Methods and devices for continuous non-invasive physiological parameter detection

Abstract

This invention relates to a method and apparatus for continuous non-invasive physiological parameter detection, belonging to the field of physiological parameter detection technology. The method is based on photoplethysmography (PPG) pulse wave detection data at multiple wavelengths, extracting absorbance-related characteristic parameters and pulse wave-related characteristic parameters. The hemoglobin concentration detection result is obtained through calculation using a hemoglobin concentration calculation model. The absorbance-related characteristic parameters include adjacent surface difference, relative difference, adjacent surface ratio, relative ratio, and cumulative surface ratio, extracted based on a normalized absorbance fluctuation amplitude curve. The pulse wave characteristic parameters include the maximum slope of the rising limb, rise time, fall time, dicrotic wave slope, dicrotic amplitude ratio, rising / falling limb time ratio, low / high frequency energy ratio, and harmonic component ratio, extracted based on a normalized pulse wave. The detection apparatus for implementing this method includes a probe and data processing circuitry. This invention can effectively improve the accuracy of non-invasive hemoglobin concentration detection results.

Description

Technical Field

[0001] This invention relates to a method and apparatus for continuous non-invasive physiological parameter detection, which can be mainly used for non-invasive detection of hemoglobin concentration and belongs to the field of physiological parameter detection technology. Background Technology

[0002] The concentration of different blood components in the human body is an important clinical biochemical indicator. For example, hemoglobin concentration is crucial for the identification and classification of anemia, clinical transfusion guidance, and the diagnosis of other diseases. Currently, hemoglobin concentration testing can be divided into invasive, minimally invasive, and non-invasive methods. Invasive and minimally invasive methods require blood sampling, which can cause pain and psychological distress to the test subject, and the wound also carries a certain risk of infection. Furthermore, they cannot achieve continuous, real-time monitoring. While existing non-invasive methods do not require blood sampling and can achieve continuous, real-time monitoring, their accuracy still has significant room for improvement due to individual differences and limitations in applicability, hindering product implementation and technology promotion. Summary of the Invention

[0003] The purpose of this invention is to improve the accuracy of non-invasive hemoglobin concentration detection results.

[0004] The technical solution of this invention is: a continuous non-invasive physiological parameter detection method, which is based on photoplethysmography pulse wave detection data under multiple wavelengths (using light of various different wavelengths), extracts absorbance-related characteristic parameters and pulse wave-related characteristic parameters, and uses a hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters to calculate and obtain the hemoglobin concentration detection result.

[0005] Preferably, based on photoplethysmography (PPG) data, the absorbance fluctuation amplitude at each wavelength is calculated. An absorbance fluctuation amplitude curve is fitted based on the absorbance fluctuation amplitude at each wavelength. Several wavelength windows of a set width (WindowSize) are set. Based on the area under the absorbance fluctuation amplitude curve within the wavelength window, some or all of the following feature parameters are extracted as absorbance-related feature parameters used in the hemoglobin concentration calculation model:

[0006] Neighbor difference: ;

[0007] Relative difference: ;

[0008] Adjacent face ratio: ;

[0009] In comparison: ;

[0010] Progressive ratio: ,

[0011] in, For adjacent face differences, For relative difference, For adjacent face ratio, For comparison, For progressive surface ratio, The area under the absorbance fluctuation amplitude curve within the j-th wavelength window is denoted as . The area under the absorbance fluctuation amplitude curve within the (j-1)th wavelength window (and so on). To select the area under the absorbance fluctuation amplitude curve within the wavelength window used to calculate the relative difference, To select the one used for calculating the relative ratio The area under the absorbance fluctuation amplitude curve within each wavelength window, or the sum of the areas under the absorbance fluctuation amplitude curves. , It is the sum of the areas under the absorbance fluctuation amplitude curves for all wavelength windows before the j-th wavelength window. The subscripts j, j-1, j-2, K, K1, K2, KM, etc., represent the area under the absorbance fluctuation curve. The corresponding wavelength window number.

[0012] The absorbance fluctuation amplitude can be considered as the peak-to-peak value of the absorbance curve (the curve showing the change of absorbance over time at a corresponding wavelength). The absorbance fluctuation amplitude is usually different at different wavelengths. Therefore, an absorbance fluctuation amplitude curve (the curve showing the change of absorbance fluctuation amplitude over wavelength) can be fitted based on the absorbance fluctuation amplitude at different wavelengths. During detection, several complete absorbance waveforms can be obtained at any wavelength, and the average value of the peak-to-peak values ​​of each waveform can be used as the absorbance fluctuation amplitude at that wavelength.

[0013] Preferably, the absorbance fluctuation amplitude curve is normalized before extracting absorbance-related features, and absorbance-related feature parameters are extracted based on the normalized absorbance fluctuation amplitude curve (the normalized absorbance fluctuation amplitude curve). For example, when the absorbance-related feature parameters include or involve the area under the absorbance fluctuation amplitude curve within the wavelength window, the area under the curve is the area under the normalized absorbance fluctuation amplitude curve within the corresponding wavelength window.

[0014] Preferably, the absorbance fluctuation amplitude curve is normalized by setting a reference wavelength. (The reference wavelength should generally be within the wavelength range used for photoplethysmography (PPG) detection. For example, one of the multiple wavelengths used in the detection can be selected as the reference wavelength.) The following formula is used to normalize the absorbance fluctuation amplitude or absorbance fluctuation amplitude curve at each wavelength:

[0015] ,

[0016] in wavelength The normalized absorbance fluctuation range at the reference wavelength is 1. wavelength The absorbance fluctuation range under the condition, Reference wavelength The fluctuation range of absorbance.

