Parkinson's disease intelligent detection method and system based on data identification

Abstract

The invention relates to the technical field of intelligent detection, in particular to a Parkinson's disease intelligent detection method and system based on data recognition, and the method comprises the following steps: collecting the physiological and motion data of a user through an intelligent watch, a wrist strap and a mobile gait detector, recording the gait speed, the stride length and the hand tremor frequency, analyzing the smoothness of the hand motion, and determining the smoothness of the user. And recording the timestamp of each data point, and organizing the data according to a time sequence to obtain a time sequence data set. According to the invention, the physiological and motion data collected by the intelligent device is subjected to time serialization processing, each small change of Parkinson's disease is tracked in detail, the time slice data is analyzed by using the neural network, normal and abnormal modes are effectively distinguished, and key behavior characteristics are accurately extracted. The Parkinson's disease early-stage judgment capability is enhanced, a doctor is helped to adjust a treatment scheme in time through continuous monitoring, the life quality of a patient is remarkably improved, and continuous medical decision support is provided.

Description

Technical Field

[0001] The present invention relates to the field of intelligent detection technology, and in particular to a Parkinson's disease intelligent detection method and system based on data recognition. Background Art

[0002] The field of intelligent detection technology covers a wide range of applications, from industrial automation to medical health monitoring. It mainly uses sensor data, image recognition, sound recognition and other forms of data input, and performs pattern recognition and data analysis through algorithms to achieve automatic identification and evaluation of specific conditions or attributes. It relies on machine learning, deep learning and artificial intelligence to improve the accuracy and efficiency of detection. It can also be used to assist in the diagnosis of diseases, monitor disease progression, and evaluate treatment effects, providing assistance for clinical decision-making.

[0003] Among them, the data-recognized Parkinson's intelligent detection method is an application example of intelligent detection technology in the medical field. It aims to detect and evaluate the presence and severity of Parkinson's disease by analyzing various patient data (including motion data, physiological parameters, etc.). It uses data recognition technology to automatically analyze the patient's behavior and physiological characteristics through algorithms to identify the specific patterns of Parkinson's disease, providing a fast, non-invasive auxiliary diagnostic tool to help doctors diagnose Parkinson's disease in the early stages, so as to start treatment and management as early as possible and improve patients' quality of life.

[0004] Existing technologies often have difficulty effectively distinguishing the subtle differences between Parkinson's pathological behaviors and daily behaviors, resulting in less reliance on clinical observations and subjective reports from patients, increasing diagnostic errors and limiting the maximization of treatment effects. In addition, existing technologies generally do not support in-depth analysis and real-time feedback of complex data in terms of data processing capabilities, limiting their application value in the rapidly changing progression of Parkinson's disease. The lack of refined and dynamic data analysis capabilities makes it difficult for doctors to obtain a comprehensive view of Parkinson's disease and implement precise disease management strategies, increasing the passivity of treatment and affecting patients' treatment effects and quality of life. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a Parkinson's disease intelligent detection method based on data recognition.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a Parkinson's intelligent detection method based on data recognition, comprising the following steps: S1: Collect user physiological and motion data through smart watches, wristbands and mobile gait detectors, record gait speed, stride length and hand tremor frequency, analyze the smoothness of hand movement, record the timestamp of each data point, organize the data in chronological order, and obtain a time series data set; S2: dividing the time series data set into time windows, counting the continuous data points in each time window, calculating the average and standard deviation of gait speed, stride length and hand tremor frequency in each time window, analyzing the statistical characteristics of each time window, and obtaining time segment data; S3: importing the time segment data into a neural network, identifying changes in gait rhythm and hand tremor frequency and amplitude in each time window, comparing them with normal patterns, extracting key behavioral features from each time window for typical physiological behaviors of Parkinson's disease, analyzing and recording the manifestation and intensity of each key behavioral feature based on the key behavioral feature extraction results, analyzing the change trend, marking key behavioral indicators, and forming physiological behavioral feature analysis results; S4: Based on the analysis results of the physiological and behavioral characteristics, each key behavioral characteristic is compared and analyzed with the known standards of Parkinson's disease severity. The key behavioral characteristics are ranked according to the degree of impact of each key behavioral characteristic on the disease. According to the ranking results, the duration, frequency and intensity of the symptoms are comprehensively analyzed. The key behavioral characteristics are classified according to the severity of the disease, and the disease stage corresponding to each behavioral characteristic is recorded to generate a Parkinson's disease severity assessment result.

