Cerebral small vascular disease cognitive impairment assessment method based on gait features and ensemble learning

Abstract

The invention discloses a brain small vascular disease cognitive disorder assessment method based on gait features and ensemble learning, and relates to the technical field of medical and industrial combination. According to the technical scheme, a wearable device is used for collecting gait data; the collected original data are preprocessed; the cognitive impairment prediction method comprises the following steps: predicting the cognitive degree of a CSVD patient by adopting a plurality of types of models, and establishing a cognitive impairment prediction model by using an integrated learning method based on a voting method and adopting a model selection method based on an elimination method; and evaluating the performance of the model by using the accuracy, the precision, the recall rate, the F1 score and the area under the curve as evaluation indexes. The method has the beneficial effects of accurate, objective and reliable evaluation, early prediction and intervention, high adaptability, simplicity and convenience in operation, capability of providing scientific basis and the like; the popularization and application of the method can provide new thoughts and tools for cognitive impairment evaluation of the CSVD patient, and the cognitive health and life quality of the patient can be improved.

Description

Technical Field

[0001] The present invention relates to the field of medical-engineering integration technology, and more specifically, to a cognitive impairment assessment method based on gait characteristics and integrated learning. Background Art

[0002] Cerebral small vessel disease (CSVD) is one of the main factors causing cognitive dysfunction and dementia in the elderly. CSVD is a pathological process of cerebral arterioles, capillaries and venules caused by multiple factors. On magnetic resonance imaging (MRI), CSVD manifests as white matter hyperintensities (WMH), cerebral microbleeds (CMB), lacunar infarctions (LI), enlarged perivascular spaces (EPVS) and brain atrophy.

[0003] Cognitive impairment is a major health problem worldwide, ranging from mild cognitive impairment (MCI) to severe dementia such as Alzheimer's disease, which seriously affects patients' ability to live independently. Vascular disease is the second most common cause of age-related cognitive impairment after Alzheimer's disease, and cerebrovascular disease may be the main cause of cognitive impairment, especially in East Asia.

[0004] Accurately assessing the cognitive function of CSVD patients is of great significance for early diagnosis of the disease, formulation of treatment plans and prognosis assessment. Currently, the Montreal Cognitive Assessment (MoCA) is widely used in clinical practice to assess the cognitive function of CSVD patients. However, traditional scale-based cognitive assessment methods usually rely on self-reports or clinical assessments of subjects and are easily affected by subjective factors. In certain specific cognitive areas or specific populations, the sensitivity of MoCA may be insufficient, and repeated assessments in a short period of time may affect the accuracy of the results. A study explored the relationship between cognitive function and gait, and proposed a cognitive quantitative assessment method based on gait parameters to predict cognitive level.

[0005] Gait analysis is a non-invasive, objective assessment method that can reflect the functional status of an individual's nervous system by quantifying various parameters during walking. Cognitive impairment caused by CSVD is the main cause of gait disorders. Gait problems in the elderly are associated with CSVD, early dementia, and dementia. Patients with cognitive impairment caused by CSVD show significant gait and balance disorders, which are proportional to the severity of cognitive decline. These gait disorders have a serious impact on daily life. The impact of CSVD on the gait of the elderly is mainly reflected in slower gait speed, shorter stride length, and balance disorders. CSVD affects gait and balance by directly damaging motor pathways, such as the prefrontal lobe and basal ganglia, or indirectly by affecting cognitive function. A systematic review of studies summarized the relationship between CSVD and gait disorders or fall risk, and found that most studies reported an association between CSVD and gait disorders. Understanding the relationship between gait, CSVD, and cognitive impairment is crucial for identifying individuals with cognitive impairment and dementia risk. Therefore, exploring how to use gait parameters to establish a predictive model for cognitive function in CSVD patients is of great research value for improving the diagnosis and management of CSVD patients.

