Method for measuring vascular elasticity modulus and blood pressure based on LC-OCT hyperelastic modeling

Abstract

The invention discloses a method for measuring vascular elasticity modulus and blood pressure based on LC-OCT hyperelastic modeling, which comprises the following steps: accurately capturing dynamic deformation of blood vessels through high-resolution three-dimensional imaging and Doppler blood flow detection of LC-OCT, calculating the elasticity modulus (error is less than or equal to + / -5%) by combining a Fung model, and constructing a mixed blood pressure prediction model based on a theory-data dual-drive framework. High-precision inversion (MAE < = 5.1 mmHg) of systolic pressure, diastolic pressure and average arterial pressure is achieved, and dynamic transfer learning is adopted to adapt to the blood vessel characteristic differences of different crowds. The method has the advantages that blood vessel full-dimension characterization is achieved through multi-mode LC-OCT, dynamic transfer learning adapts to the difference of healthy / sick people, a theory-data dual-drive model gives consideration to physical interpretability and prediction precision, and a high-precision solution is provided for early diagnosis and personalized treatment of hypertension.

Description

Technical Field

[0001] The present invention belongs to the field of medical image processing, and relates to a method for processing OCT images, specifically to a method for measuring vascular elastic modulus and blood pressure based on LC-OCT hyperelastic modeling. Background Art

[0002] Traditional blood pressure measurement techniques mainly rely on cuff pressure methods and invasive arterial catheter methods. The cuff method blocks blood flow by inflating and compressing the artery, and calculates blood pressure values using Korotkoff sounds or pressure oscillation waves. However, it has significant defects: it can only obtain blood pressure values at specific time points and cannot reflect the dynamic changes within the cardiac cycle in real time; it has no detection ability for the hemodynamic characteristics of microvessels and is difficult to evaluate early microcirculation disorders caused by hypertension; and improper cuff placement, limb movement, or excessive subcutaneous fat in obese patients may lead to reading errors (the error can reach ±10 mmHg). Although the invasive catheter method can provide high-precision continuous blood pressure monitoring, it requires arterial puncture, which has risks such as infection and thrombosis, and is also complex to operate and costly, and is only used in intensive care or during surgery.

[0003] In recent years, optical imaging technologies (such as optical coherence tomography OCT and photoacoustic imaging PAI) have made progress in the field of vascular visualization, but their application in blood pressure assessment is still limited. Although traditional OCT can achieve micron-level resolution (axial 5-15 μm), it lacks the ability to synchronously quantify vascular mechanical properties (such as elastic modulus) and hemodynamics; most optical technologies have a low frame rate (<10 fps), making it difficult to capture the instantaneous deformation of the blood vessel wall with the heartbeat (such as microsecond-level strain fluctuations); and the imaging depth of skin OCT is only 1-2 mm, making it impossible to effectively observe deep blood vessels (such as carotid arteries and femoral arteries). Although photoacoustic imaging (PAI) can provide functional information by combining blood flow absorption characteristics, its spatial resolution (50-100 μm) is much lower than that of OCT, making it difficult to meet the research needs of microvessels.

[0004] Line-field confocal optical coherence tomography (LC-OCT), as an emerging technology, combines the high lateral resolution (1-3 μm) of a confocal microscope and the depth resolution ability of OCT (axial 5-10 μm), supports real-time three-dimensional imaging (above 30 fps), and has shown advantages in the diagnosis of skin lesions. However, its existing application scenarios still have limitations: LC-OCT mainly focuses on structural imaging (such as the epidermal layer and dermal papillary layer), and has not fully developed its potential in measuring vascular mechanical parameters (elastic modulus) and hemodynamics (Doppler flow velocity); the current LC-OCT system lacks the ability to synchronously collect and fuse and analyze structural, mechanical, and blood flow parameters, and cannot construct an "anatomy-function" correlation model; although laboratory studies have confirmed the high-resolution advantages of LC-OCT, its application in cardiovascular diseases (such as hypertension and arteriosclerosis) is still in the exploratory stage.

