Evaluation method and system for diabetic retinopathy

Abstract

The invention discloses a diabetic retinopathy assessment method and system, and relates to the technical field of retina image detection.The method comprises the steps that a stepped pressure signal is generated through an intraocular pressure device, high-resolution OCT equipment is synchronously triggered, full-period response of blood vessels from compensation to decompensation is covered, and the limitation of traditional single-time pressurization detection is broken through; the change rate of the diameter of the blood vessel is calculated in real time, the pressure gradient or interval time is dynamically adjusted when the threshold value is exceeded, and the detection efficiency is improved while safety is guaranteed; constructing a stress-strain model of a pressure-diameter scatter diagram, extracting an initial elastic modulus, a hardening turning point and a hysteresis coefficient for joint analysis, and solving the problem of high misjudgment rate of a single parameter; and triggering a risk level signal when the at least two characteristic parameters exceed the limit. Through a technical closed loop of dynamic stimulation-image analysis-multi-parameter diagnosis, an innovative means is provided for early warning and accurate intervention of diabetic retinopathy.

Description

Technical Field

[0001] The present invention relates to the technical field of retinal image detection, and in particular to a method and system for evaluating diabetic retinopathy. Background Art

[0002] Diabetic retinopathy (DR), one of the most common microvascular complications of diabetes, has become a leading cause of blindness in adults worldwide. Currently, clinical diagnosis relies primarily on imaging techniques such as fundus photography, fluorescein angiography (FA), and optical coherence tomography (OCT), which assess morphological features such as microaneurysms, hemorrhages, and hard exudates for grading. However, these methods have significant limitations:

[0003] Morphological changes (such as microvascular leakage) usually appear in the middle and late stages of vascular functional damage. Staging is performed by observing structural abnormalities such as microaneurysms and bleeding points, but it is not sensitive enough to the early functional compensation stage, resulting in a missed detection rate of early lesions as high as 40%. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method and system for evaluating diabetic retinopathy.

[0005] In order to achieve the above object, the technical solution of the present invention is as follows:

[0006] In a first aspect, the present invention discloses a method for evaluating diabetic retinopathy, comprising the following steps:

[0007] Acquire a stepped pressure adjustment signal sequence generated by an intraocular pressure control device, wherein the pressure adjustment signal sequence includes at least three pressure step values ​​of a predetermined number of digits;

[0008] triggering an optical coherence tomography device based on the pressure adjustment signal sequence to obtain dynamic response image data of the target blood vessel segment under the corresponding pressure step;

[0009] Determining real-time blood vessel diameter data by processing the dynamic response image data, and generating a pressure-diameter scatter plot by associating corresponding pressure step values;

[0010] Real-time calculation of the rate of change of blood vessel diameter in dynamic response imaging data:

[0011] When the blood vessel diameter change rate exceeds the first feedback threshold, the interval time of the next pressure step is extended to the preset safety time;

[0012] When the blood vessel diameter change rate is lower than the second feedback threshold, increasing the gradient amplitude of the pressure step value to a preset maximum change rate;

[0013] Constructing a stress-strain response curve model of a pressure-diameter scatter plot, and extracting a predefined characteristic parameter group in the curve model, wherein the characteristic parameter group includes at least an initial elastic modulus, a hardening turning point, and a hysteresis coefficient;

[0014] The characteristic parameter group is compared with a preset threshold range, and when at least two characteristic parameters exceed the corresponding threshold range, a diabetic retinopathy risk level signal is generated.

[0015] In a second aspect, the present invention further discloses a diabetic retinopathy evaluation system, which implements the above-mentioned diabetic retinopathy evaluation method, comprising:

[0016] a pressure control and image acquisition module, configured to obtain a stepped pressure regulation signal sequence generated by the intraocular pressure control device, and trigger an optical coherence tomography device based on the pressure regulation signal sequence to obtain dynamic response image data of the target blood vessel segment under corresponding pressure steps;

[0017] a blood vessel dynamic tracking module, configured to determine real-time blood vessel diameter data by processing the dynamic response image data, and generate a pressure-diameter scatter plot by associating the corresponding pressure step values;

[0018] A biomechanical modeling module is used to construct a stress-strain response curve model of a pressure-diameter scatter plot and extract a predefined characteristic parameter group in the curve model;

[0019] Dynamic feedback decision module, used to calculate the rate of change of blood vessel diameter in dynamic response imaging data in real time:

[0020] When the blood vessel diameter change rate exceeds the first feedback threshold, the interval time of the next pressure step is extended to the preset safety time;

[0021] When the blood vessel diameter change rate is lower than the second feedback threshold, increasing the gradient amplitude of the pressure step value to a preset maximum change rate;

[0022] The risk assessment and output module is used to compare the characteristic parameter group with a preset threshold range, and generate a diabetic retinopathy risk level signal when at least two characteristic parameters exceed the corresponding threshold range.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. By simulating physiological intraocular pressure fluctuations and stimulating the true mechanical response of blood vessels, the stepped pressure sequence can capture the nonlinear response characteristics of blood vessels in different pressure ranges, significantly improving the detection rate of early lesions compared to traditional static or linear pressure stimulation.

[0025] 2. Extract parameters such as the initial elastic modulus, sclerosis turning point, and hysteresis coefficient from the pressure-diameter scatter plot. A decrease in the initial elastic modulus reflects early collagen degradation, an anterior shift in the sclerosis turning point indicates vascular fibrosis, and an abnormal hysteresis coefficient indicates increased blood flow resistance. The combined determination of these three parameters can reduce the false positive rate to below 5%.

[0026] 3. Through the technical path of dynamic pressure stimulation-real-time feedback-biomechanical modeling, the key defects of traditional diabetic retinopathy assessment methods in early detection, quantitative accuracy, and equipment universality are solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The disclosure of the present invention is described with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components. Among them:

[0028] Figure 1 is a flow chart of the steps of the present invention;

[0029] Figure 2 is a flow chart for calculating real-time blood vessel diameter data of the present invention;

[0030] Figure 3 A flow chart for generating anti-interference blood vessel diameter data according to the present invention;

[0031] Figure 4 Constructing a flow chart for the stress-strain response curve model of the present invention;

[0032] Figure 5 FIG. 1 is a diagram illustrating the execution process of the piecewise linear regression algorithm of the present invention;

[0033] Figure 6 This is a risk level signal triggering flow chart of the present invention;

[0034] Figure 7 This is a flow chart of the noise filtering steps of the present invention;

[0035] Figure 8 generating a flow chart for the disease progression report of the present invention;

[0036] Figure 9 This is a diagram of the system module composition of the present invention. DETAILED DESCRIPTION

[0037] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention and should not be regarded as the entire invention or as a limitation or restriction of the technical solution of the present invention.

[0038] Application Overview:

[0039] In the traditional existing diabetic retinopathy assessment system, the morphological feature monitoring methods that clinically rely on are unable to capture early functional damage to the biomechanical properties of the vascular wall. Morphological analysis methods based on fundus color photography and optical coherence tomography can only identify structural abnormalities such as vascular leakage and microaneurysms. However, changes in mechanical parameters such as decreased elastic modulus of the vascular wall and delayed stress response cannot be effectively quantified by differences in grayscale values ​​of two-dimensional images during the compensatory period. When the patient's vascular endothelial function has shown reduced compliance but no visible lesions have yet formed, the existing system's pressure loading mode and image acquisition timing lack a dynamic coordination mechanism, resulting in the inability to establish a continuous response curve of vascular diameter changes with pressure.

