CN119745378B - Blood fat monitoring method based on prediction model - Google Patents

Abstract

The present invention discloses a blood lipid monitoring method based on a prediction model, which relates to the technical field of blood lipid monitoring, and includes the following steps: using a spectrometer to emit light of an initially set wavelength to illuminate the skin of a patient's test area, and detecting the reflected light to obtain reflective spectrum data at different wavelengths; in a detection window of a set duration, using optical sensing technology to obtain basic skin information of the patient's test area. The present invention utilizes light of an initially set wavelength of a spectrometer to illuminate the skin, and evaluates the light absorption characteristics of the patient's skin through a deep learning model, intelligently controls the wavelength of light, and optimizes the high light absorption period and the low light absorption period respectively, to ensure accurate blood lipid monitoring under different skin conditions, which not only improves the measurement accuracy of blood lipid concentration, reduces the risk of misdiagnosis and missed diagnosis, but also helps to promptly detect patients with hyperlipidemia, prevent further deterioration of cardiovascular disease, and significantly improve the reliability and effectiveness of clinical diagnosis.

Description

Blood lipid monitoring method based on prediction model

Technical Field

The present invention relates to the technical field of blood lipid monitoring, and in particular to a blood lipid monitoring method based on a prediction model.

Background Art

Blood lipid monitoring based on prediction models is a method that uses data analysis and machine learning technology to predict and manage individual blood lipid levels. By collecting patients' historical blood lipid data, living habits, diet, exercise and other related information, the prediction model can analyze these data and identify the key factors that affect blood lipid levels, thereby predicting future blood lipid trends. It can not only help individuals discover potential health risks in advance, but also provide doctors with a scientific basis to formulate more accurate intervention measures.

The core of this monitoring method is to continuously update and optimize the prediction results through algorithms and models to improve the accuracy and reliability of the prediction. Compared with traditional blood lipid monitoring methods, monitoring based on predictive models can track changes in blood lipid levels in a more personalized and dynamic manner, providing opportunities for early warning and intervention, thereby more effectively preventing and managing health problems related to blood lipids such as cardiovascular disease.

Blood lipid monitoring based on predictive models usually relies on optical sensing technology and is implemented through the principle of spectral analysis. Specifically, this technology emits light of a specific wavelength to the skin and detects the reflected light to obtain concentration information of certain components in the blood. Different blood lipid components (such as cholesterol, triglycerides, etc.) have specific absorption and reflection characteristics for light of different wavelengths. By analyzing these characteristics, the concentration of blood lipids can be indirectly inferred. For example, near-infrared light (NIR) has strong penetration in skin tissue and can provide clearer blood component information.

The prior art has the following deficiencies:

When using optical sensing technology to monitor blood lipids, the wavelength of emitted light in existing technologies is usually not intelligently adjustable. Since the patient's skin is dynamically changeable, the skin's absorption and reflection characteristics of fixed wavelength light are different in different states. Therefore, the continuous use of a fixed emission wavelength may not accurately measure the blood lipid levels of all individuals. This inaccurate measurement data can lead to errors in the estimation of blood lipid concentrations, which may lead to misdiagnosis or missed diagnosis. For example, if patients with hyperlipidemia fail to detect the problem in time, it may lead to worsening of cardiovascular disease.

The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art.

Summary of the invention

The purpose of the present invention is to provide a blood lipid monitoring method based on a predictive model, which uses light of an initially set wavelength of a spectrometer to irradiate the skin, and evaluates the light absorption characteristics of the patient's skin through a deep learning model, intelligently adjusts the light wavelength, and optimizes the high light absorption period and the low light absorption period respectively to ensure accurate blood lipid monitoring under different skin conditions. This not only improves the measurement accuracy of blood lipid concentration and reduces the risk of misdiagnosis and missed diagnosis, but also helps to promptly detect patients with hyperlipidemia and prevent further deterioration of cardiovascular diseases, thereby significantly improving the reliability and effectiveness of clinical diagnosis to solve the problems in the above-mentioned background technology.

