Multi-threshold state light dynamic image rendering method for heart failure signs time series

Abstract

This invention relates to the field of image generation technology and discloses a method for rendering dynamic images of multi-threshold status lights based on the temporal sequence of heart failure vital signs. The method includes the following steps: Step S101, calculating the oxygen saturation recovery time constant after standing at dawn; Step S102, constructing a multi-signal input vector; Step S103, solving for the probability distribution of multi-threshold color bands; Step S104, calculating and obtaining smoothed color and smoothed brightness; Step S105, generating brightness modulation and radius modulation signals; and Step S106, generating a dynamic image of multi-threshold status lights. This invention uses the oxygen saturation recovery time constant after standing at dawn as the main anchor point, integrating key vital signs such as body weight, heart rate, and hemodynamics. Through monotonic integral embedding, contextual embedding, and convex optimization algorithms, it transforms complex temporal vital signs into intuitive dynamic images of multi-threshold status lights, providing reliable assistance for long-term monitoring of heart failure, thereby improving the timeliness of disease warnings and the efficiency of doctor-patient communication.

Description

Technical Field

[0001] This invention relates to the field of image generation technology, and more specifically, to a method for rendering dynamic images of multi-threshold state lights oriented towards the temporal sequence of heart failure symptoms. Background Technology

[0002] As a highly prevalent chronic cardiovascular disease, heart failure requires long-term, accurate monitoring of vital signs to reduce the risk of disease progression. Clinical studies have shown that the temporal characteristics of blood oxygen saturation recovery trajectory after standing upright in the morning, weight fluctuations, nighttime heart rate, and hemodynamic parameters can directly reflect fluid retention, sympathetic nerve load, and circulatory function stability in patients, serving as a core basis for early warning of disease fluctuations.

[0003] Current heart failure vital sign monitoring technologies face numerous unresolved issues. Traditional monitoring methods mostly present vital sign data in discrete numerical form, lacking intuitive visualization. This makes it difficult for patients and non-professionals to quickly interpret the underlying disease trends, and healthcare professionals in multi-patient management scenarios must spend considerable time integrating and analyzing the data, resulting in low efficiency. While some visualization methods attempt to indicate risk through color, the use of a single fixed threshold for color gradation fails to reflect the continuous transition characteristics of physiological states. Small fluctuations in vital sign values ​​can cause visual jumps, which do not match the actual gradual changes in heart failure patients' conditions and may lead to misdiagnosis.

[0004] Furthermore, existing technologies fail to establish a correlation between physiological timescales and visual presentation. They neither dynamically adjust the monitoring window duration based on core vital signs such as blood oxygen saturation recovery time, nor adapt to visual smoothness. This results in key physiological signals either being obscured by redundant information or missed due to excessively short windows. Simultaneously, the integration of multiple vital signs suffers from weight ambiguity, failing to highlight the dominant role of core vital signs, or reducing monitoring accuracy due to conflicting information from multiple indicators, making it impossible to accurately capture the core driving factors of disease progression. Summary of the Invention

[0005] This invention provides a multi-threshold state light dynamic image rendering method for heart failure signs time series, solving the technical problems in the background art mentioned above.

[0006] This invention provides a multi-threshold state light dynamic image rendering method for heart failure signs time series, including the following steps:

[0007] Step S101: Calculate the oxygen saturation recovery time constant after standing upright at dawn based on the blood oxygen saturation regression trajectory after standing upright in the morning.

[0008] Step S102: The oxygen saturation recovery time constant after standing upright at dawn is used as the main anchor point and combined with body weight, heart rate and hemodynamic components in a fixed order to construct a multi-sign input vector;

[0009] Step S103: Perform monotonic integral embedding and context embedding on the multi-signature input vector, solve the multi-threshold color band probability distribution by linear superposition and a differentiable convex allocation layer that satisfies the constraints, and generate the brightness attenuation constant, animation pulse frequency and halo radius ratio based on the distribution.

[0010] Step S104: The primary color is synthesized by weighting according to the probability distribution of the multi-threshold color band, and the time smoothing coefficient is determined by the oxygen saturation recovery time constant after the dawn upright. The primary color and the target brightness determined by the probability distribution of the multi-threshold color band are exponentially smoothed to obtain smooth color and smooth brightness.

[0011] Step S105: Set the duration of the morning window according to the oxygen saturation recovery time constant after standing upright at dawn, and generate a brightness modulation signal and a radius modulation signal within the window according to the ratio of animation pulse frequency to halo radius.

[0012] Step S106: Based on the smooth color, smooth brightness, and brightness attenuation constant, the main light and the halo are spatially synthesized by combining the brightness modulation signal and the radius modulation signal, and a multi-threshold state light dynamic image is output.

[0013] The beneficial effects of this invention are as follows: Using the oxygen saturation recovery time constant after standing upright at dawn as the main anchor point, this invention integrates key vital signs such as weight, heart rate, and hemodynamics. Through monotonic integral embedding, contextual embedding, and convex optimization algorithms, it transforms complex temporal vital signs into intuitive multi-threshold dynamic images. This achieves continuous transitions in risk levels, avoiding visual jumps. The temporal smoothing coefficient is linked to physiological recovery characteristics, ensuring that visual feedback aligns with human physiological patterns. The dynamic images clearly present the trend of disease progression. This invention is suitable for various scenarios, including home self-management and clinical inpatient care. It requires no complex operations, balancing practicality and convenience, providing reliable assistance for long-term heart failure monitoring, thereby improving the timeliness of disease warnings and the efficiency of doctor-patient communication. Attached Figure Description

[0014] Figure 1 This is a flowchart of the multi-threshold state lamp dynamic image rendering method for heart failure signs time sequence according to the present invention;

[0015] Figure 2 This is the multi-threshold color band probability distribution heatmap of the present invention. Detailed Implementation

[0016] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0017] It should be noted that, unless otherwise defined, the technical or scientific terms used in one or more embodiments of the present invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in one or more embodiments of the present invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" indicate that the element or object preceding the term encompasses the elements or objects listed following the term and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0018] like Figures 1-2 As shown, the multi-threshold state light dynamic image rendering method for heart failure signs time series includes the following steps:

[0019] Step S101: Calculate the oxygen saturation recovery time constant after standing upright at dawn based on the blood oxygen saturation regression trajectory after standing upright in the morning.