[0017] Before fitting the absorbance fluctuation amplitude curve, the absorbance fluctuation amplitude can be normalized as described above, and the normalized absorbance fluctuation amplitude can be used to fit the normalized absorbance fluctuation amplitude curve. Alternatively, the absorbance fluctuation amplitude can be used to fit the absorbance fluctuation amplitude curve first, and then the above normalization operation can be performed on the absorbance fluctuation amplitude curve to form the normalized absorbance fluctuation amplitude curve.

[0018] Typically, the photoplethysmography (PPG) data under multiple wavelengths originates from PPG detection performed using at least three different wavelengths of light, such as 6, 8, 10, or 12.

[0019] Preferably, light of some or all of the following wavelengths is used to detect photoplethysmography data at multiple wavelengths: 560nm, 570nm, 660nm, 920nm, 600nm, 620nm, 725nm, 950nm, 805nm, 940nm, 970nm and greater than 1100nm.

[0020] Preferably, a pulse wave (photoplethysmography pulse wave) is obtained based on photoplethysmography pulse wave detection data. Amplitude and heart rate normalization operations are then performed on the pulse wave to obtain a normalized pulse wave (normalized pulse wave). Based on the normalized pulse wave, some or all of the following feature parameters are extracted as pulse wave-related feature parameters used in the hemoglobin concentration calculation model:

[0021] Maximum slope of the ascending branch: ;

[0022] Rise time: For example, the time it takes to rise from 10% peak (10% of the peak value) to 90% peak (90% of the peak value);

[0023] Fall time: For example, the time it takes for the peak to fall to 50% of the peak (50% of the peak value);

[0024] Diphthousand wave slope: The slope of the line connecting the peak value and the trough value of the diphthous wave;

[0025] Dip-amplitude ratio: ;

[0026] Rising and falling ratio: ;

[0027] Low-to-high frequency energy ratio: ;

[0028] Harmonic component ratio: For example, the ratio of the third harmonic (amplitude) to the fundamental harmonic (amplitude).

[0029] in, The maximum slope of the ascending branch; For normalized pulse wave (normalized pulse wave function); The amplitude ratio of diabetic stroke; Main amplitude value (pulse amplitude value); The amplitude of the first diabetic wave after the main wave; When the support is raised or lowered; Ascending time; For the time of descent; The energy ratio is low to high frequency. Low-frequency component energy (area under the normalized power spectral density curve within a defined low-frequency range). High-frequency component energy (area under the normalized power spectral density curve within a defined high-frequency range).

[0030] Preferably, in cases involving (for example, the existence or computation of) normalized pulse waves at several wavelengths, the normalized pulse wave at a set wavelength is used as the normalized pulse wave for extracting pulse wave related feature parameters, or the average value of the normalized pulse waves at each related wavelength is used as the normalized pulse wave for extracting pulse wave related feature parameters.

[0031] Preferably, the hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters is constructed in the following manner: Absorbance-related characteristic parameters and pulse wave-related characteristic parameters are extracted from photoplethysmography (PPG) data at multiple wavelengths (using light of various wavelengths); based on the extracted absorbance-related characteristic parameters and pulse wave-related characteristic parameters, and the corresponding hemoglobin concentration detection data, a one-step or two-step method is used to construct the hemoglobin concentration calculation model based on the absorbance-related characteristic parameters and pulse wave-related characteristic parameters.

[0032] 1) One-step method: Using absorbance-related characteristic parameters and pulse wave-related characteristic parameters as model input data, and corresponding hemoglobin concentration detection data as model output data, one-step sample data is constructed. A hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters is then built using this one-step sample data. This hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters (the one-step hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters) can take the following forms:

[0033] ,

[0034] in, The table shows hemoglobin concentration. This indicates the various absorbance-related characteristic parameters used; This indicates the relevant characteristic parameters of each pulse wave used;

[0035] 2) Two-step method:

[0036] Step 1: Using absorbance-related characteristic parameters as model input data and corresponding hemoglobin concentration detection data as model output data, construct sample data for the first hemoglobin concentration calculation model (hemoglobin concentration calculation model based on absorbance-related characteristic parameters), and use the sample data of the first hemoglobin concentration calculation model to construct the first hemoglobin concentration calculation model; using pulse wave-related characteristic parameters as model input data and corresponding hemoglobin concentration detection data as model output data, construct sample data for the second hemoglobin concentration calculation model (hemoglobin concentration calculation model based on pulse wave-related characteristic parameters), and use the sample data of the second hemoglobin concentration calculation model to construct the second hemoglobin concentration calculation model;

[0037] The primary calculation model for hemoglobin concentration can be adopted in the following form:

[0038] ,

[0039] in, For the first calculation model, hemoglobin concentration, Indicates the characteristic parameters related to each absorbance;

[0040] The second calculation model for hemoglobin concentration can take the following form:

[0041] ,

[0042] in, For the second calculation model, hemoglobin concentration, Represents the characteristic parameters related to each pulse wave;

[0043] Step 2: Based on the first calculation model of hemoglobin concentration, the calculated value of hemoglobin concentration of the first calculation model is obtained by using absorbance-related characteristic parameters. Based on the second calculation model of hemoglobin concentration, the calculated value of hemoglobin concentration of the second calculation model is obtained by using pulse wave-related characteristic parameters. The calculated values ​​of hemoglobin concentration of the first and second calculation models are used as input data for the model, and the corresponding hemoglobin concentration detection data is used as output data for the model. This constitutes the sample data of the hemoglobin concentration calculation model based on the first and second calculation models of hemoglobin concentration. Using the sample data of the hemoglobin concentration calculation model based on the first and second calculation models of hemoglobin concentration, a hemoglobin concentration calculation model based on the first and second calculation models of hemoglobin concentration is constructed. This model is used as the hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters.

