Automatic assessment system for Parkinson's disease severity and stage based on gait data

Abstract

An automated assessment system for the severity and staging of Parkinson's disease based on gait data includes a wearable multi-source gait acquisition module, a quantitative feature analysis module, a motor symptom severity quantification module, a Parkinson's disease severity quantification module, and an automated Hoehn-Yahr staging diagnosis module. The wearable multi-source gait acquisition module collects various gait signals; the quantitative feature analysis module processes the gait signals, extracts quantitative features corresponding to MDS-UPDRS sub-items, and uses these features for quantitative assessment of the sub-items; the motor symptom severity quantification module calculates the weights of motor features, performs feature fusion, and outputs a quantitative score for the severity of motor features; the Parkinson's disease severity quantification module calculates the weights of effective features, performs feature fusion, and outputs a quantitative score for the severity of Parkinson's disease; and the automated Hoehn-Yahr staging diagnosis module incorporates the quantitative features into the judgment process and outputs the staging result. This invention automates the diagnosis, enables real-time monitoring, and facilitates mobile and convenient diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of health monitoring technology, specifically relating to an automatic assessment system for the severity and staging of Parkinson's disease based on gait data. Background Technology

[0002] The severity of Parkinson's disease symptoms gradually increases over time. Although Parkinson's disease has a low mortality rate, it has a high disability rate, greatly impacting patients' daily lives. Moreover, current medications can only slow the progression of the disease, not cure it. Therefore, monitoring the severity of Parkinson's disease symptoms is crucial. Continuously sensing changes in the severity of symptoms helps doctors develop appropriate medication plans and treatment strategies.

[0003] Currently, the diagnosis of the severity of Parkinson's disease symptoms mainly relies on clinical diagnosis by physicians and specialists. The assessment method typically uses the Parkinson's Disease Rating Scale MDS-UPDRS (Goetz CG, Tilley BC, Shaftman SR, Stebbins GT, Fahn S, Martinez-Martin P, Poewe W, Sampaio C, Stern MB, Dodel R: MovementDisorder Society-sponsored revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS): scale presentation and clinimetric testing results. MovementDisorders: official journal of the Movement Disorder). Society 2008, 23(15):2129-2170.) Expert doctors assess and score patients' symptoms one by one based on their visual observation and experience. This method has some drawbacks: doctors assess and score patients' symptoms one by one according to the MDS-UPDRS scale, which is time-consuming. The scoring depends on the expert's knowledge and experience, which is subjective and a semi-quantitative method that is easy to cause misdiagnosis. This assessment method is also difficult to achieve real-time monitoring of the severity of Parkinson's disease symptoms and cannot obtain timely feedback on the condition, which is not conducive to the mobility and convenience of patient diagnosis. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, the present invention aims to provide an automatic assessment system for the severity and staging of Parkinson's disease based on gait data. This system extracts objective quantitative features with medical interpretability based on multiple gait signals and quantitative feature analysis. By combining the quantification of motor symptom severity, the quantification of Parkinson's disease severity, and automatic Hoehn-Yahr staging diagnosis, it achieves quantitative assessment of motor symptom severity, quantitative assessment of Parkinson's disease severity, and automatic Hoehn-Yahr staging diagnosis. The entire diagnostic process is completed automatically and quickly, reducing the risk of misdiagnosis. It enables real-time monitoring of the severity of Parkinson's disease symptoms, facilitating mobile and convenient diagnosis for patients.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] An automated assessment system for the severity and staging of Parkinson's disease based on gait data includes a wearable multi-source gait acquisition module, a quantitative feature analysis module, a motor symptom severity quantification module, a Parkinson's disease severity quantification module, and an automated Hoehn-Yahr staging diagnosis module. The wearable multi-source gait acquisition module is responsible for acquiring various gait signals; the quantitative feature analysis module processes the gait signals, extracts quantitative features corresponding to MDS-UPDRS sub-items, and uses these features for quantitative assessment of the sub-items; the motor symptom severity quantification module calculates the weights of motor features, performs feature fusion, and outputs a quantitative score for the severity of motor features; the Parkinson's disease severity quantification module calculates the weights of all effective features, performs feature fusion, and outputs a quantitative score for the severity of Parkinson's disease; and the automated Hoehn-Yahr staging diagnosis module incorporates the quantitative features into the judgment process and outputs the staging result.

[0007] The wearable multi-source gait acquisition module acquired gait signals of the subject as they completed a paradigmatic movement, which included reciprocating walking, standing still, and continuous heel tapping. The acquisition device has built-in sensors of various types, including pressure sensors, strain sensors, and inertial sensors, which acquired pressure signals, dynamic deformation signals, and inertial signals of the subject's gait. The inertial signals include acceleration signals and angular velocity signals.

