Human bilirubin content detection method and device, storage medium and server

Abstract

The invention discloses a human bilirubin content detection method and device, a storage medium and a server. The human bilirubin content detection method comprises the following steps: acquiring face video data; face blood filling degree caused by blood pulsation in face video data is identified to determine spectrums with different wavelengths, and a target dynamic spectrum is extracted from the spectrums; performing model training based on the target dynamic spectrum by using a partial least squares regression algorithm to obtain a bilirubin detection model; receiving a to-be-detected face video configured by a to-be-detected user, inputting the to-be-detected face video into the bilirubin detection model, and predicting to obtain the human bilirubin content of the to-be-detected user; the technical problem that the intelligent level is not high due to the fact that the corresponding spectrum cannot be extracted through analysis of the face video is solved.

Description

Technical Field

[0001] This application relates to the field of intelligent disease detection. Specifically, it relates to a method, device, storage medium, and server for detecting the bilirubin content in the human body. Background Art

[0002] Bilirubin (BR) is one of the final products of the catabolism of heme in vertebrates and has functions such as antioxidant and anti-inflammatory effects.

[0003] Maintaining a normal bilirubin content in the body plays a very important role in human health and is considered beneficial for preventing the occurrence of diseases such as cancer, stroke, diabetes, and cardiovascular diseases. However, excessive bilirubin is considered a sign of liver dysfunction and is also a cause of severe brain damage in newborns. Therefore, rapid and accurate detection of bilirubin content in the human body has very important application value. So far, the methods for detecting bilirubin content in serum samples mainly include the diazo method, peroxidase method, fiber optic sensing detection method, fluorescence spectroscopy method, etc.

[0004] In clinical laboratories, methods such as the vanadic acid oxidation method, modified J-G method, and bilirubin oxidase method all require fasting blood sampling for patients. That is, venous blood is centrifuged to separate blood cells from serum, and serum is taken and combined with reagents, and an instrument is used to view the concentrations of total bilirubin, direct bilirubin, and indirect bilirubin. If the concentration value of bilirubin is significantly increased, it indicates that liver cells are damaged, and bilirubin also has important significance in the differential diagnosis of jaundice.

[0005] In the prior art, the patent with the application number CN202310300867.8 discloses a skin quality evaluation method based on the volume content of skin components. The method respectively collects hyperspectral images of the skin at three parts of the human face, and uses a skin hyperspectral reconstruction algorithm to obtain the 9 skin component contents of each pixel within the analysis area. The 9 skin components are: eumelanin, pheomelanin, melanin, carotene, bilirubin, blood oxygen concentration, epidermal moisture, dermal moisture, and collagen, and three types of indicators are obtained through the following methods: ① The average value of 9 skin components, which is used for quantitative skin quality analysis and product efficacy evaluation; ② 6-dimensional evaluation indexes, which are respectively: skin whiteness, skin yellowness, skin redness, moisture, collagen, and uniformity, and are evaluated from 6 visual dimensions based on the threshold of the human face skin sample database; ③ Comprehensive evaluation, based on the human face skin sample database, for grade evaluation and ranking (percentile). It supports personalized skin quality evaluation.

[0006] As can be seen from the above, although this patent can collect hyperspectral images of the skin at three parts of the human face and analyze the hyperspectral images through a skin hyperspectral reconstruction algorithm to obtain skin components, including bilirubin; however, it cannot extract the corresponding spectrum by analyzing the human face video, and the intelligent level is not high.

[0007] In the related art, there is no effective solution to the problem that the corresponding spectrum cannot be extracted by analyzing a face video, and the intelligent level is not high. Summary of the Invention

[0008] The main purpose of this application is to provide a method, device, storage medium and server for detecting the content of human bilirubin, so as to solve the problem that the corresponding spectrum cannot be extracted by analyzing a face video, and the intelligent level is not high.

[0009] To achieve the above object, according to one aspect of this application, a method for detecting the content of human bilirubin is provided.

[0010] The method for detecting the content of human bilirubin according to this application includes: obtaining face video data; determining spectra of different wavelengths by identifying the facial blood filling degree caused by blood pulsation in the face video data, and then extracting a target dynamic spectrum therefrom; using the partial least squares regression algorithm to perform model training based on the target dynamic spectrum to obtain a bilirubin detection model; receiving a face video to be measured configured by a user to be measured, inputting the face video to be measured into the bilirubin detection model, and predicting the content of human bilirubin of the user to be measured.

