Diabetic foot risk dynamic monitoring system based on wearable device

Abstract

The application relates to the technical field of diabetic foot risk monitoring, in particular to a diabetic foot risk dynamic monitoring system based on a wearable device. The system comprises the following steps: a data acquisition module acquires the skeletal key points, pressure and glycated hemoglobin detection value of the feet of target personnel; a dynamic feature extraction module constructs a skeletal proximity vector based on the proximity of the sensor nodes and the skeletal key points; the pressure diffusion connectivity between the sensor nodes is determined according to the distance between the sensor nodes and the glycated hemoglobin detection value; the optimal transmission cost calculation is carried out on the pressure and the ideal safety distribution constructed based on the skeletal proximity vector, and the local accumulation index is obtained in combination with the pressure diffusion connectivity; a risk assessment module obtains the cumulative load index according to the pressure, distribution form deviation index and local accumulation index in a gait cycle, and then outputs the corresponding risk prompt. The application improves the accuracy of the dynamic monitoring result of the diabetic foot risk.

Description

Technical Field

[0001] This invention relates to the field of diabetic foot risk monitoring technology, specifically to a dynamic monitoring system for diabetic foot risk based on wearable devices. Background Technology

[0002] Diabetic foot ulcer (DFU) is a common complication of diabetes. Its pathogenesis mainly involves sensory loss due to peripheral neuropathy and local stress concentration caused by abnormal foot anatomy. In current clinical practice, physicians primarily rely on static X-ray images to assess the presence of skeletal deformities (such as metatarsal head depression, claw toes, etc.) and combine this with biochemical indicators such as glycated hemoglobin to assess the severity of the patient's condition. However, this static clinical data is mainly used for triage diagnosis within hospitals and is difficult to directly guide patients' daily dynamic protection.

[0003] Currently, wearable monitoring devices on the market primarily monitor vertical pressure on the sole of the foot in real time through pressure sensor arrays integrated into the insoles, issuing alarms when the pressure amplitude exceeds a fixed threshold. However, this monitoring method based on a single vertical pressure threshold has significant limitations: First, it ignores the relative positional relationship between the pressure distribution pattern and the skeletal anatomy. For patients with skeletal deformities, even if the vertical pressure amplitude does not reach the traditional high-pressure threshold, if the center of pressure happens to be pressing on a bony prominence (lacking soft tissue cushioning), it can still lead to extremely high local risks. Second, it cannot detect the decreased stress dissipation capacity of the foot's soft tissues due to glycosylation and hardening. In the case of tissue hardening, the tissues around the stress point have difficulty dispersing stress laterally, easily forming local stress accumulation; thus, the accuracy of diabetic foot risk monitoring results is low. Summary of the Invention

[0004] To address the issue of low accuracy in existing methods for monitoring the risk of diabetic foot, this invention aims to provide a dynamic monitoring system for the risk of diabetic foot based on wearable devices. The specific technical solution adopted is as follows: This invention provides a wearable device-based dynamic monitoring system for the risk of diabetic foot, the system comprising: The data acquisition module is used to acquire key skeletal points, plantar pressure data, and glycated hemoglobin test values ​​of the target person's feet; The dynamic feature extraction module is used to construct a skeletal proximity vector based on the proximity of each sensor node to key skeletal points; determine the pressure diffusion connectivity between each pair of sensor nodes based on the distance between sensor nodes and the glycated hemoglobin detection value; obtain the distribution morphology deviation index by calculating the optimal transmission cost of the plantar pressure data and the ideal safe distribution constructed based on the skeletal proximity vector; and obtain the local accumulation index based on the pressure diffusion connectivity. The risk assessment module is used to obtain the cumulative load index based on plantar pressure data throughout the gait cycle, the distribution pattern deviation index, and the local accumulation index. The feedback module is used to output corresponding risk warnings based on the value of the cumulative load index.

[0005] Preferably, the step of constructing a bone proximity vector based on the proximity of each sensor node to the skeletal key points includes: The first distance between each sensor node and the nearest point among the skeletal key points is obtained respectively. The skeletal proximity coefficient of each sensor node is determined based on the first distance. The skeletal proximity coefficient is negatively correlated with the first distance. Arrange the bone proximity coefficients of all sensor nodes in order to generate a bone proximity vector.

[0006] Preferably, determining the pressure diffusion connectivity between any two sensor nodes based on the distance between the sensor nodes and the glycated hemoglobin detection value includes: A baseline hardening assessment coefficient is determined based on the glycated hemoglobin test value, and the baseline hardening assessment coefficient is positively correlated with the glycated hemoglobin test value. For any two sensor nodes: if the physical distance between any two sensor nodes is less than a preset distance threshold, then the connectivity weight between the two sensor nodes is calculated based on the physical distance and the benchmark hardening evaluation coefficient, and the connectivity weight is inversely proportional to both the physical distance and the benchmark hardening evaluation coefficient; if the physical distance between any two sensor nodes is greater than or equal to the preset distance threshold, then the connectivity weight between the two sensor nodes is set to zero. A pressure diffusion connectivity matrix is ​​constructed based on the pressure diffusion connectivity between all pairs of sensor nodes.