[0008] The quantitative feature analysis module extracts features from various types of gait signals acquired by the wearable multi-source gait acquisition module, and filters out features that are correlated with the scores of each sub-item of the MDS-UPDRS of Parkinson's disease patients for the assessment of specific symptoms.

[0009] The motor symptom severity quantification module takes the features representing motor symptoms extracted from the quantitative feature analysis module as input, calculates the weight of each motor feature through a regression algorithm, performs feature fusion according to the weight ratio, and outputs a motor symptom severity quantification score. In addition, doctors and experts, based on the motor symptom severity quantification score and their own experience and knowledge, return to correct the weights of the features, making the assessment of the severity of motor symptoms more accurate.

[0010] The Parkinson's disease severity quantification module takes all valid features selected in the quantitative feature analysis module as input, calculates the weight of each valid feature through a regression algorithm, performs feature fusion according to the weight ratio, and outputs a Parkinson's disease severity quantification score. If doctors or experts have objections to the output Parkinson's disease severity quantification score, they can return to correct the feature weights based on their experience and knowledge, making the assessment of Parkinson's disease severity more accurate.

[0011] The aforementioned automatic Hoehn-Yahr staging diagnosis module determines the subject's symptom status based on quantitative features and medical diagnostic standards, and completes the Hoehn-Yahr staging diagnosis. The entire process is an automatic workflow. Quantitative features are input, symptoms are graded and determined, and Hoehn-Yahr staging results are output. The determination of each symptom depends on a feature set composed of its related features, and the input feature set determines the direction of the determination.

[0012] Compared with the prior art, the present invention has the following beneficial technical effects:

[0013] This invention uses quantitative features corresponding to the MDS-UPDRS sub-projects to assess and analyze the severity of specific symptoms in Parkinson's disease patients. Combined with the Parkinson's disease severity quantification module, it realizes the quantitative diagnosis of Parkinson's disease severity. Combined with the automatic Hoehn-Yahr staging diagnosis module, it realizes the staging diagnosis of Parkinson's disease. This provides a quantitative and convenient diagnostic method for real-time monitoring of Parkinson's disease, improves the reliability of diagnosis and monitoring, and reduces the possibility of misdiagnosis.

[0014] This invention's wearable multi-source gait acquisition module collects rich gait information of various types, which can meet the requirements for relevance feature extraction of each sub-item of MDS-UPDRS; the quantitative feature analysis module constructs quantitative features corresponding to each sub-item of MDS-UPDRS, which can directly quantify and assess specific symptoms, and has medical interpretability and objectivity; the motor symptom severity quantification module and the Parkinson's disease severity quantification module output quantitative scores for motor symptom severity and Parkinson's disease severity, respectively. The whole process is completed quickly and in a short time. Expert knowledge is introduced to correct and update the feature weight ratios within the quantification module, which improves the accuracy and rationality of quantitative diagnosis; the automatic Hoehn-Yahr staging diagnosis module determines the patient's symptom status through grading, and can quickly, accurately and automatically output the patient's staging diagnosis results. Attached Figure Description

[0015] Figure 1 This is a system block diagram of the present invention.

[0016] Figure 2 This is a schematic box-shaped diagram illustrating the quantitative features of this invention.

[0017] Figure 3 This is a schematic diagram illustrating the effect of the quantitative score for the severity of Parkinson's disease in this invention.

[0018] Figure 4 This is a flowchart of the automatic Hoehn-Yahr staging diagnosis process of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] Reference Figure 1 An automatic assessment system for the severity and staging of Parkinson's disease based on gait data, including a wearable multi-source gait acquisition module, a quantitative feature analysis module, a motor symptom severity quantification module, a Parkinson's disease severity quantification module, and an automatic Hoehn-Yahr staging diagnosis module;

[0021] The wearable multi-source gait acquisition module is responsible for acquiring various gait signals and providing them to the quantitative feature analysis module;

[0022] The quantitative feature analysis module is responsible for processing gait signals, extracting quantitative features corresponding to each sub-item of MDS-UPDRS, assessing the specific symptoms of the subjects, and providing the features to the motor symptom severity quantification module, the Parkinson's disease severity quantification module, and the automatic Hoehn-Yahr staging diagnosis module.

[0023] The motor symptom severity quantification module is responsible for calculating the weights of motor features, performing feature fusion, and outputting a motor feature severity quantification score.

[0024] The Parkinson's disease severity quantification module is responsible for calculating the weights of all effective features, performing feature fusion, and outputting a Parkinson's disease severity quantification score.