[0011] Further, obtaining face video data includes: irradiating a face with a near-infrared and visible light binocular camera to obtain face video data; wherein, the face video data is a pulse wave signal with a sampling frequency of n Hz and a continuous acquisition time of t seconds.

[0012] Further, after obtaining the face video data and before determining spectra of different wavelengths by identifying the facial blood filling degree caused by blood pulsation in the face video data, it further includes: performing a frame splitting operation on the face video data to obtain several short signals arranged in chronological order; performing peak detection and sorting on the several short signals to obtain key peak points; classifying and feature recording the face video data according to the key peak points.

[0013] Further, by identifying the facial blood filling caused by blood pulsation in the facial video data to determine spectra of different wavelengths, and then extracting the target dynamic spectrum therefrom, including: selecting a specified number of wavelengths in the facial video data based on the wavelength ranges of near-infrared and visible light; after extracting the fundamental wave components and introducing harmonic components, obtaining the absorption spectra and fluorescence spectra of normal human serum and bilirubin-serum, and obtaining the spectral line characteristics of both and the content of bilirubin in the serum based on the absorption spectra and fluorescence spectra; forming three-dimensional optoelectronic logarithmic pulse data by taking the logarithm of the spectra and arranging them along the time axis; referring to the linear relationship between the optoelectronic logarithmic pulse wave amplitudes at different wavelengths and the blood absorbance, and sequentially extracting the optoelectronic logarithmic pulse wave amplitudes at different wavelengths based on the frequency domain extraction method with linear characteristics of Fourier transform to obtain the target dynamic spectrum.

[0014] Further, using the partial least squares regression algorithm to perform model training based on the target dynamic spectrum to obtain a bilirubin detection model, including: S001, preprocessing the target dynamic spectrum using a preprocessing method; S002, determining the optimal number of principal components for modeling within the full spectral range; S003, dividing the entire spectral region into n equally wide subintervals by the iPLS algorithm; S004, performing PLS regression on each subinterval to obtain n local regression models; S005, using the root mean square error of cross-validation RMSECV value as the accuracy measure for each model, respectively comparing the accuracies of the full-spectrum model and each local model, and taking the subinterval where the local model with the highest accuracy is located as the selected interval; S006, performing internal PLS regression within the selected interval to establish a bilirubin detection model.

[0015] Further, using the partial least squares regression algorithm to perform model training based on the target dynamic spectrum to obtain a bilirubin detection model, further including: S007, re-partitioning the spectral region, repeating S003 - S005, and comparing the modeling effects of the selected intervals in different partitions to determine the optimal interval; S008, performing internal PLS regression within the optimal interval to establish an optimal bilirubin detection model.

[0016] To achieve the above object, according to another aspect of the present application, a human bilirubin content detection device is provided.

[0017] The human bilirubin content detection device according to the present application includes: an acquisition module for acquiring facial video data; an extraction module for determining spectra of different wavelengths by identifying the facial blood filling caused by blood pulsation in the facial video data, and then extracting the target dynamic spectrum therefrom; a training module for using the partial least squares regression algorithm to perform model training based on the target dynamic spectrum to obtain a bilirubin detection model; and a prediction module for receiving the facial video to be measured configured by the user to be measured, inputting the facial video to be measured into the bilirubin detection model, and predicting the human bilirubin content of the user to be measured.

[0018] Further, after obtaining the face video data, before determining the spectra of different wavelengths by identifying the facial blood filling degree caused by blood pulsation in the face video data, it further includes: performing a frame splitting operation on the face video data to obtain several short signals arranged in chronological order; performing peak detection and sorting on the several short signals to obtain key peak points; classifying and feature recording the face video data according to the key peak points.

[0019] To achieve the above object, according to another aspect of the present application, there is provided a computer-readable storage medium.

[0020] In the computer-readable storage medium according to the present application, a computer program is stored therein, wherein the computer program is set to execute the method for detecting the content of human bilirubin when running.

[0021] To achieve the above object, according to another aspect of the present application, there is provided a server.