[0044] The hemoglobin concentration calculation model based on the first and second calculation models (a two-step hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters) can take the following form:

[0045] ,

[0046] or,

[0047] ,

[0048] in, The table shows hemoglobin concentration. The first calculation model shows the hemoglobin concentration (calculated value). The second calculation model shows the hemoglobin concentration (calculated value). and The coefficients of the first and second calculation models, respectively. It is a constant.

[0049] A continuous non-invasive physiological parameter monitoring device includes:

[0050] The probe is a multi-wavelength photoplethysmography probe, used to perform multi-wavelength photoplethysmography detection and generate optical signals for multi-wavelength photoplethysmography detection.

[0051] The host computer includes a signal sampling circuit, a data processing circuit (microprocessor), a human-computer interaction device, and a storage device. The signal sampling circuit is used to acquire optical signal data for multi-wavelength photoplethysmography (PPG) detection, forming multi-wavelength PPG detection data. The data processing circuit performs data processing using any of the continuous non-invasive physiological parameter detection methods disclosed in this invention to obtain hemoglobin concentration detection results. The human-computer interaction device is used for human-computer interaction and includes a display screen for displaying hemoglobin detection results (and usually other data). The storage device is used for data storage and model (hemoglobin concentration calculation model) storage.

[0052] The probe can employ any suitable existing technology. For example, it can be a transmissive probe, a reflective probe, or a combination of transmissive and reflective probes; it can be a probe for detecting finger areas or a probe for detecting arm areas.

[0053] The beneficial effects of this invention are as follows: Based on the influence mechanism of hemoglobin concentration on the fluctuation amplitude of absorbance at different wavelengths and on the morphology of pulse waves, a hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters is constructed. This incorporates more factors related to hemoglobin concentration into the calculation of hemoglobin density, thereby not only improving the accuracy of the calculation results but also helping to avoid or reduce deviations caused by individual factors. Since the extraction of each characteristic parameter is based on normalized correlation curves, it helps to make the comparability of related parameters between individuals, further improving the accuracy of the model and detection results.

[0054] This invention is used for processing photoplethysmography (PPG) data and, based on a corresponding computational model, calculates and obtains hemoglobin concentration detection results. Attached Figure Description

[0055] Figure 1 This invention relates to an example of a photoplethysmography (electrical signal);

[0056] Figure 2 This invention relates to examples of absorbance coefficient (as a function of wavelength) curves for different blood components;

[0057] Figure 3 This invention relates to an example of the normalized absorbance fluctuation amplitude (as a function of wavelength) curves under different hemoglobin concentrations;

[0058] Figure 4 This invention relates to examples of normalized (heart rate normalized and amplitude normalized) pulse waves at different hemoglobin concentrations;

[0059] Figure 5 This invention relates to examples of normalized power spectral density curves at different hemoglobin concentrations.

[0060] Figure 6 This is a schematic diagram of the detection device of the present invention that measures via the finger area;

[0061] Figure 7 This is a schematic diagram of the structure of the detection device (reflective type) that measures through the wrist area according to the present invention;

[0062] Figure 8 A schematic diagram of a detection device architecture capable of implementing the detection method of the present invention is shown.

[0063] The markings in the diagram are: 1. Main unit; 2. Finger tip photoelectric probe; 3. Optical cable; 4. Wristband; 5. Reflective probe. Detailed Implementation

[0064] This invention is based on photoplethysmography (PPG) pulse wave detection at multiple wavelengths. According to the different absorbance variations (fluctuation amplitude, which is the difference between the maximum and minimum values, or peak-to-peak value) and different pulse wave morphologies of different hemoglobin concentrations at different wavelengths (detection light wavelengths), based on the corresponding normalization curves, absorbance-related characteristic parameters and pulse wave-related characteristic parameters that are sensitive to hemoglobin concentration are extracted, and a regression model is established as the calculation model for hemoglobin, thereby improving the accuracy of non-invasive measurement results.

[0065] I. Detection Device

[0066] See Figure 8 and Figure 6 and Figure 7 This invention can utilize existing photoplethysmography (PPG) detection devices, employing multi-wavelength probes, and with software support, to implement continuous non-invasive physiological parameter detection. The detection device mainly includes three modules: a probe (sensor) acquisition module, a signal processing module, and a human-computer interaction module.

[0067] The probe acquisition module (or probe), taking transmission detection as an example, includes a multi-wavelength light emitter and a photodetector. The multi-wavelength light emitter is made up of multiple wavelengths of emitting lamp cores integrated together to create an emitting lamp sensor that is close to a point light source. The photodetector is a semiconductor device that converts light signals into electrical signals and can convert the light energy passing through the finger into an electric current signal.

[0068] The signal processing module is mainly responsible for acquiring clinically valuable electrical signals. The acquisition system is required to have high precision, high stability, high input impedance, high common-mode rejection ratio, and low noise.