In order to achieve the above object, the present invention provides the following technical solution: a blood lipid monitoring method based on a prediction model, comprising the following steps:

A spectrometer is used to emit light of an initially set wavelength onto the skin of the patient's test area, and the reflected light is detected to obtain reflectance spectrum data at different wavelengths;

In a detection window of set duration, optical sensing technology is used to obtain basic skin information of the patient's test area, key features are extracted from the obtained basic skin information, and the key features are preprocessed to form a physiological feature vector;

After analyzing the preprocessed physiological feature vector, the analyzed physiological feature vector is input into the trained deep learning model, and the light absorption characteristics of the patient's skin are evaluated by the deep learning model;

Based on the light absorption characteristics output by the deep learning model, the skin of the patient's test area is divided into a high light absorption period and a low light absorption period;

During the period of low light absorption, the light of the initially set wavelength continues to be irradiated onto the skin of the patient's test area for blood lipid monitoring. During the period of high light absorption, the actual wavelength of the emitted light is intelligently controlled based on the initially set wavelength to compensate for the light absorption during the period of high light absorption. The actual wavelength after control is then put into practical use to ensure efficient monitoring of blood lipids under different skin conditions.

Preferably, the key features extracted from the acquired basic skin information include pigmentation and blood flow in the patient's test area, and the pigmentation and blood flow in the patient's test area are acquired by optical coherence tomography in optical sensing technology, and the pigmentation and blood flow are preprocessed to form a physiological feature vector, wherein the physiological feature vector includes a pigmentation feature vector and a blood flow feature vector. The specific steps are as follows:

Use the OCT device to scan the patient's test area and collect tomographic images of the skin;

The acquired OCT images are processed and segmented to extract areas related to pigmentation and blood flow;

Extracting features related to pigmentation and blood flow from the segmented images;

Preprocessing the extracted pigmentation and blood flow features to form a pigmentation feature vector and a blood flow feature vector;

The preprocessed pigmentation and blood flow features are integrated to construct pigmentation feature vectors and blood flow feature vectors.

Preferably, after analyzing the preprocessed pigmentation feature vector and blood flow feature vector, a pigmentation index and an abnormal blood flow increment index are generated, and the analyzed pigmentation index and abnormal blood flow increment index are input into a trained deep learning model to generate a light absorption evaluation coefficient, and the light absorption characteristics of the patient's skin are intelligently evaluated through the light absorption evaluation coefficient.

Preferably, the light absorption evaluation coefficient generated in the detection window of the set time length is compared and analyzed with the preset light absorption evaluation coefficient reference threshold, and the light absorption state of the detection window is divided. The result of the comparison and analysis is as follows:

If the light absorption evaluation coefficient is greater than the light absorption evaluation coefficient reference threshold, the detection window is divided into a high light absorption period;

If the light absorption evaluation coefficient is less than or equal to the light absorption evaluation coefficient reference threshold, the detection window is divided into a low light absorption period.

Preferably, for the high light absorption period, the actual wavelength of the emitted light is intelligently regulated based on the initial set wavelength, the light absorption in the high light absorption period is compensated, and the specific steps of actually applying the regulated actual wavelength are as follows:

Determine the initial setting wavelength and calibrate the initial setting wavelength as ;

Based on the light absorption evaluation result, that is, the light absorption evaluation coefficient, the actual wavelength of the emitted light is intelligently controlled to compensate for the light absorption effect during the high light absorption period. The specific control formula is: , where Indicates the actual wavelength of the emitted light after intelligent control. Represents the reference threshold value of the light absorption evaluation coefficient, represents the light absorption evaluation coefficient, and , Represents the adjustment coefficient, which is used to control the amplitude of wavelength adjustment;

The actual wavelength calculated according to the intelligent control formula , adjust the emission light wavelength of the optical sensor in real time, and use the adjusted wavelength for blood lipid monitoring to ensure that the optical sensor provides accurate reflection spectrum data.

Preferably, after analyzing the preprocessed pigmentation feature vector, the specific steps of generating the pigmentation index are as follows:

Under the detection window of set duration, the pigmentation feature vector is normalized to ensure that all feature values are on the same scale. The calculation expression of normalization is: , where X is the original pigmentation feature vector, which contains feature data of multiple dimensions. is the normalized eigenvector, all eigenvalues are between 0 and 1, and are the minimum and maximum values in the eigenvector, respectively, used for normalization;

The light absorption characteristic model is used to evaluate the light absorption at different wavelengths. Considering the influence of various characteristics on light absorption, the expression of the light absorption characteristic model is: , is at wavelength The light absorption characteristics under is the influence coefficient of the i-th feature on the light absorption characteristics, is the attenuation coefficient corresponding to the i-th feature, indicating the absorption rate of light under this feature, is the normalized ith eigenvalue;

