Disease auxiliary detection method, system and equipment based on gait analysis and medium

Abstract

The invention relates to the technical field of auxiliary detection of gait abnormity related diseases, in particular to an auxiliary disease detection method, system and device based on gait analysis and a medium, the method is suitable for gait abnormity related diseases, and the method comprises the following steps: collecting multi-modal gait data of a subject, and preprocessing the multi-modal gait data; gait features related to a pathological mechanism are extracted from the gait features; taking the gait features as structured nodes, determining a conduction driving association relationship among the structured nodes, and constructing an individual gait specific feature chain; aligning the individual gait specific feature chain with the structured nodes of a pre-constructed standard gait specific feature chain, and calculating the total matching degree score and the position matching degree of the structured nodes between the chains; and obtaining a risk assessment result of the target disease according to the total matching degree score and the position matching degree, and outputting a targeted intervention suggestion. According to the method, objectivity and pathological relevance of auxiliary detection are greatly improved, and scientific assessment of various gait-related disease risks is realized.

Description

Technical Field

[0001] This invention relates to the field of auxiliary detection technology for gait abnormality-related diseases, and more specifically, to a disease auxiliary detection method, system, device, and medium based on gait analysis. Background Technology

[0002] Early diagnosis and intervention of gait abnormality-related diseases (such as motor-related neurological disorders, peripheral musculoskeletal disorders, and musculoskeletal disorders) are crucial for improving patient prognosis and slowing disease progression. These diseases often present with gait dysfunction as the core clinical manifestation, and their pathological mechanisms are diverse, encompassing abnormalities in the motor control pathways of the nervous system and structural or functional abnormalities of the musculoskeletal system. This leads to abnormal timing of the stance and swing phases in the gait cycle, abnormal joint range of motion, or decreased balance regulation. However, current clinical auxiliary detection methods for these diseases still have many limitations, making it difficult to meet the needs of early screening.

[0003] From a clinical testing perspective, traditional diagnosis relies heavily on the subjective judgment of doctors, making preliminary assessments by observing patients' gait and inquiring about symptoms and medical history, lacking objective and quantitative indicators. This approach is not only susceptible to differences in doctors' experience and the observation environment, but also struggles to detect subtle gait abnormalities in the early stages of the disease. In most gait-related diseases, gait changes are often subtle in the early stages, making it difficult to distinguish from normal aging or mild motor function decline by visual observation alone, resulting in many patients missing the opportunity for early intervention.

[0004] With technological advancements, some clinical practices have begun to incorporate motion capture devices or pressure-sensing insoles to collect multimodal gait data and extract fundamental parameters such as stride length, cadence, and gait cycle, attempting to aid diagnosis through quantitative indicators. However, existing technologies still suffer from the following shortcomings: Firstly, most systems focus only on single or isolated gait parameters, failing to consider that gait is a dynamic process involving the synergistic effects of multiple features. For instance, abnormal cadence may be correlated with limited lower limb joint mobility and decreased balance regulation. Isolated analysis of a single parameter cannot reflect the systemic impact of pathological mechanisms, leading to insufficient detection specificity. Secondly, existing methods rely solely on comparisons with healthy individuals to identify abnormalities, making it difficult to distinguish similar gait manifestations caused by different diseases.

[0005] Therefore, there is an urgent need for an auxiliary detection technology for gait abnormality-related diseases that can objectively quantify gait characteristics, reveal pathological correlations between characteristics, and match disease-specific patterns. Summary of the Invention

[0006] The purpose of this invention is to provide a disease-assisted detection method, system, device, and medium based on gait analysis, in order to solve the technical problem of how to objectively quantify gait characteristics, reveal pathological correlations between characteristics, match disease-specific patterns, and realize the auxiliary detection of gait abnormality-related diseases.