[0069] The LED driver circuit enables control over the timing and intensity of light emission. It can adjust the current gain based on the different finger sizes and structures of individuals, thereby adjusting the light intensity. The light intensity adjustment is performed at the beginning of the measurement process, and the measurement is conducted after the adjustment is completed. The incident light intensity remains constant during the measurement process.

[0070] The microcontroller controls the front-end signal controller in the signal processing module to perform gain control, and controls the LED driver, thereby controlling the light emitter to realize functions such as probe detection, ADC signal acquisition, physiological signal processing, storage and transmission.

[0071] Physiological signal processing mainly involves extracting the transmitted light intensity and waveform characteristics corresponding to different wavelengths, thereby establishing mathematical models to measure the concentration of different blood components.

[0072] Storage devices store electrical signals of different wavelengths. The stored data can be read and identified by a PC, and playback functionality can be configured based on existing technology.

[0073] The human-computer interaction module can be equipped with a touch LCD display, etc. It is low in power consumption, small in size, and can input and display patient information and corresponding physiological parameter measurement values.

[0074] II. Algorithm for Non-invasive Continuous Hemoglobin Concentration Measurement

[0075] The photoplethysmography can be detected using existing detection methods, such as transmission, emission, or a combination of transmission and reflection. The probe (light emitter) used is a multi-wavelength light emitter capable of emitting light of several different wavelengths (frequency) determined based on the detection method of this invention. Correspondingly, the photodetector can effectively detect or collect these different wavelengths of light.

[0076] See Figure 1 Based on Beer-Lambert law, the wavelength is The change in absorbance (the amplitude of absorbance fluctuation) is denoted as . This is the difference between the maximum absorbance and the minimum absorbance.

[0077] (1)

[0078] in, The incident light intensity is [value missing]. When the blood vessels dilate to their maximum, the optical path length increases to its maximum value, and the emitted light intensity is at its minimum. When the blood vessels constrict to their minimum, the optical path length decreases to its minimum value, and the emitted light intensity is at its maximum. , This represents the peak-to-peak value of the pulse wave signal. It is a DC signal.

[0079] 1. Extract absorbance-related feature parameters based on the detection data.

[0080] Based on the fact that different components in blood absorb light of the same wavelength to varying degrees, and that specific wavelengths of light cause significant differences in the absorption of corresponding light energy by different blood components, wavelengths with large differences in the absorption coefficients of different blood components, as well as some special wavelengths with isoabsorption points, can be selected. After passing through tissue, light of different wavelengths corresponding to different emitted intensities is received at the receiving end. The absorbance change at different wavelengths can be calculated (characterized by absorbance fluctuation amplitude). Based on this, an absorbance fluctuation amplitude curve (or function) as a function of wavelength is fitted. After normalization, a normalized absorbance fluctuation amplitude curve is formed. Based on this curve, the area under the normalized absorbance fluctuation amplitude curve at different wavelengths (the area under the curve) is calculated within a certain wavelength window. This area under the curve changes with the wavelength (wavelength window position) and is compared to a single ratio of existing absorbance changes at different wavelengths. (in, The results are more stable when using two different wavelengths of light, and absorbance-related characteristic parameters related to hemoglobin concentration can be extracted based on this area change.

[0081] The specific method is as follows:

[0082] 1) Select multiple wavelengths to perform multi-wavelength photoplethysmography (PPG) detection, and calculate the absorbance fluctuation amplitude (the difference between the absorbance peak and valley values) at each wavelength.

[0083] Based on the significant differences in absorbance coefficients among different blood components and specific wavelengths with isoabsorption points, a limited number of wavelengths can be selected to detect multiple wavelengths. For example:

[0084] The wavelengths of green light absorbed by oxyhemoglobin and deoxyhemoglobin in the human body are the largest, such as 560nm and 570nm.

[0085] The wavelengths at which the absorption of oxygenated hemoglobin and deoxyhemoglobin differ most in the human body are, for example, 660nm and 920nm.

[0086] The wavelengths used to distinguish between oxyhemoglobin and methemoglobin in the human body, for example: 600nm, 620nm;

[0087] Different components in human blood have different absorption rates, such as 725nm and 950nm.

[0088] The absorption points of oxyhemoglobin and deoxyhemoglobin in human blood, for example: 805nm;

[0089] The absorption points of oxyhemoglobin and carboxyhemoglobin in human blood, for example: 640nm;

[0090] The wavelengths used to distinguish the concentration of water and hemoglobin in human blood, for example: 970nm, wavelengths greater than 1100nm.

[0091] All of the wavelengths mentioned above can be selected. When appropriate, to reduce computational load, only a portion of the wavelengths mentioned above can be selected (e.g., multiple wavelengths with similar spacing). Furthermore, other wavelengths can be added to all or part of the wavelengths mentioned above, or other suitable wavelengths besides those mentioned can be used.

[0092] 2) Normalize the absorbance fluctuation amplitude based on the absorbance at the reference wavelength.

[0093] For example, the wavelength of the isoabsorption point of oxygenated and reduced hemoglobin concentrations can be selected as the reference wavelength. Based on reference wavelength Absorbance fluctuation range The absorbance fluctuation range calculated at each wavelength Normalization is performed to obtain the wavelength. Normalized absorbance fluctuation range :

[0094] .

[0095] The reference wavelength can be flexibly determined based on actual conditions. For example, a wavelength that is sensitive to water absorption can be selected as the reference wavelength, or the wavelength at which the absorption coefficient of oxyhemoglobin is at its maximum can be selected as the reference wavelength.