[0005] In the field of vascular biomechanics, classical linear models (such as Hooke's law) are difficult to accurately describe the true mechanical response of blood vessels because they neglect the nonlinear, anisotropic, and viscoelastic properties of biological tissues. Blood vessels exhibit significant nonlinear hardening behavior under large deformations (strain > 10%), and the linear model underestimates the actual stress by more than 30%; the collagen fibers in the vessel wall are oriented (mainly circumferentially), resulting in a significant difference in circumferential and axial stiffness (up to 5 - 10 times), which cannot be characterized by isotropic models; traditional models do not consider time dependence (such as stress relaxation, creep), making it difficult to simulate the transient mechanical behavior of blood vessels during the cardiac cycle. Although hyperelastic models (such as the Fung model) partially solve the nonlinear problem, their parameter fitting depends on complex experiments (biaxial tension, internal pressure - diameter tests), and they are not deeply integrated with high - resolution imaging techniques, limiting their clinical practicality.

[0006] Existing blood pressure monitoring technologies have core defects such as intermittent measurement (e.g., cuff - type sphygmomanometers cannot continuously monitor), invasive operations (e.g., arterial catheter insertion), and functional singularity (lack of synchronous analysis of vascular biomechanical parameters). Although LC - OCT has the advantage of micron - level real - time dynamic imaging, the disconnection between its multimodal data (structure + hemodynamics) and non - linear elastic mechanics modeling has not fully released the technical potential of vascular elastic modulus inversion and blood pressure dynamic assessment. This background provides a clear technical need and innovation space for the invention of an LC - OCT - based multi - parameter blood pressure assessment system. Summary of the Invention

[0007] Object of the Invention: The present invention aims to overcome the invasive, intermittent, and functionally singular defects of traditional blood pressure measurement technologies. By integrating the high - resolution real - time imaging ability of line - field confocal optical coherence tomography (LC - OCT), Doppler blood flow detection technology, and non - linear hyperelastic mechanics models (such as the Fung model), a non - invasive, multimodal - fusion vascular biomechanics assessment system is constructed. This system can simultaneously obtain the microscopic structure (micron - level three - dimensional deformation), elastic modulus (quantifying vascular hardness), and hemodynamic parameters (velocity, acceleration) of blood vessels, and establish a quantitative correlation model of "vascular mechanics - blood flow - blood pressure" based on machine learning algorithms, ultimately achieving non - invasive, continuous, and high - precision blood pressure dynamic monitoring. It is especially suitable for early microcirculation abnormality screening in hypertension, arteriosclerosis risk warning, and personalized treatment support for cardiovascular diseases, promoting the leap - forward development of optical imaging technology from structural diagnosis to functional and quantitative assessment.

[0008] Technical Solution: The present invention discloses a method for measuring vascular elastic modulus and blood pressure based on LC - OCT hyperelastic modeling, including the following steps:

[0009] 1) Dynamically capture microscopic deformation data of superficial blood vessels through the high-resolution real-time three-dimensional imaging module of LC-OCT;

[0010] 2) Combine the data in step 1) with the hyperelastic mechanical Fung model to construct a non-linear mapping of the stress-strain relationship of the blood vessel wall, and calculate the blood vessel elastic modulus;

[0011] 3) Establish a correlation model based on the elastic modulus and blood pressure, quantify the blood vessel hardness and inversely infer the blood pressure value.

[0012] Furthermore, the high-resolution real-time three-dimensional imaging module in step 1) is specifically:

[0013] 1.1) Select a suitable superficial blood vessel area for imaging, such as the wrist, fingertip, earlobe or forearm. The blood vessels in these areas (such as arterioles or veins) are close to the skin surface and are easy to detect by LC-OCT. According to the depth and diameter of the target blood vessel, adjust the scanning parameters of LC-OCT to a wavelength of 830 nm, a focal length of 15 mm, a line field width of 7 mm, and an NA of 0.25 to ensure that the blood vessel wall and blood flow are clearly visible;

[0014] 1.2) Use the LC-OCT device to perform three-dimensional scanning on the target area to generate high-resolution cross-sectional and longitudinal images, and use the real-time imaging function to dynamically observe the morphological changes of the blood vessels, especially the periodic dilation and contraction of the blood vessel wall;

[0015] 1.3) The lateral resolution of LC-OCT can reach 1-3 microns, and the axial resolution is 5-10 microns, and it can clearly show the microscopic structures such as the endothelium, smooth muscle layer and adventitia layer of the blood vessel wall.