[0007] This invention is achieved through the following technical solution: a disease-assisted detection method based on gait analysis, applicable to gait abnormality-related diseases, comprising the following steps: Multimodal gait data of the subjects were collected, and the multimodal gait data were preprocessed. Extract gait features related to pathological mechanisms from the multimodal gait data; Using the gait features as structured nodes, and determining the correlation between the transmission and driving relationships between each structured node, an individual gait-specific feature chain is constructed. Align the structured nodes of the individual gait-specific feature chain with the pre-constructed standard gait-specific feature chain, and calculate the total matching score and position matching degree of the structured nodes between the chains; The risk assessment results for the target disease are obtained based on the total matching score and the location matching score, and intervention recommendations are output.

[0008] According to a preferred embodiment, the correlation of the propagation drive between each structured node is determined, specifically including: A two-way Granger causality test based on pathological priors was used to determine the transmission-driven associations between structured nodes.

[0009] According to a preferred embodiment, the expression for the correlation relationship of the propagation drive between each structured node is as follows:

[0010] In the above formula, This represents the variance of the prediction error in a conditional prediction model. This represents the variance of the prediction error in an unconditional prediction model. Represents structured nodes For structured nodes The conduction strength; Among them, the unconditional prediction model is defined as using only... Past predictions The future can be expressed as follows:

[0011] In the above formula, express time Corresponding gait features, This represents the intercept term of the unconditional prediction model. express The lag order, In the unconditional prediction model Lag The coefficient of the period, express time Corresponding gait features, Indicates prediction error; The definition of a conditional prediction model is: and Past predictions The future can be expressed as follows:

[0012] In the above formula, This represents the intercept term of the conditional prediction model. In conditional prediction models Lag The coefficient of the period, express The lag order, express Lag The coefficient of the period, express time Corresponding gait features, This indicates the new prediction error.

[0013] According to a preferred embodiment, the method further includes: determining anomaly scores for each of the structured nodes by quantifying the gait features, as expressed below:

[0014] In the above formula, Represents disease D structured nodes Abnormal scores, , Represents structured nodes The quantized values ​​of the corresponding gait features, Represents disease D structured nodes The maximum health value. Represents disease D structured nodes Mild pathological threshold, Represents disease D structured nodes The threshold for severe pathology.

[0015] According to a preferred embodiment, the method further includes introducing a disease-specific calibration factor to calibrate the abnormal scores of each of the structured nodes, as shown in the following expression:

[0016] In the above formula, This is used to limit the maximum score for anomalies to 100 points. This represents a disease-specific calibration factor that is positively correlated with the strength of its association with gait characteristics and pathology.

[0017] According to a preferred embodiment, the total matching score of structured nodes between chains is calculated as follows: For each structured node pair after inter-chain alignment The difference in abnormal scores is calculated using the following expression:

[0018] In the above formula, The standard gait-specific feature chain is represented by structured nodes. Standard scoring; The total match score is calculated based on the difference in anomaly scores, as shown in the following expression:

[0019] In the above formula, Represents weights, in relation to structured nodes. The priorities of related pathological steps are positively correlated. Indicates the most likely penalty option. This indicates the weighted cumulative penalty item.

[0020] According to a preferred embodiment, the process for calculating the positional matching degree of structured nodes between chains is as follows:

[0021] In the above formula, Indicates the degree of positional matching. , Represents the set of abnormal nodes It consists of structured nodes in the individual gait-specific feature chain whose abnormal scores exceed the corresponding disease severity pathology threshold. This represents the set of standard aberration nodes in the standard gait-specific feature chain of disease D. The cardinality of a set.

[0022] This invention also provides a disease-assisted detection system based on gait analysis, which is applied to the disease-assisted detection method based on gait analysis described above. The system includes: A multimodal data acquisition unit is used to acquire multimodal gait data of the subject and preprocess the multimodal gait data; A gait feature extraction unit is used to extract gait features related to pathological mechanisms from the multimodal gait data; A gait-specific feature chain construction unit is used to construct an individual gait-specific feature chain by using the gait features as structured nodes and determining the correlation between the transmission and driving relationships between the structured nodes. The calculation unit is used to align the structured nodes of the individual gait-specific feature chain with the pre-constructed standard gait-specific feature chain, and to calculate the total matching score and position matching degree of the structured nodes between the chains; The risk assessment unit is used to obtain the risk assessment result of the target disease based on the total matching score and the location matching degree, and output intervention recommendations.