[0007] Preferably, the step of calculating the distribution morphology deviation index by performing optimal transmission cost calculation on the plantar pressure data and the ideal safety distribution constructed based on the skeletal proximity vector includes: The plantar pressure distribution data of each sensor node is normalized into a source probability distribution; Generate a target probability distribution based on the bone proximity vector; The optimal transport algorithm is used to calculate the transmission cost between the source probability distribution and the target probability distribution, and the transmission cost is used as the distribution shape deviation index.

[0008] Preferably, the step of calculating the transmission cost between the source probability distribution and the target probability distribution using the optimal transport algorithm includes: The Sinkhorn iterative algorithm is used to calculate the Wasserstein distance between the source probability distribution and the target probability distribution by alternately updating the scaling vector and constructing a cost kernel matrix based on the first distance, which is then used as the transmission cost.

[0009] Preferably, obtaining the local accumulation index based on the pressure diffusion connectivity includes: High-pressure areas are determined based on plantar pressure data, and these high-pressure areas are composed of a preset number of sensor nodes with the largest pressure amplitude; nodes at the physical edge of the sensor array are defined as edge areas. A graph theory network is constructed using each sensor node as a network node and the pressure diffusion connectivity as the capacity of the edges between nodes. The maximum flow algorithm is used to calculate the maximum diversion flux from the high-pressure region to the edge region. The local accumulation index is obtained based on the ratio of the sum of pressure data in the high-pressure region to the maximum diversion flux.

[0010] Preferably, obtaining the cumulative load index based on plantar pressure data throughout the gait cycle, the distribution pattern deviation index, and the local accumulation index includes: The plantar pressure data, distribution morphology deviation index, and local accumulation index at each sampling time throughout the entire gait cycle are coupled and integrated to obtain the single-step composite risk index; The cumulative load index is obtained by summing the single-step composite risk index to the cumulative load index.

[0011] Preferably, the step of coupling and integrating the plantar pressure data, distribution morphology deviation index, and local accumulation index at each sampling time throughout the entire gait cycle to obtain a single-step composite risk index includes: For any sampling moment within the entire gait cycle: Calculate the first product of the plantar pressure data and the distribution morphology deviation index at any sampling time, and the second product of the accumulation weighting coefficient and the local accumulation index; combine the first product and the second product to obtain the risk characteristic value at any sampling time. The risk characteristic values ​​of all sampling moments within the entire gait cycle are multiplied by the sensor's sampling period and then summed to obtain the single-step composite risk index.

[0012] Preferably, the risk warning output based on the value of the cumulative load index includes: Calculate the first ratio of the cumulative load index to the personalized damage threshold; When the first ratio is less than a preset first threshold, a safety alert is output; when the first ratio is greater than or equal to the preset first threshold and less than a preset second threshold, a fatigue alert is output; when the first ratio is greater than or equal to the preset second threshold, a high-risk alarm is output; the preset first threshold is less than the preset second threshold.

[0013] Preferably, obtaining the personalized damage threshold includes: The tolerance decline value is calculated based on the degree of difference between the glycated hemoglobin test value and the preset health reference threshold; Based on the tolerance decline value and the standard tolerance value of healthy people, a personalized injury threshold is determined; the personalized injury threshold is not lower than the preset minimum physiological limit threshold.

[0014] The present invention has at least the following beneficial effects: The data acquisition module of the dynamic monitoring system for diabetic foot risk provided by this invention collects key skeletal points, plantar pressure data, and glycated hemoglobin levels from the feet of target individuals. It organically integrates skeletal spatial information reflecting anatomical structure, plantar pressure distribution information reflecting real-time biomechanical state, and biochemical indicators reflecting tissue metabolism and hardening degree, laying a multi-dimensional data foundation for subsequent risk feature extraction. This overcomes the information bias problem caused by relying solely on a single vertical pressure threshold in existing technologies. The dynamic feature extraction module constructs a skeletal proximity vector based on the proximity of sensor nodes to key skeletal points. This vector accurately characterizes the spatial distribution of soft tissue thickness in different areas of the foot. By calculating the optimal transmission cost between plantar pressure data and the ideal safety distribution constructed based on this skeletal proximity vector, it can accurately reflect the distribution morphology deviation index, which reflects the degree of deviation of the pressure distribution from the skeletal anatomy. The dynamic feature extraction module also considers the distance between sensor nodes and... Glycated hemoglobin (HbA1c) levels determine the pressure diffusion connectivity between each pair of sensor nodes. This connectivity fully considers the inhibitory effect of soft tissue hardening caused by glycosylation on the lateral transmission of pressure. Furthermore, a local accumulation index is obtained based on the pressure diffusion connectivity. This index quantitatively characterizes the difficulty of pressure dissipating from high-pressure areas to surrounding areas under the constraints of hardened tissue, thus accurately identifying the stress island effect caused by tissue hardening. This overcomes the deficiency of existing technologies in sensing the impact of changes in soft tissue hardness on risk. The risk assessment module integrates plantar pressure data, distribution morphology deviation index, and local accumulation index throughout the entire gait cycle to obtain a cumulative load index. This index combines the mechanical load reflecting pressure amplitude, the morphological deviation reflecting the rationality of pressure location, and the accumulation effect reflecting tissue dissipation capacity to form a quantitative characterization of the comprehensive damage suffered by the foot soft tissue in each gait cycle, providing a scientific and accurate assessment basis for the dynamic monitoring of diabetic foot risk. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a structural block diagram of a wearable device-based dynamic monitoring system for diabetic foot risk, provided in an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of a dynamic monitoring system for diabetic foot risk based on a wearable device, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details a specific solution for a wearable device-based dynamic monitoring system for diabetic foot risk provided by the present invention.