[0006] In order to intervene in time, delay or prevent the progression of the disease and prevent further deterioration of cognitive function, early diagnosis and prediction of cognitive impairment are crucial. Therefore, there is an urgent need to develop new screening methods to identify patients at risk of cognitive decline caused by CSVD. Summary of the invention

[0007] In order to solve the technical problems existing in the above-mentioned prior art, the present invention establishes an evaluation method for the cognitive function of CSVD patients using gait parameters, proposes an integrated prediction model based on voting, and adopts a model selection method based on elimination method, which is applied to the prediction of cognitive function of CSVD patients and explores the optimal model combination; it aims to provide an early, objective and accurate cognitive impairment risk assessment tool to better identify patients, provide early intervention and improve cognitive health.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for evaluating cognitive impairment in cerebral small vessel disease based on gait characteristics and ensemble learning, the steps are as follows:

[0010] S1. Collect gait data using wearable devices;

[0011] S2, preprocessing the collected raw data;

[0012] S3. Several types of models were used to predict the cognitive level of CSVD patients, using an ensemble learning method based on voting and a model selection method based on elimination to establish a cognitive impairment prediction model.

[0013] S4. Use accuracy, precision, recall, F1 score and area under the curve as evaluation metrics to evaluate the performance of the model.

[0014] Furthermore, in step S1, four gait tasks are completed using the wearable device, including two single tasks and two dual tasks.

[0015] Furthermore, in step S2, before model training, the original data is preprocessed, and the missing values ​​are filled with the average value of the gait data of each task to remove abnormal data;

[0016] The symmetry of the data is evaluated by comparing the gait parameters of the left and right legs. The calculation method is as shown in formula (1):

[0017]

[0018] The dual-task cost is calculated by comparing the gait parameters under single-task and dual-task conditions. The calculation method is as shown in formula (2):

[0019]

[0020] All features are standardized and the cognitive level is used as the label.

[0021] Furthermore, in step S3, six different types of classifiers were used to predict the MoCA scores of CSVD patients, including logistic regression, support vector machine, decision tree, random forest, gradient boosting, and adaptive boosting;

[0022] Four evaluation indicators are used to evaluate the classification model: accuracy, precision, recall, and F1 score. The calculation method is as follows:

[0023]

[0024] Among them, TP represents true positive, TN represents true negative, FP represents false positive, and FN represents false negative;

[0025] The area under the curve is used for evaluation to quantify the classification performance of the model at different thresholds. AUC provides a comprehensive evaluation of the overall performance of the model, avoiding possible bias at a single threshold.

[0026] Furthermore, soft voting is used as a model integration method to integrate the prediction results of multiple base models and improve the overall prediction performance; the prediction probabilities of each base model are weighted averaged to obtain the final prediction result;

[0027] Furthermore, the elimination method is used as the model integration method. The specific steps are as follows:

[0028] T1. Initialize a set containing all base models;

[0029] T2. For each base model in the current model set, eliminate the model, calculate new weights and generate a new integrated model;

[0030] T3, select the base model with the best performance after elimination and eliminate the model from the set;

[0031] T4. Repeat the above steps until the optimal result is obtained;

[0032] T5. Output the final retained base model set as the integrated model.

[0033] Furthermore, the SMOTE algorithm is used for oversampling: new synthetic samples are generated by interpolating between samples of the minority class, thereby increasing the sample size of the minority class; and the grid search method is used to find the optimal model parameters.

[0034] Furthermore, the single task was to walk freely at a habitual speed; the dual task included a walking task with a plate and a walking task for a verbal fluency test.

[0035] Furthermore, in the task of walking with a plate, the participant walks with the plate in both hands and avoids dropping the cup in the plate during walking;

[0036] The verbal fluency test walking task: participants start walking while saying the name of any animal;

[0037] Furthermore, the collected gait parameters include stride length, stride speed, stride frequency, stance phase, swing phase, stride time, stance time, swing time, toe-off angle, and heel-strike angle of the left and right feet.

[0038] Beneficial effects of the present invention:

[0039] Compared with the prior art, the method for assessing cognitive impairment of cerebral small vessel disease based on gait characteristics and integrated learning of the present invention has the following technical features or beneficial effects:

[0040] (1) Improved evaluation accuracy: This paper constructs a high-precision cognitive impairment prediction model by integrating multiple machine learning models and gait features. The voting method and elimination method are used for model integration, which effectively reduces the overfitting risk of a single model and improves the overall prediction performance. The experimental results show that the AUC of the optimized integrated model reaches 0.9942, which significantly improves the accuracy of the evaluation.

[0041] (2) Objectivity and reliability: Traditional cognitive assessment methods rely on self-reports or clinical assessments by subjects and are easily affected by subjective factors. The present invention uses gait parameters as evaluation indicators, which is a non-invasive and objective evaluation method. Data collection through wearable devices ensures the objectivity and reliability of the data and reduces the interference of subjective factors.