[0020] Step S102: The oxygen saturation recovery time constant after standing upright at dawn is used as the main anchor point and combined with body weight, heart rate and hemodynamic components in a fixed order to construct a multi-sign input vector;

[0021] Step S103: Perform monotonic integral embedding and context embedding on the multi-signature input vector, solve the multi-threshold color band probability distribution by linear superposition and a differentiable convex allocation layer that satisfies the constraints, and generate the brightness attenuation constant, animation pulse frequency and halo radius ratio based on the distribution.

[0022] Step S104: The primary color is synthesized by weighting according to the probability distribution of the multi-threshold color band, and the time smoothing coefficient is determined by the oxygen saturation recovery time constant after the dawn upright. The primary color and the target brightness determined by the probability distribution of the multi-threshold color band are exponentially smoothed to obtain smooth color and smooth brightness.

[0023] Step S105: Set the duration of the morning window according to the oxygen saturation recovery time constant after standing upright at dawn, and generate a brightness modulation signal and a radius modulation signal within the window according to the ratio of animation pulse frequency to halo radius.

[0024] Step S106: Based on the smooth color, smooth brightness, and brightness attenuation constant, the main light and the halo are spatially synthesized by combining the brightness modulation signal and the radius modulation signal, and a multi-threshold state light dynamic image is output.

[0025] In one embodiment of the present invention, the time of standing up is determined by recording the blood oxygen saturation time sequence as follows: ,in Represents any time The measured blood oxygen saturation value; the attitude label obtained by recognizing the body position signal and acceleration signal is denoted as . ,in The value can be either supine or upright, corresponding to the supine and upright postures of the human body, respectively; let the time of standing up be... satisfy:

[0026]

[0027] in The time point on the day when the body first changes from a supine to an upright position; stand indicates the upright position of the human body, and lie indicates the supine position of the human body.

[0028] Determine the daytime steady-state baseline: Define the set of daytime periods over the past three days as... , It consists of all times from 13:00 to 17:00 each day for the past three days; making the daytime steady-state baseline satisfy:

[0029]

[0030] in This represents the median of all measured blood oxygen saturation values ​​during the daytime period over the past three days.

[0031] Establishing the fitting window and regression model: Let the preset fitting window length be... , A fixed numerical value with a time dimension; within a time interval Within this model, a single exponential recovery model is used to describe the regression trajectory of blood oxygen saturation. This model satisfies the following:

[0032]

[0033] in For any time The estimated blood oxygen saturation value, The instantaneous blood oxygen saturation in an upright posture relative to the daytime steady-state baseline The magnitude of the gap, Let be the oxygen saturation recovery time constant after standing upright at dawn, which is to be determined.

[0034] Parameter estimation and output results: Solving by the least squares criterion and The solution formula is:

[0035]

[0036] in The result is the calculated oxygen saturation recovery time constant after standing upright at dawn. This is the fitting result for the gap amplitude.

[0037] It should be noted that posture label data represents time-series data containing human body position information, with each time point corresponding to either supine or upright positions, used to identify real-time changes in human posture. The regression starting point represents the specific time node at which the body first transitions from a supine to an upright position that day, serving as the starting time anchor for fitting the subsequent blood oxygen saturation regression trajectory. Historical blood oxygen saturation data represents the sequence of actual blood oxygen saturation values ​​continuously and in real-time collected by a blood oxygen detection device within a preset daytime period over the past three days. The daytime steady-state baseline represents the value obtained by calculating the median of all historical blood oxygen saturation data within the preset daytime period over the past three days, serving as the target steady-state reference standard for blood oxygen saturation recovery. The single-exponential recovery model is a mathematical model used to quantitatively describe the gradual regression trajectory of blood oxygen saturation towards the daytime steady-state baseline after standing upright in the morning; its output value gradually approaches the daytime steady-state baseline over time. The gap amplitude parameter represents the magnitude of the decrease in measured blood oxygen saturation relative to the daytime steady-state baseline at the instant of standing upright in the morning, and is only a non-negative value. The recovery time constant parameter is a core parameter regulating the rate at which blood oxygen saturation recovers to the daytime steady-state baseline; a larger value indicates a slower recovery rate. The preset fitting duration represents a fixed time length predetermined after the regression start point, used to limit the data range fitted by the single-exponential recovery model. The single-exponential recovery model estimate represents the predicted blood oxygen saturation value calculated by substituting the corresponding time point, gap amplitude parameter, and recovery time constant parameter into the single-exponential recovery model. The dawn standing oxygen saturation recovery time constant represents the recovery time constant parameter obtained after solving the single-exponential recovery model parameters using the least squares criterion; it is the core output reflecting the rate at which blood oxygen saturation recovers to the daytime steady-state after standing in the morning.

[0038] It should be noted that the core of the least squares criterion is to minimize the sum of squared errors between the measured values ​​and the model estimates. First, the range of the gap amplitude parameter is set to 0 to 10% of the daytime steady-state baseline, and the range of the recovery time constant parameter is set to 10 to 300 seconds. Then, the ranges of the two parameters are divided by a fixed step size (0.1% for the gap amplitude parameter and 1 second for the recovery time constant parameter), resulting in all possible parameter combinations. Next, for each parameter combination, the model estimate is calculated for all sampling time points within the preset fitting time. Then, the difference between the measured blood oxygen saturation and the model estimate is calculated for each time point. The squares of each difference are summed to obtain the total error for that parameter combination. Finally, the total errors of all parameter combinations are compared, and the recovery time constant parameter corresponding to the parameter combination with the smallest total error is selected as the oxygen saturation recovery time constant after standing upright at dawn. If multiple parameter combinations have the same total error, the average value of the recovery time constant parameters among these combinations is taken as the result.

[0039] It should be noted that the single-exponential recovery model is based on the physiological law of the instantaneous decrease and gradual recovery of blood oxygen saturation after standing upright in the morning. An exponential function is used to describe the recovery process. Because the exponential function has the characteristic of gradually approaching the steady-state value, it highly matches the physiological trend of blood oxygen saturation recovery. A gap amplitude parameter is introduced to quantify the initial decrease, and a recovery time constant parameter is introduced to quantify the recovery speed, thus achieving a fit to the recovery trajectory. The preset daytime period for the past three days is 13:00 to 17:00. During this period, the human body is usually in a state of non-strenuous activity, and blood oxygen saturation is less affected by factors such as exercise and emotions, objectively reflecting the steady-state level of blood oxygen in the resting state and ensuring the reliability of the daytime steady-state benchmark. The preset fitting duration is preferentially set to 5 minutes, meaning that the recovery process of blood oxygen saturation after standing upright in the morning is mainly concentrated within five minutes. After five minutes, blood oxygen saturation basically stabilizes near the daytime steady-state benchmark. This duration can fully cover the core stage of recovery, balancing fitting accuracy and computational efficiency, and will not be elaborated further here.