[0096] 3) Fit the curve to obtain the normalized absorbance fluctuation range.

[0097] The normalized absorbance fluctuation amplitude curve can be obtained by using traditional interpolation signal processing methods (interpolation method) or other suitable methods.

[0098] 4) Extract (obtain by calculation) absorbance-related characteristic parameters from the normalized absorbance fluctuation amplitude curve.

[0099] Taking transmission as an example, different hemoglobin concentrations have different absorption levels to the same wavelength of light source. Therefore, the absorbance fluctuation amplitude curves corresponding to different wavelengths are different. By extracting the characteristic parameters of the normalized absorbance fluctuation amplitude curve, the measurement of different hemoglobin concentrations can be achieved.

[0100] ① The area under the curve of absorbance fluctuation amplitude within a certain wavelength range (wavelength window):

[0101] The area under the line can usually be calculated using integration. Considering computational convenience, especially when using interpolation to generate discrete data, and when sampling absorbance fluctuations at fixed wavelength intervals or when the sampling points are roughly evenly distributed, it is possible to calculate the area of ​​each fluctuation within the wavelength window. The discrete data is summed instead of integrated; that is, the area under the curve of the absorbance fluctuation amplitude of the j-th fixed wavelength window is calculated using the following formula. :

[0102] , or can be expressed as: .

[0103] For example: the area corresponding to the normalized absorbance fluctuation amplitude curve within the wavelength window [600nm, 650nm], where n is the starting wavelength of the fixed wavelength window j (600nm) and N is the wavelength window length (Window Size, 50nm).

[0104] Since integrals over continuous functions and sums over discrete data are equivalent in the parameter calculations or calculation results involved in this invention, the above summation formula can also be understood as (equivalently replaced by) the corresponding formula for the area integral under the line.

[0105] ② Other absorbance-related characteristic parameters:

[0106] Extract (obtained through calculation) feature parameters that reflect the changes in the area below the line within the window, including:

[0107] Feature parameters (adjacent difference): ;

[0108] Feature parameters (Relatively poor): ;

[0109] Feature parameters (Ratio of adjacent faces): ;

[0110] Feature parameters (Comparative comparison):

[0111] ;

[0112] Feature parameters (Progressive face ratio):

[0113] .

[0114] Depending on the actual situation, wavelength window K for calculating the relative difference and wavelength windows K1, K2, ..., KM for calculating the relative ratio can be selected, where... This refers to the total number of wavelength windows selected for calculating the relative difference. Typical or representative wavelength windows can be selected as the wavelength windows used for calculating the relative difference, such as wavelength windows located in the middle, wavelength windows located at one end, or other wavelength windows that are typical based on experience or statistical results. A portion (one or more) of the wavelength windows can be selected as the wavelength windows used for calculating the relative difference, such as three wavelength windows located at both ends and in the middle, or other equally spaced wavelength windows, or one or more wavelength windows that are typical based on experience or statistical results.

[0115] Depending on the actual situation, other absorbance-related characteristic parameters can also be set.

[0116] Based on the above absorbance-related characteristic parameters, it can be expressed in functional form. This indicates the concentration of any type of hemoglobin or total hemoglobin.

[0117] The aforementioned characteristic parameters and related calculations (data processing) can also be obtained using reflectance detection data. By calculating the normalized absorbance fluctuation amplitude curve as a function of wavelength, the aforementioned characteristic parameters can be obtained. The wavelength / wavelength range used for detection can be selected according to actual use, for example, selecting a certain portion of the characteristic wavelengths within the band [540nm, 810nm].

[0118] The aforementioned characteristic parameters and corresponding calculation (data processing) methods can also be obtained by combining detection data and calculating the normalized absorbance fluctuation amplitude curve as a function of wavelength, thereby obtaining the aforementioned characteristic parameters. The wavelength range used for detection can be selected according to actual use, for example, selecting several characteristic wavelengths within the band [540nm, 1400nm].

[0119] 2. Extract pulse wave related feature parameters based on the detection data.

[0120] Under normal conditions of other blood components (such as plasma volume, plasma proteins, and platelets), an increase in hemoglobin concentration usually indicates an increase in the number of red blood cells, leading to an increase in hematocrit (HCT) and consequently, increased blood viscosity. Blood viscosity affects hemodynamics and vascular elasticity, and the pulse wave is the result of the combined effects of cardiac contraction and vascular elasticity. Therefore, increased viscosity leads to increased blood flow resistance, thereby altering the morphology of the pulse wave. Thus, pulse wave morphology can be used as a basis for calculating blood hemoglobin concentration.

[0121] The pulse wave obtained from the detection data undergoes pulse rate and amplitude normalization operations to form a normalized pulse wave with normalized frequency and amplitude. Heart rate normalization can be achieved by normalizing the acquired (measured) pulse wave rate to a standard frequency (standard heart rate), for example, 75 bpm, to reduce the influence of different heart rates on model construction and blood pressure measurement results. Amplitude normalization can be achieved by normalizing the acquired (measured) pulse wave amplitude to a standard amplitude (fixed value), obtaining a normalization coefficient k, and then using this normalization coefficient to perform a normalization operation on the entire pulse wave (multiplying by k) to achieve amplitude normalization of the pulse wave, thus facilitating the comparability of pulse wave morphology between different individuals / samples.

[0122] The pulse wave used to extract pulse wave related feature parameters can be a normalized pulse wave at a set wavelength (which can be set according to existing technology), or it can be the cumulative average of normalized pulse waves at multiple wavelengths.