[0022] The server according to the present application includes: a memory and a processor, wherein a computer program is stored in the memory, and the processor is set to run the computer program to execute the method for detecting the content of human bilirubin.

[0023] In the embodiments of the present application, a method for detecting the content of human bilirubin is adopted. By obtaining face video data; by identifying the facial blood filling degree caused by blood pulsation in the face video data to determine spectra of different wavelengths, and then extracting a target dynamic spectrum therefrom; using the partial least squares regression algorithm to perform model training based on the target dynamic spectrum to obtain a bilirubin detection model; receiving a face video to be measured configured by a user to be measured, and inputting the face video to be measured into the bilirubin detection model to predict the content of human bilirubin of the user to be measured; the purpose of extracting corresponding spectra through the analysis of the face video is achieved, thereby achieving the technical effect of improving the intelligent level, and further solving the technical problem of low intelligent level due to the inability to extract corresponding spectra through the analysis of the face video. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings constituting a part of the present application are used to provide a further understanding of the present application, making other features, objects, and advantages of the present application more obvious. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings:

[0025] Figure 1 is a flowchart of the method for detecting the content of human bilirubin according to the embodiments of the present application;

[0026] Figure 2It is a schematic structural diagram of a human bilirubin content detection device according to an embodiment of the present application; Detailed implementation manners

[0027] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0028] It should be noted that the terms "first", "second", etc. in the specification, claims and above-mentioned drawings of the present application are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so as to implement the embodiments of the present application described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0029] In the present application, the orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", "front", "rear", "top", "bottom", "inner", "outer", "middle", "vertical", "horizontal", "lateral", "longitudinal", etc. is based on the orientation or positional relationship shown in the accompanying drawings. These terms are mainly used to better describe the present invention and its embodiments, and are not used to limit that the indicated device, element or component must have a specific orientation, or be constructed and operated in a specific orientation.

[0030] Moreover, in addition to being used to represent an orientation or positional relationship, some of the above terms may also be used to represent other meanings. For example, the term "upper" may also be used to represent a certain attachment relationship or connection relationship in some cases. For those of ordinary skill in the art, the specific meanings of these terms in the present invention can be understood according to specific circumstances.

[0031] In addition, the terms "installed", "set up", "provided with", "connected", "connected to", "socketed" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral structure; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or there can be internal communication between two devices, elements or components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0032] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments 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.

[0033] According to an embodiment of the present invention, a method for detecting the human body bilirubin content is provided, as Figure 1 shown, the method includes the following steps S101 to step S104:

[0034] Step S101, obtain face video data;

[0035] Specifically, obtaining face video data includes: irradiating a human face with a near-infrared and visible light binocular camera to obtain face video data; wherein, the face video data is a pulse wave signal with a sampling frequency of n Hz and a continuous acquisition time of t seconds.

[0036] It should be understood that the selection of the facial area stems from its rich capillary distribution and its proximity to the carotid artery, which creates the characteristic of higher blood signal intensity here. To ensure the accuracy and reliability of video acquisition, the convenience of the facial area is particularly considered, and it is ensured that there is sufficient light during the acquisition process, and the use of excessive makeup or contouring products is avoided as much as possible. Such operations can ensure uniform and bright facial light to reduce the influence of light source fluctuations on weak skin color changes.

[0037] It should also be understood that during the video acquisition process, the person being measured needs to maintain a normal breathing state, while minimizing head and body movements and avoiding large facial expressions. This series of operation steps helps to ensure the high quality and reliability of the acquired face video dataset. By maintaining the stability of light conditions and avoiding excessive external interference, more accurate and detailed facial data can be obtained. These detailed data are crucial for subsequent analysis and cholesterol level prediction.