[0023] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the disease-assisted detection method based on gait analysis as described above.

[0024] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the disease-assisted detection method based on gait analysis as described above.

[0025] The technical solution of the disease-assisted detection method, system, device and medium based on gait analysis provided by this invention has at least the following advantages and beneficial effects: This invention uses gait features as structured nodes and constructs feature chains, which can systematically capture the transmission and driving relationship between gait features, better fit the systemic influence mechanism of diseases on motor control pathways, and greatly improve the objectivity and pathological correlation of detection; at the same time, by aligning individual feature chains with standard feature chains and calculating the total matching degree score and position matching degree, the degree of fit between individual gait abnormalities and disease features can be accurately quantified, realizing the scientific assessment of disease risk. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the disease-assisted detection method based on gait analysis provided in Embodiment 1 of the present invention. Figure 2 This is a structural block diagram of the disease-assisted detection system based on gait analysis provided in Embodiment 2 of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0028] Example 1 This invention provides a disease-assisted detection method based on gait analysis, which is applicable to gait abnormality-related diseases. Figure 1 This is a flowchart illustrating the disease-assisted detection method based on gait analysis. (See attached diagram) Figure 1 As shown, this disease-assisted detection method based on gait analysis includes the following steps: Step 1: Collect multimodal gait data from the subjects and preprocess the multimodal gait data; In some embodiments of this example, an inertial measurement unit, pressure sensor, and encoder on a wearable device can be used to achieve seamless real-time acquisition of motion data. The data acquisition frequency can be set according to different sensor types and required data accuracy, and no specific limitations are made here.

[0029] The inertial measurement unit may include an accelerometer and a gyroscope to obtain acceleration and angular velocity information. In this embodiment, the accelerometer and gyroscope adopt microelectromechanical systems technology to achieve miniaturization and low power consumption. The accelerometer and gyroscope have a sampling frequency of 100Hz-200Hz. Pressure sensors can be placed on the hands and / or soles of the wearable device to acquire pressure data at the corresponding locations. The pressure sensor acquisition frequency is 50Hz-100Hz. Preferably, the pressure sensor on the sole of the foot can be a thin-film pressure sensor, which is evenly distributed on the bottom of the wearable device in contact with the ground to accurately measure the pressure distribution on the sole of the foot. The pressure sensor on the hand can be a six-dimensional force sensor to acquire all the forces and torques acting on the hand in space, enabling comprehensive monitoring of the forces acting on the hand. The encoder is installed on the rotating parts of each joint of the wearable device to accurately measure the joint angle and rotation speed. The encoder's acquisition frequency is related to the joint movement speed to ensure that it can accurately capture the subtle changes in the subject's movements during walking.

[0030] Furthermore, regarding the preprocessing of multimodal gait data, the preprocessing used in this embodiment includes denoising, normalization, data binning, and standardization; Specifically, for denoising the data acquired by the accelerometer and gyroscope, this embodiment uses the Kalman filter algorithm to remove high-frequency noise in real time; for denoising the pressure sensor data, this embodiment uses a moving average filter for smoothing; for normalizing the motion data, this embodiment scales the data from different types of sensors to a uniform range to eliminate the impact of differences in data magnitude on subsequent analysis; for binning the motion data, this embodiment divides the data according to the gait cycle, and the start and end points of the gait cycle are determined by changes in plantar pressure or acceleration data characteristics; for standardizing the motion data, this embodiment uses Z-score standardization to easily eliminate the influence of dimensions.