[0016] Furthermore, the dynamic deformation data acquisition in step 1) is specifically:

[0017] 2.1) During the cardiac pulsation cycle, continuously record the periodic movements of the blood vessel wall diameter change and wall displacement. Through time-series imaging at 30 frames per second, capture the instantaneous deformation characteristics of the blood vessel with blood pressure fluctuations, and record the changes in blood vessel diameter (Δr), wall displacement (Δx, Δy) and blood flow pulse waveform;

[0018] 2.2) Integrate the Doppler LC-OCT module to measure the blood flow velocity and direction, and synchronously record the blood vessel geometric parameters (initial radius r0, wall thickness h) and environmental parameters (temperature, contact pressure);

[0019] 2.3) Use a motion artifact correction algorithm (such as non-rigid image registration) to eliminate the interference of breathing or limb movement, and apply wavelet denoising technology to improve the signal-to-noise ratio of the blood vessel boundary and blood flow signal (SNR>20 dB).

[0020] Further, the Doppler LC-OCT module in step 2.2) is specifically as follows:

[0021] 3.1) The Doppler LC-OCT module realizes quantitative measurement of blood flow velocity by detecting the optical frequency shift caused by scattering particles (such as red blood cells) in flowing blood. When the laser irradiates the moving red blood cells, the frequency of the reflected light will shift (Δf), and the magnitude of the frequency shift is proportional to the blood flow velocity v. The formula is:

[0022]

[0023] where λ is the laser wavelength and θ is the angle between the light beam and the blood flow direction;

[0024] 3.2) LC-OCT analyzes the frequency shift through the phase change (Δφ) of the interference signal. n is the tissue refractive index, T is the time interval between adjacent A-lines, and the relationship between the phase change and the flow velocity is:

[0025]

[0026] And calculate the phase difference Δφ between adjacent A-lines. I1 and are the complex interference signals of adjacent A-lines. The formula is:

[0027]

[0028] Judge the blood flow direction according to the positive or negative of Δφ, and generate a color Doppler blood flow map in combination with the blood vessel structure image.

[0029] Further, step 2) is specifically as follows:

[0030] 4.1) Obtain the blood vessel wall displacement field data through LC-OCT image segmentation and calculate the Green-Lagrange strain:

[0031]

[0032] E 11 ,E 22 are the circumferential and axial Green-Lagrange strain components, u r is the radial displacement of the blood vessel wall, R is the original radius of the blood vessel, u z is the axial displacement of the blood vessel wall, is the axial displacement gradient, and construct the strain tensor E to characterize the three-dimensional deformation state of the blood vessel wall;

[0033] 4.2) Use the Fung strain energy function to describe the nonlinear mechanical behavior of the blood vessel:

[0034]

[0035] Among them, W is the strain energy density. By fitting the material parameters c, a1, a2, a4 of the blood vessel through experimental data, the stress tensor is calculated

[0036] 4.3) Since the blood vessel wall structure is divided into three layers: the intima, the media, and the adventitia, and the arrangement directions of the collagen fibers in each layer are different, a hierarchical model is constructed so that each layer uses independent Fung model parameters:

[0037] W total = W intima + W media + W adventitia ,

[0038] At the same time, to simulate the instantaneous response of the blood vessel during the cardiac pulsation cycle, a viscoelastic term is superimposed in the Fung model:

[0039]

[0040] Among them, G(t) is the relaxation modulus, which describes the decay of stress with time, and τ is the relaxation time constant,

[0041] The strain is input into the Fung model, the stress distribution is iteratively solved, and the instantaneous tangent modulus is calculated according to the stress-strain curve Generate the dynamic curve of blood vessel stiffness with the change of blood pressure, and output the spatio-temporal distribution map of the blood vessel elastic modulus to quantify the local hardened or softened areas.

[0042] Furthermore, the specific content of step 3) is as follows:

[0043] 5.1) Combining hemodynamics and blood vessel mechanics theories, establish the theoretical equation of elastic modulus - blood pressure:

[0044]

[0045] Among them, k1, k2, k3 are weight coefficients fitted by the least squares method or clinical data, is the blood flow acceleration, is the wall thickness - radius ratio, and C is an individualized calibration constant (such as the baseline blood pressure);

[0046] 5.2) Input the LC-OCT multimodal data (elastic modulus E tan , blood flow velocity v, blood vessel geometry ) into the neural network (NN) to train the end-to-end blood pressure prediction model, and adopt the transfer learning strategy to adapt to the differences in blood vessel characteristics of different populations (healthy, hypertensive, diabetic patients);

[0047] Use the theoretical model to provide the initial parameter relationship and machine learning to correct the residuals of the theoretical model, expressed as:

[0048]

[0049] Thus, the final hybrid model is constructed.