[0038] Step S102, determine spectra of different wavelengths by identifying the facial blood filling degree caused by blood pulsation in the face video data, and then extract the target dynamic spectrum from them;

[0039] Specifically, determining spectra of different wavelengths by identifying the facial blood filling degree caused by blood pulsation in the face video data, and then extracting the target dynamic spectrum from them, includes:

[0040] Based on the wavelength ranges of near-infrared and visible light, a specified number of wavelengths are selected from the face video data; after extracting the fundamental wave components and introducing harmonic components, the absorption spectra and fluorescence spectra of normal human serum and bilirubin-serum are obtained, and the spectral line characteristics of the two and the content of bilirubin in the serum are obtained based on the absorption spectra and fluorescence spectra; by taking the logarithm of the spectra and arranging them along the time axis, three-dimensional optoelectronic logarithmic pulse data is formed; referring to the linear relationship between the optoelectronic logarithmic pulse wave amplitude at different wavelengths and the blood absorbance, the optoelectronic logarithmic pulse wave amplitudes at different wavelengths are sequentially extracted based on the frequency-domain extraction method with linear characteristics of Fourier transform to obtain the target dynamic spectra.

[0041] The absorption and fluorescence spectra of bilirubin serum and normal serum are different. It is not only higher than normal serum in terms of absorption rate and fluorescence intensity, but also presents new absorption peaks and fluorescence peaks. Comparing the spectra of the serum to be tested can preliminarily determine the content of bilirubin in the serum. After appropriately introducing harmonic components, the correlation of the dynamic spectral data extracted from different parts of the same body increases, while the correlation of the data extracted from the same part of different individuals decreases. This improves the ability to distinguish data from different sampling parts of the same body and different individuals. Based on the frequency-domain extraction method with linear characteristics of Fourier transform, using the fundamental wave amplitude of the logarithmic pulse frequency domain to replace the absorbance can overcome the influence of the baseline drift of the pulse wave on the result and improve the extraction accuracy of the dynamic spectra.

[0042] Step S103: Use the partial least squares regression algorithm to perform model training based on the target dynamic spectra to obtain a bilirubin detection model;

[0043] In a specific implementation manner, using the partial least squares regression algorithm to perform model training based on the target dynamic spectra to obtain a bilirubin detection model includes:

[0044] S001: Perform preprocessing on the target dynamic spectra using a preprocessing method;

[0045] S002: Determine the optimal number of principal components for modeling within the full spectral range;

[0046] S003: Divide the entire spectral region into n equally wide sub-intervals through the iPLS algorithm;

[0047] S004: Perform PLS regression on each sub-interval to obtain n local regression models;

[0048] S005: Using the root mean square error of cross-validation RMSECV value as the accuracy measurement standard for each model, compare the accuracies of the full-spectrum model and each local model respectively, and select the sub-interval where the local model with the highest accuracy is located as the selected interval;

[0049] S006: Perform internal PLS regression in the selected interval to establish a bilirubin detection model.

[0050] Further, using the partial least squares regression algorithm, based on the target dynamic spectrum for model training to obtain a bilirubin detection model, further including:

[0051] S007. Re-partition the spectral region, repeat S003 - S005, and compare the modeling effects of the selected intervals in different partitions to determine the optimal interval;

[0052] S008. Perform internal PLS regression in the optimal interval to establish an optimal bilirubin detection model.

[0053] Here, partial least squares regression (PLS) provides a multi - to - multi linear regression modeling method, which can better solve many problems that could not be solved by ordinary multiple regression in the past, and can realize the comprehensive application of various data analysis methods. For example, it can be applied to: perform regression modeling under the condition that there is a serious multi - collinearity among independent variables, and allow the regression model to be established under the condition that the number of sample points is less than the number of variables.

[0054] It should be understood that the principle of interval partial least squares (ipls) is to first divide the entire spectrum into several equal - width sub - regions, and then perform partial least squares regression in each sub - interval to establish a local regression model to be measured. The leave - one - out cross - validation is used to show that this sub - interval is the optimal modeling interval. That is, first divide the full spectrum into several intervals, and there is a band in each interval. Perform PLS regression on each band interval, and calculate the RMSECV (root mean square error of cross - validation) of the corresponding variables for each band. Find the band corresponding to the relatively small RMSECV. Combine the spectral absorption matrices corresponding to these band intervals into a new matrix. Then use Monte Carlo sampling to sample the new spectral absorption matrix and the corresponding concentration matrix multiple times. Each time, randomly select a certain proportion of samples from the sample set as the calibration set, and the remaining are the validation set. Use the spectral matrix and concentration matrix of the selected samples to establish a PLS regression model. In each sampling, use the exponential decay function to forcibly remove the wavelength points with relatively small regression coefficients.