The nonlinear combination of features is constructed through the sine function to capture the complex interactive relationship between features. The expression for constructing the nonlinear combination feature is: , where Z is a nonlinear combination feature, which is used to comprehensively reflect the complex effects of pigmentation. is a weight parameter, which indicates the influence of the combination of the i-th and j-th features on the total features. is a sinusoidal nonlinear combination of the i-th and j-th eigenvalues;

The final pigmentation index is generated by weighted summing and taking the logarithm of the light absorption characteristics. The calculation expression is: , where Represents the pigmentation index, which comprehensively reflects the light absorption characteristics of the pigmentation feature vector. is the weight coefficient of the kth wavelength, is the light absorption characteristic at the kth wavelength, T is the duration of the detection window, is the kth wavelength.

Preferably, after analyzing the preprocessed blood flow characteristic vector, the specific steps of generating the abnormal blood flow increment index are as follows:

Under the detection window of set duration, obtain the blood flow feature vector of time series ,in represents the fth blood flow characteristic at time t;

Calculate the rate of change of each feature. The expression is: , where is the blood flow rate change vector at time t, represents the rate of change of the fth blood flow characteristic at time t;

Calculate the rate of change of blood flow rate (i.e. acceleration), the calculation expression is: , where is the blood flow acceleration vector at time t, represents the acceleration of the fth blood flow characteristic at time t;

Calculate the abnormal blood flow increment index, the calculation expression is: , where Indicates the abnormal blood flow increment index, and is the weighting coefficient of each blood flow feature, which is used to adjust the impact of different features on the comprehensive index. is the rate of change of the fth blood flow characteristic at time t, is the acceleration of the fth blood flow characteristic at time t, and Indicates the start and end time of the detection window.

In the above technical solution, the technical effects and advantages provided by the present invention are:

The present invention utilizes light of an initially set wavelength of the spectrometer to irradiate the skin, and evaluates the light absorption characteristics of the patient's skin through a deep learning model, intelligently adjusts the light wavelength, and optimizes the high light absorption period and the low light absorption period respectively, to ensure accurate blood lipid monitoring under different skin conditions. This not only improves the measurement accuracy of blood lipid concentration and reduces the risk of misdiagnosis and missed diagnosis, but also helps to promptly detect patients with hyperlipidemia and prevent further deterioration of cardiovascular diseases, thereby significantly improving the reliability and effectiveness of clinical diagnosis.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

FIG1 is a flow chart of a method for monitoring blood lipids based on a prediction model according to the present invention.

DETAILED DESCRIPTION

Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

The present invention provides a blood lipid monitoring method based on a prediction model as shown in FIG1 , comprising the following steps:

A spectrometer is used to emit light of an initially set wavelength onto the skin of the patient's test area, and the reflected light is detected to obtain reflectance spectrum data at different wavelengths;

A spectrometer is an instrument that can emit and detect light of a specific wavelength. The light of the initial set wavelength means that during the spectral analysis process, the spectrometer is first set to emit light of a specific wavelength. This wavelength is used as a reference to illuminate the skin of the area to be tested in order to begin collecting initial reflectance spectrum data. When selecting the initial wavelength, the needs of the target measurement and the characteristics of the skin are usually considered to ensure that the selected wavelength can effectively interact with the various components in the skin and provide useful reflection signals, laying the foundation for subsequent spectral analysis and prediction of blood lipid levels.

When light hits the skin, it interacts with various components on the surface and inside the skin. Different components have different absorption and reflection characteristics for light. This step is the basis for collecting skin reflection spectrum data to ensure reliable original lighting conditions for subsequent testing and analysis. The sensor of the spectrometer collects the light signal reflected from the skin. The sensor is usually highly sensitive and can detect the intensity of light at different wavelengths. By detecting the reflected light, the reflection characteristic data of the skin to light of different wavelengths can be collected. These data will be used to analyze the optical properties of the skin.

The spectrometer generates a complete reflection spectrum by scanning light of different wavelengths and recording the reflection intensity at each wavelength. The reflection spectrum data contains the reflection characteristics of the skin at each wavelength and can be used to extract characteristic information of the skin. These spectral data will be input into the prediction model to analyze the optical properties of the skin and ultimately predict blood lipid levels.

