level, and a high-risk level. For example, the low-risk level (when the triglyceride content value is <1.69mmol / L), the normal critical level (the triglyceride content value is 1.70-2.25mmol / L), the medium-risk level (the triglyceride content value is 2.26-5.63mmol / L), and the high-risk level (the triglyceride content value is >5.64mmol / L).
[0065] S250: Constructing a spectral sample data set according to the preprocessed face spectral sample and the risk level of triglyceride content corresponding to the subject.
[0066] After obtaining the spectral sample data set, the basic conditions for building a risk prediction model are met. Before building the risk prediction model, special attention should be paid to the selection of the benchmark model because the wavelength range of the spectral sample is relatively large, 350nm-1650nm, and there may be insufficient sample size and uneven distribution. S300: Build a risk prediction model based on the spectral sample data set. Specifically:
[0067] S310: Selecting a support vector machine SVM as a benchmark model. By selecting the support vector machine SVM, an optimal classification boundary can be established in multiple classifications, so that data between different classifications can be separated from each other with a maximum classification interval.
[0068] S320: Divide the spectral sample data set into a training set and a test set.
[0069] The training set accounts for 80% of the total data set, and the test set accounts for 20% of the total data set. For the spectral sample data set, each sample wavelength feature vector is represented by h i Indicates that i represents the i-th spectral sample, and the risk level category corresponding to the spectral sample is y j ∈{0,1,2,3}, which are low risk level, normal critical level, medium risk level, and high risk level respectively. Therefore, the above spectral sample data set can be expressed as (h i ,y j ).
[0070] S330: Construct multiple binary sample classifiers.
[0071] Since the standard support vector machine can only complete the binary classification task, and the risk level in this embodiment is 4, it is necessary to construct multiple binary sample classifiers, that is, by constructing 4*(4-1) / 2 binary sample classifiers in the 4 types of samples.
[0072] S340: Training the training set using a plurality of the binary classifiers, and pre-generating a risk prediction model.
[0073] Specifically, after constructing 6 sample classifiers in step S330, the 6 sample classifiers are used to train various samples in the training set to generate a support vector machine SVM classification model for risk identification.
[0074] Furthermore, an optimal classification line is established between the triglyceride content data at different risk levels, specifically including:
[0075] Get M risk-to-predict face spectrum samples, and the characteristic vector of the wavelength of each face spectrum sample is h i (i=1,2,…,M); where the four risk levels corresponding to the M face spectrum samples are y j ∈{0,1,2,3}, respectively low risk level, normal critical level, medium risk level and high risk level;
[0076] Construct a sample set (h) based on the face spectrum samples to be tested and the risk level i ,y j ), where i = 1, 2, ..., M, j = 0, 1, 2, 3;
[0077] When the risk level of face spectrum samples is classified into linear inseparable categories, the kernel function k(h m ,h n ) Input vector h m and h n Mapped to a high-dimensional feature space, where the input vector h m and hn are two arbitrary data samples without label y, h m ∈h i 、h n ∈h i ;
[0078] There exists a nonlinear mapping φ:R d →H, for h after nonlinear mapping m and h n Perform dot product operation<φ(h m )·φ(h n )>Construct the optimal classification line.
[0079] Among them, kernel functions are mostly linear kernel functions, polynomial functions and radial basis functions. Due to the significant characteristics of radial basis functions in high-dimensional nonlinear space classification, the kernel function uses radial basis functions, i.e. k(h i ,y i )=exp(-||h m -h n || 2 / 2σ 2 ), σ is the normal coefficient of the function. At this time, the nonlinear classification model expression is where a i is the Lagrange multiplier, 0≤a i ≤C, where C is the penalty coefficient.
[0080] S350: Testing and optimizing the pre-generated risk prediction model through a test set, and finally outputting the constructed risk prediction model.