[0040] For example, when dynamic intraocular pressure monitoring is implemented in an ophthalmology diagnosis and treatment center, the use of a loading method with a fixed pressure step interval may cause non-steady-state changes in blood vessel diameter. When the intraocular pressure control device applies a step pressure of 20mmHg to 24mmHg at 5-second intervals, if the elastic reserve of the blood vessel wall is insufficient, the diameter change rate will show a nonlinear jump after the third pressure step. Due to the lack of a real-time feedback mechanism for the rate of change of blood vessel diameter, the existing system is unable to dynamically adjust the pressure loading parameters according to the biomechanical response, resulting in excessive inter-frame displacement or a decrease in the signal-to-noise ratio in the key turning area of ​​the collected dynamic image data, which in turn causes the coordinate drift of the hardening turning point to exceed the order of 0.3mmHg when fitting the stress-strain curve.

[0041] If these issues are not addressed, measurement errors in vascular biomechanical parameters will directly impact the accuracy of diabetic retinopathy staging. Clinically, the progression of endothelial dysfunction cannot be determined by the rate of elastic modulus decay, leading to high-risk patients missing the optimal intervention window. The lack of a quantitative assessment system also results in a lack of comparability of test results under different pressure-loading protocols, hindering the establishment of standardized diagnostic reference intervals across medical institutions.

[0042] When faced with the above problems, this application first considers how to achieve dynamic coordination between pressure loading and vascular biomechanical response. Traditional fixed-interval pressure step loading causes the rate of change of vascular diameter to exceed the effective monitoring range at the critical stage, and a real-time feedback mechanism needs to be established. This application analyzed and found that the sudden change in the rate of change of vascular diameter is directly related to the adaptability of the pressure step parameters. When the rate of change is too high, the loading process needs to be delayed to avoid non-steady-state response, while a low rate of change indicates that the stimulation intensity needs to be increased. To this end, this application introduces a closed-loop control logic of the rate of change of vascular diameter and pressure parameters, and dynamically adjusts the pressure loading timing and gradient amplitude by setting a dual feedback threshold to ensure that the blood vessels are in a resolvable mechanical response range.

[0043] like Figure 1As shown, in this regard, the present application proposes a method for evaluating diabetic retinopathy, comprising the following steps:

[0044] A stepped pressure regulation signal sequence generated by an intraocular pressure control device is obtained, where the pressure regulation signal sequence includes at least three predetermined-digit pressure step values. A stepped pressure regulation signal sequence refers to a control instruction sequence including multiple different pressure values. Specifically, this can be achieved by using an intraocular pressure control device to generate a pressure waveform according to preset pressure step values ​​and time intervals. By applying different pressures in stages to stimulate vascular responses, a quantitative basis is provided for evaluating vascular elasticity.

[0045] The optical coherence tomography device is triggered based on the pressure regulation signal sequence to obtain dynamic response image data of the target blood vessel segment under the corresponding pressure step; dynamic response image data refers to the morphological image record of the blood vessel changing over time under different pressures. Specifically, it can be achieved by using the optical coherence tomography device to perform high-speed continuous shooting when pressure is triggered, capturing the real-time deformation process of the blood vessel under pressure changes.

[0046] By processing dynamic response image data, real-time vascular diameter data is determined and associated with corresponding pressure step values ​​to generate a pressure-diameter scatter plot. A pressure-diameter scatter plot refers to a data set of associated pressure values ​​and corresponding vascular diameters. Specifically, the pressure step values ​​can be matched with the synchronously measured vascular diameter values ​​through coordinate mapping to establish a quantitative relationship model between pressure and vascular deformation.

[0047] Real-time calculation of the rate of change of blood vessel diameter in dynamic response imaging data:

[0048] When the blood vessel diameter change rate exceeds the first feedback threshold, the interval time of the next pressure step is extended to the preset safety time;

[0049] When the blood vessel diameter change rate is lower than the second feedback threshold, increasing the gradient amplitude of the pressure step value to a preset maximum change rate;

[0050] The vascular diameter change rate refers to the fluctuation in vascular diameter per unit time. This can be calculated by dividing the diameter difference between adjacent time points by the time interval and is used to determine the stability of the vascular dynamic response in real time. The first and second feedback thresholds are preset critical values ​​for vascular response rates, specifically set within the normal range of vascular contraction and expansion rates determined by clinical statistics. When these thresholds are exceeded, the pressure application strategy is automatically adjusted to balance test safety and data validity.

[0051] A stress-strain response curve model of the pressure-diameter scatter plot is constructed, and a predefined characteristic parameter group in the curve model is extracted. The characteristic parameter group at least includes the initial elastic modulus, the stiffening turning point, and the hysteresis coefficient. The stress-strain response curve model refers to a mathematical model that describes the relationship between pressure and vascular deformation. Specifically, it can be constructed through piecewise linear regression or nonlinear fitting algorithms to extract biomechanical parameters such as the elastic modulus and turning point to quantify the vascular functional status.

[0052] The characteristic parameter set is compared with a preset threshold range. When at least two characteristic parameters exceed the corresponding threshold range, a diabetic retinopathy risk level signal is generated. The characteristic parameter set refers to a collection of key indicators reflecting the mechanical properties of blood vessels, specifically including the initial elastic modulus, the sclerosis turning point, and the hysteresis coefficient. By comparing to the preset threshold, abnormal blood vessel function can be identified, compensating for the lag of morphological detection.

[0053] The core innovation of this application lies in the simultaneous acquisition of stepped pressure stimulation and vascular dynamic response images, combined with the biomechanical model to extract vascular functional parameters, thereby realizing the early functional compensation stage detection of diabetic retinopathy, breaking through the traditional morphological method's reliance on mid- and late-stage structural abnormalities.

[0054] The working process and principle of the present application are as follows: the method for evaluating diabetic retinopathy includes obtaining a stepped pressure regulation signal sequence generated by an intraocular pressure control device, triggering an optical coherence tomography device based on the pressure regulation signal sequence to obtain dynamic response image data, processing the image data to determine the blood vessel diameter data and generate a pressure-diameter scatter plot, calculating the blood vessel diameter change rate in real time and dynamically adjusting the pressure loading parameters, constructing a stress-strain response curve model and extracting characteristic parameters, and comparing the characteristic parameters with the threshold range to generate a risk level signal.

[0055] First, the intraocular pressure control device generates a stepped pressure adjustment signal sequence consisting of at least three pressure steps. The optical coherence tomography device is triggered based on this signal sequence to acquire dynamic response image data of the target vessel segment under different pressure steps. This image data is processed to determine real-time vessel diameter data and correlated with the corresponding pressure step values ​​to generate a pressure-diameter scatter plot.

[0056] During data acquisition, the system calculates the rate of change of vascular diameter in real time. When the rate of change exceeds the first feedback threshold, the interval between the next pressure step is extended to a preset safety duration to prevent non-steady-state responses in the blood vessels. When the rate of change falls below the second feedback threshold, the gradient amplitude of the pressure step is increased to the preset maximum rate of change to enhance the stimulation intensity. This dynamic adjustment mechanism ensures that the blood vessels remain within a resolvable mechanical response range.

[0057] Based on the obtained pressure-diameter scatter plot, a stress-strain response curve model was constructed. A predefined set of characteristic parameters, including the initial elastic modulus, the stiffening turning point, and the hysteresis coefficient, were extracted from this model. These parameters reflect the biomechanical properties of the vascular wall.

[0058] Finally, the extracted feature parameter set is compared with a preset threshold range. If at least two feature parameters exceed their corresponding thresholds, the system generates a diabetic retinopathy risk level signal. This biomechanical parameter-based assessment method can detect early changes in vascular function and improve the accuracy of early diagnosis of diabetic retinopathy.