[0031] Step 2: Extract gait features related to pathological mechanisms from the multimodal gait data; It should be noted that gait features can be one or more of the following: basic kinematic parameters, spatiotemporal coupling features, and dynamic response features. They can be set according to the specific disease type that needs to be assisted in detection. This implementation does not impose any specific restrictions. Basic kinematic parameters reflect the macroscopic performance of gait and can be directly acquired through inertial sensors, plantar membrane pressure sensors, and hand pressure sensors. These parameters include spatiotemporal gait parameters, dynamic characteristics, and kinematic features. Spatiotemporal gait parameters can be further subdivided into temporal and spatial parameters. Temporal parameters include, but are not limited to, gait cycle time, stance phase time, swing phase time, and double stance time. Spatial parameters include, but are not limited to, stride length, stride width, stride speed, and cadence. Dynamic characteristics reflect the changes in force and torque during walking and can be subdivided into ground reaction force, joint torque, muscle activation patterns, and plantar pressure distribution. Ground reaction force characteristics include, but are not limited to, the time curves and peak values ​​of vertical, anterior-posterior, and lateral ground reaction forces. Joint torque characteristics include, but are not limited to, torque changes in the hip, knee, and ankle joints. Muscle activation pattern characteristics include, but are not limited to, the activation sequence and intensity of major lower limb muscles. Plantar pressure distribution characteristics include, but are not limited to, peak pressure on the heel or forefoot. Kinematic characteristics describe the movement trajectory and angular changes of joints and body parts, and can be further subdivided into joint motion parameters, body center of gravity trajectory, and limb swing patterns. Among them, joint motion parameters include, but are not limited to, the range of flexion and extension angles of the hip, knee, or ankle joints and joint angular velocities; body center of gravity trajectory characteristics are mainly the three-dimensional motion trajectory of the body center of gravity during walking; and limb swing patterns are mainly the coordinated swing patterns of the upper and lower limbs.

[0032] Spatiotemporal coupling features reflect the dynamic changes of basic kinematic parameters and need to be extracted from the original multimodal gait data using algorithms. In this embodiment, spatiotemporal coupling features include symmetry features, periodic features, and event sequence features. Symmetry features include, but are not limited to, left-right stride length ratio, left-right joint angle peak difference, and left-right plantar pressure curve similarity. Periodic features include, but are not limited to, stride frequency, gait asymmetry index, and gait variability. Event sequence features include, but are not limited to, the time interval variation of gait events.

[0033] Dynamic response features are obtained by introducing active intervention prediction stimulus signals, which can reflect the response patterns of gait features and further amplify the differences between diseases. This is the key to distinguishing diseases with similar mechanisms but different subtypes. In this embodiment, dynamic response features include visual cue response, auditory beat response and perturbation response, which will not be elaborated on here.

[0034] To further clarify the extraction of gait features related to different pathological mechanisms, the following examples are provided for illustration, using typical gait abnormality-related diseases as examples: Parkinson's disease patients experience increased and more volatile transmission delays in the transmission of brain motor commands to muscle activation and limb movement due to damage to the basal ganglia-cortical circuit. This embodiment uses motion data collected by a synchronous inertial measurement unit and muscle activation data collected by an electromyography (EMG) sensor to quantify multimodal dynamic synchronicity and calculate the motion-EMG delay and variability. Based on this, a motion-EMG delay coupling coefficient is constructed, expressed as follows:

[0035]

[0036]

[0037]

[0038] In the above formula, Indicates the motion-electromyography delay coupling coefficient. Indicates the average delay. The coefficient of variation, representing the delay, reflects the basal ganglia's ability to maintain stable motion. This represents the total number of gait cycles collected. Indicates the first Motor-electromyographic delay per gait cycle Indicates the signal transmission efficiency of the associated basal ganglia. Indicates the inertial measurement unit in the first... The time point at which knee extension begins, detected in each gait cycle. Indicates the electromyography sensor at the first Quadriceps activation time points detected during gait cycles, under normal conditions and synchronous; It should be noted that the motor-electromyography delay in Parkinson's patients is characterized by increased and fluctuating speed. This synergistic change is pathologically specific. The motor-electromyography delay coupling coefficient obtained above can be used to quantify the risk of Parkinson's disease. The larger the value of the motor-electromyography delay coupling coefficient, the greater the risk of Parkinson's disease.

[0039] Step 3: Using the gait features as structured nodes, determine the transmission and driving relationships between each structured node, and construct an individual gait-specific feature chain; It should be noted that the individual gait-specific feature chain constructed in this embodiment represents a multi-node, chain-like feature response sequence based on pathological damage exhibited in multimodal gait data. Each structured node corresponds to the quantitative representation of a pathological link, and there is a pathological transmission-driven correlation between structured nodes, which can be represented as follows: ,in This represents the number of structured nodes in the individual gait-specific feature chain. The first element in the individual gait-specific feature chain represents the... A structured node.