[0020] Example of a dynamic monitoring system for diabetic foot risk based on wearable devices: This embodiment proposes a dynamic monitoring system for diabetic foot risk based on wearable devices, such as... Figure 1 As shown in this embodiment, a dynamic monitoring system for diabetic foot risk based on wearable devices is provided. The system includes a data acquisition module, a dynamic feature extraction module, a risk assessment module, and a feedback module.

[0021] The data acquisition module is used to acquire key skeletal points, plantar pressure data, and glycated hemoglobin test values ​​of the target person's feet. The dynamic feature extraction module is used to construct a skeletal proximity vector based on the proximity of each sensor node to key skeletal points; determine the pressure diffusion connectivity between each pair of sensor nodes based on the distance between sensor nodes and the glycated hemoglobin detection value; obtain the distribution morphology deviation index by calculating the optimal transmission cost of the plantar pressure data and the ideal safe distribution constructed based on the skeletal proximity vector; and obtain the local accumulation index based on the pressure diffusion connectivity. The risk assessment module is used to obtain the cumulative load index based on plantar pressure data throughout the gait cycle, the distribution pattern deviation index, and the local accumulation index. The feedback module is used to output corresponding risk warnings based on the value of the cumulative load index.

[0022] The data acquisition module is used to acquire key skeletal points, plantar pressure data, and glycated hemoglobin test values ​​of the target person's feet.

[0023] During the dynamic monitoring of diabetic foot risk in target individuals, wearable devices are provided. In this embodiment, the wearable device is a smart insole, which is equipped with sensors to collect various monitoring data.

[0024] The data acquisition module of the wearable device-based dynamic monitoring system for diabetic foot risk provided in this embodiment acquires X-ray image data of the target person's foot under weight-bearing conditions through a data interface. It then uses image processing algorithms to identify and extract the coordinates of key skeletal points from the X-ray image data. These key skeletal points include the centers of the first to fifth metatarsal heads, the center of the calcaneal tuberosity, and high-density points (i.e., bony protrusions / deformities) in the X-ray image with gray values ​​higher than a preset gray value threshold. In this embodiment, the preset gray value threshold is 200; however, in specific applications, the implementer can set it according to the specific circumstances.

[0025] The patient is guided to wear the device and maintain a standard upright posture for a preset time, which can be 5 seconds. During this time, the pressure sensor of the data acquisition module collects pressure data from the sensor array, that is, it acquires the plantar pressure data of each sensor node, and calculates the coordinates of the measured pressure center and the effective distribution contour of the plantar pressure. The specific method for obtaining the effective distribution contour is as follows: the pressure data of each sensor node during the static standing period is averaged over time to obtain the average pressure data corresponding to each sensor node. An effective pressure threshold is set, and sensor nodes with average pressure data greater than the effective pressure threshold are marked as effective force points. The coordinate set of the outer boundary nodes of all adjacent effective force points is extracted, and this coordinate set is used as the effective distribution contour of the plantar pressure. In this embodiment, the effective pressure threshold is 5 kPa. In specific applications, the implementer can set it according to the specific situation. In this embodiment, the number of sensor nodes is 50. In specific applications, it is determined according to the specific situation.

[0026] Furthermore, the system's data acquisition module calculates the affine transformation matrix from the image coordinate system to the sensor's physical coordinate system. This calculation aims to maximize the overlap between the theoretical pressure center in the image and the measured pressure center, and between the image contour and the measured pressure contour. The system uses this transformation matrix to map the extracted skeletal keypoint coordinates to the sensor coordinate system, generating a calibrated set of skeletal keypoints. This calibration step eliminates initial spatial errors caused by differences in wearing position and establishes a unified computational coordinate system. It should be noted that the coordinates used subsequently in calculating physical distances are all calibrated coordinates.

[0027] In addition, the data acquisition module also acquires the glycated hemoglobin test values ​​of the target personnel.

[0028] The dynamic feature extraction module is used to construct a skeletal proximity vector based on the proximity of each sensor node to the skeletal key points; determine the pressure diffusion connectivity between each pair of sensor nodes based on the distance between sensor nodes and the glycated hemoglobin detection value; obtain the distribution morphology deviation index by calculating the optimal transmission cost of the plantar pressure data and the ideal safe distribution constructed based on the skeletal proximity vector; and obtain the local accumulation index based on the pressure diffusion connectivity.

[0029] The dynamic feature extraction module first obtains the physical distance between each sensor node and the nearest point among the skeletal keypoints based on the coordinates of each sensor node and each skeletal keypoint. This physical distance is recorded as the first distance; this distance is used to reflect the soft tissue thickness margin below the sensor position. Then, the skeletal proximity coefficient of each sensor node is determined based on the first distance. The skeletal proximity coefficient is negatively correlated with the first distance.

[0030] As a concrete example, the specific formula for calculating the bone proximity coefficient is given. The bone proximity coefficient of the i-th sensor node can be expressed as: in, This represents the skeletal proximity coefficient of the i-th sensor node. This represents the preset gain constant. This represents the distance between the i-th sensor node and the nearest point among the skeletal keypoints. This represents the smoothing constant.