[0059] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0060] The intraocular pressure control device generates a stepped pressure adjustment signal sequence, with the pressure step value range set to 10 mmHg to 25 mmHg, and the initial interval between adjacent pressure steps configured to be 4 seconds. The optical coherence tomography device is triggered based on this signal sequence to obtain dynamic response imaging data of the target blood vessel segment.

[0061] Three characteristic vascular segments with diameters ranging from 100 to 200 μm were located from the dynamic response image data. The contour boundary points of each characteristic vascular segment were determined based on the gradient changes in the vessel wall reflection intensity. The displacement vectors of the contour boundary points in adjacent image frames were calculated, and abnormal frames with displacement amplitudes exceeding a preset motion threshold were eliminated. This method of image data processing determined real-time vascular diameter data and associated corresponding pressure step values ​​to generate a pressure-diameter scatter plot.

[0062] Calculates the rate of change in blood vessel diameter in real time. When the rate of change exceeds 0.5% / mmHg, the interval between pressure steps is extended to 6 seconds. When the rate of change is less than 0.1% / mmHg, the pressure step gradient is increased to 2 mmHg / step.

[0063] A piecewise linear regression algorithm was used to fit the pressure-diameter scatter plot and construct a stress-strain response curve model. The initial elastic modulus was determined by the slope of the curve in the pressure range of 10-15 mmHg. The hardening turning point was determined by detecting a slope change of more than 20% between adjacent intervals. The hysteresis coefficient was calculated by calculating the difference in the average slope between the pressurization and decompression paths.

[0064] The extracted characteristic parameter set is compared with the preset threshold range. When the initial elastic modulus drops by more than 30% and the hardening turning point is below 16 mmHg, a high-risk level signal is output. When the hysteresis coefficient is greater than 0.25, a medium-risk level signal is output.

[0065] Through the above scheme, this application achieves dynamic coordination between pressure loading and vascular biomechanical response. By introducing closed-loop control logic for the rate of change of vascular diameter and pressure parameters, the timing and gradient amplitude of pressure loading are dynamically adjusted to ensure that the vessels are within a resolvable mechanical response range. This approach overcomes the problem of traditional fixed-interval pressure step loading causing the rate of change of vascular diameter to exceed the effective monitoring range during critical stages.

[0066] Assessment methods based on biomechanical parameters can capture early changes in vascular function and improve the accuracy of early diagnosis of diabetic retinopathy. By quantifying changes in mechanical parameters such as decreased elastic modulus and stress response hysteresis, it is possible to effectively identify reduced endothelial compliance before visible lesions develop. This early functional assessment overcomes the limitations of traditional morphological analysis methods and reduces the rate of missed detection of early lesions.

[0067] Furthermore, this application establishes a standardized pressure loading protocol and biomechanical parameter extraction method, improving the comparability of different test results. This lays the foundation for establishing standardized diagnostic reference intervals across medical institutions, helping to improve the consistency and accuracy of diabetic retinopathy diagnosis.

[0068] In some of the above-mentioned solutions of the present application, there is a problem of insufficient blood vessel tracking stability when processing dynamic response image data. When the blood vessel wall moves with a large amplitude or the image quality fluctuates, the contour boundary point recognition is easily disturbed, resulting in abnormal jumps in the blood vessel diameter data, affecting the accuracy of the pressure-diameter scatter plot.

[0069] like Figure 2 As shown, this is a flow chart for calculating real-time blood vessel diameter data. The present application further proposes a scheme for determining real-time blood vessel diameter data by processing the dynamic response image data, comprising the following steps: locating at least three characteristic blood vessel segments from the dynamic response image data, with the diameter range of the characteristic blood vessel segments set to 100 to 200 μm; determining the contour boundary points of each characteristic blood vessel segment based on the gradient change of the blood vessel wall reflection intensity; calculating the displacement vectors of the contour boundary points in adjacent frame images, and eliminating abnormal frame data whose displacement amplitude exceeds a preset motion threshold.

[0070] The selection of characteristic vessel segments was based on the arteriolar region with a diameter range of 100–200 μm, which corresponds to the tertiary branch vessels of the retina and has a relatively straight anatomy and few motion artifacts.

[0071] The detection of the gradient change of the reflection intensity of the blood vessel wall adopts the gray value difference algorithm of adjacent pixels. The blood vessel cross-sectional image is scanned along the direction perpendicular to the blood flow, and the position where the reflection intensity drops by more than 50% is determined as the vessel wall boundary.

[0072] The displacement vector is calculated by matching the coordinate difference of boundary points with the same spatial coordinates in adjacent frames, and the preset motion threshold is set to the threshold condition that the displacement of a single frame exceeds 15% of the blood vessel diameter.

[0073] Specifically, when locating characteristic vascular segments, priority is given to straight vascular segments with bifurcated structures, and capillaries and main veins are excluded through diameter screening to ensure the measurability of the mechanical response of the vascular wall. The identification of changes in reflection intensity gradients uses a dynamic threshold adjustment mechanism to automatically compensate for the impact of grayscale value fluctuations on boundary judgment when blood flow velocity changes. During the calculation of the displacement vector, the optical flow method is used to track the motion trajectory of the boundary points. When the displacement amplitude of a frame exceeds the threshold, the frame is automatically marked as abnormal data and excluded, retaining the continuity of the change in vascular diameter. This solution controls the diameter measurement error within ±3μm through parallel processing of multiple vascular segments and an abnormal data filtering mechanism, ensuring that the accuracy of the pressure-diameter scatter plot construction meets the requirements of biomechanical modeling.

[0074] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0075] Three characteristic vascular segments were located from the dynamic response image data, with diameters ranging from 100 to 200 μm. The contour boundary points of each characteristic vascular segment were determined based on the gradient of the vessel wall reflection intensity. The displacement vectors of the contour boundary points in adjacent image frames were calculated, and abnormal frames with displacement amplitudes exceeding a preset motion threshold were eliminated.

[0076] Specifically, an image analysis algorithm is first used to identify and locate three characteristic vessel segments that meet diameter requirements within the dynamic response image data. Next, an edge detection algorithm analyzes the gradient changes in the vessel wall's reflection intensity to precisely locate the contour boundary points of each characteristic vessel segment. Next, the displacement vectors of the contour boundary points in adjacent image frames are calculated, and a preset motion threshold is set. If the displacement vector amplitude in a frame exceeds this threshold, the frame is identified as an abnormal frame and removed to ensure data quality for subsequent analysis.

[0077] Through the above technical solution, this application can accurately track the dynamic changes of characteristic vascular segments, effectively remove interference data caused by factors such as micro-movement of the eyeball, and improve the accuracy and reliability of vascular diameter measurement. This provides high-quality basic data for subsequent pressure-diameter relationship analysis and biomechanical modeling, thereby improving the accuracy of early assessment of diabetic retinopathy.

[0078] In some of the above-mentioned schemes of the present application, in the process of determining real-time blood vessel diameter data by processing dynamic response image data, when the gradient change of the blood vessel wall reflection intensity causes unstable tracking of the contour boundary points, the confidence score may be reduced due to fluctuations in image quality or blood vessel motion artifacts, resulting in abnormal jumps or accumulated deviations in the blood vessel diameter data, thereby affecting the accuracy of the subsequent pressure-diameter scatter plot.