[0050] Beneficial effects

[0051] The beneficial effects of the present invention compared with the prior art are that through the combination of multimodal data fusion and high-resolution imaging technology, the limitations of traditional blood pressure measurement are broken through, and non-invasive, continuous and high-precision vascular function evaluation is realized. First, the integration of the micron-level resolution of LC-OCT (1-3 μm laterally and 5-10 μm axially) and Doppler technology can synchronously obtain the instantaneous deformation, blood flow velocity and direction of the blood vessel wall, making it possible to perform dynamic mechanical analysis at the microvascular level and solving the problem of the single function of traditional OCT. Second, the calculation of elastic modulus based on the Fung hyperelastic model and non-linear mechanical modeling can accurately quantify the vascular stiffness (error ≤ ±5%), overcoming the error accumulation problem of Hooke's law of linear models under large deformations, and providing highly reliable biomechanical parameters for blood pressure inversion.

[0052] From the perspective of clinical practicability, the non-invasive characteristic of the present invention significantly reduces the infection risk and patient discomfort, and is especially suitable for long-term monitoring of children, the elderly and critically ill patients; its dynamic continuous measurement ability (≥30 fps) can capture the instantaneous fluctuations of blood pressure within the cardiac cycle (such as systolic pressure peak and diastolic pressure trough), providing richer time-series data support for hypertension typing and drug efficacy evaluation. In addition, the multi-parameter fusion model (elastic modulus, blood flow velocity, blood vessel geometry) realizes a blood pressure prediction error ≤ ±5 mmHg through the neural network algorithm of machine learning, which is better than the ±10 mmHg error of the traditional cuff method, and is more stable in obese or hypotensive patients.

[0053] In the field of early disease screening and personalized treatment, the present invention can warn of the risk of hypertension or atherosclerosis 3-6 months in advance by combining abnormal microvascular elastic modulus (such as Etan > 200 kPa in the initial stage of sclerosis) with hemodynamic parameters (such as decreased flow velocity and abnormal acceleration), providing a window period for early intervention. In terms of technology transformation potential, the modular architecture of this system can be extended to deep vascular imaging (such as carotid plaque stability assessment) or work in cooperation with other medical imaging devices (such as ultrasound, MRI), promoting the rapid implementation of vascular biomechanics research from the laboratory to the clinic. Description of the drawings

[0054] Figure 1 is the flowchart of the method of the present invention;

[0055] Figure 2 is the LC-OCT principle diagram of the method of the present invention. Detailed implementation manners

[0056] This application relates to multiple scientific and technological fields such as medicine, OCT, and image processing, with a wide coverage. Therefore, unless otherwise defined, the meanings of all technical and scientific terms used herein are the same as those commonly understood by those skilled in the art of this application field; the terms described herein are only used for the purpose of describing specific embodiments and do not mean a limitation to this application; the terms "including" and "having" used in this application, and any variants thereof, are intended to cover a non-exclusive meaning.

[0057] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0058] As Figure 1 shown, a specific embodiment of the present invention discloses a vascular elasticity module based on line-field confocal optical coherence tomography (LC-OCT) technology, including the following steps:

[0059] 1) First, conduct an overall stratification based on the type of hypertension, divide the samples into primary hypertension (accounting for 90%-95%) and secondary hypertension (accounting for 5%-10%), and focus on ensuring that the population over 55 years old accounts for 60% of the total sample to cover the main high-incidence groups. For the young population (<40 years old), a subgroup is separately set within the secondary hypertension group, requiring its proportion to reach 15%-20%, and further refine the sampling strategy - use stratified random sampling for this subgroup, so that renovascular hypertension accounts for 50%-60% and endocrine hypertension accounts for 30%-40%.