[0042] (3) Early prediction and intervention: The assessment method provided by the present invention can identify the risk of cognitive impairment in CSVD patients at an early stage, providing the possibility for timely intervention, delaying or preventing the progression of the disease. This is of great significance for improving the cognitive health of patients and improving their quality of life.

[0043] (4) Strong adaptability: The present invention is not only applicable to the assessment of cognitive impairment in CSVD patients, but can also be expanded to other types of cognitive impairment assessments. By adjusting gait parameters and model parameters, it can meet the needs of different diseases and populations.

[0044] (5) Easy operation: Wearable devices are used to collect gait data, which is easy to operate and promote. At the same time, five-fold cross-validation and SMOTE technology are used for data preprocessing to improve the stability and reliability of the model and reduce the dependence on data quality.

[0045] (6) Providing scientific basis: This invention provides a scientific basis for early diagnosis, treatment plan formulation and prognosis evaluation of the disease by exploring the relationship between gait parameters and cognitive impairment in CSVD patients. It is helpful to promote research progress in related fields and promote the innovation and development of medical technology.

[0046] In summary, the present invention proposes a method for assessing cognitive impairment in cerebral small vessel disease based on gait characteristics and integrated learning, which has the beneficial effects of accurate assessment, objective and reliable, early prediction and intervention, strong adaptability, simple operation and provision of scientific basis. The promotion and application of this method will provide new ideas and tools for the assessment of cognitive impairment in CSVD patients, and help improve patients' cognitive health and quality of life. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0048] Figure 1 is a flow chart of determining the best model combination by the elimination method of the present invention;

[0049] Figure 22 is the ROC curve diagram of different models of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Figure 1-2 The method for assessing cognitive impairment in cerebral small vessel disease based on gait characteristics and integrated learning is further explained, and the technical solutions in the embodiments of the present application are clearly and completely described; obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments, and all other embodiments obtained by ordinary technicians in this field without making creative work based on the embodiments in the present application are within the scope of protection of the present application.

[0051] Example 1

[0052] A method for evaluating cognitive impairment in cerebral small vessel disease based on gait characteristics and ensemble learning mainly includes the following steps:

[0053] (1) Data collection: using wearable devices to collect gait data;

[0054] (2) Data processing: preprocessing the collected raw data;

[0055] (3) Model establishment: Six different types of models were used to predict the cognitive level of CSVD patients, including logistic regression, support vector machine (SVM), decision tree (DecisionTree), random forest (RandomForest), gradient boosting regression (GBR) and adaptive boosting (AdaBoost). An ensemble learning method based on voting and a model selection method based on elimination were used to establish a cognitive impairment prediction model.

[0056] (4) Model evaluation: Accuracy, precision, recall, F1 score and area under the curve (AUC) were used as evaluation indicators to evaluate the performance of the model.

[0057] The present invention uses gait parameters to predict cognitive impairment in CSVD patients and evaluates the performance of multiple machine learning models and ensemble methods. The ensemble model shows high accuracy in identifying cognitive impairment in CSVD patients.

[0058] Example 2

[0059] This embodiment is a new embodiment or a supplement to Embodiment 1. A method for evaluating cognitive impairment of cerebral small vessel disease based on gait characteristics and integrated learning, comprising the following steps:

[0060] (1) Data Collection

[0061] Wearable devices were used to collect multiple gait parameters such as distance, angle, and time during walking. Wearable devices were used to complete four gait tasks, including two single tasks and two dual tasks, with a walking length of more than 10 meters. The single task was to walk freely at a habitual speed; the dual tasks included a plate-carrying walking task and a verbal fluency test (VFT) walking task. Plate-carrying walking task: Participants walked with plates in both hands and avoided dropping the cups in the plates during walking. VFT walking task: Participants started walking and said the name of any animal at the same time (fruit names can be used instead for multiple repetitions). The collected gait parameters included the stride length, speed, frequency, support phase, swing phase, stride time, support time, swing time, toe-off angle, and heel-strike angle of the left and right feet.

[0062] (2) Data preprocessing

[0063] Before model training, the original data needs to be preprocessed first. Missing values ​​are filled with the average value of the gait data of each task to remove abnormal data.