[0079] like Figure 3 FIG. 1 is a flowchart for generating anti-interference blood vessel diameter data; the present application further proposes that determining real-time blood vessel diameter data by processing the dynamic response image data also includes:

[0080] The confidence score of the contour boundary points of each characteristic vascular segment is calculated in real time. When the confidence score is lower than the preset tracking threshold, the current vascular diameter value is reconstructed based on the displacement trajectory data of the bifurcation points of adjacent vascular segments; the reconstructed data is weightedly fused with the original tracking data to generate interference-resistant vascular diameter data.

[0081] The confidence score is calculated by analyzing the positional standard deviation of the contour boundary points across five consecutive image frames, converting the standard deviation into a normalized score ranging from 0 to 1. The bifurcation point displacement trajectory data is derived by extracting the motion vector field of the bifurcation points of two adjacent vascular segments in three-dimensional space and using a polynomial interpolation algorithm to generate the predicted displacement at the current time point. During the weighted fusion process, the weight coefficients between the reconstructed and original tracking data are dynamically adjusted based on the linear attenuation function of the confidence score.

[0082] Specifically, when the confidence score is lower than the preset tracking threshold (configured to 0.6), the bifurcation point displacement trajectory reconstruction process is started. The displacement trajectory data of the bifurcation point is established by tracking the three-dimensional motion path of the bifurcation point during the pressurization process, and a trajectory model is established at intervals of 0.2 seconds. The reconstructed vascular diameter value is obtained by reversely calculating the spatial geometric relationship between the bifurcation point trajectory model and the current vascular segment to obtain the compensated diameter estimate. The reconstructed data is then fused with the original tracking data at a weight ratio of 0.3:0.7, where the weight of the reconstructed data increases as the confidence score decreases. This fusion process effectively suppresses the mutations caused by single-frame image noise while retaining the real-time response characteristics of the original data, so that the fluctuation amplitude of the final anti-interference vascular diameter data is reduced by more than 42%.

[0083] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0084] Calculate the confidence score of each characteristic vessel segment's contour boundary point in real time. The confidence score is determined by the correlation coefficient between the grayscale gradient value of the boundary point and the preset standard gradient template. When the confidence score falls below the preset tracking threshold, perform the following steps:

[0085] First, the current vessel diameter is reconstructed based on the displacement trajectory data of the bifurcation points of adjacent vessel segments. Specifically, the coordinates of the bifurcation points of adjacent vessel segments are extracted, and the displacement vector of the bifurcation points between consecutive frames is calculated. This displacement vector is then used to interpolate and estimate the boundary points of the current vessel segment to obtain the reconstructed vessel diameter value.

[0086] Next, the reconstructed data is weighted and fused with the original tracking data to generate interference-resistant vessel diameter data. The weighting coefficient is dynamically adjusted based on the confidence score; the lower the confidence score, the higher the weight given to the reconstructed data. The fused vessel diameter data retains the accuracy of the original tracking while improving robustness to interference.

[0087] Through the above technical solution, this application effectively improves the accuracy and stability of vascular diameter measurement. When the quality of the original tracking data degrades, the vessel diameter is reconstructed using bifurcation point displacement information and fused with the original data, overcoming the limitations of a single tracking method. This multi-source data fusion method significantly enhances the resistance of vascular diameter measurement to interference factors such as imaging noise and vascular motion, providing a more reliable data foundation for subsequent biomechanical analysis.

[0088] In some of the above-mentioned schemes of the present application, when constructing the stress-strain response curve model of the pressure-diameter scatter plot, the traditional fixed-interval linear fitting method cannot adapt to the nonlinear change characteristics of the vascular mechanical response in different pressure intervals, resulting in the identification error of the initial elastic modulus and the hardening turning point increasing significantly with the change of the discreteness of the scatter plot.

[0089] like Figure 4 As shown, it is a flow chart for constructing a stress-strain response curve model; the present application further proposes to construct a stress-strain response curve model for a pressure-diameter scatter plot, including: using a piecewise linear regression algorithm to fit the pressure-diameter scatter plot, and the interval division of the piecewise linear regression algorithm is dynamically adjusted according to the real-time calculated scatter plot discreteness; when the scatter plot discreteness is lower than the preset discrete threshold, a three-segment linear division is adopted; when the scatter plot discreteness is higher than the preset discrete threshold, it is switched to a nonlinear polynomial fitting mode; the initial elastic modulus is the slope of the curve of the preset pressure interval with configurable endpoints; the hardening turning point is determined by detecting that the slope change rate of adjacent intervals exceeds the preset turning threshold; the hysteresis coefficient is obtained by calculating the average slope difference between the pressurization path and the decompression path.

[0090] The interval division of the piecewise linear regression algorithm is based on the dynamic adjustment of the discreteness of the scatter plot, and the discreteness is obtained by calculating the ratio of the standard deviation of adjacent data points to the mean; the three-segment linear division divides the pressure interval into three continuous segments: low, medium, and high, and the boundaries of each segment are automatically optimized by minimizing the overall fitting residual; the nonlinear polynomial fitting mode uses a quadratic or cubic polynomial function, and the fitting order increases according to the preset rules based on the discreteness value; the endpoint configuration range of the initial elastic modulus is set to the first 30% to 50% of the pressure step value, and the specific endpoint position is automatically calibrated according to the fitting residual; the preset turning threshold of the hardening turning point is set to twice the standard deviation of the difference in slopes of adjacent intervals; in the calculation of the hysteresis coefficient, the average slopes of the pressurization path and the decompression path are fitted separately by the least squares method.

[0091] Specifically, when the scatter plot dispersion falls below a preset threshold, a three-segment linear partitioning scheme is employed. A sliding window algorithm traverses all possible interval split points and selects the partitioning scheme with the smallest residual sum of squares, achieving precise segmentation of low-discrete data. When the dispersion exceeds the preset threshold, a quadratic polynomial fitting is employed, utilizing a nonlinear model to capture the viscoelastic deformation characteristics of blood vessels under high pressure. The endpoint configuration range of the initial elastic modulus is dynamically adjusted based on the trend of vessel diameter changes. For example, when a vessel exhibits rapid expansion during the initial pressure phase, the endpoint is automatically shifted back to avoid encompassing the nonlinear response region. To detect the stiffening turning point, the slope difference between adjacent intervals is calculated by calculating the slope difference of two regression lines and verifying its statistical significance with confidence intervals. The hysteresis coefficient is further calculated by incorporating a path integral method, quantifying the slope difference between pressurization and decompression as the hysteresis of vessel deformation per unit pressure change. By dynamically adjusting the fitting model and parameter extraction rules, the system can adapt to the biomechanical characteristics of different vessel segments, significantly improving the sensitivity of the stiffening turning point and hysteresis coefficient detection.

[0092] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0093] The stress-strain response curve model for the pressure-diameter scatter plot is constructed using a piecewise linear regression algorithm. The algorithm's interval divisions are dynamically adjusted based on the real-time calculated scatter plot dispersion. When the scatter plot dispersion falls below a preset dispersion threshold, a three-segment linear partitioning scheme is used. When the scatter plot dispersion exceeds the preset dispersion threshold, a nonlinear polynomial fitting mode is used.

[0094] The initial elastic modulus is calculated by calculating the slope of the curve within a pre-set pressure range with configurable endpoints. The hardening turning point is determined by detecting when the rate of change in the slope of adjacent intervals exceeds a pre-set turning threshold. The hysteresis coefficient is calculated by calculating the average slope difference between the pressurization path and the decompression path.

[0095] In specific implementations, the preset discrete threshold can be set to 0.05 or 0.15. The preset pressure range can be set to 10-20 mmHg. The preset turning threshold can be set to 30%. The average slope difference threshold between the pressurization and decompression paths can be set to 0.1.