[0055] In this embodiment, preferably, the standard algorithm of the partial least squares regression model includes the following steps:

[0056] The first step: Standardize the original data X and Y to obtain X0 and Y0, where X is m - dimensional data and Y is p - dimensional data; Select the column with the largest variance from Y0 as u1 for convenience in subsequent calculations; Because selecting the column with the largest variance means that this column can best reflect the information of the original data, that is, according to the idea of principal component analysis, this column vector is generally called the first principal component, and the correlation between X and Y is maximized.

[0057] The standardized matrix is:

[0058] Step 2: Iteratively solve for the transformation weights (w1, c1) and the comprehensive factors (t1, u1) of X and Y until convergence;

[0059] Assume that the principal components extracted from X and Y are t1 and u1, and t1 is a linear combination of the independent variable set : Y max = max(Y), and u1 is a linear combination of the dependent variable set Y min = min(Y): Range = Y max - Y min ; For the needs of regression analysis, two requirements need to be met: t1 and u1 each extract as much variation information of the variable group they belong to as possible; the correlation between t1 and u1 reaches the maximum. The calculation formula is:

[0060] Use the columns in Y selected in Step 1 to solve for the transformation weight factors of X CSS = SSY - SMY 2 / W

[0061] Use the information t1 of X to solve for the transformation weight c1 of Y and update the value of the factor u1

[0062] Judge whether a reasonable solution has been found. Otherwise, continue to search.

[0063] The estimation equations for t1 and u1 are:

[0064]

[0065] Step 3: Calculate the residual matrices of X and Y; The calculation formula is:

[0066] 1) Calculate the load P1 of X. The load reflects the direct relationship between X0 and the factor T1;

[0067]

[0068] 2) Calculate the residual X1 of X0. The residual represents the part of X0 information that u1 cannot reflect;

[0069]

[0070] 3) Calculate the load Q1 of Y;

[0071]

[0072] 4) Establish a regression model between the X factor t1 and the Y factor u1, and use t1 to predict the information of u1;

[0073]

[0074] 5) Calculate the residual Y1 of Y0, which represents the information in Y0 that cannot be predicted by X and factor t1;

[0075]

[0076] Step 4: Use X1 and Y1 to repeat the above steps to solve the parameters of the partial least squares of the next principal component.

[0077] Finally, obtain the regression equation of the partial least squares regression model. In addition, the regression coefficients should also be tested. Generally, it can be determined through cross-validation. Cross-validation is achieved by solving the ratio of the sum of squared prediction errors to the sum of squared errors. The smaller this ratio, the better. The generally set limit value is 0.05. Therefore, when this ratio is smaller, adding a new principal component is beneficial to improving the accuracy of the model; conversely, it is considered that adding a new component has no obvious improvement effect on reducing the prediction error of the equation.

[0078] Define cross-validation: In this way, before the end of each step of calculation during model building, cross-validation is performed. If at the h-th step the model has reached the accuracy requirement, the extraction of components can be stopped. If represents the marginal contribution of the component extracted at the h-th step is significant, the calculation of the (h + 1)-th step should be continued.

[0079] Based on bilirubin in human pulsating blood and clinical trial data, we construct a model to calculate the bilirubin content in human pulsating blood. This model considers the influence of various factors on serum bilirubin levels. In addition to factors such as diseases and medications, daily lifestyle also affects bilirubin, including excessive fatigue, staying up late, drinking alcohol, and unhealthy diet, etc.

[0080] Step S104: Receive the to-be-tested face video configured by the to-be-tested user, input the to-be-tested face video into the bilirubin detection model, and predict the bilirubin content of the to-be-tested user.

[0081] Use a binocular camera to obtain the facial video of an individual, and then use the aforementioned bilirubin detection model to predict the bilirubin content in the facial pulsating blood of the person, thereby obtaining a non-invasive bilirubin detection result.