[0081] When optimizing the model, the PSO particle swarm optimization algorithm is used to optimize the parameters of the risk prediction model. For the PSO algorithm, in the D-dimensional search space, there are m particles forming a population. Suppose the position of the i-th particle at time t is The speed is V i t , the optimal position of an individual is The global optimal position is Then the velocity V of the particle at time t+1 is i t+1 and position X i t+1 The formula is:
[0082]
[0083] Where: ω is the inertia weight; c1 and c2 are learning factors; r1 and r2 are random numbers uniformly distributed in [0,1]. For the above-mentioned PSO algorithm, it is easy to fall into the local optimal solution during the parameter optimization process. This embodiment improves the algorithm as follows:
[0084] (1) Dynamically adjust the inertia weight ω. The inertia weight ω mainly keeps the particles inertia and has the ability to expand the search space. It is an important parameter for estimating the global and local search capabilities of the algorithm. Therefore, the inertia weight allocation strategy is particularly important. Compared with the commonly used linear decreasing allocation strategy, the cosine function is used to control the inertia weight ω, and a random adjustment strategy that obeys the beta distribution is added to achieve dynamic adjustment of the inertia weight. The calculation formula is as follows:
[0085]
[0086] Where: max and ω min are the maximum and minimum values of the inertia weight respectively; T max is the maximum number of iterations of the particle; σ is the inertia adjustment factor; betarnd generates random numbers that obey the Beta distribution, and a and b are the two parameters of the Beta distribution.
[0087] (2) Dynamically adjust the learning factors c1 and c2. The learning factors c1 and c2 are mainly used to adjust the step size of the particle moving to the individual optimum and the global optimum. The traditional PSO algorithm is usually set to c1 = c2, which obviously cannot meet the multi-scenario requirements of the learning factor. Therefore, in order to achieve multiple iterations to obtain a large number of random solutions and realize the dynamic change of parameters, the learning factor calculation method of the fused cosine function is used. The formula is as follows:
[0088]
[0089] It can be seen that with the increase of the number of iterations, the absolute value of the cosine function also increases accordingly, c1 can decrease nonlinearly, and on the contrary, the c2 value can increase nonlinearly, thereby realizing the dynamic change of the learning factor in the iteration process.
[0090] (3) Particle position update: To avoid premature phenomenon in the algorithm and speed up the convergence of the algorithm, the position processing operation needs to be simplified. When the particle velocity update is deleted and only the position update is used, the idea of the mean particle swarm algorithm is combined. By using the linear combination of the individual optimal position and the global optimal position to replace the individual optimal position and the global optimal position in the algorithm, the deviation between the flight direction of the particle and the current position direction can be better controlled, ensuring the efficient optimization process and thus speeding up its convergence. The final formula is as follows:
[0091]
[0092] Combining the improvements at the above three levels, the PSO algorithm is optimized, which solves the problem that the PSO algorithm is prone to falling into local optimality. At the same time, it can speed up the early search speed, improve the global search capability and optimization accuracy, and improve the accuracy of the SVM model to a high extent.
[0093] S400: Obtain a face spectrum sample to be tested, and input it into the risk prediction model to obtain a corresponding triglyceride risk prediction result.
[0094] In summary, the introduction of the PSO particle swarm optimization algorithm to calculate the optimal parameter solution can significantly improve the accuracy of the SVM basic classifier, and accelerate the convergence speed of the model during the training process, saving the computing time cost, and ultimately improving the generalization ability of the prediction model. It can be seen that by collecting and analyzing facial spectral data through near-infrared spectroscopy equipment for necessary processing, and using appropriate machine learning modeling methods, the risk level status corresponding to the triglyceride content value of the population can be quickly identified. Therefore, a triglyceride risk prediction method based on near-infrared facial spectroscopy is of great significance in determining the quantitative detection range of substances and identifying risk levels. It can provide screening recommendations for the risk level of related diseases caused by abnormal triglyceride content, so as to achieve the purpose of early prevention and early management.
[0095] refer to Figure 2 This embodiment further discloses a triglyceride risk prediction system based on near-infrared face spectrum, including:
[0096] The face scanning module 21 is used to scan the face of the person being tested and collect face spectrum samples, and detect the content of triglycerides in the blood of the person being tested; specifically:
[0097] Scan the faces of the tested people at least once, use the Dlib model to detect the face target area in the scanned image, and generate a face detection box (x1, x2, x3, x4), where x i is a tuple of pixel i, indicating the row and column pixel value of the pixel;
[0098] Crop the core area of the face according to the face detection frame, average and convert at least one scan data of the core area to obtain a face spectrum sample;
[0099] The triglyceride content in the blood of the tested population is tested by using a blood test method to obtain the triglyceride content;
[0100] Establish the correspondence between the facial spectrum sample data and triglyceride content of the tested population.