[0096] Through the above technical solution, this application achieves adaptive curve fitting for pressure-diameter scatter plots. Based on the actual degree of dispersion of the scatter plot, the appropriate fitting method is dynamically selected, improving the accuracy and robustness of the model. By extracting key characteristic parameters such as the initial elastic modulus, the hardening turning point, and the hysteresis coefficient, a comprehensive assessment of the mechanical properties of the vascular wall can be achieved, providing more reliable quantitative indicators for the early diagnosis of diabetic retinopathy.

[0097] In some of the above-mentioned schemes of this application, when a piecewise linear regression algorithm is used to fit the pressure-diameter scatter plot, when the sum of squared residuals in the initial fitting interval exceeds the preset error range, it may cause the curve model to mismatch the actual biomechanical response, thereby affecting the accurate extraction of characteristic parameters such as the initial elastic modulus.

[0098] like Figure 5 As shown in FIG, it is a diagram of the execution process of the piecewise linear regression algorithm; the present application further proposes that the execution process of the piecewise linear regression algorithm includes:

[0099] Calculate the residual sum of squares of the initial fitting interval; when the residual sum of squares exceeds the preset fitting error threshold, automatically reduce the fitting interval range and recalculate the slope; update the reduced interval range as the new preset pressure interval to the characteristic parameter group.

[0100] The residual sum of squares is calculated by comparing the sum of the squares of the vertical distances between the actual scattered points and the fitted line. The preset fitting error threshold is set to 15% to 20% of the total dispersion of the interval based on the statistical distribution of historical data. The automatic reduction of the fitting interval range uses an iterative algorithm that gradually eliminates endpoint data, with each reduction step configured to 5% to 10% of the current interval length. The least squares method is used to update the linear equation parameters when recalculating the slope, and the final confirmed effective pressure interval coverage range is synchronized with the preset pressure interval variable in the characteristic parameter group.

[0101] Specifically, in the linear regression process of the pressure-diameter scatter plot, the sum of squared residuals between each data point and the fitted straight line is first calculated within the original pressure range. When the value exceeds the preset threshold, the system automatically removes edge data points from both ends of the current interval according to the preset step size to form a new pressure range. Re-execute linear fitting within the reduced interval until the sum of squared residuals meets the error requirements. The optimized pressure interval is used as the new effective calculation range to replace the original preset interval for feature parameter extraction. This process eliminates the interference of abnormal pressure points on the curve model by dynamically adjusting the data fitting range, thereby ensuring the measurement accuracy of parameters such as the initial elastic modulus. For example, when there is abnormal fluctuation data in the blood vessel diameter within the initial 20 mmHg pressure interval, the system can automatically reduce the effective interval to the range of 16-22 mmHg, thereby improving the accuracy of sclerosis turning point detection.

[0102] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0103] The execution process of the piecewise linear regression algorithm includes the following steps:

[0104] First, calculate the residual sum of squares for the initial fitting interval. Specifically, select an initial interval in the pressure-diameter scatter plot, such as the pressure range of 10 mmHg to 15 mmHg, and perform a linear regression fit on the data points within this interval to obtain the equation of the fitted line. Then, calculate the square of the vertical distance from each data point to the fitted line and add all the squared values ​​to obtain the residual sum of squares.

[0105] Next, determine whether the residual sum of squares exceeds a preset fitting error threshold. For example, the preset fitting error threshold can be set to 0.05. If the calculated residual sum of squares exceeds 0.05, proceed to the next step; otherwise, maintain the current fitting interval.

[0106] When the residual sum of squares exceeds the preset fitting error threshold, the fitting interval is automatically reduced and the slope is recalculated. Specifically, the fitting interval can be reduced by 0.5 mmHg, that is, adjusted to 10.5 mmHg to 15 mmHg. The linear regression fit is re-performed within the new interval and a new slope value is calculated.

[0107] Finally, the reduced range is updated as the new preset pressure range to the characteristic parameter group. For example, the pressure range of 10.5 mmHg to 15 mmHg and the corresponding slope value are updated to the characteristic parameter group for subsequent elastic modulus calculation and hardening turning point determination.

[0108] Through the above technical solution, the present application can improve the fitting accuracy of the pressure-diameter scatter plot. By dynamically adjusting the fitting interval, the changing trend of vascular elastic properties can be more accurately captured, reducing the impact of abnormal data points on the fitting results. At the same time, the adaptive interval adjustment mechanism can adapt to the differences in vascular characteristics of different patients, improving the universality and reliability of the assessment method. In addition, the updated preset pressure interval provides a more accurate data basis for subsequent feature parameter extraction, which helps to improve the accuracy of diabetic retinopathy risk assessment.

[0109] In some of the above-mentioned schemes of the present application, the parameter setting of the pressure regulation signal sequence lacks clear guidance, which may cause the pressure change rate to exceed the range of the physiological response capacity of the blood vessels, resulting in excessive stretching of the blood vessel wall or insufficient pressure step response during the dynamic imaging data acquisition process, affecting the construction accuracy of the stress-strain model.

[0110] The present application further proposes that the pressure step value range of the stepped pressure regulation signal sequence is set to 10 mmHg to 25 mmHg, the interval time between adjacent pressure step values ​​is configured to be 3 to 5 seconds, and the pressure change rate is limited to less than 0.5 mmHg / second.

[0111] The pressure step value range is limited to 10 mmHg to 25 mmHg, which covers the normal physiological pressure fluctuation range of the retinal artery while preventing excessive pressure from causing vascular spasm. The interval between adjacent pressure steps is set to 3 to 5 seconds to provide a pressure adaptation period for the smooth muscle of the vascular wall and ensure quasi-static measurement conditions for changes in vascular diameter. The pressure change rate is strictly limited to below 0.5 mmHg / second to prevent sudden pressure changes from causing inertial deformation of the vascular wall and interfering with the calculation of the elastic modulus.

[0112] Specifically, the pressure step value range matches the critical collapse pressure of the retinal microvessels. When the pressure is lower than 10 mmHg, it cannot effectively stimulate the elastic deformation of the blood vessel wall, while exceeding 25 mmHg may cause endothelial damage. When the interval time is configured to 3 seconds, the stable plateau period of blood vessel diameter change can be captured; when the blood vessel diameter fluctuates greatly, the interval time is extended to 5 seconds to improve measurement stability. The pressure change rate limit is achieved through the stepper motor of the pressure controller, and the PID algorithm is used to control the slope of the pressure gradient rise so that the pressure regulation curve has a smooth slope shape. This parameter combination ensures that within a single pressure loading cycle, sufficient blood vessel diameter change data points can be collected and the biomechanical response of the blood vessel wall can be maintained in the linear range, providing high-quality raw data for subsequent feature parameter extraction.

[0113] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0114] The pressure step value range of the stepped pressure regulation signal sequence is set to 10 mmHg to 25 mmHg, the interval time between adjacent pressure step values ​​is configured to be 3 to 5 seconds, and the pressure change rate is limited to less than 0.5 mmHg / second. Specifically, the pressure regulation signal sequence can include multiple pressure step values, such as 10 mmHg, 15 mmHg, 20 mmHg and 25 mmHg. Each pressure step lasts for 3 to 5 seconds, during which the pressure remains constant. The pressure change rate between adjacent pressure steps is controlled to below 0.5 mmHg / second to ensure a smooth transition. In this way, precisely controlled pressure stimulation can be applied to the target blood vessel segment while avoiding potential damage to the blood vessel caused by excessive pressure changes.