[0040] In one embodiment of the present invention, the basic input and timing data are determined: the stand-up time is invoked. Oxygen saturation recovery time constant after standing upright at dawn ; Record the weight time sequence as ,in Represents any time The actual measured weight value; the heart rate time sequence is recorded as... ,in Represents any time The measured heart rate value; denoted as systolic blood pressure timing. ,in Represents any time The measured systolic blood pressure value; the pulmonary artery pressure time sequence is denoted as ,in Represents any time The measured pulmonary artery pressure value;

[0041] Calculate three-day weight drift: First, determine the time of weight collection on that day. , satisfy ,in For the moment of standing up The moment when the first valid weight data is obtained on that day; let the weight value for that day be... Then determine the baseline weight for the next three days. , ,in Indicates the time of standing up The previous three-day timeframe, This represents the median of all valid weight data within the specified interval; finally, the three-day weight drift is calculated. ;

[0042] Calculate the average nighttime heart rate: Let the average nighttime heart rate be... ,in Indicates the time of standing up The previous three-hour time interval, This represents the average of all valid heart rate data within that interval.

[0043] Calculate morning systolic blood pressure: Determine the time for collecting morning systolic blood pressure data. , ,in Indicates the time of standing up The next three hours This refers to the moment when valid systolic blood pressure data is first obtained within this interval; let the morning systolic blood pressure be... ;

[0044] Calculate morning pulmonary artery pressure offset: Determine the time of morning pulmonary artery pressure collection for the current day. , ,in for The moment when valid pulmonary artery pressure data is first obtained within the interval; let the morning pulmonary artery pressure of that day be... ;

[0045] Determine the assembly time for the seven days prior. ,in Indicates the current standing time The previous seven consecutive days of the first The time point of the first supine to upright rotation on each day; for each Determine the corresponding morning pulmonary artery pressure collection time. , The corresponding morning pulmonary artery pressure ;

[0046] Calculate the baseline morning pulmonary artery pressure over the past seven days:

[0047] ,in This represents the median of the morning pulmonary artery pressure data for the seven corresponding days;

[0048] Calculate morning pulmonary artery pressure shift ;

[0049] Constructing multi-sign input vectors: Multi-sign input vectors are obtained by combining them in a fixed order. .

[0050] It should be noted that the multi-sign input vector represents a column vector formed by integrating core physiological sign parameters in a fixed order. "Time of first standing up upon waking" indicates the specific time point at which the individual first transitions from a supine to an upright position that day. "Weight of the day" refers to the first weight value collected after the first time of standing up upon waking that day that meets data quality requirements. "Preset time period before the time of first standing up upon waking in the past three days" refers to a fixed time period pre-set each day before the time of first standing up upon waking for the three consecutive days prior to the current date. "Baseline weight" represents the median value calculated from all valid weight data within the preset time period before the time of first standing up upon waking in the past three days, serving as a benchmark for measuring daily weight change. "Three-day weight drift" represents the difference between the current day's weight and the baseline weight, used to quantify recent weight trends and reflect fluid retention or consumption.

[0051] It should be noted that the preset nighttime period refers to a fixed nighttime period set before the time of standing up in the morning, used to extract heart rate data reflecting nighttime sympathetic load. The nighttime mean heart rate represents the arithmetic mean of all valid heart rate data within the preset nighttime period, used to reflect the nighttime cardiac sympathetic nerve activity load. The preset morning window represents a fixed morning time period set after the time of standing up in the morning. Morning systolic blood pressure represents the first systolic blood pressure value collected within the preset morning window that meets data quality requirements, reflecting the morning vasoconstrictive pressure status. The morning pulmonary artery pressure represents the first pulmonary artery pressure value collected within the preset morning window that meets data quality requirements, reflecting the morning pulmonary blood circulation pressure status. The baseline pulmonary artery pressure represents the value obtained by calculating the median of the daily morning pulmonary artery pressure data over the past seven days, serving as a long-term benchmark for measuring the daily morning pulmonary artery pressure change. The morning pulmonary artery pressure offset represents the difference between the daily morning pulmonary artery pressure and the baseline pulmonary artery pressure, used to quantify the fluctuation of the daily morning pulmonary artery pressure relative to the long-term benchmark.

[0052] It should be noted that the selection of five components—the time constant of oxygen saturation recovery after standing at dawn, the three-day weight drift, the mean nighttime heart rate, the morning systolic blood pressure, and the morning pulmonary artery pressure offset—is an optimized choice based on the physiological and pathological characteristics of heart failure patients. Specifically, the time constant of oxygen saturation recovery after standing at dawn dominates the position-related oxygenation recovery status; the three-day weight drift is associated with fluid retention (a core complication of heart failure); the mean nighttime heart rate reflects sympathetic load (a key indicator for monitoring heart failure); and the morning systolic blood pressure and pulmonary artery pressure offset directly reflect hemodynamic stability. These five components complement each other in reflecting the state of heart failure from different dimensions. The fixed order is to ensure the consistency of the input vectors and the stability of subsequent model training and inference, avoiding parameter mapping confusion caused by changes in the order.

[0053] In one embodiment of the invention, monotonic integral embedding is performed by calling the dawn-on-standing oxygen saturation recovery time constant from the multi-signal input vector. Constructing embedded scalars ,satisfy:

[0054]

[0055] in It is a positive smoothing function. It is a basis function vector composed of predefined basis functions (such as piecewise polynomials or radial basis functions). The learnable weight vector is the basis function vector. For learnable bias parameters, For learnable constant terms, For integration variables;

[0056] Execution context embedding: Extracting from multi-sign input vectors except The external components form a vector. Calculate the context vector ,satisfy:

[0057]

[0058] in , For learnable weight matrix, , For learnable bias vectors, , The predefined differentiable activation function (such as the tanh function or the positive smoothing function).