[0031] In this embodiment, the preset gain constant is 100 and the smoothing constant is 5mm. The smoothing constant is introduced to prevent the denominator from being 0. The dimension of the distance between the sensor node and the nearest point in the skeletal key point is consistent with the dimension of the smoothing constant. In specific applications, the implementer can set the values ​​of the preset gain constant and smoothing constant according to the specific situation.

[0032] The closer the i-th sensor node is to the nearest point among the skeletal key points, the closer the i-th sensor node is to the bone, the thinner the soft tissue, and the less suitable it is to withstand high pressure. The larger the bone proximity coefficient of the i-th sensor node is.

[0033] Using the above method, the skeletal proximity coefficient of each sensor node can be obtained, and the skeletal proximity coefficients of all sensor nodes can be arranged in order to generate a skeletal proximity vector.

[0034] To simulate the hindering effect of tissue hardening on pressure conduction dispersion, a weighted network describing the connection relationships between sensor nodes needs to be constructed.

[0035] Specifically, a baseline sclerosis assessment coefficient is determined based on the patient's glycated hemoglobin test value, and the baseline sclerosis assessment coefficient is positively correlated with the glycated hemoglobin test value.

[0036] As a specific example, the formula for calculating the baseline hardening evaluation coefficient is given. The baseline hardening evaluation coefficient can be expressed as: in, Indicates the benchmark hardening evaluation coefficient. Indicates the hardening sensitivity coefficient. This indicates the patient's glycated hemoglobin test value. This indicates a preset health reference threshold. This represents the function that takes the maximum value.

[0037] The hardening sensitivity coefficient is a preset value determined based on statistical regression analysis of large-sample clinical data of diabetic foot and tissue elastic modulus. In this embodiment, the preset healthy reference threshold is 6.0%, and the hardening sensitivity coefficient is 0.2. In other embodiments, the system can also adaptively fine-tune the baseline hardening assessment coefficient according to the steepness of the pressure gradient in historical monitoring. The steeper the gradient, the higher the hardness, and the higher the baseline hardening assessment coefficient should be.

[0038] The baseline hardening assessment coefficient is used to characterize the degree of tissue hardening; the larger the baseline hardening assessment coefficient, the higher the degree of tissue hardening.

[0039] For any two sensor nodes: if the physical distance between the two sensor nodes is less than a preset distance threshold, then the connectivity weight between the two sensor nodes is calculated based on the physical distance between them and the baseline hardening evaluation coefficient. The connectivity weight is inversely proportional to both the physical distance and the baseline hardening evaluation coefficient. If the physical distance between the two sensor nodes is greater than or equal to the preset distance threshold, then the connectivity weight between the two sensor nodes is set to zero. In this embodiment, the preset distance threshold is 30mm. In specific applications, the implementer can set it according to the specific circumstances.

[0040] As a concrete example, the specific formula for calculating the connectivity weight is given. The connectivity weight between the i-th sensor node and the j-th sensor node can be expressed as: in, This represents the connectivity weight between the i-th sensor node and the j-th sensor node. Indicates the reference connectivity gain. This represents the physical distance between the i-th sensor node and the j-th sensor node. Indicates the benchmark hardening evaluation coefficient. This indicates the preset distance threshold.

[0041] The baseline connectivity gain characterizes the maximum allowable conduction flux between adjacent nodes in standard healthy tissue. Its value is determined based on clinical trials. In this embodiment, the baseline connectivity gain is set to 100.

[0042] When the physical distance between two sensor nodes is less than a preset distance threshold, the two corresponding sensor nodes are considered connected, which conforms to the rule that stress can only be transmitted within the neighborhood. The larger the baseline hardening assessment coefficient, the more severe the target person's condition may be. The smaller the connectivity weight between sensor nodes, the narrower the channel or the greater the resistance between the sensor nodes.

[0043] Using the above method, the connectivity weights between each pair of sensor nodes can be obtained. Based on the pressure diffusion connectivity between all pairs of sensor nodes, a pressure diffusion connectivity matrix is ​​constructed, with the number of rows and columns being equal to 1. , The number of sensor nodes is represented by the element in the pressure diffusion connectivity matrix. The elements in the matrix represent the connectivity weights between the sensor nodes. Specifically, the element in the i-th row and j-th column of the pressure diffusion connectivity matrix represents the connectivity weight between the i-th sensor node and the j-th sensor node.

[0044] To assess whether the current pressure distribution pattern is dangerous, i.e. whether it is excessively concentrated in the bony protrusion area, this embodiment will use the Sinkhorn algorithm in optimal transport theory. The logical essence of this algorithm is to calculate a transport cost: how much it would cost to shift the current actual pressure distribution to an ideal safe distribution.

[0045] Specifically, the plantar pressure distribution data of each sensor node is normalized into a source probability distribution. The source probability distribution of the i-th sensor node can be expressed as: in, Let represent the source probability distribution of the i-th sensor node. This represents the plantar pressure distribution data of the i-th sensor node at the k-th time. This represents the plantar pressure distribution data of the j-th sensor node at time k. Indicates the number of sensor nodes. This indicates a parameter for preventing zeroing.

[0046] In this embodiment, the value of the zero-prevention parameter is 0.001. In specific applications, the implementer can set it according to the specific situation.

[0047] The target probability distribution is generated based on the bone proximity vector. The target probability distribution of the i-th sensor node can be expressed as: in, Let represent the target probability distribution of the i-th sensor node. Represents the distribution focusing coefficient. This represents the skeletal proximity coefficient of the i-th sensor node. This represents the skeletal proximity coefficient of the j-th sensor node. This represents an exponential function with the natural constant as its base.