[0060] 2) Through the high-resolution acquisition step of the real-time three-dimensional imaging module of the samples in step 1) by LC-OCT, dynamically capture the microscopic deformation data of superficial blood vessels;

[0061] 2.1) Select a suitable superficial blood vessel area for imaging, such as the wrist, fingertip, earlobe, or forearm. The blood vessels in these areas (such as arterioles or veins) are close to the skin surface and are easy to detect by LC-OCT. According to the depth and diameter of the target blood vessel, adjust the scanning parameters of LC-OCT to a wavelength of 830 nm, a focal length of 15 mm, a line-field width of 7 mm, and an NA of 0.25 to ensure that the blood vessel wall and blood flow are clearly visible;

[0062] 2.2) Use the LC-OCT device to perform three-dimensional scanning on the target area. The system uses a superluminescent diode (SLD) light source with a central wavelength of 830 nm and a bandwidth of 70 nm, and combines a high-speed galvanometer (scanning frequency ≥2 kHz) to achieve synchronous scanning in the transverse (B-scan) and longitudinal (C-scan) directions, generate high-resolution cross-sectional and longitudinal-sectional images, and use the real-time imaging function to dynamically observe the morphological changes of blood vessels, especially the periodic dilation and contraction of the blood vessel wall;

[0063] 2.3) By integrating a three-dimensional imaging module, the lateral resolution of LC-OCT can reach 1-3 microns, and the axial resolution is 5-10 microns. It can clearly display the multi-layer microscopic structure of the blood vessel wall, including the cell arrangement and surface microvilli of the endothelium, the layered distribution of the smooth muscle layer, and the collagen fiber and elastic fiber structures of the adventitia layer.

[0064] 3) The specific acquisition of the dynamic deformation data of the three-dimensional imaging module in step 2) is as follows:

[0065] 3.1) During the cardiac pulsation cycle, continuously record the periodic motion of the blood vessel wall diameter change and wall displacement. Through time-series imaging at 30 frames per second, capture the instantaneous deformation characteristics of the blood vessel with blood pressure fluctuations, and record the change in blood vessel diameter (Δr), wall displacement (Δx, Δy), and blood flow pulse waveform;

[0066] 3.2) The integrated Doppler LC-OCT module realizes the accurate measurement of blood flow velocity and direction by detecting the frequency shift effect of moving red blood cells on the incident light. The system adopts the Phase-Resolved Doppler OCT algorithm, and calculates the flow velocity based on the phase difference (Δφ) between adjacent A-scans. Where λ is the central wavelength of the light source (830 nm), n is the tissue refractive index (~1.38), T is the A-scan time interval (5 ms). Its core performance parameters include a flow velocity measurement range of 0.1-20 mm / s, a velocity resolution of ±0.05 mm / s, and a direction recognition accuracy of ±5°. In addition, the Doppler LC-OCT module combines the synchronous recording of blood vessel geometric parameters (initial radius r0, wall thickness h) and environmental parameters (temperature, contact pressure) to comprehensively evaluate the hemodynamic state;

[0067] 3.3) At the same time, use a rigid image registration algorithm of the free deformation model based on non-uniform B-spline to perform motion correction on consecutive frame images. This algorithm constructs an elastic deformation field, aligns the blood vessel structures at different time points pixel by pixel, locally matches the blood vessel dilation morphology during cardiac systole with the diastolic morphology, compensates for the global translation and local distortion caused by respiration or slight movement, and enables the displacement correction accuracy to reach the sub-pixel level (<1μm).

[0068] Meanwhile, to address the impact of high-frequency speckle noise and random noise on the clarity of blood vessel boundaries, wavelet denoising technology is introduced. The noise components and real signals are separated through five-level decomposition using the Daubechies4 wavelet basis: the macroscopic deformation characteristics of the blood vessel wall are retained in the low-frequency subbands, while the speckle noise is suppressed in the high-frequency subbands using an adaptive threshold (BayesShrink). Finally, the signal-to-noise ratio (SNR) of the reconstructed image is increased to over 20 dB. This combined strategy not only improves the resolution of the inner and outer membrane boundaries of blood vessels by approximately 30% (the boundary is sharpened to 2 - 3 μm), but also clearly retains the instantaneous details of blood flow pulses (such as microsecond-level fluctuations in peak flow velocity), providing a high-fidelity data basis for subsequent elastic modulus calculation and blood pressure inversion.