[0101] The data preprocessing module 22 is used to preprocess the face spectrum samples of the tested person, and divide the risk level according to the triglyceride content in the blood of the tested person, and construct a spectrum sample data set according to the face spectrum samples and the risk level; specifically:
[0102] Using a sliding average filter method or a Savitzky-Golay filter method to eliminate noise in the face spectrum sample;
[0103] Correcting the baseline drift of the face spectrum sample by using polynomial baseline correction or least squares baseline correction;
[0104] Normalizing the face spectrum samples and scaling them to the same range to obtain preprocessed face spectrum samples;
[0105] According to the triglyceride content in the blood of the subjects, the subjects are divided into low-risk level, normal critical level, medium-risk level and high-risk level;
[0106] A spectral sample data set is constructed based on the preprocessed facial spectral samples and the risk level of triglyceride content corresponding to the subject.
[0107] The model building module 23 is used to build a risk prediction model according to the spectral sample data set; specifically:
[0108] Support vector machine SVM is selected as the benchmark model;
[0109] Dividing the spectral sample data set into a training set and a test set;
[0110] Construct multiple binary sample classifiers;
[0111] The training set is trained by using a plurality of the binary classifiers, and a risk prediction model is pre-generated; specifically:
[0112] Establishing the optimal classification line between the triglyceride content data at different risk levels, specifically including:
[0113] Get M risk-to-predict face spectrum samples, and the characteristic vector of the wavelength of each face spectrum sample is h i (i=1,2,…,M); where the four risk levels corresponding to the M face spectrum samples are y j ∈{0,1,2,3}, respectively low risk level, normal critical level, medium risk level and high risk level;
[0114] Construct a sample set (h) based on the face spectrum samples to be tested and the risk level i ,y j ), where i = 1, 2, ..., M, j = 0, 1, 2, 3;
[0115] When the risk level of face spectrum samples is classified into linear inseparable categories, the kernel function k(h m ,h n ), where the input vector h m and h n are two arbitrary data samples without label y, h m ∈h i 、h n ∈h i ;
[0116] After nonlinear mapping, h m and h n Perform dot product operation<φ(h m )·φ(h n )>construct an optimal classification line; test and optimize the pre-generated risk prediction model through a test set, and finally output the constructed risk prediction model.
[0117] The risk prediction module 24 is used to obtain a face spectrum sample to be tested and input it into the risk prediction model to obtain a corresponding triglyceride risk prediction result.
[0118] Figure 3 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as Figure 3 As shown, the electronic device 50 includes: a processor 501 (processor), a memory 502 (memory) and a bus 503;
[0119] The processor 501 and the memory 502 communicate with each other via the bus 503 ; the processor 501 is used to call program instructions in the memory 502 to execute the methods provided by the above-mentioned method embodiments.
[0120] This embodiment provides a non-transitory computer-readable storage medium, which stores computer instructions. The computer instructions enable a computer to execute the methods provided by the above method embodiments.
[0121] A person skilled in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various storage media that can store program codes.
[0122] The device embodiments described above are merely illustrative, wherein 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, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.
[0123] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.
[0124] The specific manner of executing the operation of each unit in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.
[0125] The above are only preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. Triglyceride risk prediction method based on near-infrared facial spectrum, It is characterized in that The method comprises: Scan the face of the person being tested and collect the face spectrum sample, and detect the triglyceride content in the blood of the person being tested; Preprocessing the face spectrum samples of the tested persons, dividing the risk levels according to the triglyceride content in the blood of the tested persons, and constructing a spectrum sample data set according to the face spectrum samples and the risk levels; constructing a risk prediction model based on the spectral sample data set; Obtain a face spectrum sample to be tested, and input it into the risk prediction model to obtain a corresponding triglyceride risk prediction result.
2. The triglyceride risk prediction method based on near infrared face spectrum according to claim 1, It is characterized in that The method of scanning the face of the person being tested, collecting the face spectrum sample, and detecting the content of triglycerides in the blood of the person being tested comprises: Scan the faces of the tested people at least once, use the Dlib model to detect the face target area in the scanned image, and generate a face detection box (x 1 ,x 2 ,x 3 ,x 4 ), where x i is a tuple of pixel i, indicating the row and column pixel value of the pixel; Crop the core area of the face according to the face detection frame, average and convert at least one scan data of the core area to obtain a face spectrum sample; The triglyceride content in the blood of the tested population is tested by using a blood test method to obtain the triglyceride content; Establish the correspondence between the facial spectrum sample data and triglyceride content of the tested population.