[0115] Through the above technical solution, the present application realizes precise pressure stimulation of fundus blood vessels. The design of the stepped pressure regulation signal sequence ensures the smoothness and controllability of pressure changes, and avoids the potential damage to blood vessels caused by sudden pressure changes. At the same time, the reasonable pressure range and interval time setting allow the blood vessels to have enough time to fully respond to each pressure step, thereby improving the accuracy of subsequent vascular dynamic response analysis. The limitation of the pressure change rate further enhances the safety and comfort of the measurement process. This precise and controllable pressure stimulation scheme lays the foundation for the subsequent vascular elasticity assessment and helps to improve the sensitivity of early diagnosis of diabetic retinopathy.

[0116] In some of the above-mentioned schemes of this application, during the risk assessment of diabetic retinopathy, due to the lack of clear characteristic parameter threshold judgment standards and risk level mapping rules, it is impossible to accurately distinguish different stages of the disease in clinical applications, and the assessment results lack operational clinical guidance value.

[0117] like Figure 6 As shown, it is a risk level signal triggering flow chart; this application further proposes that the preset threshold range of the characteristic parameter group includes:

[0118] The first abnormal sign is triggered when the baseline value of the initial elastic modulus drops by more than 30%, the second abnormal sign is triggered when the hardening turning point is lower than 16 mmHg, and the third abnormal sign is triggered when the hysteresis coefficient is greater than 0.25.

[0119] Generating a diabetic retinopathy risk level signal includes:

[0120] When the first abnormal flag and the second abnormal flag are triggered at the same time, a high-risk level signal is output; when only the third abnormal flag is triggered, a medium-risk level signal is output; and the risk level signal is matched with a preset clinical treatment recommendation library to generate an evaluation report including a review cycle recommendation.

[0121] Among them, the threshold value for a decrease of more than 30% in the initial elastic modulus baseline value is set based on the critical value of vascular wall collagen fiber degradation, the hardening turning point is lower than 16 mmHg, which corresponds to the critical pressure at which the capillary self-regulation function fails, and the hysteresis coefficient is greater than 0.25, which reflects the quantitative indicator of vascular endothelial cell dysfunction. The risk level signal is generated using a multi-marker combined triggering mechanism. The combination of the first abnormal mark and the second abnormal mark corresponds to the superposition state of microvascular structural damage and functional decompensation, and the independent triggering of the third abnormal mark indicates early hemodynamic abnormalities. It is recommended that the review cycle be dynamically adjusted according to the risk level. The high risk level corresponds to a review interval of 3 months, and the medium risk level corresponds to a review interval of 6 months.

[0122] Specifically, when the initial elastic modulus drops by more than 30%, the cross-linked structure of the collagen in the vascular wall is destroyed, resulting in a significant decrease in the ability to resist deformation, triggering the first abnormal sign. When the sclerosis turning point is lower than 16 mmHg, the vascular smooth muscle cells undergo irreversible contraction in the low-pressure stage, triggering the second abnormal sign. When the hysteresis coefficient is greater than 0.25, the energy dissipation during vasodilation increases abnormally, triggering the third abnormal sign. When the first and second abnormal signs are triggered at the same time, it indicates that vascular structural damage and functional decompensation coexist, and a high-risk level signal is output at this time. When only the third abnormal sign is triggered, it indicates that the endothelium-dependent vasodilation function of the blood vessels is impaired, and a medium-risk level signal is output. The preset clinical treatment recommendation library associates the high-risk level with a 3-month optical coherence tomography review requirement and the medium-risk level with a 6-month review requirement, and clearly marks the review time node through the evaluation report.

[0123] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0124] The preset threshold ranges of the characteristic parameter group include: triggering the first abnormal sign when the baseline value of the initial elastic modulus drops by more than 30%; triggering the second abnormal sign when the hardening turning point is lower than 16 mmHg; triggering the third abnormal sign when the hysteresis coefficient is greater than 0.25; generating a diabetic retinopathy risk level signal includes: when the first abnormal sign and the second abnormal sign are triggered at the same time, outputting a high-risk level signal; when only the third abnormal sign is triggered, outputting a medium-risk level signal; matching the risk level signal with a preset clinical treatment recommendation library to generate an assessment report including review cycle recommendations.

[0125] Specifically, the initial elastic modulus baseline value can be set to 1.5 MPa / mmHg. A drop to 1.05 MPa / mmHg or lower triggers the first abnormality flag. The hardening turning point threshold is set to 16 mmHg; a value below this threshold triggers the second abnormality flag. The hysteresis coefficient threshold is set to 0.25; exceeding this value triggers the third abnormality flag.

[0126] The risk level determination can be further refined as follows: a high-risk signal is output when the first and second abnormality flags are triggered simultaneously; a medium-risk signal is output when only the third abnormality flag is triggered; and a low-risk signal is output when either the first or second abnormality flag is triggered.

[0127] The pre-set clinical management recommendation library can include review interval recommendations for different risk levels, such as review within one month for high-risk patients, within three months for medium-risk patients, and within six months for low-risk patients. The assessment report will include the risk level determination and the corresponding review interval recommendations.

[0128] Through the above technical solution, this application can comprehensively assess the risk of diabetic retinopathy based on multiple biomechanical characteristic parameters, improving the accuracy and reliability of the assessment. By setting multiple risk levels and corresponding clinical treatment recommendations, it provides a scientific basis for doctors to formulate personalized follow-up plans, which helps to promptly detect changes in the condition and take intervention measures. In addition, the assessment report containing the review cycle recommendation can standardize the follow-up process of patients with diabetic retinopathy and improve the efficiency and compliance of follow-up.

[0129] In some of the above-mentioned schemes of the present application, when calculating real-time vascular diameter data based on dynamic response imaging data, non-physiological deviations may exist in the data between multiple feature vascular segments. This deviation comes from micro-movements of the eye or image acquisition noise, resulting in instantaneous mutations in the vascular diameter data, thereby affecting the accuracy of the subsequent pressure-diameter scatter plot construction.

[0130] like Figure 7 As shown, this is a flow chart of the noise filtering step; the present application further proposes that after determining the real-time vascular diameter data by processing the dynamic response image data, a noise filtering step is also included: calculating the correlation coefficient between the real-time vascular diameter data of each characteristic vascular segment; when the correlation coefficient is lower than a preset consistency threshold, the vascular diameter data at the corresponding time point is eliminated.

[0131] The noise filtering step is implemented through a collaborative validation mechanism for multi-vessel segment data. Correlation coefficients are calculated using the Pearson linear correlation coefficient algorithm, with a preset consistency threshold set between 0.85 and 0.95. Only transient outlier data points are removed, retaining the preceding and succeeding consecutive valid data segments. This step, along with the pre-stage pressure step control and vessel tracking algorithm, forms a closed-loop data verification loop. After outlier data is removed, the system automatically triggers a re-acquisition command for the corresponding pressure value at that time.

[0132] Specifically, after obtaining the diameter data of multiple characteristic vascular segments, the system synchronously calculates the correlation index of the data of each vascular segment. When the data of a single vascular segment deviates significantly from the trend of other vascular segments at a certain time point, the correlation coefficient calculation result will be lower than the preset threshold, indicating that there is a motion artifact or tracking failure at that time point. At this time, the system automatically marks the time point as invalid data and notifies the pressure control module to supplement the missing data in the next pressure step maintenance phase. This mechanism effectively eliminates the systematic deviation caused by local vascular segment tracking errors, ensures that the data input into the biomechanical model is consistent in time and space, and thus improves the calculation accuracy of the hardening turning point and hysteresis coefficient.