[0048] In this embodiment, the distribution focusing coefficient is set to 0.5. In specific applications, the implementer can set it according to the specific circumstances.

[0049] The target probability distribution describes an ideal safety condition where, in such a case, the more prominent the bone, the less pressure should be distributed; conversely, the deeper the bone, the greater the pressure should be distributed. The larger the bone proximity coefficient of the i-th sensor node, the closer the i-th sensor node is to the bone. The smaller the value, the smaller the target probability distribution of the i-th sensor node.

[0050] Next, the optimal transport algorithm is used to calculate the transmission cost between the source probability distribution and the target probability distribution, and the transmission cost is used as the distribution shape deviation index.

[0051] Specifically, two auxiliary scaling vectors are initialized. and These two scaling vectors are all 1s. Construct the cost kernel matrix; the elements of the cost kernel matrix can be represented as: in, This represents the element in the i-th row and j-th column of the cost kernel matrix. This represents the physical distance between the i-th sensor node and the j-th sensor node, which is also the first distance between the i-th sensor node and the j-th sensor node. Represents the regularization parameter. This represents an exponential function with the natural constant as its base.

[0052] In this embodiment, the regularization parameter is set to 0.8 times the average physical distance between all sensor nodes. In specific applications, the implementer can set it according to the specific situation.

[0053] The Sinkhorn iterative algorithm is employed, with the system performing a fixed number of alternating iterations to update the scaling vector. The Wasserstein distance between the source and target probability distributions is calculated based on the constructed cost kernel matrix and used as the transmission cost. In this embodiment, the fixed number of iterations is 20; however, in specific applications, the implementer can adjust this number according to their specific circumstances. The alternating iterative update can be represented as: in, This indicates element-wise division. Represents the cost kernel matrix, This indicates transpose.

[0054] As a concrete example, the distribution morphology deviation index can be expressed as: in, This represents the distribution shape deviation index at the k-th sampling time. and This represents the scaling vector generated by the Sinkhorn iterative algorithm to satisfy the edge constraints of the source and target distributions. This represents the element in the i-th row and j-th column of the cost kernel matrix. This represents the physical distance between the i-th sensor node and the j-th sensor node.

[0055] The magnitude of the distribution morphology deviation index reflects the degree of misalignment between the actual and ideal distribution. If the current pressure is mainly distributed in areas with smaller bone proximity vector values ​​(i.e., the safe zone with thick flesh), then the actual distribution... Compared with the ideal distribution The cost of transportation is very small. A smaller value indicates a safer morphology. Conversely, if pressure is concentrated in an area with a high bone proximity vector value—the danger zone of a prominent bone—it requires a greater distance to transport the pressure to a safe area, resulting in a higher transportation cost. The value has increased significantly.

[0056] In this embodiment, to accurately quantify the conductivity of tissue surrounding the stress point, the sensor array is treated as a flow network defined by graph theory. Unlike existing technologies that cannot sense tissue stiffness, this embodiment uses the pressure diffusion connectivity matrix as the topological basis of this network.

[0057] First, for the k-th sampling time, a high-pressure region is determined based on the plantar pressure data at that sampling time. This high-pressure region consists of a preset number of sensor nodes with the largest pressure amplitude. These nodes represent the injection center of the current pressure, i.e., the high-pressure area that needs to be relieved. The set of these nodes is denoted as the source point set. In this embodiment, the preset number is 3; in specific applications, the implementer can set it according to the specific situation. Nodes at the physical edge of the sensor array are defined as edge regions. A ring of nodes at the physical edge of the sensor array is forcibly defined as sink points, representing stress that ultimately needs to dissipate towards the foot boundary or non-contact area. All sink points constitute the sink point set.

[0058] Then, the system executes a maximum flow algorithm (e.g., the Edmonds-Karp algorithm) in the network defined by the pressure diffusion connectivity matrix. This algorithm simulates the process of pressure flow diffusing from the source to the sink, specifically including: (1) Network construction: The nodes in the network are sensor units, and the edges are the connections between nodes. The capacity of the edges is determined by the elements in the pressure diffusion connectivity matrix. The higher the glycated hemoglobin detection value, the harder the tissue; the smaller the connectivity weight between sensor nodes, that is, the smaller the capacity of the edges, the thinner the channels.

[0059] (2) Path search: The algorithm searches for all feasible paths from the source set to the sink set in the network, i.e. augmenting paths.

[0060] (3) Flow calculation: For each path, the maximum flow that can pass through is limited by the edge with the smallest capacity on that path (the bottleneck edge). The algorithm accumulates the flow of all paths until no new path can be found, and obtains the maximum throughput of the network.

[0061] Furthermore, the local accumulation index is obtained based on the ratio of the sum of pressure data in the high-pressure area to the maximum evacuation flux.

[0062] As a concrete example, the specific formula for calculating the local accumulation index is given. The local accumulation index at the k-th sampling time can be expressed as: in, Represents the local accumulation index at the k-th sampling time. This represents the number of nodes in the source set. This represents the plantar pressure data of the nth node in the source point set at the kth sampling time. This represents the maximum diversion throughput.