In a detection window of set duration, optical sensing technology is used to obtain basic skin information of the patient's test area, key features are extracted from the obtained basic skin information, and the key features are preprocessed to form a physiological feature vector;

The key features extracted from the basic skin information include pigmentation and blood flow in the patient's test area. The pigmentation and blood flow in the patient's test area are obtained by optical coherence tomography (OCT) in optical sensing technology. The pigmentation and blood flow are preprocessed to form a physiological feature vector, wherein the physiological feature vector includes a pigmentation feature vector and a blood flow feature vector. The specific steps are as follows:

Use the OCT device to scan the patient's test area and collect tomographic images of the skin;

These images contain information about the structure of the skin, including pigmentation and blood flow. OCT can provide high-resolution cross-sectional images, and through optical interferometry, it can obtain skin reflection signals at different depths and generate detailed images of skin tissue.

The acquired OCT images are processed and segmented to extract areas related to pigmentation and blood flow;

Image processing includes steps such as denoising, contrast enhancement and edge detection. Segmentation is the process of identifying and separating pigmented and vascular areas in the skin through image segmentation algorithms such as threshold segmentation, region growing or machine learning methods.

Extracting features related to pigmentation and blood flow from the segmented images;

For pigmentation, the extracted features may include the distribution density, concentration, and area of the pigment, etc. For blood flow, the extracted features may include blood flow velocity, vascular density, and blood flow intensity, etc. These features reflect the quantitative information of skin pigment and blood flow.

Preprocessing the extracted pigmentation and blood flow features to form a pigmentation feature vector and a blood flow feature vector;

Preprocessing steps include data standardization, normalization, and dimensionality reduction. Standardization and normalization help eliminate data scale differences, and dimensionality reduction techniques (such as PCA) can reduce the dimension of feature vectors and retain the most important feature information.

Integrate the preprocessed pigmentation and blood flow features to construct a pigmentation feature vector and a blood flow feature vector;

The pigmentation feature vector contains the data of various pigment-related features, and the blood flow feature vector contains the data of various blood flow-related features. These feature vectors will be used for subsequent machine learning model training and analysis to provide a detailed quantitative description of skin conditions.

After analyzing the preprocessed physiological feature vector, the analyzed physiological feature vector is input into the trained deep learning model, and the light absorption characteristics of the patient's skin are evaluated by the deep learning model;

After analyzing the preprocessed pigmentation feature vector and blood flow feature vector, the pigmentation index and blood flow abnormal increment index are generated. The analyzed pigmentation index and blood flow abnormal increment index are input into the trained deep learning model to generate a light absorption evaluation coefficient. The light absorption evaluation coefficient is used to intelligently evaluate the light absorption characteristics of the patient's skin.

Pigmentation in the patient's lipid monitoring area will cause the wavelength of the emitted light to be partially absorbed, thus affecting the accuracy of lipid monitoring. Pigmentation is mainly due to the increase in the content of melanin and other pigments in the skin, which have a strong absorption effect on light of specific wavelengths. When light shines on the pigmented area, a portion of the light energy is absorbed by the pigment instead of being reflected or transmitted back for detection. This light absorption will cause the reflected light signal received by the optical sensor to be weakened or distorted, which in turn affects the accuracy of the measurement data. Especially in lipid monitoring based on spectral analysis, the intensity and characteristics of the reflected light signal are directly related to the prediction and analysis of blood lipid concentration.

In addition, light absorption caused by pigmentation may not only change the intensity of the reflected light, but also the spectral characteristics of the reflected light. This means that the spectral data received by the sensor may no longer accurately reflect the actual blood composition under the skin, but instead contain spectral features caused by pigments. This mixed spectral information can interfere with the input data of the prediction model, causing the model to make errors when assessing blood lipid levels. For example, the spectral characteristics of blood lipid concentration may be masked by the absorption characteristics of pigments, resulting in the model being unable to correctly identify and separate the spectral signals of blood lipids. In this case, it is difficult to accurately assess blood lipid concentrations even with advanced machine learning or deep learning models.