[0040] In this embodiment, a bidirectional Granger causality test based on pathological priors is used to determine the transmission-driven association relationships between each structured node; the expression for the transmission-driven association relationships between each structured node is as follows:

[0041] In the above formula, This represents the variance of the prediction error in a conditional prediction model. This represents the variance of the prediction error in an unconditional prediction model. Represents structured nodes For structured nodes The conduction strength; Among them, the unconditional prediction model is defined as using only... Past predictions The future can be expressed as follows:

[0042] In the above formula, express time Corresponding gait features, This represents the intercept term of the unconditional prediction model. express The lag order, In the unconditional prediction model Lag The coefficient of the period, express time Corresponding gait features, Indicates prediction error; The definition of a conditional prediction model is: and Past predictions The future can be expressed as follows:

[0043] In the above formula, This represents the intercept term of the conditional prediction model. In conditional prediction models Lag The coefficient of the period, express The lag order, express Lag The coefficient of the period, express time Corresponding gait features, This indicates the new prediction error.

[0044] Furthermore, the prediction errors of the unconditional prediction model and the conditional prediction model are compared using the F-test. If... The variance is significantly smaller than The variance indicates the introduction of The lag term can be significantly improved The prediction accuracy is assessed, and the correlation direction is determined by combining the reverse test; if both sides are insignificant, there is no correlation; based on the transmission strength... The quantification results are sorted in descending order, and strong correlation directional links are selected first to construct individual gait-specific feature chains.

[0045] Furthermore, since pathological damage is progressive, this embodiment also includes determining the anomaly score of each structured node by quantifying the gait features to achieve quantitative grading, as shown in the following expression:

[0046] In the above formula, Represents disease D structured nodes Abnormal scores, , Represents structured nodes The quantized values ​​of the corresponding gait features, Represents disease D structured nodes The maximum health value. Represents disease D structured nodes Mild pathological threshold, Represents disease D structured nodes The threshold for severe pathology.

[0047] It should be noted that this embodiment uses a piecewise nonlinear mapping function to fit the nonlinear law of pathological progression, directly corresponding to the pathological grade. The output abnormal score is back-mapped to the pathological damage grade 0 to 2, where 0 represents normal, 1 represents mild, and 2 represents severe. , and These three thresholds will structure the disease D nodes. The corresponding gait characteristics are quantitatively divided into healthy range, mildly abnormal range, and severely abnormal range, corresponding to pathological damage levels 0 to 2.

[0048] Furthermore, to avoid the mismatch in correlation strength that could cause abnormal scores to deviate from the true pathological state and thus generate quantification bias when calculating abnormal scores using a unified quantification function, this embodiment further introduces a disease-specific calibration factor to calibrate the abnormal scores of each structured node, as shown in the following expression:

[0049] In the above formula, This is used to limit the maximum score for anomalies to 100 points. This represents a disease-specific calibration factor that is positively correlated with the strength of its association with gait characteristics and pathology.

[0050] Step 4: Align the individual gait-specific feature chain with the structured nodes of the pre-constructed standard gait-specific feature chain, and calculate the total matching score and position matching degree of the structured nodes between the chains; In this embodiment, the calculation process for the total matching score of the structured nodes between chains is as follows: For each structured node pair after inter-chain alignment The difference in abnormal scores is calculated using the following expression:

[0051] In the above formula, Represents structured node pairs Euclidean distance, The standard gait-specific feature chain is represented by structured nodes. Standard scoring; The smaller the value, the more structured nodes in the individual gait-specific feature chain correlate with the disease. The higher the matching degree; The total match score is calculated based on the difference in anomaly scores, as shown in the following expression:

[0052] In the above formula, Represents weights, in relation to structured nodes. The priorities of related pathological steps are positively correlated. This represents the maximum possible penalty term, which, through the sum of the products of the weights and the maximum distance, unifies feature chains of different lengths to the same dimension, ensuring that the total matching score is within a certain range. Comparable within the interval, This represents the weighted cumulative penalty term, and the structured node pairs. The greater the Euclidean distance, the higher the cumulative penalty value; The closer the value is to 1, the higher the matching degree of the structured nodes between the individual gait-specific feature chain and the pre-constructed standard gait-specific feature chain.