[0025] The automatic Hoehn-Yahr staging diagnosis module is responsible for incorporating quantitative features into the judgment process and outputting staging results.

[0026] The wearable multi-source gait acquisition module collects gait signals of the subject as they complete paradigmatic movements. The paradigmatic movement process includes, but is not limited to, reciprocating walking, standing still, and continuous heel tapping. The gait signal acquisition process is completed within 3-5 minutes. The acquisition device has built-in various types of sensors, including pressure sensors, strain sensors, and inertial sensors, and collects the pressure signals, dynamic deformation signals, and inertial signals of the subject's gait. The inertial signals include acceleration signals and angular velocity signals. All collected signals are input into the quantitative feature analysis module.

[0027] The quantitative feature analysis module extracts features from various types of gait signals acquired by the wearable multi-source gait acquisition module. It extracts features correlated with the Parkinson's Disease MDS-UPDRS sub-project. Therefore, these features directly map to the presence and severity of specific symptoms, possessing medical interpretability and can be directly used for symptom assessment. The feature distribution is shown in the attached diagram. Figure 2 , Figure 2 This embodiment uses a box plot to represent the extracted quantitative features 40 of the MDS-UPDRS sub-items 40. Each MDS-UPDRS sub-item has a score of 0-4, where 0 represents normal and 4 represents severe. Higher scores indicate more severe symptoms for that sub-item. The distribution of quantitative features among subjects with different scores shows significant differences and is correlated with the sub-item scores. Quantitative features are used to directly assess the specific symptoms of the subjects. When building this module, a sufficient number of subjects' data were collected, and each quantitative feature was calculated to form a database, completing the module's construction. In this embodiment, the scores of the subjects' sub-items can be quickly assessed based on the range of the subjects' feature values, and the features are output to the motor symptom severity quantification module, the Parkinson's disease severity quantification module, and the automatic Hoehn-Yahr staging diagnosis module.

[0028] The motor symptom severity quantification module uses features representing motor symptoms as input to the quantitative feature analysis module. It calculates the weight of each motor feature using a regression algorithm, fuses features according to their weights, and outputs a motor symptom severity quantification score. The choice of regression algorithm is not unique or fixed, and includes, but is not limited to, linear regression, multinomial regression, support vector machine regression, decision tree regression, random forest regression, LASSO regression, Ridge regression, Elastic Net regression, and XGBoost regression. Furthermore, medical experts combine the motor symptom severity quantification score with their own experience and knowledge to adjust the weights of the returned features, making the module's assessment of motor symptom severity more accurate.

[0029] The Parkinson's disease severity quantification module takes all valid features from the quantitative feature analysis module as input, calculates the weight of each feature using a regression algorithm, and fuses the features according to their weights to output a Parkinson's disease severity quantification score. Similar to the motor symptom severity quantification module, the regression algorithms include, but are not limited to, linear regression, multinomial regression, support vector machine regression, decision tree regression, random forest regression, LASSO regression, Ridge regression, ElasticNet regression, and XGBoost regression. If doctors or experts disagree with the module's output score, they can adjust the feature weights based on their experience and knowledge to make the module's assessment of Parkinson's disease severity more accurate. The assessment effect diagram is shown below. Figure 3 , Figure 3 The graph shows the effect of the Parkinson's disease severity quantification score. The horizontal axis represents the subject's MDS-UPDRS scale score, and the vertical axis represents the Parkinson's disease severity quantification score output by this module. The higher the score, the more severe the subject's disease. Most of the data points are located on the diagonal, showing a strong linear relationship. Correlation analysis between the Parkinson's disease severity quantification score output by the Parkinson's disease severity quantification module and the scale score revealed a high correlation between them.