After analyzing the preprocessed pigmentation feature vector, the specific steps to generate the pigmentation index are as follows:

Under the detection window of set duration, the pigmentation feature vector is normalized to ensure that all feature values are on the same scale. The calculation expression of normalization is: , where X is the original pigmentation feature vector, which contains feature data of multiple dimensions. is the normalized eigenvector, all eigenvalues are between 0 and 1, and are the minimum and maximum values in the eigenvector, respectively, used for normalization;

The light absorption characteristic model is used to evaluate the light absorption at different wavelengths. Considering the influence of various characteristics on light absorption, the expression of the light absorption characteristic model is: , is at wavelength The light absorption characteristics under is the influence coefficient of the i-th feature on the light absorption characteristics, is the attenuation coefficient corresponding to the i-th feature, indicating the absorption rate of light under this feature, is the normalized ith eigenvalue;

The influence coefficient of pigmentation characteristics on light absorption characteristics It can be obtained through experimental data and regression analysis. First, the skin spectrum data of different individuals is measured experimentally to obtain the light absorption at different wavelengths. Then, the pigmentation characteristics of these individuals, such as pigment concentration, distribution density, etc., are recorded. Using these data, a relationship model between pigmentation characteristics and light absorption characteristics is established through multiple linear regression or nonlinear regression analysis methods. Influence coefficient The coefficients of the regression model can be directly extracted, reflecting the contribution of each pigmentation feature to light absorption.

Attenuation coefficient corresponding to pigmentation features Obtained through light absorption experiments and spectral analysis methods. When conducting spectral experiments, a spectrometer is used to illuminate the skin sample at different wavelengths and measure the intensity of the reflected light. Combined with the data of pigmentation characteristics, the law of light absorption changing with wavelength is analyzed through nonlinear fitting methods (such as exponential decay model fitting). Specifically, the nonlinear least squares method can be used to fit the absorption curve and extract the attenuation coefficient These attenuation coefficients describe the light absorption rate of each pigmentation feature at different wavelengths, reflecting the attenuation characteristics of light when it propagates in the skin.

The nonlinear combination of features is constructed through the sine function to capture the complex interactive relationship between features. The expression for constructing the nonlinear combination feature is: , where Z is a nonlinear combination feature, which is used to comprehensively reflect the complex effects of pigmentation. is a weight parameter, which indicates the influence of the combination of the i-th and j-th features on the total features. is a sinusoidal nonlinear combination of the i-th and j-th eigenvalues;

The final pigmentation index is generated by weighted summing and taking the logarithm of the light absorption characteristics. The calculation expression is: , where Represents the pigmentation index, which comprehensively reflects the light absorption characteristics of the pigmentation feature vector. is the weight coefficient of the kth wavelength, is the light absorption characteristic at the kth wavelength, T is the duration of the detection window, is the kth wavelength.

It can be seen from the pigmentation index that within the set detection window, the larger the value of the pigmentation index, the more severe the pigmentation in the patient's blood lipid monitoring area, and the greater the impact on light absorption, resulting in poorer accuracy of blood lipid monitoring. On the contrary, the smaller the value of the pigmentation index, the lighter the pigmentation and the smaller the impact on light absorption, so the higher the accuracy of blood lipid monitoring. A high pigmentation index means that light will be absorbed and scattered more when propagating in the skin, affecting the intensity and spectral characteristics of the reflected light, and thus interfering with the accurate measurement of blood lipid concentration; a low pigmentation index means less light absorption and scattering, a clearer reflected light signal, and a more accurate measurement result.

Increased blood flow in the patient's blood lipid monitoring area will affect the wavelength of the emitted light being absorbed, thereby affecting the accuracy of blood lipid monitoring. Increased blood flow means that the blood content in the area has increased, and the components in the blood (such as red blood cells, hemoglobin, etc.) have significant light absorption characteristics. Specifically, hemoglobin has a high absorption rate for light of specific wavelengths (especially near-infrared light and visible light). When blood flow increases, the concentration of these components rises, causing more light to be absorbed, and the reflected light signal is weakened, affecting the signal strength and quality received by the optical sensor.

Since the light absorption properties of blood can interfere with optical signals, increased blood flow may mask or confuse the light absorption properties of other components in the skin, such as fat or pigment. This interference makes it difficult for spectral analysis to accurately distinguish and quantify blood lipid concentrations, resulting in deviations in the input data of the prediction model, which in turn affects the final monitoring results. In addition, dynamically changing blood flow results in different conditions for each measurement, increasing data inconsistency and measurement errors.