3. The triglyceride risk prediction method based on near infrared face spectrum according to claim 1, It is characterized in that The method of preprocessing the face spectrum samples of the tested person, dividing the risk level according to the triglyceride content in the blood of the tested person, and constructing the spectrum sample data set according to the face spectrum samples and the risk level includes: Using a sliding average filter method or a Savitzky-Golay filter method to eliminate noise in the face spectrum sample; Correcting the baseline drift of the face spectrum sample by using polynomial baseline correction or least squares baseline correction; Normalizing the face spectrum samples and scaling them to the same range to obtain preprocessed face spectrum samples; According to the triglyceride content in the blood of the subjects, the subjects are divided into low-risk level, normal critical level, medium-risk level and high-risk level; A spectral sample data set is constructed based on the preprocessed facial spectral samples and the risk level of triglyceride content corresponding to the subject.
4. The triglyceride risk prediction method based on near infrared face spectrum as claimed in claim 3, It is characterized in that The method for constructing a risk prediction model according to the spectral sample data set comprises: Support vector machine SVM is selected as the benchmark model; Dividing the spectral sample data set into a training set and a test set; Construct multiple binary sample classifiers; Training the training set by using a plurality of the binary classifiers and pre-generating a risk prediction model; The pre-generated risk prediction model is tested and optimized through the test set, and finally the constructed risk prediction model is output.
5. The triglyceride risk prediction method based on near infrared face spectrum according to claim 4, It is characterized in that The method of training the training set by using a plurality of the binary classifiers and pre-generating a risk prediction model includes: Establishing the optimal classification line between the triglyceride content data at different risk levels, specifically including: Get M risk-to-predict face spectrum samples, and the characteristic vector of the wavelength of each face spectrum sample is h i (i=1,2,…,M); where the four risk levels corresponding to the M face spectrum samples are y j ∈{0,1,2,3}, respectively low risk level, normal critical level, medium risk level and high risk level; Construct a sample set (h) based on the face spectrum samples to be tested and the risk level i ,y j ), where i = 1, 2, ..., M, j = 0, 1, 2, 3; When the risk level of face spectrum samples is classified into linear inseparable categories, the kernel function k(h m ,h n ) Input vector h m and h n Mapped to a high-dimensional feature space, where the input vector h m and h n are two arbitrary data samples without label y, h m ∈h i 、h n ∈h i ; After nonlinear mapping, h m and h n Perform dot product operation<φ(h m )·φ(h n )>Construct the optimal classification line.
6. The triglyceride risk prediction method based on near infrared face spectrum according to claim 4, It is characterized in that The method of testing and optimizing the pre-generated risk prediction model through a test set and finally outputting the constructed risk prediction model comprises: The PSO particle swarm optimization algorithm is used to optimize the parameters of the support vector machine SVM to complete the optimization of the risk prediction model.
7. The triglyceride risk prediction method based on near infrared face spectrum according to claim 1, It is characterized in that The triglyceride risk prediction results include low-risk level, normal critical level, medium-risk level and high-risk level.
8. Triglyceride risk prediction system based on near-infrared facial spectrum, It is characterized in that include: A face scanning module is used to scan the face of the person being tested and collect face spectrum samples, and detect the content of triglycerides in the blood of the person being tested; A data preprocessing module is used to preprocess the face spectrum samples of the tested person, divide the risk level according to the triglyceride content in the blood of the tested person, and construct a spectrum sample data set based on the face spectrum samples and risk levels; A model building module, used to build a risk prediction model based on the spectral sample data set; The risk prediction module is used to obtain the face spectrum sample to be tested and input it into the risk prediction model to obtain the corresponding triglyceride risk prediction result.
9. A computer-readable storage medium, It is characterized in that The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.
Citation Information
Cited By
Method and device for detecting creatinine content information of blood
CN119770037A
METHOD FOR PREDICTING LIKELIHOOD OF TRIGLYCERIDE LEVELS EXCEEDING 1.7 mmol / l IN PATIENTS WITH PSYCHIATRIC DISORDERS
RU2866481C1