[0123] Based on the normalized pulse wave, the following feature parameters are extracted (obtained through computation) as pulse wave related feature parameters:

[0124] 1) Time-domain characteristic parameters

[0125] Considering that changes in blood viscosity can lead to changes in the shape of the ascending limb (pulse wave ascending limb) and descending limb (pulse wave descending limb), for example, increased viscosity leads to increased peripheral resistance, resulting in changes in the maximum slope and rise time (10%-90% peak time) of the ascending limb. Increased viscosity also leads to increased resistance to diastolic blood flow return, restricts vascular elastic recoil, affects the attenuation of the descending limb, and may cause the dicrotic wave to become less prominent or disappear. The following time-domain feature parameters are extracted:

[0126] Feature parameters (Maximum slope of the ascending branch): , This is the pulse wave;

[0127] Feature parameters (Rise time): The time it takes for the pulse wave value to rise from 10% peak value (10% of the peak value) to 90% peak value (90% of the peak value);

[0128] Feature parameters (Fall time): The time it takes for the pulse wave value to drop from its peak value to 50% of its peak value;

[0129] Feature parameters (Diphthousand-wave slope): The slope of the line connecting the peak value and the trough value of the diphthous wave;

[0130] Feature parameters (Occurrence amplitude ratio): ,in It is the amplitude of the main wave (pulse wave amplitude). It is the amplitude of the first diphtheria wave after the main wave;

[0131] Feature parameters (Ratio of rising to falling time: the ratio of rising time to falling time) ,in It is the rise time (the time it takes for the rising phase to go from the trough to the peak). It is the descent time (the time it takes for the descent to go from the peak to the trough).

[0132] 2) Frequency domain characteristic parameters

[0133] Since changes in blood viscosity affect the power spectral density in the pulse wave frequency domain, the normalized power spectral density curve (distribution function) can be calculated based on the normalized pulse wave. Based on the normalized power spectral density curve, the following frequency domain feature parameters are extracted:

[0134] Feature parameters (Low-to-high frequency energy ratio, the ratio of the energy of the low-frequency component to the energy of the high-frequency component): For example, low-frequency components (frequency range) can be defined as 0Hz ≤ f ≤ 2Hz, and high-frequency components (frequency range) can be defined as 5Hz ≤ f ≤ 10Hz. In this case, , ;

[0135] Feature parameters (Harmonic component ratio): The ratio that reflects the passivation of the waveform, specifically the third harmonic (amplitude) to the fundamental harmonic (amplitude).

[0136] Depending on the actual situation, other pulse wave related characteristic parameters can also be set.

[0137] Therefore, it can be expressed in functional form. This indicates the concentration of any type of hemoglobin or total hemoglobin.

[0138] 3. Establishment of a model for calculating hemoglobin concentration

[0139] Since pulse waves do not change drastically in a short period of time under normal circumstances, continuous photoplethysmography (PPG) detection can be performed within a fixed time window. Multiple complete measured pulse waves are obtained within the window. The frequency and amplitude of these measured pulse waves are normalized to form a normalized measured pulse wave. The normalized measured pulse waves within the time window are accumulated and averaged to obtain a single normalized pulse wave, which is used as the normalized pulse wave for this detection (the detection result of the normalized pulse wave) for subsequent calculations. This includes constructing the absorbance fluctuation amplitude curve (function) / normalized absorbance fluctuation amplitude curve, and using the characteristic parameters obtained from the calculation as the characteristic parameters of this detection, which are included in a sample data.

[0140] The cumulative averaging method can be to accumulate and superimpose all complete pulse waves (normalized and centered on the peak point of the main wave) within the corresponding time window, and divide by the number of accumulated pulse waves.

[0141] In continuous detection data processing, a fixed time length (e.g., 15s) can be set as a time window, and the time window can be slid with a set sliding step size (e.g., 1.2s). Each time window is used as one detection, and the feature parameters and other detection results of that detection are calculated.

[0142] When collecting sample data for constructing the hemoglobin concentration calculation model, subjects should be selected to provide sample data according to relevant standards. During photoplethysmography (PPG) testing, a standard or sufficiently accurate hemoglobin concentration detection method / instrument should be used to detect hemoglobin concentration at the same time interval. Feature parameters for model construction are selected, and relevant feature parameters for each test are obtained through PPG data calculation. These are used as the model's input data (or independent variables). The hemoglobin concentration results obtained at the same time interval using a standard or sufficiently accurate hemoglobin concentration detection method / instrument are used as the model's input data (or dependent variables). A set of corresponding model input data (at the same time / within the same time window) and the model input are used as a sample data set. With several sample data sets that meet the specifications or requirements, the hemoglobin concentration calculation model is constructed.

[0143] There are two methods for calculating hemoglobin concentration: the one-step method and the two-step method.

[0144] 1. Two-step method

[0145] 1) Based on the correlation between absorbance-related characteristic parameters and pulse wave-related characteristic parameters and hemoglobin concentration, respectively, a hemoglobin calculation model based on absorbance-related characteristic parameters and a hemoglobin calculation model based on pulse wave-related characteristic parameters are constructed.

[0146] The hemoglobin calculation model based on absorbance-related characteristic parameters can be expressed as:

[0147]

[0148] The model input data used for model construction consists of absorbance-related feature parameters and hemoglobin concentration (hemoglobin concentration obtained using standard or sufficiently high-precision hemoglobin concentration detection methods / instruments). This represents the hemoglobin concentration of the model (the hemoglobin concentration predicted based on absorbance-related characteristic parameters).