[0069] 4) The Doppler LC-OCT module in step 3.2) is specifically as follows:

[0070] 4.1) The Doppler LC-OCT module quantitatively measures blood flow velocity by detecting the optical frequency shift caused by scattering particles (such as red blood cells) in flowing blood. When the laser irradiates moving red blood cells, the reflected light frequency will shift (Δf), and the magnitude of the frequency shift is proportional to the blood flow velocity v. The formula is:

[0071]

[0072] where λ is the laser wavelength and θ is the angle between the light beam and the blood flow direction;

[0073] 4.2) LC-OCT analyzes the frequency shift through the phase change (Δφ) of the interference signal. n is the tissue refractive index, T is the time interval between adjacent A-lines, and the relationship between the phase change and the flow velocity is:

[0074]

[0075] And the phase difference Δφ between adjacent A-lines is calculated. I1 and are the complex interference signals of adjacent A-lines. The formula is:

[0076]

[0077] The blood flow direction is judged based on the sign of Δφ, and combined with the blood vessel structure image, a color Doppler blood flow map is generated.

[0078] 5) Combine the dynamic deformation data in step 3) with the hyperelastic mechanical Fung model to construct a non-linear mapping of the stress-strain relationship of the blood vessel wall and calculate the blood vessel elastic modulus;

[0079] 5.1) Obtain the blood vessel wall displacement field data through LC-OCT image segmentation and calculate the Green-Lagrange strain:

[0080]

[0081] E 11 ,E 22 are the circumferential and axial Green-Lagrange strain components, u r is the radial displacement of the vessel wall, R is the original radius of the vessel, u z is the axial displacement of the vessel wall, is the axial displacement gradient. The strain tensor E is constructed to characterize the three-dimensional deformation state of the vessel wall;

[0082] 5.2) The Fung strain energy function is used to describe the nonlinear mechanical behavior of the blood vessel:

[0083]

[0084] where W is the strain energy density. By fitting the material parameters of the blood vessel with experimental data, c = 15.2 kPa, a1 = 0.87, a2 = 0.34, a4 = 0.16, the stress tensor

[0085] 5.3) Since the blood vessel wall structure is divided into three layers: intima, media, and adventitia, and the arrangement directions of collagen fibers in each layer are different, a layered model is constructed so that each layer uses independent Fung model parameters:

[0086] W total = W intima + W media + W adventitia ,

[0087] At the same time, to simulate the instantaneous response of the blood vessel during the cardiac cycle, a viscoelastic term is superimposed on the Fung model:

[0088]

[0089] where G(t) is the relaxation modulus, which describes the decay of stress with time, and τ is the relaxation time constant,

[0090] The strain is input into the Fung model, the stress distribution is iteratively solved, and the instantaneous tangent modulus is calculated according to the stress-strain curve Generate the dynamic curve of blood vessel stiffness with blood pressure change and output the spatio-temporal distribution map of blood vessel elastic modulus to quantify local hardening or softening areas.

[0091] 6) Establish a correlation model with blood pressure based on the elastic modulus in step 5) to quantify blood vessel stiffness and back-calculate blood pressure values;

[0092] 6.1) Combine hemodynamics and blood vessel mechanics theories to establish an elastic modulus-blood pressure theoretical equation:

[0093]

[0094] where k1, k2, and k3 are weight coefficients fitted by the least squares method or clinical data. is the blood flow acceleration. is the wall thickness - radius ratio, and C is an individualized calibration constant (such as baseline blood pressure);

[0095] 6.2) Normalize (Z - score normalization) and temporally align (segment by cardiac cycle) the LC - OCT multimodal data (elastic modulus E tan , blood flow velocity v, vascular geometry ), and perform data augmentation by adding Gaussian noise (SNR = 30 dB) and random scaling (±5%) to simulate measurement errors and individual differences, and simultaneously collect systolic blood pressure (SBP), diastolic blood pressure (DBP), and mean arterial pressure (MAP) as labeled data, which are obtained by the calibrated cuff method;

[0096] 6.3) Input the data into a neural network (NN). The neural network adopts a heterogeneous multi - branch structure to separately process scalar features (E tan , ) and temporal features (E tan (t) at 30 frames per second). The scalar features are input into a fully - connected layer (128 nodes, ReLU activation). The temporal features are flattened after extracting features through a one - dimensional convolutional layer (1D - CNN, 16 filters, kernel size = 3). The outputs of the two branches are concatenated at the fusion layer and then pass through a fully - connected layer (64 nodes, ReLU activation) for feature abstraction. Finally, the systolic blood pressure (SBP), diastolic blood pressure (DBP), and mean arterial pressure (MAP) are predicted through a linear output layer. The network is optimized by mean squared error (MSE) and L2 regularization (λ = 0.01), and trained using the Adam optimizer (learning rate = 1×10 -4 , batch size = 32) to obtain an end - to - end blood pressure prediction model;