[0133] As a preferred embodiment, the solution of the present application is specifically implemented as follows: After acquiring real-time vessel diameter data for three characteristic vessel segments, the Pearson correlation coefficient is simultaneously calculated for each 0.2-second time window. When the correlation coefficient for the diameter data of the superior tributary vein, nasal arteriole, and macular venule at a given time point is less than 0.85, the vessel diameter data at that time point is determined to have measurement interference. In this case, the three sets of diameter data at that time point are marked as low-confidence data, and data are removed before constructing the pressure-diameter scatter plot. Only time series data with correlation coefficients consistently above 0.85 are retained for subsequent analysis.

[0134] Through the above technical solution, the present application effectively solves the problem of local abnormal fluctuations in vascular diameter data caused by micro-movements of the eye or equipment noise. Through the multi-vascular segment data correlation verification mechanism, the measurement deviation caused by single-point tracking error can be eliminated, thereby improving the data consistency of the pressure-diameter scatter plot, providing high signal-to-noise ratio input data for the construction of the stress-strain model, and reducing the probability of misjudgment of lesion risk assessment due to data noise.

[0135] like Figure 8 As shown, it is a flow chart for generating a disease progress report; this application further proposes that it also includes:

[0136] Mapping the characteristic parameters of each characteristic blood vessel segment to the coordinates of the retinal anatomical position to generate a three-dimensional elastic distribution map;

[0137] The elastic modulus decay rate of each characteristic blood vessel segment is calculated based on historical detection data, and when the decay rate exceeds a preset rate threshold, a disease progression warning signal is triggered.

[0138] As a preferred embodiment, the solution of the present application is specifically implemented as follows: a retinal vascular coordinate mapping engine is configured in an ophthalmic image processing server. The engine receives vascular segment elastic modulus data and its corresponding retinal position code output by an optical coherence tomography device. Specifically, the anatomical coordinates of characteristic vascular segments at the fovea, superior temporal arterial arch, and optic disc edge are obtained from a retinal vascular topology database, and a mapping relationship is established between the elastic modulus values ​​and three-dimensional spatial coordinates. A voxel interpolation algorithm is used to spatially fill the discrete vascular segment data, generating a three-dimensional elastic modulus heat map covering the posterior pole of the retina, where different color gradients represent the elastic modulus distribution within the range of 10-50 kPa.

[0139] The biomechanical analysis module accesses elastic modulus test records from a historical database and calculates the modulus attenuation of the target vascular segment on a semi-annual basis. If the elastic modulus of the inferior temporal vein branches decreases by more than 15 kPa / year between two consecutive tests, the system automatically compares the attenuation to a preset lesion progression threshold parameter table. If the attenuation rate exceeds the threshold critical value for the temporal vascular zone, a warning signal for retinal microcirculatory decompensation is triggered, and a disease progression report, including a recommended time window for photocoagulation treatment, is delivered to the medical imaging workstation.

[0140] Through the above technical solution, this application achieves visual monitoring of the spatial heterogeneity of retinal vascular biomechanical properties, enabling accurate identification of early elasticity decline trends in specific anatomical regions. By combining time series analysis with historical data, it effectively distinguishes physiological elastic fluctuations from pathological compensatory failure, providing clinically actionable criteria for determining the timing of disease interventions, and significantly improving the spatiotemporal resolution of early warning of diabetic retinopathy progression.

[0141] like Figure 9 As shown, the present application further proposes an evaluation system for diabetic retinopathy, including a pressure regulation and image acquisition module, a vascular dynamic tracking module, a biomechanical modeling module, a dynamic feedback decision module, and a risk assessment and output module.

[0142] Among them, the pressure control and image acquisition module is configured to control the intraocular pressure changes according to the step-by-step pressure adjustment signal sequence, and synchronously trigger the optical coherence tomography device to collect vascular images during the pressure step stabilization stage.

[0143] The vascular dynamic tracking module locates characteristic vascular segments with a diameter of 100-200 μm by analyzing the vascular wall reflection intensity gradient in the dynamic image sequence, calculates the displacement vector of the contour boundary points between adjacent frames, and uses a weighted fusion algorithm to generate interference-resistant vascular diameter data.

[0144] The biomechanical modeling module uses a piecewise linear regression algorithm to fit the pressure-diameter scatter plot, automatically switches the fitting mode according to the dispersion of the scatter points, and extracts a characteristic parameter group including the initial elastic modulus, hardening turning point, and hysteresis coefficient.

[0145] The dynamic feedback decision module monitors the rate of change of blood vessel diameter in real time, and dynamically adjusts the pressure step interval time or gradient amplitude when the rate of change exceeds a threshold. For example, when the rate of change exceeds 0.15 μm / s, the interval time is extended to 8 seconds.

[0146] The risk assessment and output module compares the characteristic parameters with the preset thresholds, generates a high-risk signal when the initial elastic modulus drops by more than 30% and the hardening turning point is lower than 16 mmHg, and matches the clinical treatment recommendation library to output an assessment report including a review period.

[0147] Specifically, during system operation, the pressure control module generates a stepped pressure signal of 10-25 mmHg, with each step maintained for 3-5 seconds. The optical coherence tomography device triggers image acquisition after pressure stabilizes, acquiring dynamic response data for the vascular segments. The vascular tracking module eliminates abnormal frames with displacement amplitudes exceeding 5 μm, calculates the average diameter of three characteristic vascular segments, and generates a pressure-diameter scatter plot. The biomechanical modeling module selects a fitting algorithm based on the scatter plot's dispersion. When the dispersion is less than 0.02, a three-segment linear partitioning is used. The slope value is selected within the range of 10-15 mmHg when calculating the initial elastic modulus. When the diameter change rate falls below 0.08 μm / s, the dynamic feedback module increases the pressure gradient from 2 mmHg to 4 mmHg to enhance stimulation intensity. The risk assessment module integrates the elastic modulus decay rate with anatomical location coordinates to generate a three-dimensional elastic distribution map. A multi-level warning signal is triggered when more than two characteristic parameters are abnormal. Each module achieves closed-loop control through real-time data stream interaction. For example, after the pressure regulation module receives dynamic feedback instructions to adjust parameters, the biomechanical modeling module synchronously updates the characteristic parameter threshold range, thereby improving the detection sensitivity of early microvascular functional damage.

[0148] As a preferred embodiment, the solution of the present application is specifically implemented as follows: the pressure control and image acquisition module is configured to control the intraocular pressure device to generate a stepped pressure signal sequence, which includes pressure values ​​in the range of 10 mmHg to 25 mmHg. The interval time between adjacent pressure steps is initially set to 4 seconds, and the pressure change rate is limited to 0.4 mmHg / second. Under the triggering of this module, the optical coherence tomography device collects dynamic images of the superior temporal retinal artery branches at a rate of 12 frames per second, and the image data is transmitted to the vascular dynamic tracking module in real time via Gigabit Ethernet. The vascular dynamic tracking module uses an adaptive boundary detection algorithm to locate a vascular segment with a diameter of approximately 150 μm in continuous image frames, and generates a diameter change curve by calculating the displacement of the vascular wall reflection peak in adjacent frames. The curve is time-synchronized with the corresponding pressure step value to generate a pressure-diameter scatter plot containing 23 data points. The biomechanical modeling module uses an automatic segmented fitting algorithm to process scatter plot data, where the initial elastic modulus is obtained by calculating the linear regression slope in the 10-15 mmHg range. When the slope change rate in the 15-18 mmHg range exceeds 20%, it is determined that a hardening turning point exists. The dynamic feedback decision module monitors the diameter change rate in real time. When the change rate exceeds 0.8 μm / s, the next pressure step interval is automatically extended to 7 seconds. When the change rate is lower than 0.3 μm / s, the pressure gradient amplitude is increased to 2 mmHg. The risk assessment and output module compares the calculated elastic modulus of 0.52 kPa, the hardening turning point of 14.8 mmHg, and the hysteresis coefficient of 0.28 with the preset thresholds, generates a high-risk level signal, and outputs an assessment report containing a three-month review recommendation.