After analyzing the preprocessed blood flow feature vector, the specific steps of generating the abnormal blood flow increment index are as follows:

Under the detection window of set duration, obtain the blood flow feature vector of time series ,in represents the fth blood flow characteristic at time t;

Calculate the rate of change of each feature. The expression is: , where is the blood flow rate change vector at time t, represents the rate of change of the fth blood flow characteristic at time t;

Calculate the rate of change of blood flow rate (i.e. acceleration), the calculation expression is: , where is the blood flow acceleration vector at time t, represents the acceleration of the fth blood flow characteristic at time t;

Calculate the abnormal blood flow increment index, the calculation expression is: , where Indicates the abnormal blood flow increment index, and is the weighting coefficient of each blood flow feature, which is used to adjust the impact of different features on the comprehensive index. is the rate of change of the fth blood flow characteristic at time t, is the acceleration of the fth blood flow characteristic at time t, and Indicates the start time and end time of the detection window;

The weighting coefficient of each blood flow feature can be determined by using a multi-objective optimization algorithm (such as genetic algorithm, particle swarm optimization algorithm) or a machine learning model (such as multivariate regression analysis, random forest regression). These algorithms use historical data sets and labeled data to automatically adjust and determine the optimal weighting coefficient by optimizing the objective function (such as minimizing the prediction error), thereby improving the prediction accuracy and robustness of the model.

It can be seen from the abnormal blood flow increment index that, within the set time window, the larger the performance value of the abnormal blood flow increment index in the patient's blood lipid monitoring area, the more obvious the abnormal change in blood flow, which will interfere with the detection signal of the optical sensor, thereby reducing the accuracy of blood lipid monitoring; on the contrary, the smaller the performance value of the abnormal blood flow increment index, the more stable the blood flow change is, and the less interference to the optical sensor is, thereby improving the accuracy of blood lipid monitoring.

The deep learning model is not specifically limited here, and can achieve the pigmentation index Abnormal blood flow increment index Comprehensive analysis to generate light absorption evaluation coefficients In order to realize the technical solution of the present invention, the present invention provides a specific implementation method;

Light absorption evaluation coefficient The resulting calculation formula is: , where , Pigmentation Index Abnormal blood flow increment index The preset scaling factor of , Both are greater than 0.

It can be seen from the light absorption evaluation coefficient that, within the detection window of set time length, the greater the performance value of the pigmentation index generated after analyzing the preprocessed pigmentation feature vector, the greater the performance value of the blood flow abnormal increment index generated after analyzing the preprocessed blood flow feature vector, that is, the greater the performance value of the light absorption evaluation coefficient generated within the detection window of set time length, the more serious the light absorption, and vice versa.

Based on the light absorption characteristics output by the deep learning model, the skin of the patient's test area is divided into a high light absorption period and a low light absorption period;

The light absorption evaluation coefficient generated in the detection window of the set time is compared and analyzed with the preset light absorption evaluation coefficient reference threshold, and the light absorption state of the detection window is divided. The results of the comparison and analysis are as follows:

If the light absorption evaluation coefficient is greater than the light absorption evaluation coefficient reference threshold, the detection window is divided into a high light absorption period;

If the light absorption evaluation coefficient is less than or equal to the light absorption evaluation coefficient reference threshold, the detection window is divided into a low light absorption period.

During the period of low light absorption, the light of the initially set wavelength continues to be irradiated onto the skin of the patient's test area to monitor blood lipids. During the period of high light absorption, the actual wavelength of the emitted light is intelligently regulated based on the initially set wavelength to compensate for the light absorption during the period of high light absorption. The actual wavelength after regulation is actually used to ensure efficient monitoring of blood lipids under different skin conditions.

During the period of low light absorption, the initially set wavelength of light continues to be used to illuminate the patient's skin in the area to be tested for blood lipid monitoring. The purpose is to ensure that a stable light source and wavelength are used for monitoring when light absorption is low to obtain consistent and reliable data. This method takes advantage of the fact that the skin has less interference with light during the period of low light absorption, ensuring that the optical sensor can accurately detect the reflected light signal, thereby improving the accuracy of blood lipid monitoring.