[0063] It should be noted that when calculating the local accumulation index, if the denominator of the calculation formula is 0, a preset zero-prevention parameter is added to the denominator of the 0 part, and then substituted into the above formula for calculation. In this embodiment, the preset zero-prevention parameter is 0.0001. In specific applications, the implementer can set it according to the specific situation.

[0064] A higher maximum drainage flux and lower plantar pressure data indicate a stronger network drainage capacity and a smaller local accumulation index, meaning the pressure can be dispersed more smoothly without accumulation. Conversely, a lower maximum drainage flux and higher plantar pressure data indicate that the network is riddled with bottlenecks, and even with low input pressure, the local accumulation index will increase significantly. This signifies the formation of stress islands that can enter but cannot dissipate, resulting in a higher risk of accumulation.

[0065] Using the above methods, the dynamic feature extraction module can obtain the local accumulation index at each sampling time.

[0066] The risk assessment module is used to obtain the cumulative load index based on plantar pressure data throughout the gait cycle, the distribution pattern deviation index, and the local accumulation index.

[0067] After the dynamic feature extraction module extracts the local accumulation index at each sampling time, the data is accumulated throughout the entire gait cycle.

[0068] The plantar pressure data, distribution morphology deviation index, and local accumulation index at each sampling time within the entire gait cycle (i.e., from heel touch to toe lift) are coupled and integrated to obtain the single-step composite risk index.

[0069] Specifically, for any sampling moment within the entire gait cycle: calculate the product of the plantar pressure data and the distribution pattern deviation index at that sampling moment, and record this product as the first product; calculate the product of the accumulation weighting coefficient and the local accumulation index, and record this product as the second product; combine the first product and the second product to obtain the risk characteristic value at that sampling moment; multiply the risk characteristic values ​​of all sampling moments within the entire gait cycle by the sensor's sampling period and then sum them up to obtain the single-step composite risk index. The gait cycle is obtained by taking the time period consisting of consecutive moments where the total plantar pressure is greater than a preset resting threshold as a complete gait cycle. In this embodiment, the preset resting threshold is 5N; in specific applications, the implementer can set it according to the specific situation. It should be noted that: to prevent prolonged static standing from causing an infinite extension of the single gait cycle, an upper limit threshold for the gait cycle duration is set, for example, 2 seconds; if the continuous force application time exceeds this upper limit threshold, the current gait cycle division is interrupted, and the time period is divided into multiple static load-bearing time windows of fixed duration for calculation.

[0070] As a concrete example, the specific formula for calculating the single-step compound risk index is given. The single-step compound risk index can be expressed as: in, This indicates a single-step compound risk index. This indicates the first sampling moment of the entire gait cycle. This represents the last sampling moment of the entire gait cycle. This represents the plantar pressure data at the k-th sampling time during the entire gait cycle. This represents the distribution morphology deviation index at the k-th sampling time within the entire gait cycle. Represents the accumulation weighting coefficient. Indicates the sampling period of the sensor. It represents the local accumulation index at the k-th sampling time.

[0071] The method for obtaining plantar pressure data at the kth sampling time within the entire gait cycle is as follows: the sum of the plantar pressure data of all sensor nodes at the kth sampling time within the entire gait cycle is taken as the plantar pressure data at the kth sampling time within the entire gait cycle.

[0072] In this embodiment, the accumulation weighting coefficient is 2.0. In specific applications, implementers can set it according to specific circumstances.

[0073] Represents the first product. This represents the second product. The formula for calculating the single-step composite risk index is a multiplicative coupling model used to characterize the superposition effect of risk factors. The distribution morphology deviation index acts as a multiplier. If the pressure is on the bone, the distribution morphology deviation index is large, and the basic risk is amplified; if the pressure is on the flesh, the distribution morphology deviation index is small, and the risk is reduced.

[0074] When tissue hardening leads to accumulation, and the local accumulation index at the k-th sampling time is large, the risk is further amplified significantly. The single-step composite risk index represents the cumulative damage equivalent to the foot soft tissues caused by that gait cycle.

[0075] To simulate the physiological characteristics of soft tissue injury and recovery during rest, the single-step composite risk index is accumulated to the cumulative load index to obtain a new cumulative load index, which characterizes the cumulative load on the patient's foot that day. This variable is automatically reset to zero at midnight each day. The system performs bidirectional updates to this variable based on the current activity status: whenever the system detects the end of a gait cycle and receives the single-step composite risk index for the corresponding gait cycle, it performs an accumulation operation. in, This represents the new cumulative load index. The above method accumulates the equivalent microscopic damage caused to each pair of tissues. Because... The weights for morphological misalignment and local accumulation have already been included, so the cumulative value here can truly reflect the damage process under pathological conditions.

[0076] To demonstrate that unloading and resting contribute to prolonging tissue lifespan, the total pressure value of the sensor array is monitored in real time. When the total pressure is detected to be consistently below a preset resting noise threshold for more than the effective resting time, the system initiates a repair mechanism. At this point, the system adjusts the pressure according to an exponential law. Attenuation is specifically represented as follows: in, The current duration of continuous stillness (in hours). For the repair rate constant, This represents an exponential function with the natural constant as its base.

[0077] The longer the rest period, the greater the decrease in the accumulated load value, indicating that the fatigue level of the tissue has been relieved. In this embodiment, the preset resting noise threshold is 5 kPa, the effective resting time is 15 minutes, and the repair rate constant is 0.1. In specific applications, implementers can set these values ​​according to specific circumstances.