[0059] Linear superposition forms the target probability centroid: based on context vectors With embedded scalar The target probability centroid is calculated sequentially using the following formula. :

[0060] ,in For learnable weight matrix, A learnable bias vector;

[0061] ,in Let it be a vector of all 1s, where all elements are 1. express The sum of all elements;

[0062] ,in For learnable weight matrix, This represents the L2 norm, used to normalize a vector to a unit vector.

[0063] ,in These are learnable parameters;

[0064] ;

[0065] Solving the multi-threshold color band probability distribution using a differentiable convex allocation layer: Let... ,in Let be a learnable parameter; It is a first-order difference operator that satisfies ( For vectors The (element); solve for the multi-threshold color band probability distribution that satisfies the constraints. ,satisfy:

[0066]

[0067] The constraints are ( (All elements are non-negative) ( The sum of all elements is 1), where express and The square of the second norm, express The first norm;

[0068] Mapping generates rendering parameters: Let the number of multi-threshold color bands be... ( (where is a positive integer greater than 1), define the linear scale vector of the color band. , its first Each component satisfies ( From 1 to (positive integers); the brightness attenuation constant is calculated sequentially using the following formula. Animation pulse frequency Ratio to the radius of the halo :

[0069] ,in express and The inner product;

[0070] ,in , These are learnable parameters;

[0071] ,in , These are learnable parameters;

[0072] ,in It is an S-shaped function. , These are learnable parameters.

[0073] It should be noted that monotonic integral embedding represents a dedicated embedding transformation designed for the oxygen saturation recovery time constant after standing upright at dawn. It uses definite integral operations to ensure a structural constraint that the output scalar monotonically increases with this time constant. The smooth positive function represents the fundamental function used for monotonic integral embedding and subsequent parameter calculations. It features a consistently positive output, continuous differentiability, and smoothness without abrupt changes, avoiding numerical anomalies. The embedding scalar represents the output of the monotonic integral embedding, a real number that monotonically increases with the oxygen saturation recovery time constant after standing upright at dawn, used to modulate the target probability centroid. Context embedding represents a nonlinear transformation performed on the non-core components of the multi-signal input vector. A multilayer perceptron extracts the correlation features of each component, generating a context vector of uniform dimension. The multilayer perceptron is a feedforward neural network composed of two linear transformation layers and two nonlinear activation functions cascaded together, used to achieve nonlinear extraction of context information. The context vector represents the output of the context embedding, a fixed-dimensional real vector containing the correlation features of non-core sign components, used to generate a daytime steady-state template and risk migration direction.

[0074] It should be noted that the daytime steady-state template vector represents the baseline probability distribution vector generated based on the context vector, with each component being non-negative and summing to one, reflecting the baseline of the color band probability distribution when there is no obvious risk. The risk migration direction vector represents the unit vector generated based on the context vector, indicating the direction of migration of the color band probability distribution towards high-risk areas, with each component summing to one. The target probability centroid vector represents the vector obtained by superimposing the daytime steady-state template vector with the risk migration direction vector after embedding scalar weighting, and is the target reference vector of the multi-threshold color band probability distribution. The multi-threshold color band probability distribution represents a vector composed of the probability components corresponding to K color bands, with each component being non-negative and summing to one, used for weighted synthesis of primary colors and calculation of scalar risk, where K is the number of color bands (preferably 5). The total variation regularization term represents the regularization term used to suppress abrupt changes in adjacent components of the multi-threshold color band probability distribution, which is the product of the sum of the absolute values ​​of the differences between adjacent components and the smoothing coefficient. Linear constraints represent constraints on the probability distribution of multi-threshold color bands, including two core requirements: first, all components must be greater than or equal to 0; second, the sum of all components must be equal to 1, ensuring that it is a legal probability distribution.

[0075] It should be noted that the convex quadratic programming problem represents an optimization problem with Euclidean distance combined with total variation regularization as the objective function and linear constraints as the limiting condition. Its optimal solution is the valid multi-threshold color band probability distribution. The scalar risk is a real number obtained by weighted summation of the multi-threshold color band probability distributions, ranging from 0 to 1; a larger value indicates a higher risk related to heart failure. The affine transformation represents a linear transformation that maps the scalar risk to rendering parameters. The slope and intercept of the mapping are adjusted through learnable parameters to establish the correlation between the risk and rendering parameters. The brightness decay constant represents a parameter controlling the decay rate of the halo of the status light; a larger value indicates slower halo decay. It is obtained by mapping the scalar risk through an affine transformation and a smooth activation function. The animation pulse frequency represents a parameter controlling the blinking rhythm of the status light, obtained by mapping the scalar risk through an affine transformation and a smooth activation function; a larger value indicates more frequent blinking. The halo radius ratio represents a parameter controlling the magnification ratio of the status light halo, ranging from 0 to 1. It is obtained by mapping the scalar risk through an affine transformation and a sigmoid activation function.

[0076] It should be noted that solving the multi-threshold color band probability distribution is transformed into a constrained convex quadratic programming problem. A dual-objective design, using an Euclidean distance term to ensure the distribution is close to the target centroid and a total variation regularization term to ensure distribution smoothness, ensures the dominant influence of the oxygen saturation recovery time constant after standing upright at dawn on the distribution, while avoiding jumps in adjacent color band probabilities. The characteristics of convex optimization guarantee the uniqueness and numerical stability of the optimal solution, while differentiability supports end-to-end training. Specifically, the multi-threshold color band probability distribution is first initialized as a uniform distribution (each component takes a value of one divided by the number of color bands K); then, the Euclidean distance term and the total variation regularization term of the current distribution are calculated and summed to obtain the objective function value; then, under the premise of satisfying linear constraints, the components of the distribution are fine-tuned (adjacent components are adjusted synchronously to reduce abrupt changes), and the objective function value is recalculated; until the objective function value no longer decreases (the decrease is less than one ten-thousandth), the distribution at this point is the optimal solution to the convex quadratic programming problem, i.e., the final multi-threshold color band probability distribution.