[0064] Several derived features were calculated based on gait parameters, including step length, step speed, step frequency, stance phase, swing phase, stride time, stance time, swing time, heel strike angle, and toe-off angle under single-task and dual-task conditions, as well as their symmetry and dual-task cost.

[0065] Asymmetry index (ASI) can be evaluated by comparing the gait parameters of the left and right legs. The calculation method is as shown in Equation 1.

[0066]

[0067] Dual Task Cost (DTC) usually refers to the impact on the performance of the primary task (such as walking) when performing a second task. The dual task cost can be calculated by comparing the gait parameters under single-task and dual-task conditions. The calculation method is shown in Equation 2.

[0068]

[0069] Finally, 46 gait features were obtained. Each participant took multiple steps in the gait test, and each step was taken as an independent sample. In order to eliminate the influence of different dimensions between features, all features were standardized. Cognitive level (cognitive impairment, normal cognition) was used as a label.

[0070] (3) Model establishment

[0071] Six different types of classifiers were used to predict the MoCA scores of CSVD patients, including logistic regression, support vector machine (SVM), decision tree, random forest, gradient boosting, and adaptive boosting (AdaBoost). Four evaluation indicators were used to evaluate the classification model: accuracy, precision, recall, and F1 score, which are calculated as follows:

[0072]

[0073] Note: TP (True Positive, true positive), TN (True Negative, true negative), FP (False Positive, false positive), FN (False Negative, false negative)

[0074] The area under the curve (AUC) was also used for evaluation to quantify the classification performance of the model at different thresholds. AUC provides a comprehensive evaluation of the overall performance of the model, avoiding possible bias at a single threshold.

[0075] (4) Model integration

[0076] Voting is a commonly used ensemble learning method. Its basic idea is to integrate the prediction results of multiple learners and determine the final prediction by voting. This mechanism can effectively reduce the overfitting problem that may occur in a single model, thereby improving the overall generalization performance.

[0077] In the present invention, soft voting is used as an integration method to integrate the prediction results of multiple base models and improve the overall prediction performance. Soft voting obtains the final prediction result by weighted averaging the prediction probabilities of each base model. This method can make full use of the advantages of each base model and enhance the robustness and accuracy of the model.

[0078] The elimination method is a method for determining the best ensemble model combination. Its basic idea is to recursively eliminate base models with poor prediction performance and ultimately retain a streamlined high-performance model combination. The specific steps are as follows:

[0079] 1. Initialize a set containing all base models;

[0080] 2. For each base model in the current model set, eliminate the model, calculate new weights and generate a new integrated model;

[0081] 3. Select the base model with the best performance after elimination and eliminate it from the set;

[0082] 4. Repeat the above steps until no better result is found;

[0083] 5. Output the final retained base model set as the integrated model. Figure 1 Flowchart for determining the best model combination for the elimination method.

[0084] (5) Model training

[0085] Five-fold cross validation was used to ensure the stability and reliability of the model. Due to the imbalance of data categories, SMOTE (Synthetic Minority Over-sampling Technique) was used for oversampling. This method generates new synthetic samples by interpolating between samples of the minority class, thereby increasing the sample size of the minority class and improving the imbalance of categories during training. Grid Search was used to find the optimal model parameters.

[0086] (6) Model evaluation

[0087] The ensemble model was established based on the voting method, and then the elimination method was used to determine the best model combination. The six base models were used as the initial model combination, and the best model combination was determined after multiple iterations.

[0088] Table 1 Prediction performance of model combinations determined by elimination method

[0089]

[0090] Table 1 shows the performance of the model combination determined by the elimination method. In the initial stage, the integrated model includes logistic regression, SVM, decision tree, random forest, GBDT and AdaBoost, and the initial AUC is 0.9892; after eliminating the decision tree in the first round of iteration, the AUC increased to 0.9910, indicating that the decision tree has a significant negative impact on the prediction performance of the integrated model; in the second round of iteration, the initial AUC is 0.9910, and after eliminating logistic regression, the AUC rises to 0,9930; in the third round of iteration, SVM is removed, and the AUC rises to 0.9942; in the fourth round of iteration, the performance of the integrated model no longer increases after eliminating any model, and the elimination process ends. The best model combination is determined to be random forest, GBDT and AdaBoost.