[0078] Using the methods described above, the risk assessment module obtains the cumulative load index.

[0079] The feedback module is used to output corresponding risk warnings based on the value of the cumulative load index.

[0080] After receiving the cumulative load index from the risk assessment module, the feedback module performs a risk assessment based on the cumulative load index.

[0081] First, the tolerance decline value is calculated based on the difference between the glycated hemoglobin test value and the preset health reference threshold. Based on the tolerance decline value and the standard tolerance value of healthy people, a personalized injury threshold is determined. The personalized injury threshold is not lower than the preset minimum physiological limit threshold.

[0082] As a concrete example, the specific formula for calculating the personalized damage threshold is given. The personalized damage threshold can be expressed as: in, Indicates the individualized damage threshold. This represents the preset minimum physiological limit. This represents the standard tolerance value for healthy individuals. Indicates the tolerance decrease coefficient. This indicates the patient's glycated hemoglobin test value. This indicates a preset health reference threshold. This represents the function that takes the maximum value.

[0083] This represents the tolerance decrease value. In the above formula for calculating the personalized injury threshold, a preset minimum physiological limit is introduced to prevent the calculated result from being negative or too small at extremely high glycated hemoglobin levels.

[0084] In this embodiment, the standard tolerance value for healthy individuals is... The preset minimum physiological limit threshold is 0.2 times the standard tolerance value of healthy people; the tolerance reduction coefficient is 0.13. In specific applications, implementers can set it according to specific circumstances.

[0085] Then, the ratio of the cumulative load index to the personalized injury threshold is calculated and recorded as the first ratio. When the first ratio is less than a preset first threshold, it indicates a safe state, and the system's feedback module outputs a safety warning. When the first ratio is greater than or equal to the preset first threshold and less than a preset second threshold, it indicates a period of fatigue accumulation, and the system's feedback module outputs a fatigue warning, advising the patient to rest. When the first ratio is greater than or equal to the preset second threshold, it indicates a high-risk critical period, and the system's feedback module outputs a high-risk alarm, prompting the patient to immediately stop bearing weight. The preset first threshold is less than the preset second threshold. In this embodiment, the preset first threshold is... The preset second threshold is In specific applications, implementers can configure it according to the specific circumstances.

[0086] Thus, the dynamic monitoring system for diabetic foot risk provided in this embodiment has achieved accurate monitoring of diabetic foot risk.

[0087] The data acquisition module of the dynamic monitoring system for diabetic foot risk provided in this embodiment collects key skeletal points, plantar pressure data, and glycated hemoglobin levels from the feet of target individuals. It organically integrates skeletal spatial information reflecting anatomical structure, plantar pressure distribution information reflecting real-time biomechanical state, and biochemical indicators reflecting tissue metabolism and hardening degree, laying a multi-dimensional data foundation for subsequent risk feature extraction. This overcomes the information bias problem caused by relying solely on a single vertical pressure threshold in existing technologies. The dynamic feature extraction module constructs a skeletal proximity vector based on the proximity of sensor nodes to key skeletal points. This vector accurately characterizes the spatial distribution of soft tissue thickness in different areas of the foot. By calculating the optimal transmission cost between plantar pressure data and the ideal safety distribution constructed based on this skeletal proximity vector, it can accurately reflect the distribution morphology deviation index, which reflects the degree of deviation of the pressure distribution from the skeletal anatomy. The dynamic feature extraction module also considers the distance between sensor nodes and... Glycated hemoglobin (HbA1c) levels determine the pressure diffusion connectivity between each pair of sensor nodes. This connectivity fully considers the inhibitory effect of soft tissue hardening caused by glycosylation on the lateral transmission of pressure. Furthermore, a local accumulation index is obtained based on the pressure diffusion connectivity. This index quantitatively characterizes the difficulty of pressure dissipating from high-pressure areas to surrounding areas under the constraints of hardened tissue, thus accurately identifying the stress island effect caused by tissue hardening. This overcomes the deficiency of existing technologies in sensing the impact of changes in soft tissue hardness on risk. The risk assessment module integrates plantar pressure data, distribution morphology deviation index, and local accumulation index throughout the entire gait cycle to obtain a cumulative load index. This index combines the mechanical load reflecting pressure amplitude, the morphological deviation reflecting the rationality of pressure location, and the accumulation effect reflecting tissue dissipation capacity to form a quantitative characterization of the comprehensive damage suffered by the foot soft tissue in each gait cycle, providing a scientific and accurate assessment basis for the dynamic monitoring of diabetic foot risk.

[0088] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dynamic monitoring system for diabetic foot risk based on wearable devices, characterized in that, The system includes: The data acquisition module is used to acquire key skeletal points, plantar pressure data, and glycated hemoglobin test values ​​of the target person's feet; The dynamic feature extraction module is used to construct a skeletal proximity vector based on the proximity of each sensor node to key skeletal points; determine the pressure diffusion connectivity between each pair of sensor nodes based on the distance between sensor nodes and the glycated hemoglobin detection value; obtain the distribution morphology deviation index by calculating the optimal transmission cost of the plantar pressure data and the ideal safe distribution constructed based on the skeletal proximity vector; and obtain the local accumulation index based on the pressure diffusion connectivity. The risk assessment module is used to obtain the cumulative load index based on plantar pressure data throughout the gait cycle, the distribution pattern deviation index, and the local accumulation index. The feedback module is used to output corresponding risk warnings based on the value of the cumulative load index.