[0077] It should be noted that all learnable weight matrices are initialized to normally distributed random values ​​with a mean of zero and a standard deviation of 0.01; all learnable bias vectors are initialized to zero vectors; this ensures stable model output during the initial training phase and avoids training failure due to improper parameter initialization. The preferred value for the number of color bands, K, is 5, corresponding to five levels of risk: low risk, relatively low risk, medium risk, relatively high risk, and high risk, with the color band numbers increasing from one to five. Three radial basis functions are selected to form the basis function vector, with centers at 50 seconds, 150 seconds, and 300 seconds (covering the range of common clinical recovery time constants), and a width parameter of 100 seconds for each. This setting effectively fits individual differences in recovery speed, ensuring the flexibility and stability of monotonic integral embedding, which will not be elaborated upon here.

[0078] In one embodiment of the present invention, the basic input parameters are determined by calling the multi-threshold color band probability distribution. (in For the current rendering moment, (The number of multi-threshold color bands); call the oxygen saturation recovery time constant after dawn uprighting. Call the linear scale vector of the color band. Let the set of color centers for the multi-threshold color band be... ,in The first color space defined in the preset color space (HSV or CIELAB) The color center of each color band; assuming the rendering frame interval is... ,in The time interval between rendering images of two adjacent frames;

[0079] Weighted composite primary colors: Calculate the time using the following formula base color :

[0080]

[0081] in For a moment Multi-threshold color band probability distribution The One portion, For a moment The base color;

[0082] Determine the target brightness: Calculate the time sequentially using the following formulas. Scalar risk and target brightness:

[0083] ,in Represents the linear scale vector of the color band With multi-threshold color band probability distribution The inner product, For a moment scalar risk level;

[0084] ,in The constant term coefficients for brightness mapping, The coefficients of the first-order term of the brightness mapping are... , and The system configuration parameters are preset or determined through training. For a moment Target brightness;

[0085] Determine the time smoothing factor: Calculate the time smoothing factor using the following formula. :

[0086]

[0087] in A preset numerical stability constant greater than 0 (preferably set to 10 seconds) is used to avoid a denominator of 0. Based on the oxygen saturation recovery time constant after standing upright at dawn A defined time smoothing coefficient;

[0088] Perform exponential time smoothing: Apply exponential time smoothing to the primary color and target brightness using the following formulas to obtain a smoothed color. With smooth brightness :

[0089] Color smoothing: ,in For a moment Smooth color, initial condition is ( The starting time of the morning window, i.e., the time of standing up. );

[0090] Brightness smoothing: ,in For a moment Smooth brightness, initial conditions are .

[0091] It should be noted that the color center set represents a preset set of K fixed color values, with each color center corresponding to a baseline color for a risk level. The CIELAB color space is used to ensure perceptual uniformity. The base color represents the fundamental color obtained at the current moment based on the multi-threshold color band probability distribution and the weighted sum of the color centers. The preset mapping coefficients represent fixed coefficients used to linearly map the scalar risk level to the target brightness, including intercept and slope coefficients, ensuring that the brightness mapping meets visual perception requirements. The target brightness represents the normalized brightness value obtained by linearly mapping the scalar risk level, ranging from 0 to 1; a larger value indicates higher visual brightness. The exponential function model is used to calculate the temporal smoothing coefficient, with the natural constant as the base. The exponent term consists of the negative of the ratio of the rendering frame interval to the oxygen saturation recovery time constant after dawn, ensuring that the output smoothing coefficient is between 0 and 1. The rendering frame interval represents the time interval between two adjacent dynamic images, used to control the temporal granularity of the smoothing process, ensuring a natural visual transition. The time smoothing coefficient controls the strength of exponential time smoothing, ranging from 0 to 1. A larger value indicates a stronger influence of historical smoothed values ​​on the current result. Historical weights represent the weights used in exponential time smoothing to weight the smoothed value from the previous time step, and are equal to the time smoothing coefficient. Current weights represent the weights used in exponential time smoothing to weight the original value (primary color / target brightness) at the current time step, and are summed with historical weights.

[0092] It should be noted that the oxygen saturation recovery time constant after standing upright at dawn (reflecting the physiological recovery timescale) is directly mapped to a time smoothing coefficient (reflecting the visual smoothing timescale), achieving an isomorphic mapping from the physiological timescale to the visual timescale. That is, the slower the physiological recovery (the larger the time constant), the closer the visual smoothing coefficient is to one, and the stronger the smoothing effect, avoiding sudden color / brightness changes in a short period and ensuring a high degree of consistency between visual performance and the dynamic characteristics of the physiological state. The CIELAB color space is chosen to define the color centers of the color bands because this space approximates the uniformity of human visual perception. The linearly weighted composite primary colors can accurately reflect the continuous transition of risk levels, avoiding perceptual jumps caused by the characteristics of the color space. The traditional RGB space, due to its non-linear perception, is unsuitable for this type of continuous risk mapping. For example, the preferred value for the number of color bands, K, is 5, corresponding to five color centers. The CIELAB space coordinates are (80, -10, -20), (70, 10, -30), (60, 30, -10), (50, 40, 20), and (40, 20, 40), respectively, corresponding to color bands from low risk to high risk. The intercept coefficient ranges from 0.4 to 0.6 (ensuring brightness is not too dark at the lowest risk) and the slope coefficient ranges from 0.4 to 0.6 (ensuring brightness is not too overexposed at the highest risk). These values ​​can be fine-tuned according to the brightness range of the display device, with a default value of 0.5 for both, balancing versatility and visual comfort. Furthermore, the smooth color in the initial frame (first frame) is assigned the current base color, and the smooth brightness is assigned the current target brightness, avoiding rendering anomalies in the first frame due to missing initial values. The rendering frame interval is preferentially set to 0.04 seconds, which can be adjusted to 0.05 seconds for high-performance devices. Frame rates higher than 0.02 seconds offer no significant visual improvement, only increasing hardware overhead, and will not be discussed further here.

[0093] In one embodiment of the present invention, the basic input parameter is determined by calling the oxygen saturation recovery time constant after standing upright at dawn. With the moment of standing up Call the animation pulse frequency (Frequency dimension) ratio to halo radius (Dimensionless); Let the preset upper limit of the morning window be... , Let the constant be a fixed value with the dimension of time; let the proportionality constant be... , It is a fixed value with dimensionless dimensions, and the time dimension is the same as... match;

[0094] Set the morning window duration and window indicator function:

[0095] Calculate the morning window duration using the following formula. :

[0096]

[0097] in The duration of the morning window, with a value of [value]. and The smaller value in;

[0098] Define window indicator function :

[0099]

[0100] in For the current moment, The time interval for the morning wake-up window;

[0101] Generate the luminance modulation signal: Calculate the time using the following formula. Brightness modulation signal :

[0102]

[0103] in Angular frequency, For the current moment Relative to the time of standing up Time difference, It is a sine function with coefficients Used to standardize the range of values ​​for the sine function to 0 to 1. For a moment The brightness modulation signal;

[0104] Generate radius-modulated signal: Calculate the time using the following formula radius modulation signal :

[0105]

[0106] in For a moment The radius-modulated signal.