[0149] The hemoglobin calculation model based on pulse wave related characteristic parameters can be expressed as:

[0150]

[0151] The model input data used for model construction consists of pulse wave related feature parameters and hemoglobin concentration (hemoglobin concentration obtained using standard or sufficiently high-precision hemoglobin concentration detection methods / instruments). This represents the hemoglobin concentration of the model (the hemoglobin concentration predicted based on pulse wave related characteristic parameters).

[0152] The two computational models mentioned above can be established using their respective sample data and employing artificial neural networks, nonlinear regression, partial least squares regression, multi-level model architecture, signal analysis, or statistical methods.

[0153] 2) Calculation results based on the above two models and A hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters was constructed:

[0154]

[0155] That is,

[0156]

[0157] in, Hemoglobin concentration, and The coefficients and constants of the two types of feature parameter regression models (corresponding hemoglobin concentration calculation results) are respectively.

[0158] The model input data in the sample data used to build this model is and The model input data is hemoglobin concentration (hemoglobin concentration obtained using standard or sufficiently high-precision hemoglobin concentration detection methods / instruments), where... This represents the hemoglobin concentration of the model (the hemoglobin concentration predicted based on absorbance-related features and pulse wave-related features).

[0159] The optimal regression coefficients can be found using the least squares method or other suitable methods. This allows for the establishment of a hemoglobin calculation model, enabling non-invasive hemoglobin measurement.

[0160] 2. One-step method

[0161] Based on the correlation between absorbance-related characteristic parameters and pulse wave-related characteristic parameters and hemoglobin concentration, a hemoglobin calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters is constructed.

[0162] The hemoglobin calculation model based on absorbance-related characteristic parameters can be expressed as:

[0163]

[0164] The model input data used for model construction consists of absorbance-related feature parameters and pulse wave-related feature parameters. The model input data is hemoglobin concentration (hemoglobin concentration obtained using a standard or sufficiently high-precision hemoglobin concentration detection method / instrument). This indicates the concentration of hemoglobin.

[0165] The computational model can be established using artificial neural networks, nonlinear regression, partial least squares regression, multilevel model architecture, signal analysis, or statistical methods.

[0166] The construction and updating of relevant calculation models can be implemented using corresponding computer platforms. The terminal (detection device) used for detection can send the detection data to the platform, and the platform will perform relevant calculations. The detection device will then obtain the calculation results of hemoglobin concentration obtained by the platform. Alternatively, the calculation model can be stored locally on the detection device, and the detection device can perform calculations on hemoglobin concentration to generate the calculation results of hemoglobin concentration.

[0167] Figure 6This invention illustrates a detection device for measuring hemoglobin via a fingertip (performing photoplethysmography). This continuous non-invasive hemoglobin measurement device is worn on the fingertip to perform measurements. The external host unit (including a microprocessor, memory, and interface / data acquisition circuitry, etc.) 1 performs control, data storage, signal processing, and display functions. The fingertip photoelectric probe (photoelectric PPG probe) collects multi-wavelength photoelectric pulse waves and can be a transmissive, reflective, or combined probe. The optical cable 3 transmits the optical signals obtained by the probe to the interface circuitry (data acquisition circuitry) of the main unit.

[0168] Figure 7 This invention illustrates a detection device that measures hemoglobin via the wrist. This continuous non-invasive hemoglobin measuring device is worn on the wrist to perform measurements. The external host 1 (including a microprocessor, memory, and interface circuits / data acquisition circuits, etc.) can perform functions such as control, data storage, signal processing, and display. The reflective probe 5 is an integrated multi-wavelength light source and photoelectric sensor, embedded in the wristband 4. When the wristband is worn on the wrist, it can collect photoelectric pulse waves of multiple wavelengths to measure parameters such as hemoglobin.

[0169] Figure 8 An optical circuit structure for a detection device capable of implementing the present invention is shown, which can be performed using suitable prior art with the support of software for implementing the detection method of the present invention.

[0170] Unless otherwise specified, the preferred and optional technical means disclosed in this invention can be arbitrarily combined to form several different specific embodiments when one preferred or optional technical means is a further limitation of another technical means.

Claims

1. A method for continuous non-invasive physiological parameter detection, characterized in that... Based on photoplethysmography pulse wave detection data under multiple wavelengths, absorbance-related feature parameters and pulse wave-related feature parameters are extracted. A hemoglobin concentration calculation model based on absorbance-related feature parameters and pulse wave-related feature parameters is used to calculate the hemoglobin concentration detection results. Based on photoplethysmography (PPG) data, the absorbance fluctuation amplitude at each wavelength is calculated. An absorbance fluctuation amplitude curve is then fitted based on this amplitude. Several wavelength windows of predetermined widths are set. Based on the area under the absorbance fluctuation amplitude curve within each wavelength window, some or all of the following feature parameters are extracted as absorbance-related feature parameters used in the hemoglobin concentration calculation model: Neighborhood differences: ; Relative difference: ; Adjacent face ratio: ; In comparison: ; Progressive ratio: , in, For adjacent face differences, For relative difference, For adjacent face ratio, For comparison, For progressive surface ratio, For the first j The area under the curve of absorbance fluctuation within a wavelength window. For the first j- The area under the curve of absorbance fluctuation within one wavelength window To select the area under the absorbance fluctuation amplitude curve within the wavelength window used to calculate the relative difference, To select the one used for calculating the relative ratio The sum of the areas under the absorbance fluctuation curve within each wavelength window. , For the first j The sum of the areas under the absorbance fluctuation amplitude curves within all wavelength windows before the current wavelength window; The pulse wave is obtained by processing the photoplethysmography pulse wave detection data. The amplitude and heart rate are normalized to obtain the normalized pulse wave. Based on the normalized pulse wave, some or all of the following feature parameters are extracted as pulse wave related feature parameters used in the hemoglobin concentration calculation model: maximum slope of the ascending limb, rise time, fall time, dicrotic wave slope, dicrotic amplitude ratio, ascending-descending limb time ratio, low-frequency energy ratio, and harmonic component ratio.