[0007] As a further solution of the present invention, the steps of acquiring the time series data set are: S111: Collect user physiological and motion data through smart watches, wristbands and mobile gait detectors, extract gait speed, stride length and hand tremor frequency, record the timestamp of each data point, and perform preliminary integration of the data to obtain a multidimensional data set; S112: Based on the multidimensional data set, gait speed, stride length and hand tremor frequency are arranged in time order according to timestamps, using the formula: ; Calculating smoothness index of hand tremor movement , and at the same time, the continuous change amplitude of the tremor frequency is counted to generate a preliminary evaluation result of smoothness, where Indicates hand tremor frequency data points, Indicates hand tremor frequency data points, Indicates the total number of data points; S113: In combination with the preliminary smoothness evaluation result, the smoothness characteristics of the hand tremor movement are analyzed according to the time series distribution law, and the gait speed, stride length, and hand tremor frequency are arranged in chronological order in combination with the timestamp to generate a time series data set.

[0008] As a further solution of the present invention, the step of acquiring the time segment data is: S211: dividing the time series by a set time window length according to the timestamp in the time series data set, extracting all data points in the time window according to the start time and end time of each time window, screening the gait speed, stride length and hand tremor frequency parameter data, and classifying them according to the time window to generate a parameter classification data set in the time window; S212: Based on the parameter classification data set within the time window, for each data point within the time window, respectively calculate the mean value and standard deviation of the gait speed, stride length and hand tremor frequency data points, and correspond the statistical characteristics to the time window to establish a time window statistical characteristic set; S213: calling the time window statistical feature set, combining the mean and standard deviation of gait speed, stride length and hand tremor frequency with the timestamp range of each time window, integrating all time window data in chronological order to form time segment data.

[0009] As a further solution of the present invention, the step of extracting the key behavior features is: S311: importing the time segment data into a neural network, performing fine-tuning through LoRA, synchronously pairing the gait speed and stride length in combination with the start and end time of the time window, performing peak and trough detection on the tremor frequency data in each time window, and generating paired data of the gait speed and stride length in each time window; S312: Based on the paired data of gait speed and stride length in each time window, the changing trend of gait speed and stride length is analyzed using the formula: ; Calculate the gait rhythm change value in each time window , generate gait rhythm analysis results within the time window, where, and Indicates and Gait speed at each time point, and Indicates and The stride length at each time point, Represents the total number of data points in the time window; S313: Based on the gait rhythm analysis results within the time window, typical physiological behavioral characteristics of Parkinson's disease in each time window are extracted, and the key behavioral characteristics of each time window are summarized in combination with the abnormal amplitude of the gait rhythm, the degree of change of the tremor frequency and the fluctuation of the tremor amplitude, to generate the key behavioral characteristics extraction results.

[0010] As a further solution of the present invention, the steps of obtaining the physiological behavior characteristic analysis results are: S321: calling the key behavior feature extraction result, sorting the key behavior feature expression forms and corresponding values ​​according to the time window sequence for the key behavior feature extracted in each time window, and recording the key behavior feature type and intensity of each time window to generate a key behavior feature expression form and intensity record; S322: Based on the key behavior feature expression form and strength record, perform time series trend analysis on the strength of each key behavior feature in time window order, analyze the change trend of the feature strength over time, and generate trend analysis results of the key behavior features in correspondence with the behavior feature type; S323: In combination with the trend analysis results of the key behavioral features, the feature types and intensities in the time window are screened, the key behavioral features that exceed the normal threshold range are extracted, and the time positions and abnormal indicators are marked accordingly to generate physiological behavioral feature analysis results.