For the high light absorption period, the actual wavelength of the emitted light is intelligently regulated based on the initial set wavelength to compensate for the light absorption during the high light absorption period, and the specific steps for the actual application of the regulated wavelength are as follows:

Determine the initial setting wavelength and calibrate the initial setting wavelength as ;

Based on the light absorption evaluation result, that is, the light absorption evaluation coefficient, the actual wavelength of the emitted light is intelligently controlled to compensate for the light absorption effect during the high light absorption period. The specific control formula is: , where Indicates the actual wavelength of the emitted light after intelligent control. Represents the reference threshold value of the light absorption evaluation coefficient, represents the light absorption evaluation coefficient, and , It represents the adjustment coefficient, which is used to control the amplitude of wavelength adjustment and determine the sensitivity of adjustment;

Adjustment coefficient The setting of is usually determined through experiments and optimization methods, including regression analysis based on historical data or using optimization algorithms (such as genetic algorithms, particle swarm optimization) to adjust to minimize the prediction error or maximize the measurement accuracy. In these processes, the adjustment coefficient It was repeatedly adjusted and tested until the optimal value was found that provided the best lipid monitoring performance under various light absorption conditions.

The actual wavelength calculated according to the intelligent control formula , adjust the emission light wavelength of the optical sensor in real time, and use the adjusted wavelength for blood lipid monitoring to ensure that the optical sensor provides accurate reflection spectrum data.

The present invention utilizes light of an initially set wavelength of the spectrometer to irradiate the skin, and evaluates the light absorption characteristics of the patient's skin through a deep learning model, intelligently adjusts the light wavelength, and optimizes the high light absorption period and the low light absorption period respectively, to ensure accurate blood lipid monitoring under different skin conditions. This not only improves the measurement accuracy of blood lipid concentration and reduces the risk of misdiagnosis and missed diagnosis, but also helps to promptly detect patients with hyperlipidemia and prevent further deterioration of cardiovascular diseases, thereby significantly improving the reliability and effectiveness of clinical diagnosis.

The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims (5)

1. The blood fat monitoring method based on the prediction model is characterized by comprising the following steps of:

The method comprises the steps of emitting light with an initial set wavelength to the skin of a region to be detected of a patient by using a spectrometer, and detecting the reflected light to obtain reflection spectrum data under different wavelengths;

under a detection window with a set duration, acquiring basic skin information of a region to be detected of a patient by using an optical sensing technology, extracting key features from the acquired basic skin information, and preprocessing the key features to form physiological feature vectors;

after analyzing the preprocessed physiological feature vector, inputting the analyzed physiological feature vector into a trained deep learning model, and evaluating the light absorption characteristic of the skin of a patient through the deep learning model;

dividing the skin of the region to be detected of the patient into a high light absorption period and a low light absorption period based on the light absorption characteristic result output by the deep learning model;

For the low light absorption period, continuously keeping the light with the initial set wavelength to irradiate the skin of the region to be detected of the patient for blood fat monitoring, for the high light absorption period, intelligently regulating and controlling the actual wavelength of the emitted light based on the initial set wavelength, compensating the light absorption of the high light absorption period, actually applying the regulated and controlled actual wavelength, and ensuring the high-efficiency monitoring of the blood fat under different skin states;

The key features extracted from the acquired basic skin information comprise pigmentation and blood flow of the region to be detected of the patient, the pigmentation and the blood flow of the region to be detected of the patient are acquired through optical coherence tomography in an optical sensing technology, and the pigmentation and the blood flow are preprocessed to form physiological feature vectors, wherein the physiological feature vectors comprise the pigmentation feature vectors and the blood flow feature vectors, and the specific steps are as follows:

Scanning the region to be detected of the patient by using OCT equipment, and collecting a tomographic image of the skin;

Processing and segmenting the acquired OCT image to extract regions related to pigmentation and blood flow;

Extracting features related to pigmentation and blood flow from the segmented image;

Preprocessing the extracted pigmentation and blood flow characteristics to form a pigmentation characteristic vector and a blood flow characteristic vector;

Integrating the pretreated pigmentation and blood flow characteristics to construct a pigmentation characteristic vector and a blood flow characteristic vector;

after the pre-processed pigmentation characteristic vector and blood flow characteristic vector are analyzed, a pigmentation index and a blood flow abnormal increment index are generated, the analyzed pigmentation index and blood flow abnormal increment index are input into a trained deep learning model to generate a light absorption evaluation coefficient, and the light absorption characteristics of the skin of a patient are intelligently evaluated through the light absorption evaluation coefficient.