2. The continuous non-invasive physiological parameter detection method as described in claim 1, characterized in that... Before extracting absorbance-related features, the absorbance fluctuation amplitude curve is normalized, and absorbance-related feature parameters are extracted based on the normalized absorbance fluctuation amplitude curve.

3. The continuous non-invasive physiological parameter detection method as described in claim 2, characterized in that... The method for normalizing the absorbance fluctuation curve is as follows: set a reference wavelength. The absorbance fluctuation amplitude or absorbance fluctuation amplitude curve at each wavelength is normalized using the following formula: ， in wavelength The normalized absorbance fluctuation range is below. wavelength The absorbance fluctuation range under the condition, Reference wavelength The fluctuation range of absorbance.

4. The continuous non-invasive physiological parameter detection method as described in claim 1, characterized in that... Photoplethysmography (PPG) data were obtained using some or all of the following wavelengths: 560nm, 570nm, 660nm, 920nm, 600nm, 620nm, 725nm, 950nm, 805nm, 940nm, 970nm, and wavelengths greater than 1100nm.

5. The continuous non-invasive physiological parameter detection method as described in claim 4, characterized in that... In cases involving normalized pulse waves at several wavelengths, the normalized pulse wave at a set wavelength is used as the normalized pulse wave for extracting pulse wave-related feature parameters, or the average value of the normalized pulse waves at each relevant wavelength is used as the normalized pulse wave for extracting pulse wave-related feature parameters.

6. The method for continuous non-invasive physiological parameter detection as described in any one of claims 1-5, characterized in that... A hemoglobin concentration calculation model based on absorbance-related and pulse wave-related characteristic parameters is constructed using the following method: Absorbance-related and pulse wave-related characteristic parameters are extracted from multi-wavelength photoplethysmography (PPG) pulse wave detection data. Based on the extracted absorbance-related and pulse wave-related characteristic parameters and the corresponding hemoglobin concentration detection data, a one-step or two-step method is used to construct the hemoglobin concentration calculation model based on these parameters. One-step method: Using absorbance-related characteristic parameters and pulse wave-related characteristic parameters as model input data, and corresponding hemoglobin concentration detection data as model output data, one-step sample data is constructed, and a hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters is constructed using the one-step sample data; Two-step method: Step 1: Using absorbance-related characteristic parameters as model input data and corresponding hemoglobin concentration detection data as model output data, construct sample data for the first hemoglobin concentration calculation model, and then construct the first hemoglobin concentration calculation model using the sample data of the first hemoglobin concentration calculation model; using pulse wave-related characteristic parameters as model input data and corresponding hemoglobin concentration detection data as model output data, construct sample data for the second hemoglobin concentration calculation model, and then construct the second hemoglobin concentration calculation model using the sample data of the second hemoglobin concentration calculation model. Step 2: Based on the first calculation model of hemoglobin concentration, the calculated value of hemoglobin concentration of the first calculation model is obtained by using absorbance-related characteristic parameters. Based on the second calculation model of hemoglobin concentration, the calculated value of hemoglobin concentration of the second calculation model is obtained by using pulse wave-related characteristic parameters. The calculated values ​​of hemoglobin concentration of the first and second calculation models are used as input data for the model, and the corresponding hemoglobin concentration detection data is used as output data for the model. This constitutes the sample data of the hemoglobin concentration calculation model based on the first and second calculation models of hemoglobin concentration. Using the sample data of the hemoglobin concentration calculation model based on the first and second calculation models of hemoglobin concentration, a hemoglobin concentration calculation model based on the first and second calculation models of hemoglobin concentration is constructed. This model serves as the hemoglobin concentration calculation model based on absorbance-related characteristic parameters and pulse wave-related characteristic parameters.

7. A continuous non-invasive physiological parameter detection device, characterized in that... include: The probe is a multi-wavelength photoplethysmography probe, used to perform multi-wavelength photoplethysmography detection and generate optical signals for multi-wavelength photoplethysmography detection. The host computer includes a signal sampling circuit, a data processing circuit, a human-computer interaction device, and a storage device. The signal sampling circuit is used to acquire optical signal data for multi-wavelength photoplethysmography (PPG) detection, forming multi-wavelength PPG detection data. The data processing circuit performs data processing using the continuous non-invasive physiological parameter detection method described in any one of claims 1-6 to obtain hemoglobin concentration detection results. The human-computer interaction device is used for human-computer interaction and includes a display screen for displaying hemoglobin detection results. The storage device is used for data storage and model storage.

8. The continuous non-invasive physiological parameter detection device as described in claim 7, characterized in that... The probe is a transmission probe, a reflection probe, or a combination of transmission and reflection probe; the probe is used for detection of finger parts or for detection of arm parts.

Images