[0107] It should be noted that the preset scaling factor is a fixed coefficient used to convert the oxygen saturation recovery time constant after sunrise into the window duration. The calculation duration is the product of the oxygen saturation recovery time constant after sunrise and the preset scaling factor, and is a candidate value for the morning window duration. The effective time interval represents the time range starting from the sunrise moment and lasting for a duration equal to the morning window duration; non-zero modulation signals are only generated within this interval. The current rendering moment represents the generation moment of each frame of the dynamic image, used to determine whether it is within the effective time interval. The window indicator value represents a binary value used to control the modulation signal's activation status; it is one only within the effective time interval and zero at other times. The animation pulse frequency is used to control the flashing rhythm of the brightness modulation signal; the larger the value, the more frequent the flashing. The normalized pulse component represents the pulse signal after offset and scaling processing, with a value range of 0 to 1, ensuring that the brightness modulation signal meets visual output requirements. The brightness modulation signal is used to control the pulse of the status light brightness; there is pulse output only within the effective time interval, and zero at other times. The halo radius ratio controls the amplification ratio of the status light halo, with a value ranging from 0 to 1. The radius modulation signal represents the signal that controls the radius of the status light halo; halo amplification is only enabled during the effective time interval, and there is no additional amplification at other times.

[0108] It should be noted that the preset scaling factor is preferably set to 3. Based on clinical data, the core cycle of blood oxygen saturation recovery in heart failure patients after standing upright in the morning is approximately 3 to 4 times the recovery time constant. A value of 3 covers the effective recovery phase for most patients, balancing sensitivity and specificity. The fine-tuning range of 2.5 to 3.5 can be adapted to different age groups (e.g., 3.5 for elderly patients, 2.5 for younger patients). The preset upper limit of the window duration is preferably set to 15 minutes. The rebalancing process of physiological indicators (blood oxygen, blood pressure, etc.) after standing upright in the morning rarely exceeds 15 minutes; exceeding this time adds no additional clinical significance to the modulated signal and increases hardware power consumption. The current rendering interval is preferably set to 0.04 seconds to ensure that the granularity of time judgment matches the rendering frame rate, avoiding signal flickering abnormalities caused by inconsistent frame intervals.

[0109] In one embodiment of the present invention, the basic input parameter is determined: the call time. Smooth color With smooth brightness Call the brightness attenuation constant ; Call time Brightness modulation signal With radius modulation signal Let the radius of the main light base be... , Let the reference attenuation coefficient be a fixed value with the dimension of length. , Let the value be a fixed value greater than 0; let any pixel in the image be... pixel The distance to the center of the image circle is , It is a numerical value with the dimension of length;

[0110] Synthetic instantaneous brightness: Calculate the time using the following formula instantaneous brightness :

[0111]

[0112] in For a moment The instantaneous brightness scalar;

[0113] Calculate the outer radius of the halo: Calculate the time using the following formula. outer radius of the halo :

[0114]

[0115] in For a moment The outer radius of the halo has the dimension of length.

[0116] Determine the radial attenuation coefficient: Calculate the radial attenuation coefficient using the following formula. :

[0117]

[0118] in A coefficient for controlling the radial attenuation intensity of the halo;

[0119] Perform spatial composition and output of the main light and halo: Define the timing. pixel The color vector is According to the following piecewise formula Assign values ​​to generate and output a dynamic image of the multi-threshold status lights:

[0120]

[0121] The first segment corresponds to the main light area assignment ( For pixels The judgment condition located in the main light area), the second segment corresponding to the halo area is assigned a value ( For pixels The determination condition located in the halo region), the assignment outside the corresponding region of the third segment ( For pixels (Judgment criteria for locations outside the main light and halo area) It is a natural exponential function. It is a zero-based color vector where all elements are 0.

[0122] It should be noted that the brightness gain factor is used to adjust the smoothed brightness, converting the brightness modulation signal into a brightness amplification ratio, with a value ranging from 1 to 2. Smoothed brightness represents the final brightness value after exponential time smoothing, avoiding sudden brightness changes, with a value ranging from 0 to 1. The instantaneous brightness scalar represents the final brightness value of the previous frame, fusing smoothed brightness with brightness pulsation characteristics. The radius expansion factor is used to adjust the main light's base radius, converting the radius modulation signal into a radius amplification ratio. The main light's base radius represents the preset reference radius of the main light's core area. The halo outer radius represents the maximum radius of the halo outside the main light's core area. The reference attenuation coefficient represents the preset radial attenuation reference value of the halo, controlling the overall attenuation intensity of the halo, and is a positive number. The brightness attenuation constant is used to control the rate of halo attenuation; the larger the value, the slower the halo attenuation. The radial attenuation coefficient is used to adjust the rate of radial attenuation of the halo, inversely proportional to the brightness attenuation constant, and is a positive number. The smoothed color is the final color value after exponential time smoothing. The pixel value of the main light region is the pixel color value of the core region of the main light (Euclidean distance less than or equal to the base radius of the main light), obtained by directly multiplying the smooth color by the instantaneous brightness scalar. The normalized distance ratio is used to control the gradual decay of the halo. The exponential function value is the intermediate value used to achieve radial exponential decay of the halo, and the value decreases with increasing distance. The pixel value of the halo region represents the pixel color value of the halo region, which decays exponentially with increasing distance. The zero vector represents the zero value vector of the pixel color, corresponding to (zero, zero, zero) in CIELAB space, indicating no color output (black). The multi-threshold state light dynamic image represents a continuous frame image composed of the pixel values ​​of the main light region, the pixel values ​​of the halo region, or the zero vector of all pixels, dynamically reflecting the heart failure-related risk status.