2. The blood lipid monitoring method based on a predictive model according to claim 1, wherein the light absorption evaluation coefficient generated under the detection window of a set duration is compared with a preset reference threshold value of the light absorption evaluation coefficient for analysis, the light absorption state of the detection window is divided, and the comparison result is as follows:

If the light absorption evaluation coefficient is greater than the light absorption evaluation coefficient reference threshold, dividing the detection window into a high light absorption period;

if the light absorption evaluation coefficient is equal to or less than the light absorption evaluation coefficient reference threshold, the detection window is divided into a low light absorption period.

3. The blood lipid monitoring method based on a predictive model according to claim 2, wherein for the period of high light absorption, the actual wavelength of the emitted light is intelligently regulated and controlled based on the initial set wavelength, the light absorption in the period of high light absorption is compensated, and the specific steps of actually applying the regulated actual wavelength are as follows:

determining an initial set wavelength and calibrating the initial set wavelength as ;

Based on the light absorption evaluation result, namely the light absorption evaluation coefficient, the actual wavelength of the emitted light is intelligently regulated so as to compensate the light absorption influence of the high light absorption period, and a specific regulation formula is as follows: In which, in the process, Indicating the actual wavelength of the intelligently regulated emitted light,Representing the light absorption evaluation coefficient reference threshold value,Represents the light absorption evaluation coefficient, an,Representing an adjustment factor for controlling the amplitude of the wavelength adjustment;

actual wavelength calculated according to intelligent regulation formula And the wavelength of the emitted light of the optical sensor is adjusted in real time, and the adjusted wavelength is used for blood fat monitoring, so that the optical sensor is ensured to provide accurate reflection spectrum data.

4. The method for monitoring blood lipid based on predictive model as set forth in claim 1, wherein the specific steps of generating a pigmentation index after analyzing the pre-processed pigmentation eigenvector are as follows:

Under a detection window with a set duration, carrying out normalization processing on the pigmentation eigenvector, and ensuring that all eigenvalues are on the same scale, wherein the calculation expression of the normalization processing is as follows: wherein X is an original pigmentation feature vector comprising feature data of multiple dimensions, Is a normalized eigenvector, all eigenvalues are between 0 and 1,AndRespectively the minimum value and the maximum value in the feature vector for normalization;

Light absorption characteristics models are used for evaluating light absorption conditions at different wavelengths, and the light absorption characteristics models are expressed as follows, taking the influence of each characteristic on light absorption into consideration: , Is at the wavelength of The light absorption characteristics of the light source are that,Is the coefficient of influence of the ith feature on the light absorption characteristics,Is the attenuation coefficient corresponding to the ith feature, which represents the absorption rate of light at that feature,Is the ith characteristic value after normalization;

The complex interaction relation among the features is captured through nonlinear combination of the sine function construction features, and the expression of the nonlinear combination feature construction is as follows: wherein Z is a nonlinear combination characteristic for comprehensively reflecting the complex influence of pigmentation, Is a weight parameter indicating the degree of influence of the combination of the ith and jth features on the total features,Is the sinusoidal nonlinear combination of the ith and jth eigenvalues;

the final pigmentation index is generated by weighted summing the light absorption characteristics and taking the logarithm, the expression calculated is: In which, in the process, Representing the pigmentation index, comprehensively reflecting the light absorption characteristics of the characteristic vector of pigmentation,Is the weight coefficient of the kth wavelength,Is the light absorption characteristic at the kth wavelength, T is the duration of the detection window,Is the kth wavelength.

5. The blood lipid monitoring method based on a predictive model according to claim 1, wherein the specific steps of generating the abnormal blood flow increment index after analyzing the preprocessed blood flow feature vector are as follows:

Under a detection window with a set duration, acquiring a time series blood flow characteristic vector WhereinRepresenting the f-th blood flow characteristic at time t;

Calculating the change rate of each feature, wherein the calculated expression is: In which, in the process, Is the blood flow change rate vector at time t,Representing the rate of change of the f-th blood flow characteristic at time t;

Calculating the change rate of the blood flow rate change rate, wherein the calculated expression is: In which, in the process, Is the blood flow acceleration vector at time t,Acceleration of the f-th blood flow characteristic at time t is represented;

Calculating an abnormal increment index of blood flow, wherein the calculated expression is: In which, in the process, Represents an index of abnormal increment of blood flow,AndIs a weighting coefficient for each blood flow characteristic, is used to adjust the impact of different characteristics on the composite index,Is the rate of change of the f-th blood flow characteristic at time t,Is the acceleration of the f-th blood flow feature at time t,AndIndicating the start time and end time of the detection window.