[0030] The automated Hoehn-Yahr staging diagnostic module completes the Hoehn-Yahr staging diagnosis by classifying and determining the subject's symptom status. The entire process is automated. The input consists of features selected by the quantitative feature analysis module, which filters features relevant to the symptoms determined at each level to form feature sets for each level. Based on the specific values ​​of each feature set, the symptom assessment flow for each level is determined, and the Hoehn-Yahr staging result is output. Figure 4 , Figure 4The flowchart for the automated Hoehn-Yahr staging diagnosis is as follows: the value of the feature set determines the judgment result of each symptom level. Feature set 1 consists of features representing all symptoms of the subject; feature set 2 consists of features representing unilateral limb symptoms; feature set 3 consists of features representing trunk symptoms; feature set 4 consists of features representing severe bilateral limb symptoms; feature set 5 consists of features representing no balance disorder; feature set 6 consists of features representing recoverable after a pull test; and feature set 7 consists of features representing independent walking or standing. The quantitative features extracted from the subjects are incorporated into the diagnostic process, automatically generating feature sets at various levels for symptom assessment: First, it is determined whether the subject has symptoms. If no symptoms are present, the subject is at level 0, and the process terminates. If symptoms are present, the next step is to determine whether the subject has unilateral limb symptoms. If unilateral limb symptoms are present, the next step is to determine whether the subject has trunk symptoms. If no trunk symptoms are present, the subject is at level 1, and the process terminates. If trunk symptoms are present, the subject is at level 1.5, and the process terminates. If no unilateral limb symptoms are present, i.e., bilateral limb symptoms are present, the next step is to determine whether the subject has severe bilateral limb symptoms. For lateral limb symptoms, if there are no severe bilateral limb symptoms, the next step is to determine if the subject has a balance disorder. If there is no balance disorder, the subject is classified as Level 2, and the process terminates. If there is a balance disorder, the next step is to determine if the subject can recover from the balance disorder by performing a pull-back test. If the subject can recover, the subject is classified as Level 2.5, and the process terminates; if the subject cannot recover, the subject is classified as Level 3, and the process terminates. If there are severe bilateral limb symptoms, the next step is to determine if the subject can walk or stand independently. If the subject can walk or stand independently, the subject is classified as Level 4, and the process terminates; if the subject cannot walk or stand independently, the subject is classified as Level 5, and the process terminates. The entire process is automated. By inputting the subject's quantitative characteristics, the Hoehn-Yahr staging diagnosis can be obtained quickly and accurately.

Claims

1. An automatic assessment system for the severity and staging of Parkinson's disease based on gait data, characterized in that: It includes a wearable multi-source gait acquisition module, a quantitative feature analysis module, a motor symptom severity quantification module, a Parkinson's disease severity quantification module, and an automatic Hoehn-Yahr staging diagnosis module; the wearable multi-source gait acquisition module is responsible for acquiring multiple gait signals; The quantitative feature analysis module is responsible for processing gait signals and extracting quantitative features corresponding to MDS-UPDRS sub-items for quantitative evaluation of the sub-items. The motor symptom severity quantification module is responsible for calculating the weights of motor features, performing feature fusion, and outputting a motor feature severity quantification score. The Parkinson's disease severity quantification module is responsible for calculating the weights of all effective features, performing feature fusion, and outputting a Parkinson's disease severity quantification score; the automatic Hoehn-Yahr staging diagnosis module is responsible for bringing the quantified features into the judgment process and outputting the staging results. The wearable multi-source gait acquisition module acquired gait signals of the subject as they completed a paradigmatic movement, which included reciprocating walking, standing still, and continuous heel tapping. The acquisition device has built-in sensors of various types, including pressure sensors, strain sensors, and inertial sensors, which acquired pressure signals, dynamic deformation signals, and inertial signals of the subject's gait. The inertial signals include acceleration signals and angular velocity signals. The quantitative feature analysis module extracts features from various types of gait signals acquired by the wearable multi-source gait acquisition module, and filters out features that are correlated with the scores of each sub-item of the MDS-UPDRS of Parkinson's disease patients for the assessment of specific symptoms. The aforementioned automatic Hoehn-Yahr staging diagnosis module determines the subject's symptom status based on quantitative features and medical diagnostic standards, and completes the Hoehn-Yahr staging diagnosis. The entire process is an automatic workflow. Quantitative features are input, symptoms are graded and determined, and Hoehn-Yahr staging results are output. The determination of each symptom depends on a feature set composed of its related features, and the input feature set determines the direction of the determination.

2. The system according to claim 1, characterized in that: The motor symptom severity quantification module takes the features representing motor symptoms extracted by the quantitative feature analysis module as input, calculates the weight of each motor feature through a regression algorithm, performs feature fusion according to the weight ratio, and outputs a motor symptom severity quantification score. In addition, based on the quantitative scores of motor symptom severity and their own experience and knowledge, medical experts return to adjust the weights of the features, making the assessment of the severity of motor symptoms more accurate.

3. The system according to claim 1, characterized in that: The Parkinson's disease severity quantification module takes all the effective features selected in the quantitative feature analysis module as input, calculates the weight of each effective feature through a regression algorithm, performs feature fusion according to the weight ratio, and outputs a Parkinson's disease severity quantification score. If doctors or experts disagree with the output score quantifying the severity of Parkinson's disease, they can adjust the weights of the features based on their experience and knowledge to make the assessment of the severity of Parkinson's disease more accurate.

Images