[0123] It should be noted that the area is divided into three categories based on the distance from the pixel to the center: the main light area, the halo area, and the outer area. The main light area is a pure color with no attenuation (ensuring the clarity of core risk signals), the halo area experiences exponential attenuation (avoiding visual abrupt changes), and the outer area has zero vector (reducing invalid calculations). The division and assignment rules of these three areas conform to the visual perception habits of the human eye. The basic radius of the main light is preferably set to 100 pixels to adapt to mainstream display devices (such as mobile phones and smartwatches with screen resolutions of 1000 to 2000 pixels), ensuring that the main light has a reasonable proportion on the screen (approximately 1 / 5 to 1 / 10 of the screen width). The resolution adaptation rules are clear to avoid imbalances in the size of the main light due to differences in devices. The baseline attenuation coefficient is preferably set to 5. At this value, the brightness attenuation ratio of the halo from the inside to the outside is approximately 98% (the brightness of the outermost layer is 2% of that of the inner layer), which ensures the layering of the halo without making the outer layer too dark to be recognized. The fine-tuning range of 3 to 7 can adapt to the visual sensitivity of different users. The zero vector (0, 0, 0) in CIELAB space corresponds to (0, 0, 0) (black) in RGB space. The accurate conversion between the two spaces can be achieved through the IEC61966-2-1 standard color conversion algorithm, ensuring that pixels outside the region present a uniform black background without color interference. This will not be elaborated on here.

[0124] It should be noted that this invention is applicable to long-term vital sign monitoring and status alerts for heart failure patients, with applications including home self-management, community medical follow-up, clinical inpatient care, and rehabilitation tracking. In home settings, patients can use a portable vital sign data collection device. The device automatically collects blood oxygen saturation, weight, heart rate, systolic blood pressure, and pulmonary artery pressure data upon waking and standing upright, generating dynamic status light images without requiring professional operation. Patients and their families can intuitively understand physiological status trends by observing the color, brightness, pulse frequency, and halo changes of the status lights, promptly identifying abnormalities and contacting medical staff. During community medical follow-up, medical staff can remotely assess the patient's condition control through the monitoring data and status light images uploaded by the patient, eliminating the need for frequent visits to community health service centers, reducing medical costs, and improving follow-up efficiency. During clinical inpatient care, this invention can be integrated into ward monitoring systems to collect patients' relevant vital sign data in real time upon waking and generate dynamic images. Medical staff can quickly grasp the dynamic physiological status of multiple patients, assisting in optimizing nursing plans and timely intervention for potential risks. In the context of rehabilitation follow-up, patients use the dynamic status light image continuously during rehabilitation training. This image serves as a direct feedback on the rehabilitation effect, helping medical staff adjust the rehabilitation plan and allowing patients to clearly perceive changes in their own status, thereby improving rehabilitation compliance.

[0125] It should be noted that, as Figure 2As shown, the horizontal axis of the multi-threshold color band probability distribution heatmap represents time, and the vertical axis represents the center of the risk color band. From the data distribution in the figure, we can observe the temporal changes in the system's assessment of heart failure risk. During low-risk periods, high probability values ​​are mainly concentrated in the color band area below the vertical axis. When the monitoring data, such as the blood oxygen recovery time constant, deteriorates, the high probability area on the heatmap will gradually migrate to the 0.7 or 0.9 high-risk color band above the vertical axis.

[0126] It should be noted that all user vital signs data (blood oxygen saturation, weight, heart rate, etc.) involved in this invention have been explicitly authorized and consented to by the users before collection, and strictly comply with relevant data security and privacy protection regulations. Furthermore, the heart failure-related status monitoring results output by this invention are for medical auxiliary reference only, used to indicate trends in physiological changes, and cannot replace professional medical diagnosis, treatment, and medical advice. For any related health decisions, please consult a licensed physician.

[0127] It should be noted that the interval and threshold sizes are set for ease of comparison. The size of the threshold depends on the amount of sample data and the base number set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, and the formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0128] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A multi-threshold state light dynamic image rendering method for heart failure signs time series, characterized in that, Includes the following steps: Step S101: Calculate the oxygen saturation recovery time constant after standing upright at dawn based on the blood oxygen saturation regression trajectory after standing upright in the morning. Step S102: The oxygen saturation recovery time constant after standing upright at dawn is used as the main anchor point and combined with body weight, heart rate and hemodynamic components in a fixed order to construct a multi-sign input vector; Step S103: Perform monotonic integral embedding and context embedding on the multi-signature input vector, solve the multi-threshold color band probability distribution by linear superposition and a differentiable convex allocation layer that satisfies the constraints, and generate the brightness attenuation constant, animation pulse frequency and halo radius ratio based on the distribution. Step S104: The primary color is synthesized by weighting according to the probability distribution of the multi-threshold color band, and the time smoothing coefficient is determined by the oxygen saturation recovery time constant after the dawn upright. The primary color and the target brightness determined by the probability distribution of the multi-threshold color band are exponentially smoothed to obtain smooth color and smooth brightness. Step S105: Set the duration of the morning window according to the oxygen saturation recovery time constant after standing upright at dawn, and generate a brightness modulation signal and a radius modulation signal within the window according to the ratio of animation pulse frequency to halo radius. Step S106: Based on the smooth color, smooth brightness, and brightness attenuation constant, the main light and the halo are spatially synthesized by combining the brightness modulation signal and the radius modulation signal, and a multi-threshold state light dynamic image is output.

2. The multi-threshold state light dynamic image rendering method for heart failure vital signs according to claim 1, characterized in that, Based on posture label data containing body position information, the time point at which the body first changes from a supine to an upright position on that day is identified and marked as the regression starting point. Extract historical blood oxygen saturation data within a preset daytime period over the past three days, calculate the statistical median of this historical blood oxygen saturation data, and mark this statistical median as the daytime steady-state baseline; A single exponential recovery model is constructed, which uses the intraday steady-state benchmark as the asymptotic final value and includes a gap magnitude parameter relative to the intraday steady-state benchmark and a recovery time constant parameter that controls the recovery rate. Within a preset fitting time after the regression starting point, the difference between the measured blood oxygen saturation and the estimated value of the single exponential recovery model is minimized according to the least squares criterion. The gap amplitude parameter and the recovery time constant parameter are numerically solved, and the solved recovery time constant parameter is output as the oxygen saturation recovery time constant after standing upright at dawn.