[0011] As a further solution of the present invention, the step of sorting the key behavior features is: S411: Based on the physiological behavior feature analysis results, the data is compared with the known Parkinson's disease severity standards item by item, the difference between the actual value and the standard value of the key behavior feature is calculated, and the difference size and direction of each feature are recorded to generate the key behavior feature comparison difference result; S412: According to the comparison difference result of the key behavior feature, the formula is used: ; Calculate the comprehensive impact of key behavioral characteristics on the condition , generate comprehensive impact assessment results of key behavioral characteristics, among which, is the characteristic intensity difference, is the characteristic frequency difference, is the characteristic duration difference; S413: Arrange the key behavior characteristics in descending order according to the comprehensive impact assessment results, record the sorting results and the corresponding comprehensive impact values, sort all the characteristics in order according to the degree of impact of the disease, and generate the key behavior characteristics sorting results.

[0012] As a further solution of the present invention, the steps for obtaining the Parkinson's disease severity assessment result are: S421: According to the ranking results of the key behavior features, the manifestation form, intensity, frequency and duration information of the features are sorted item by item in combination with the ranking positions, and the corresponding impact level is marked for each type of feature to generate a key behavior feature impact record; S422: According to the key behavior feature impact record, the feature expression form is matched with the corresponding disease state, and the change trend of each feature in the time dimension is summarized according to the significance of frequency and duration, and the key behavior feature severity classification result is generated in combination with the representativeness of the feature for Parkinson's disease; S423: Based on the severity classification results of the key behavioral characteristics, the corresponding disease stage is marked according to the comprehensive manifestation and classification results of each type of key behavioral characteristics to generate a Parkinson's disease severity assessment result.

[0013] A Parkinson's intelligent detection system based on data recognition, comprising: The data acquisition module collects the user's physiological and motion data through smart watches, wristbands and mobile gait detectors, records the timestamp of each data point, calculates the smoothness index of hand tremor movement, and counts the continuous change amplitude of tremor frequency. Combined with the timestamp, it is arranged in chronological order to generate a time series data set; The time partitioning module partitions the time series by the set time window length according to the timestamps in the time series data set, and classifies according to the time windows, calculates the mean and standard deviation of each time window and combines them with the timestamp range of each time window to form time segment data; The behavior analysis module imports the time segment data into a neural network to synchronously pair the gait speed and stride length, calculates the gait rhythm change value in each time window, extracts the typical physiological behavior characteristics of Parkinson's disease in each time window, records the key behavior characteristic type and intensity of each time window, analyzes the change trend of the characteristic intensity over time and corresponds it with the behavior characteristic type, extracts the key behavior characteristics that exceed the normal threshold range, and marks the time position and abnormal indicators accordingly, and generates the physiological behavior characteristic analysis results; The assessment generation module compares the data with the known standards of Parkinson's disease severity item by item based on the physiological and behavioral characteristics analysis results, calculates the comprehensive impact value of the key behavioral characteristics on the disease according to the difference and arranges them in descending order, sorts out the manifestation, intensity, frequency and duration information of the characteristics item by item based on the sorting position, and marks the corresponding impact level of each type of feature. The key behavioral characteristics are classified according to the representativeness of Parkinson's disease based on the characteristics, and the corresponding disease stage is simultaneously marked to generate a Parkinson's disease severity assessment result.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, the physiological and motion data collected by smart devices are processed through time series processing, and every tiny change of Parkinson's disease is carefully tracked. The time segment data is analyzed by using a neural network combined with LoRA to effectively distinguish between normal and abnormal patterns and accurately extract key behavioral features. This not only enhances the ability to make early judgments about Parkinson's disease, but also helps doctors adjust treatment plans in a timely manner through continuous monitoring, significantly improving the quality of life of patients and providing continuous medical decision-making support. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a main step flow chart of the present invention; Figure 2 The flowchart of obtaining the time series data set of the present invention; Figure 3 This is a flow chart for obtaining time segment data of the present invention; Figure 4 It is a flow chart for extracting key behavior features of the present invention; Figure 5 The flowchart of obtaining the physiological behavior characteristic analysis results of the present invention; Figure 6 A flowchart of the sorting of the key behavioral features of the present invention; Figure 7 The present invention is a flow chart for obtaining the Parkinson's disease severity assessment results. DETAILED DESCRIPTION