2. The dynamic monitoring system for diabetic foot risk based on wearable devices according to claim 1, characterized in that, The process of constructing a bone proximity vector based on the proximity of each sensor node to key bone points includes: The first distance between each sensor node and the nearest point among the skeletal key points is obtained respectively. The skeletal proximity coefficient of each sensor node is determined based on the first distance. The skeletal proximity coefficient is negatively correlated with the first distance. Arrange the bone proximity coefficients of all sensor nodes in order to generate a bone proximity vector.

3. The dynamic monitoring system for diabetic foot risk based on wearable devices according to claim 1, characterized in that, The determination of pressure diffusion connectivity between any two sensor nodes based on the distance between them and the glycated hemoglobin detection value includes: A baseline hardening assessment coefficient is determined based on the glycated hemoglobin test value, and the baseline hardening assessment coefficient is positively correlated with the glycated hemoglobin test value. For any two sensor nodes: if the physical distance between any two sensor nodes is less than a preset distance threshold, then the connectivity weight between the two sensor nodes is calculated based on the physical distance and the benchmark hardening evaluation coefficient, and the connectivity weight is inversely proportional to both the physical distance and the benchmark hardening evaluation coefficient; if the physical distance between any two sensor nodes is greater than or equal to the preset distance threshold, then the connectivity weight between the two sensor nodes is set to zero. A pressure diffusion connectivity matrix is ​​constructed based on the pressure diffusion connectivity between all pairs of sensor nodes.

4. The wearable device-based dynamic monitoring system for diabetic foot risk according to claim 2, characterized in that, The step of calculating the optimal transmission cost of the plantar pressure data and the ideal safe distribution constructed based on the bone proximity vector to obtain the distribution morphology deviation index includes: The plantar pressure distribution data of each sensor node is normalized into a source probability distribution; Generate a target probability distribution based on the bone proximity vector; The optimal transport algorithm is used to calculate the transmission cost between the source probability distribution and the target probability distribution, and the transmission cost is used as the distribution shape deviation index.

5. The wearable device-based dynamic monitoring system for diabetic foot risk according to claim 4, characterized in that, The step of calculating the transmission cost between the source probability distribution and the target probability distribution using the optimal transport algorithm includes: The Sinkhorn iterative algorithm is used to calculate the Wasserstein distance between the source probability distribution and the target probability distribution by alternately updating the scaling vector and constructing a cost kernel matrix based on the first distance, which is then used as the transmission cost.

6. The dynamic monitoring system for diabetic foot risk based on wearable devices according to claim 1, characterized in that, The step of obtaining the local accumulation index based on the pressure diffusion connectivity includes: High-pressure areas are determined based on plantar pressure data. These high-pressure areas consist of a preset number of sensor nodes with the largest pressure amplitude. Nodes at the physical edge of the sensor array are defined as edge areas. A graph theory network is constructed using each sensor node as a network node and the pressure diffusion connectivity as the capacity of the edges between nodes. The maximum flow algorithm is used to calculate the maximum diversion flux from the high-pressure region to the edge region. The local accumulation index is obtained based on the ratio of the sum of pressure data in the high-pressure region to the maximum diversion flux.

7. The wearable device-based dynamic monitoring system for diabetic foot risk according to claim 1, characterized in that, The step of obtaining the cumulative load index based on plantar pressure data throughout the gait cycle, the distribution morphology deviation index, and the local accumulation index includes: The plantar pressure data, distribution morphology deviation index, and local accumulation index at each sampling time throughout the entire gait cycle are coupled and integrated to obtain the single-step composite risk index. The cumulative load index is obtained by summing the single-step composite risk index to the cumulative load index.

8. The wearable device-based dynamic monitoring system for diabetic foot risk according to claim 7, characterized in that, The single-step composite risk index is obtained by coupling and integrating the plantar pressure data, distribution morphology deviation index, and local accumulation index at each sampling time throughout the entire gait cycle, including: For any sampling moment within the entire gait cycle: Calculate the first product of the plantar pressure data and the distribution morphology deviation index at any sampling time, and the second product of the accumulation weighting coefficient and the local accumulation index; combine the first product and the second product to obtain the risk characteristic value at any sampling time. The risk characteristic values ​​of all sampling moments within the entire gait cycle are multiplied by the sensor's sampling period and then summed to obtain the single-step composite risk index.

9. The dynamic monitoring system for diabetic foot risk based on wearable devices according to claim 1, characterized in that, The risk warning output based on the value of the cumulative load index includes: Calculate the first ratio of the cumulative load index to the personalized damage threshold; When the first ratio is less than a preset first threshold, a safety alert is output; when the first ratio is greater than or equal to the preset first threshold and less than a preset second threshold, a fatigue alert is output; when the first ratio is greater than or equal to the preset second threshold, a high-risk alarm is output; the preset first threshold is less than the preset second threshold.

10. The wearable device-based dynamic monitoring system for diabetic foot risk according to claim 9, characterized in that, The acquisition of the personalized damage threshold includes: The tolerance decline value is calculated based on the degree of difference between the glycated hemoglobin test value and the preset health reference threshold; Based on the tolerance decline value and the standard tolerance value of healthy people, a personalized injury threshold is determined; the personalized injury threshold is not lower than the preset minimum physiological limit threshold.

Images