[0091] Through the elimination method, several base models that had a negative impact on the overall performance of the model under certain conditions were successfully identified and eliminated. This process significantly improved the prediction accuracy of the model. The final optimized integrated model had an AUC of 0.9942, which was significantly improved compared to the preliminary integrated model, proving the effectiveness of optimizing the model combination based on the elimination method.

[0092] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for evaluating cognitive impairment in cerebral small vessel disease based on gait characteristics and ensemble learning, characterized in that: Here are the steps: S1. Collect gait data using wearable devices; S2, preprocessing the collected raw data; S3. Several types of models were used to predict the cognitive level of CSVD patients, using an ensemble learning method based on voting and a model selection method based on elimination to establish a cognitive impairment prediction model. S4. Use accuracy, precision, recall, F1 score and area under the curve as evaluation metrics to evaluate the performance of the model.

2. The method for evaluating cognitive impairment of cerebral small vessel disease based on gait characteristics and ensemble learning according to claim 1, characterized in that: In step S1, four gait tasks are completed using a wearable device, including two single tasks and two dual tasks.

3. The method for evaluating cognitive impairment of cerebral small vessel disease based on gait characteristics and ensemble learning as claimed in claim 1, characterized in that: In step S2, before model training, the original data is preprocessed, and the missing values ​​are filled with the average value of the gait data of each task to remove abnormal data; The symmetry of the data is evaluated by comparing the gait parameters of the left and right legs. The calculation method is as shown in formula (1): The dual-task cost is calculated by comparing the gait parameters under single-task and dual-task conditions. The calculation method is as shown in formula (2): All features are standardized and the cognitive level is used as the label.

4. The method for evaluating cognitive impairment of cerebral small vessel disease based on gait characteristics and ensemble learning as claimed in claim 1, characterized in that: In step S3, six different types of classifiers were used to predict the MoCA scores of CSVD patients, including logistic regression, support vector machine, decision tree, random forest, gradient boosting, and adaptive boosting; Four evaluation indicators are used to evaluate the classification model: accuracy, precision, recall, and F1 score. The calculation method is as follows: Among them, TP represents true positive, TN represents true negative, FP represents false positive, and FN represents false negative; The area under the curve is used for evaluation to quantify the classification performance of the model at different thresholds. AUC provides a comprehensive evaluation of the overall performance of the model, avoiding possible bias at a single threshold.

5. The method for evaluating cognitive impairment of cerebral small vessel disease based on gait characteristics and ensemble learning as claimed in claim 4, characterized in that: Soft voting is used as a model integration method to integrate the prediction results of multiple base models and improve the overall prediction performance; the prediction probabilities of each base model are weighted averaged to obtain the final prediction result.

6. The method for evaluating cognitive impairment of cerebral small vessel disease based on gait characteristics and ensemble learning as claimed in claim 4, characterized in that: Use elimination method as the model integration method. The specific steps are as follows: T1. Initialize a set containing all base models; T2. For each base model in the current model set, eliminate the model, calculate new weights and generate a new integrated model; T3, select the base model with the best performance after elimination and eliminate the model from the set; T4. Repeat the above steps until the optimal result is obtained; T5. Output the final retained base model set as the integrated model.

7. The method for evaluating cognitive impairment in cerebral small vessel disease based on gait characteristics and ensemble learning according to any one of claims 1 to 6, characterized in that: The SMOTE algorithm is used for oversampling: new synthetic samples are generated by interpolating between samples of the minority class, thereby increasing the sample size of the minority class; the grid search method is used to find the optimal model parameters.

8. The method for evaluating cognitive impairment of cerebral small vessel disease based on gait characteristics and ensemble learning as claimed in claim 2, characterized in that: The single task was to walk freely at a habitual speed; the dual task included a walking task with a plate and a walking task for a verbal fluency test.

9. The method for evaluating cognitive impairment of cerebral small vessel disease based on gait characteristics and ensemble learning as claimed in claim 8, characterized in that: The task of walking with a plate: participants walked with a plate in both hands and avoided dropping the cup on the plate during walking; The verbal fluency test walking task: the participant starts walking while saying the name of any animal.

10. The method for evaluating cognitive impairment of cerebral small vessel disease based on gait characteristics and ensemble learning according to claim 8, characterized in that: The gait parameters collected included stride length, stride speed, stride frequency, stance phase, swing phase, stride time, stance time, swing time, toe-off angle, and heel strike angle of the left and right feet.

Images