3. The multi-threshold state light dynamic image rendering method for heart failure signs according to claim 1, characterized in that, The oxygen saturation recovery time constant after standing upright at dawn is taken as the first component of the multi-sign input vector; The first valid weight data collected after standing upright in the morning is selected as the weight of that day. The median of all valid weight data within a preset time period before standing upright in the morning for the past three days is calculated as the baseline weight. The difference between the weight of that day and the baseline weight is marked as the three-day weight drift, which is used as the second component of the multi-sign input vector.

4. The multi-threshold state light dynamic image rendering method for heart failure signs according to claim 3, characterized in that, Calculate the arithmetic mean of all valid heart rate data within a preset nighttime period before the time of standing upright in the morning, and mark this arithmetic mean as the nighttime heart rate mean, which is used as the third component of the multi-sign input vector; Select the first effective systolic blood pressure data collected within the preset morning window after the time of standing upright on the same day, and mark this effective systolic blood pressure data as the morning systolic blood pressure, as the fourth component of the multi-sign input vector; The effective pulmonary artery pressure data collected for the first time in the preset morning window after standing upright on the same day is selected as the morning pulmonary artery pressure of that day, and the median value of the morning pulmonary artery pressure data of the past seven days is obtained as the baseline pulmonary artery pressure. The difference between the morning pulmonary artery pressure of that day and the baseline pulmonary artery pressure is marked as the morning pulmonary artery pressure offset, which is used as the fifth component of the multi-sign input vector. Based on the order of the oxygen saturation recovery time constant after standing upright at dawn, the three-day weight drift, the nighttime mean heart rate, the morning systolic blood pressure, and the morning pulmonary artery pressure offset, a multi-sign input vector in column vector form is generated.

5. The multi-threshold state light dynamic image rendering method for heart failure signs according to claim 1, characterized in that, Monotonically integrally embedding is performed on the oxygen saturation recovery time constant after dawn uprighting in the multi-sign input vector. A monotonically increasing embedded scalar is generated by definite integral operation with a smooth positive value function as the integrand. Context embedding is performed on the components of the multi-sign input vector except for the oxygen saturation recovery time constant after standing upright at dawn. The context vector is generated by linear transformation and nonlinear activation of a multilayer perceptron.

6. The multi-threshold state light dynamic image rendering method for heart failure signs according to claim 5, characterized in that, The intraday steady-state template vector and the risk migration direction vector are generated based on the context vector. The embedded scalar is weighted along the risk migration direction vector and superimposed on the intraday steady-state template vector to generate the target probability centroid vector. With the objective function of minimizing the Euclidean distance term and the total variation regularization term of the multi-threshold color band probability distribution, and under the linear constraint that the multi-threshold color band probability distribution is non-negative and sums to one, a convex quadratic programming problem is solved to obtain the multi-threshold color band probability distribution. The scalar risk degree is obtained by weighted summation of the probability distribution of multi-threshold color bands, and then mapped to the brightness attenuation constant, animation pulse frequency and halo radius ratio by affine transformation and smoothing activation function.

7. The multi-threshold state light dynamic image rendering method for heart failure signs according to claim 1, characterized in that, Using each probability component in the multi-threshold color band probability distribution as a weight, a linear weighted synthesis is performed on the preset color band color center set to calculate the primary color at the current moment; The scalar risk level is calculated based on the probability distribution of the multi-threshold color band, and then converted into the target brightness through a linear function containing preset mapping coefficients. An exponential function model with the natural constant as its base is constructed. The exponent of this exponential function model consists of the negative of the ratio of the rendering frame interval to the oxygen saturation recovery time constant after standing upright at dawn. The time smoothing coefficient is calculated using this exponential function model. The time smoothing coefficient is used as the historical weight, and the difference between the value 1 and the time smoothing coefficient is used as the current weight. The smooth color and the base color of the previous rendering moment are weighted and summed, and the smooth brightness of the previous rendering moment is used as the target brightness of the current moment. The smooth color and smooth brightness of the current moment are then output.

8. The multi-threshold state light dynamic image rendering method for heart failure signs according to claim 1, characterized in that, The calculation time is obtained by multiplying the oxygen saturation recovery time constant after standing upright at dawn by a preset proportional coefficient. The minimum value between the calculation time and the preset upper limit of the window is selected and determined as the morning window duration. Define a valid time interval starting from the moment of standing upright in the morning and continuing for the duration of the morning window. Determine whether the current rendering time is within the valid time interval. If it is within the valid time interval, set the window indicator value to one; otherwise, set the window indicator value to zero. A time-varying sine function is constructed based on the animation pulse frequency. An offset and halving scaling operation is performed on the sine function to generate a normalized pulse component. The normalized pulse component is multiplied by the window indicator value to output a brightness modulation signal. The radius ratio of the halo is multiplied by the window indicator value to output a radius modulation signal.

9. The multi-threshold state light dynamic image rendering method for heart failure signs according to claim 1, characterized in that, The brightness gain factor is obtained by adding the value one to the brightness modulation signal, and the smooth brightness is multiplied by the brightness gain factor to output the instantaneous brightness scalar. The radius expansion factor is obtained by adding the value one to the radius modulation signal. The radius expansion factor is then multiplied by the preset base radius of the main lamp to output the outer radius of the halo. Divide the preset baseline attenuation coefficient by the luminance attenuation constant to output the radial attenuation coefficient.

10. The multi-threshold state light dynamic image rendering method for heart failure signs according to claim 9, characterized in that, Calculate the Euclidean distance from the current rendered pixel to the center of the image. If the Euclidean distance is less than or equal to the base radius of the main light, multiply the smooth color by the instantaneous brightness scalar to obtain the pixel value of the main light area. If the Euclidean distance is greater than the base radius of the main light and less than or equal to the outer radius of the halo, then the difference between the Euclidean distance and the base radius of the main light is calculated as the numerator, and the difference between the outer radius of the halo and the base radius of the main light is calculated as the denominator. The numerator is divided by the denominator to obtain the normalized distance ratio. The normalized distance ratio is multiplied by the radial attenuation coefficient and the negative value is taken as the exponent. The exponential function value with the natural constant as the base is calculated. The smooth color, the instantaneous brightness scalar, and the exponential function value are multiplied together to obtain the pixel value of the halo region. If the Euclidean distance is greater than the outer radius of the halo, the zero vector is assigned to the currently rendered pixel, and the final multi-threshold state light dynamic image is composed of all pixels.

Images