[0097] 6.4) Adopt a transfer learning strategy to adapt to the differences in vascular characteristics among different populations (healthy, hypertensive, diabetic patients). First, pre-train a basic model on healthy population data (1000 cases) to learn general features. Then, for the target population (500 hypertensive patients), adopt a fine-tuning strategy: freeze the underlying weights (the first two layers), and only train the top fully connected layer (from 32 nodes to the output layer) to retain the general feature extraction ability and adapt to the target domain-specific patterns. To further improve cross-domain performance, introduce domain adaptation technology, and force the network to learn domain-invariant features through a gradient reversal layer (GRL). The loss function is the weighted sum of the blood pressure prediction MSE and the domain classification cross-entropy (weight ratio = 1:0.5). The dynamic learning rate adjustment adopts a cosine annealing schedule (initial learning rate = 1×10 -5 , period = 10 epochs).

[0098] Use the theoretical model to provide the initial parameter relationship and the residual of the machine learning corrected theoretical model, expressed as:

[0099]

[0100] Thus, the final hybrid model is constructed.

[0101] 7) Through the above steps, the method for measuring vascular elastic modulus established by the LC-OCT technology realizes the accurate measurement of human blood pressure.

[0102] Obviously, the embodiments described above only constitute part of the implementation schemes of this application, rather than all implementation schemes. The preferred embodiments of this application are shown in the drawings, but are not limited thereto. The patent scope of this application is not limited to the content shown in the drawings. This application can be implemented in many different forms. The purpose of providing these embodiments is to more comprehensively clarify the content of this application. Although the foregoing embodiments have been described in detail, for those skilled in the art, the specific implementation manners can still be modified, or some technical features can be equivalently replaced. Any equivalent structure directly or indirectly applied in other related technical fields using the content of the specification and drawings of this application should be included in the patent protection scope of this application.

Claims

1. A method for measuring vascular elastic modulus and blood pressure based on LC-OCT hyperelastic modeling, characterized in that, It includes the following steps: 1) Dynamically capture the microscopic deformation data of superficial blood vessels through the high-resolution real-time three-dimensional imaging module of LC-OCT; 2) Combine the data in step 1) with the hyperelasticity mechanics Fung model to construct a non-linear mapping of the stress-strain relationship of the blood vessel wall, and calculate the blood vessel elastic modulus; 3) Establish a correlation model based on the elastic modulus and blood pressure, quantify the blood vessel stiffness and inversely deduce the blood pressure value.

2. The method for measuring vascular elastic modulus and blood pressure based on LC-OCT hyperelastic modeling according to claim 1, wherein In step 1), the high-resolution real-time three-dimensional imaging module is: Step 2.1) Select a suitable superficial blood vessel area for imaging, such as the wrist, fingertip, earlobe or forearm. The blood vessels in these areas (such as arterioles or veins) are close to the skin surface and are easy to detect by LC-OCT. According to the depth and diameter of the target blood vessel, adjust the scanning parameters of LC-OCT to a wavelength of 830 nm, a focal length of 15 mm, a line field width of 7 mm, and an NA of 0.25 to ensure that the blood vessel wall and blood flow are clearly visible; Step 2.2) Use the LC-OCT device to perform three-dimensional scanning on the target area to generate high-resolution cross-sectional and longitudinal images, and use the real-time imaging function to dynamically observe the morphological changes of the blood vessels, especially the periodic dilation and contraction of the blood vessel wall; Step 2.3) The lateral resolution of LC-OCT can reach 1-3 microns, and the axial resolution is 5-10 microns, and it can clearly show the microscopic structures such as the endothelium, smooth muscle layer and adventitia layer of the blood vessel wall.