[0149] Through the above-mentioned technical solution, this application achieves dynamic quantitative assessment of vascular biomechanical properties, ensures detection safety through real-time feedback regulation of the pressure loading process, and effectively identifies early-stage vascular dysfunction using a multi-parameter combined criterion. This solution addresses the problem of traditional morphological testing being insensitive to early functional changes such as vascular wall elasticity loss and stress response hysteresis, and can provide early warning of vascular lesion risks before vascular structural damage occurs.

[0150] The technical scope of the present invention is not limited to the contents of the above description. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of ​​the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.

Claims

1. A method for evaluating diabetic retinopathy, characterized by: The steps include: Acquire a stepped pressure adjustment signal sequence generated by an intraocular pressure control device, wherein the pressure adjustment signal sequence includes at least three pressure step values ​​of a predetermined number of digits; triggering an optical coherence tomography device based on the pressure adjustment signal sequence to obtain dynamic response image data of the target blood vessel segment under the corresponding pressure step; Determining real-time blood vessel diameter data by processing the dynamic response image data, and generating a pressure-diameter scatter plot by associating corresponding pressure step values; Real-time calculation of the rate of change of blood vessel diameter in dynamic response imaging data: When the blood vessel diameter change rate exceeds the first feedback threshold, the interval time of the next pressure step is extended to the preset safety time; When the blood vessel diameter change rate is lower than the second feedback threshold, increasing the gradient amplitude of the pressure step value to a preset maximum change rate; Constructing a stress-strain response curve model of a pressure-diameter scatter plot, and extracting a predefined characteristic parameter group in the curve model, wherein the characteristic parameter group includes at least an initial elastic modulus, a hardening turning point, and a hysteresis coefficient; The characteristic parameter group is compared with a preset threshold range, and when at least two characteristic parameters exceed the corresponding threshold range, a diabetic retinopathy risk level signal is generated.

2. The method for evaluating diabetic retinopathy according to claim 1, wherein: Determining the real-time blood vessel diameter data by processing the dynamic response image data includes: Locating at least three characteristic blood vessel segments from the dynamic response image data, wherein the diameter range of the characteristic blood vessel segments is set to be 100-200 μm; Determine the contour boundary points of each characteristic blood vessel segment based on the gradient change of the blood vessel wall reflection intensity; Calculate the displacement vectors of the contour boundary points in adjacent frame images and eliminate abnormal frame data whose displacement amplitude exceeds the preset motion threshold.

3. The method for evaluating diabetic retinopathy according to claim 2, wherein: The determining of real-time blood vessel diameter data by processing the dynamic response image data further comprises: The confidence score of the contour boundary points of each characteristic blood vessel segment is calculated in real time. When the confidence score is lower than the preset tracking threshold: Reconstruct the current blood vessel diameter value based on the bifurcation point displacement trajectory data of adjacent blood vessel segments; The reconstructed data is weightedly fused with the original tracking data to generate interference-resistant vascular diameter data.

4. The method for evaluating diabetic retinopathy according to claim 1, wherein: The stress-strain response curve model for constructing a pressure-diameter scatter plot includes: The pressure-diameter scatter plot is fitted using a piecewise linear regression algorithm, where the interval division of the piecewise linear regression algorithm is dynamically adjusted according to the discreteness of the scatter plot calculated in real time: When the dispersion of the scatter plot is lower than the preset discrete threshold, a three-segment linear division is adopted; When the dispersion of the scatter plot is higher than the preset discrete threshold, it switches to the nonlinear polynomial fitting mode; in: The initial elastic modulus is the slope of the curve in the preset pressure range of the configurable endpoints; The hardening turning point is determined by detecting that the slope change rate of adjacent intervals exceeds a preset turning threshold; The hysteresis coefficient is obtained by calculating the difference in average slopes between the pressurization path and the decompression path.

5. The method for evaluating diabetic retinopathy according to claim 4, wherein: The execution process of the piecewise linear regression algorithm includes: Calculate the residual sum of squares for the initial fitting interval; When the residual sum of squares exceeds the preset fitting error threshold, the fitting interval range is automatically reduced and the slope is recalculated; The reduced interval range is updated to the characteristic parameter group as a new preset pressure interval.

6. The method for evaluating diabetic retinopathy according to claim 1, wherein: The pressure step value range of the stepped pressure regulation signal sequence is set to 10 mmHg to 25 mmHg, the interval time between adjacent pressure step values ​​is configured to be 3 to 5 seconds, and the pressure change rate is limited to less than 0.5 mmHg / second.

7. The method for evaluating diabetic retinopathy according to claim 1, wherein: The preset threshold range of the characteristic parameter group includes: When the baseline value of the initial elastic modulus drops by more than 30%, a first abnormality flag is triggered; When the hardening turning point is lower than 16 mmHg, a second abnormality flag is triggered; When the hysteresis coefficient is greater than 0.25, a third abnormal flag is triggered; Generating a diabetic retinopathy risk level signal comprises: When the first abnormality flag and the second abnormality flag are triggered at the same time, a high risk level signal is output; When only the third abnormal flag is triggered, a medium risk level signal is output; The risk level signal is matched with the preset clinical treatment recommendation library to generate an assessment report including review cycle recommendations.

8. The method for evaluating diabetic retinopathy according to claim 1, wherein: After determining the real-time blood vessel diameter data by processing the dynamic response image data, the method further includes a noise filtering step: Calculate the correlation coefficient between the real-time vascular diameter data of each characteristic vascular segment; When the correlation coefficient is lower than the preset consistency threshold, the vascular diameter data at the corresponding time point are eliminated.

9. The method for evaluating diabetic retinopathy according to claim 1, wherein: Also includes: Mapping the characteristic parameters of each characteristic blood vessel segment to the coordinates of the retinal anatomical position to generate a three-dimensional elastic distribution map; The elastic modulus decay rate of each characteristic blood vessel segment is calculated based on historical detection data, and when the decay rate exceeds a preset rate threshold, a disease progression warning signal is triggered.

10. A system for evaluating diabetic retinopathy, characterized by: The method for evaluating diabetic retinopathy according to any one of claims 1 to 9 is implemented, comprising: a pressure control and image acquisition module, configured to obtain a stepped pressure regulation signal sequence generated by the intraocular pressure control device, and trigger an optical coherence tomography device based on the pressure regulation signal sequence to obtain dynamic response image data of the target blood vessel segment under corresponding pressure steps; a blood vessel dynamic tracking module, configured to determine real-time blood vessel diameter data by processing the dynamic response image data, and generate a pressure-diameter scatter plot by associating the corresponding pressure step values; A biomechanical modeling module is used to construct a stress-strain response curve model of a pressure-diameter scatter plot and extract a predefined characteristic parameter group in the curve model; Dynamic feedback decision module, used to calculate the rate of change of blood vessel diameter in dynamic response imaging data in real time: When the blood vessel diameter change rate exceeds the first feedback threshold, the interval time of the next pressure step is extended to the preset safety time; When the blood vessel diameter change rate is lower than the second feedback threshold, increasing the gradient amplitude of the pressure step value to a preset maximum change rate; The risk assessment and output module is used to compare the characteristic parameter group with a preset threshold range, and generate a diabetic retinopathy risk level signal when at least two characteristic parameters exceed the corresponding threshold range.

Images