[0053] Considering that relying solely on the total matching score for evaluation may result in similar scores but illogical pathologies, leading to misjudgments, this embodiment further calculates the positional matching degree of structured nodes between chains, as shown in the following expression:

[0054] In the above formula, Indicates the degree of positional matching. , Represents the set of abnormal nodes It consists of structured nodes in the individual gait-specific feature chain whose abnormal scores exceed the corresponding disease severity pathology threshold. This represents the set of standard aberration nodes in the standard gait-specific feature chain of disease D. The cardinality of a set.

[0055] Step 5: Obtain the risk assessment results of the target disease based on the total matching score and the location matching score, and output intervention recommendations.

[0056] In some implementations of this embodiment, weights are assigned to the total matching score and the location matching score, with the location matching score having a larger weight. Furthermore, the final matching score for each disease is output by weighted summation of the total matching score and the location matching score. The final matching score is then classified into three levels by designing a grading threshold, such as low risk, medium risk, and high risk. This risk assessment is used as the final output, and intervention recommendations corresponding to the risk are given. Specific details are not elaborated here.

[0057] In summary, this invention uses gait features as structured nodes and constructs feature chains to systematically capture the transmission and driving relationships between gait features, which is more in line with the systemic impact mechanism of diseases on motor control pathways, and greatly improves the objectivity and pathological correlation of detection. At the same time, by aligning individual feature chains with standard feature chains and calculating the total matching score and position matching degree, the degree of fit between individual gait abnormalities and disease features can be accurately quantified, enabling a scientific assessment of disease risk.

[0058] Example 2 This embodiment, based on the technical solution provided in Embodiment 1, provides a disease-assisted detection system based on gait analysis. This system is applied to the disease-assisted detection method based on gait analysis as described in Embodiment 1. (See also...) Figure 2 As shown, the system includes: A multimodal data acquisition unit is used to acquire multimodal gait data of the subject and preprocess the multimodal gait data; A gait feature extraction unit is used to extract gait features related to pathological mechanisms from the multimodal gait data; A gait-specific feature chain construction unit is used to construct an individual gait-specific feature chain by using the gait features as structured nodes and determining the correlation between the transmission and driving relationships between the structured nodes. The calculation unit is used to align the structured nodes of the individual gait-specific feature chain with the pre-constructed standard gait-specific feature chain, and to calculate the total matching score and position matching degree of the structured nodes between the chains; The risk assessment unit is used to obtain the risk assessment result of the target disease based on the total matching score and the location matching degree, and output intervention recommendations.

[0059] The functions of each module of the disease-assisted detection system based on gait analysis in this embodiment are the same as those in the embodiment of the disease-assisted detection method based on gait analysis, and the technical effects are the same. Therefore, they will not be repeated here.

[0060] Example 3 This embodiment is based on the technical solution provided in Embodiment 1, and provides an electronic device. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the disease-assisted detection method based on gait analysis as described in Embodiment 1.

[0061] Example 4 This embodiment, based on the technical solution provided in Embodiment 1, provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the disease-assisted detection method based on gait analysis as described in Embodiment 1.

[0062] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A disease-assisted detection method based on gait analysis, applicable to gait abnormality-related diseases, characterized in that, Includes the following steps: Multimodal gait data of the subjects were collected, and the multimodal gait data were preprocessed. Gait features related to pathological mechanisms are extracted from the preprocessed multimodal gait data; Using the gait features as structured nodes, and determining the correlation between the transmission and driving relationships between each structured node, an individual gait-specific feature chain is constructed. Align the structured nodes of the individual gait-specific feature chain with the pre-constructed standard gait-specific feature chain, and calculate the total matching score and position matching degree of the structured nodes between the chains; The risk assessment results for the target disease are obtained based on the total matching score and the location matching score, and intervention recommendations are output.