[0082] It should be understood that according to the usual reference range of serum bilirubin, if the measured value is between 2 - 20 μmol / L (0.1 - 1.2 mg / dl), it is generally considered to be in the normal range; if the measured value is less than 2 or greater than 20 μmol / L (0.1 - 1.2 mg / dl), it is abnormal and there is a certain risk. Here, (1) If we want to judge the presence and degree of jaundice: Jaundice less than 34 μmol / L is not easily detected by visual inspection and is called latent jaundice; 34 - 170 μmol / L is mild jaundice; 170 - 340 μmol / L is moderate jaundice; > 340 μmol / L is severe jaundice; (2) If we want to judge the nature of jaundice: Complete obstructive jaundice is 340 - 510 μmol / L; Incomplete obstruction is 170 - 265 μmol / L; Hepatocellular jaundice is 17 - 200 μmol / L; Hemolytic jaundice is < 85 μmol / L.

[0083] From the above description, it can be seen that the present invention achieves the following technical effects:

[0084] In the embodiment of the present application, by using the method of detecting the human body bilirubin content, we obtain face video data; by identifying the facial blood filling degree caused by blood pulsation in the face video data to determine spectra of different wavelengths, and then extracting the target dynamic spectrum from them; using the partial least squares regression algorithm to perform model training based on the target dynamic spectrum to obtain a bilirubin detection model; receiving the face video to be measured configured by the user to be measured, inputting the face video to be measured into the bilirubin detection model, and predicting the human body bilirubin content of the user to be measured; achieving the purpose of extracting the corresponding spectrum by analyzing the face video, thus realizing the technical effect of improving the intelligent level, and further solving the technical problem of low intelligent level due to the inability to extract the corresponding spectrum by analyzing the face video.

[0085] According to the embodiment of the present invention, preferably, after obtaining the face video data and before determining the spectra of different wavelengths by identifying the facial blood filling degree caused by blood pulsation in the face video data, it further includes:

[0086] Performing a frame splitting operation on the face video data to obtain several short signals arranged in chronological order; performing peak detection and sorting on the several short signals to obtain key peak points; classifying and feature recording the face video data according to the key peak points.

[0087] The collected ppg signal will be subjected to a frame splitting operation. This means that for the pulse wave signal of t seconds, we use the sliding window method to intercept it to obtain m short signals. Here, the size of the sliding window and the moving step length are fixed, and are set to window_size and step_time seconds respectively.

[0088] Secondly, several short signals obtained through frame segmentation are arranged in chronological order and enter the key point detection module for subsequent processing. The main task of this module is to perform peak detection and sorting to obtain all key peak points. This detection module includes three sub-modules: filtering, peak detection, and detection supplementation, to ensure that the obtained peak points are accurate and complete.

[0089] Finally, after the processing of key point detection, the obtained peak points and the entire pulse wave signal are input into the classification module. The main task of this classification module is to classify the pulse wave signal according to the intensity, volatility, and flatness of the signal, and record its characteristics. This classification process can perform a detailed analysis of the signal to further understand the characteristics of the pulse wave signal.

[0090] This set of technical methods can divide the entire acquisition and processing process into clear steps through phased processing. The settings and operations of each step play a crucial role in the accuracy of the final pulse wave signal and feature extraction. In this way, the pulse wave signal can be collected and analyzed more precisely, providing strong support for subsequent classification and recognition.

[0091] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0092] According to an embodiment of the present invention, there is also provided an apparatus for implementing the above-mentioned method for detecting the human body bilirubin content, as Figure 2 shown, the apparatus includes: an acquisition module 10 for acquiring face video data; an extraction module 20 for determining spectra of different wavelengths by identifying the facial blood filling degree caused by blood pulsation in the face video data, and then extracting the target dynamic spectrum therefrom; a training module 30 for performing model training based on the target dynamic spectrum using the partial least squares regression algorithm to obtain a bilirubin detection model; and a prediction module 40 for receiving the face video to be measured configured by the user to be measured, inputting the face video to be measured into the bilirubin detection model, and predicting the human body bilirubin content of the user to be measured. This apparatus achieves the same technical effect as the method for detecting the human body bilirubin content.

[0093] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. In this way, the present invention is not limited to any specific combination of hardware and software.

[0094] The above are only the preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for detecting bilirubin content in human body, characterized in that: include: Get face video data; By identifying the facial blood fullness caused by blood pulsation in face video data, the spectra of different wavelengths are determined, and then the target dynamic spectrum is extracted from them; The partial least squares regression algorithm was used to train the model based on the target dynamic spectrum to obtain the bilirubin detection model; Receive the face video to be tested configured by the user to be tested, input the face video to be tested into the bilirubin detection model, and predict the human bilirubin content of the user to be tested.