3. A method for measuring vascular elastic modulus and blood pressure based on LC-OCT hyperelastic modeling according to claim 1, characterized in that In step 1), the dynamic deformation data acquisition is: Step 3.1) During the cardiac pulsation cycle, continuously record the periodic movement of the blood vessel wall diameter change and the wall displacement. Through time-series imaging at 30 frames per second, capture the instantaneous deformation characteristics of the blood vessel with blood pressure fluctuations, and record the change in blood vessel diameter (Δr), the wall displacement (Δx, Δy) and the blood flow pulse waveform; Step 3.2) Integrate the Doppler LC-OCT module to measure the blood flow velocity and direction, and synchronously record the blood vessel geometric parameters (initial radius r0, wall thickness h) and environmental parameters (temperature, contact pressure); Step 3.3) Use a motion artifact correction algorithm (such as non-rigid image registration) to eliminate the interference of breathing or limb movement, and apply wavelet denoising technology to improve the signal-to-noise ratio of the blood vessel boundary and blood flow signal (SNR>20 dB).

4. A method for measuring vascular elastic modulus and blood pressure based on LC-OCT hyperelastic modeling according to claim 1, characterized in that In step 3.2), the Doppler LC-OCT module is: Step 4.1) The Doppler LC-OCT module realizes the quantitative measurement of blood flow velocity by detecting the optical frequency shift caused by scattering particles (such as red blood cells) in flowing blood. When the laser irradiates the moving red blood cells, the reflected light frequency will shift (Δf), and the magnitude of the frequency shift is proportional to the blood flow velocity v. The formula is: where λ is the laser wavelength and θ is the angle between the light beam and the blood flow direction; Step 4.2) LC-OCT analyzes the frequency shift through the phase change (Δφ) of the interference signal. n is the tissue refractive index, T is the time interval between adjacent A-lines, and the relationship between the phase change and the flow velocity is: And calculate the phase difference Δφ between adjacent A-lines, where I1 and is the complex interference signal of adjacent A-lines, and the formula is: Judge the blood flow direction according to the positive and negative of Δφ, and combine with the blood vessel structure image to generate a color Doppler blood flow map.

5. A method for measuring vascular elastic modulus and blood pressure based on LC-OCT hyperelastic modeling according to claim 1, characterized in that, In step 2), the blood vessel elastic modulus is: Step 5.1) Obtain the vascular wall displacement field data through LC-OCT image segmentation and calculate the Green-Lagrange strain: E 11 , E 22 are the circumferential and axial Green-Lagrange strain components, u r is the radial displacement of the vessel wall, R is the original radius of the vessel, u z is the axial displacement of the vessel wall, is the axial displacement gradient, and the strain tensor E is constructed to characterize the three-dimensional deformation state of the vessel wall; Step 5.2) Use the Fung strain energy function to describe the nonlinear mechanical behavior of blood vessels: Where W is the strain energy density, and the material parameters of the blood vessel are fitted through experimental data as c = 15.2 kPa, a1 = 0.87, a2 = 0.34, a4 = 0.16 to calculate the stress tensor Step 5.3) Since the vascular wall structure is divided into three layers: intima, media, and adventitia, and the arrangement directions of collagen fibers in each layer are different, a layered model is constructed so that each layer uses independent Fung model parameters: W total = W intima + W media + W adventitia , At the same time, to simulate the instantaneous response of blood vessels during the cardiac cycle, a viscoelastic term is superimposed on the Fung model: where G(t) is the relaxation modulus, which describes the decay of stress over time, and τ is the relaxation time constant; Input the strain into the Fung model, iteratively solve the stress distribution, and calculate the instantaneous tangent modulus according to the stress-strain curve Generate the dynamic curve of vessel stiffness with blood pressure change of elastic modulus, output the spatio-temporal distribution map of vessel elastic modulus, and quantify the local hardening or softening regions 6. A method for measuring vascular elastic modulus and blood pressure based on LC-OCT hyperelastic modeling according to claim 1, characterized in that, The correlation model established in step 3) is: Step 6.1) Combine hemodynamics and vascular mechanics theories to establish an elastic modulus-blood pressure theoretical equation: where k1, k2, and k3 are weight coefficients fitted by the least squares method or clinical data, is the blood flow acceleration, is the wall thickness-radius ratio, and C is an individualized calibration constant (such as baseline blood pressure); Step 6.2) Input LC-OCT multimodal data (elastic modulus E tan , blood flow velocity v, vascular geometry ) into the neural network (NN) to train an end-to-end blood pressure prediction model, and adopt a transfer learning strategy to adapt to the differences in vascular characteristics of different populations (healthy, hypertensive, diabetic patients); Use the theoretical model to provide the initial parameter relationship and machine learning to correct the residuals of the theoretical model, expressed as: Thus, the final hybrid model is constructed.

Images