2. The disease-assisted detection method based on gait analysis as described in claim 1, characterized in that, Determine the propagation relationships between structured nodes, specifically including: A two-way Granger causality test based on pathological priors was used to determine the transmission-driven associations between structured nodes.

3. The disease-assisted detection method based on gait analysis as described in claim 2, characterized in that, The expression for the propagation and driving relationships between the structured nodes is as follows: In the above formula, This represents the variance of the prediction error in a conditional prediction model. This represents the variance of the prediction error in an unconditional prediction model. Represents structured nodes For structured nodes The conduction strength; Among them, the unconditional prediction model is defined as using only... Past predictions The future can be expressed as follows: In the above formula, express time Corresponding gait features, This represents the intercept term of the unconditional prediction model. express The lag order, In the unconditional prediction model Lag The coefficient of the period, express time Corresponding gait features, Indicates prediction error; The definition of a conditional prediction model is: and Past predictions The future can be expressed as follows: In the above formula, This represents the intercept term of the conditional prediction model. In conditional prediction models Lag The coefficient of the period, express The lag order, express Lag The coefficient of the period, express time Corresponding gait features, This indicates the new prediction error.

4. The disease-aided detection method based on gait analysis as described in any one of claims 1 to 3, characterized in that, The method further includes: determining the anomaly score of each structured node by quantifying the gait features, as shown in the following expression: In the above formula, Represents disease D structured nodes Abnormal scores, , Represents structured nodes The quantized values ​​of the corresponding gait features, Represents disease D structured nodes The maximum health value. Represents disease D structured nodes Mild pathological threshold, Represents disease D structured nodes The threshold for severe pathology.

5. The disease-assisted detection method based on gait analysis as described in claim 4, characterized in that, The method also includes introducing a disease-specific calibration factor to calibrate the abnormal scores of each of the structured nodes, as shown in the following expression: In the above formula, This is used to limit the maximum score for anomalies to 100 points. This represents a disease-specific calibration factor that is positively correlated with the strength of its association with gait characteristics and pathology.

6. The disease-assisted detection method based on gait analysis as described in claim 5, characterized in that, The calculation process for the total matching score of structured nodes between chains is as follows: For each structured node pair after inter-chain alignment The difference in abnormal scores is calculated using the following expression: In the above formula, The standard gait-specific feature chain is represented by structured nodes. Standard scoring; The total match score is calculated based on the difference in anomaly scores, as shown in the following expression: In the above formula, Represents weights, in relation to structured nodes. The priorities of related pathological steps are positively correlated. Indicates the most likely penalty option. This indicates the weighted cumulative penalty item.

7. The disease-assisted detection method based on gait analysis as described in claim 6, characterized in that, The calculation process for the position matching degree of structured nodes between chains is as follows: In the above formula, Indicates the degree of positional matching. , Represents the set of abnormal nodes It consists of structured nodes in the individual gait-specific feature chain whose abnormal scores exceed the corresponding disease severity pathology threshold. This represents the set of standard aberration nodes in the standard gait-specific feature chain of disease D. The cardinality of a set.

8. A disease-assisted detection system based on gait analysis, characterized in that, The system applies the disease-assisted detection method based on gait analysis as described in any one of claims 1 to 7, and the system comprises: A multimodal data acquisition unit is used to acquire multimodal gait data of the subject and preprocess the multimodal gait data; A gait feature extraction unit is used to extract gait features related to pathological mechanisms from the preprocessed multimodal gait data; A gait-specific feature chain construction unit is used to construct an individual gait-specific feature chain by using the gait features as structured nodes and determining the correlation between the transmission and driving relationships between the structured nodes. The calculation unit is used to align the structured nodes of the individual gait-specific feature chain with the pre-constructed standard gait-specific feature chain, and to calculate the total matching score and position matching degree of the structured nodes between the chains; The risk assessment unit is used to obtain the risk assessment result of the target disease based on the total matching score and the location matching degree, and output intervention recommendations.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the disease-assisted detection method based on gait analysis as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the disease-assisted detection method based on gait analysis as described in any one of claims 1 to 7.

Images