2. The method for detecting human bilirubin content according to claim 1, characterized in that: Get face video data, including: A near-infrared and visible light binocular camera is used to illuminate the human face to obtain face video data; wherein the face video data is a pulse wave signal with a sampling frequency of n Hz and a continuous acquisition time of t seconds.

3. The method for detecting human bilirubin content according to claim 1, characterized in that: After acquiring the face video data, before identifying the facial blood fullness caused by blood pulsation in the face video data to determine the spectrum of different wavelengths, the method further includes: The face video data is framed to obtain a number of short signals arranged in time sequence; the several short signals are peak detected and sorted to obtain key peak points; the face video data is classified and feature recorded according to the key peak points.

4. The method for detecting human bilirubin content according to claim 1, characterized in that: By identifying the facial blood fullness caused by blood pulsation in face video data, the spectra of different wavelengths are determined, and then the target dynamic spectrum is extracted from them, including: Based on the wavelength range of near-infrared and visible light, a specified number of wavelengths are selected in the face video data; after extracting the fundamental component and introducing the harmonic component, the absorption spectrum and fluorescence spectrum of normal human serum and bilirubin-serum are obtained, and the spectral line characteristics of the two and the content of bilirubin in serum are obtained based on the absorption spectrum and fluorescence spectrum; by taking the logarithm of the spectrum and arranging it according to the time axis, the photoelectric logarithmic pulse three-dimensional data is formed; referring to the linear relationship between the amplitude of the photoelectric logarithmic pulse wave at different wavelengths and the blood absorbance, the frequency domain extraction method based on the linear characteristics of Fourier transform is used to extract the amplitude of the logarithmic pulse wave at different wavelengths in turn to obtain the target dynamic spectrum.

5. The method for detecting human bilirubin content according to claim 1, characterized in that: The partial least squares regression algorithm is used to train the model based on the target dynamic spectrum to obtain the bilirubin detection model, including: S001, preprocessing the target dynamic spectrum using a preprocessing method; S002. Determine the optimal number of principal components for modeling within the entire spectrum; S003, dividing the entire spectrum region into n sub-intervals of equal width by using the iPLS algorithm; S004, perform PLS regression on each sub-interval to obtain n local regression models; S005. Using the mean square error of interactive validation RMSECV as the accuracy measure of each model, the accuracy of the full spectrum model and each local model is compared respectively, and the sub-interval where the local model with the highest accuracy is located is selected as the selected interval; S006. Perform internal PLS regression in the selected interval to establish a bilirubin detection model.

6. The method for detecting human bilirubin content according to claim 5, characterized in that: The partial least squares regression algorithm is used to train the model based on the target dynamic spectrum to obtain the bilirubin detection model, which also includes: S007, re-partition the spectral region, repeat S003-S005, and compare the modeling effects of the selected intervals of different partitions to determine the optimal interval; S008. Perform internal PLS regression in the optimal interval to establish the optimal bilirubin detection model.

7. A device for detecting bilirubin content in human body, characterized in that: include: An acquisition module is used to acquire face video data; An extraction module is used to identify the facial blood fullness caused by blood pulsation in the face video data to determine the spectrum of different wavelengths, and then extract the target dynamic spectrum from it; A training module, used to perform model training based on the target dynamic spectrum using a partial least squares regression algorithm to obtain a bilirubin detection model; The prediction module is used to receive the face video to be tested configured by the user to be tested, input the face video to be tested into the bilirubin detection model, and predict the human bilirubin content of the user to be tested.

8. The human bilirubin content detection device according to claim 7, characterized in that: After acquiring the face video data, before identifying the facial blood fullness caused by blood pulsation in the face video data to determine the spectrum of different wavelengths, the method further includes: The face video data is framed to obtain a number of short signals arranged in time sequence; the several short signals are peak detected and sorted to obtain key peak points; the face video data is classified and feature recorded according to the key peak points.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the method for detecting the bilirubin content in the human body according to any one of claims 1 to 7 when running.

10. A server, comprising: A memory and a processor, characterized in that a computer program is stored in the memory, wherein the processor is configured to run the computer program to execute the method for detecting human bilirubin content according to any one of claims 1 to 7.

Images