[0016] 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.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] See also Figure 1 , a Parkinson's intelligent detection method based on data recognition, comprising the following steps: S1: Collect user physiological and motion data through smart watches, wristbands and mobile gait detectors, record gait speed, stride length and hand tremor frequency, analyze the smoothness of hand movement, record the timestamp of each data point, organize the data in chronological order, and obtain a time series data set; S2: Divide the time series data set into time windows, count the continuous data points in each time window, calculate the mean and standard deviation of gait speed, stride length and hand tremor frequency in each time window, analyze the statistical characteristics of each time window, and obtain time segment data; S3: Import the time segment data into the neural network, fine-tune it through LoRA, identify the changes in gait rhythm and hand tremor frequency and amplitude in each time window, compare them with the normal pattern, extract key behavioral features from each time window for the typical physiological behavior of Parkinson's disease, analyze and record the manifestation and intensity of each key behavioral feature based on the key behavioral feature extraction results, analyze the change trend, mark the key behavioral indicators, and form the physiological behavioral feature analysis results; S4: Based on the analysis results of physiological and behavioral characteristics, each key behavioral characteristic is compared and analyzed with the known standards of Parkinson's disease severity. The key behavioral characteristics are ranked according to the degree of impact of each key behavioral characteristic on the disease. According to the ranking results, the duration, frequency and intensity of the symptoms are comprehensively analyzed. The key behavioral characteristics are classified according to the severity of the disease, and the disease stage corresponding to each behavioral characteristic is recorded to generate the Parkinson's disease severity assessment results.

[0019] The time series data set includes gait speed data, stride length data, hand tremor frequency data and timestamp records; the time segment data includes average speed, average stride length, average tremor frequency and statistical standard deviation; the physiological and behavioral characteristics analysis results include behavioral rhythm change analysis results, tremor frequency change analysis results, tremor amplitude change analysis results, behavioral intensity and behavioral manifestation records; the Parkinson's disease severity assessment results include key behavioral characteristics ranking results, disease classification and behavioral characteristics disease stage records.

[0020] See also Figure 2 , the steps to obtain the time series data set are: S111: Collect user physiological and motion data through smart watches, wristbands and mobile gait detectors, extract gait speed, stride length and hand tremor frequency, record the timestamp of each data point, and perform preliminary integration of the data to obtain a multidimensional data set; The user's physiological and motion data are obtained through smart watches, wristbands and mobile gait detectors. First, the original data is obtained according to the built-in sensors of the device, including parameters such as gait speed, stride length and hand tremor frequency. Each collected data point is accompanied by a timestamp record to ensure the accurate distribution of the data on the time axis. Then the original data points are filtered to eliminate missing data or erroneous data caused by signal interference or equipment abnormalities during the collection process. The filtered data are further sorted with timestamp as the primary key so that the time sequence between each data point is clearly visible. On this basis, the integrity of the data is checked to confirm that the three parameters of gait speed, stride length and hand tremor frequency have corresponding values ​​at each timestamp. If there is a single missing data, the missing value is filled by linear interpolation according to the trend of adjacent time points to ensure the continuity of the data. Finally, the processed data is integrated in chronological order to form an initial multidimensional data set containing four dimensions: timestamp, gait speed, stride length and hand tremor frequency.

[0021] S112: Based on the multidimensional data set, the gait speed, stride length and hand tremor frequency are arranged in time order according to the timestamp, using the formula: ; Calculating smoothness index of hand tremor movement , and at the same time, the continuous change amplitude of the tremor frequency is counted to generate a preliminary evaluation result of smoothness, where Indicates hand tremor frequency data points, Indicates hand tremor frequency data points, Indicates the total number of data points; In the time series, the collected hand tremor frequency data points 10Hz, 12Hz, 14Hz, 15Hz and 16Hz respectively, total data points , put the data into the formula and calculate as follows: Compute the sum of squared changes in frequency for each pair of adjacent points:

[0022] Compute the absolute change for each pair of adjacent points:

[0023] Combine the above results and divide by the number of data points:

[0024] The results showed that the smoothness index of hand tremor movement was 3.2, reflecting the comprehensive characteristics of tremor frequency changes. This result was directly used to judge the smoothness of tremor frequency during movement and provide a basis for further analysis of tremor abnormalities.

[0025] S113: Combined with the preliminary smoothness evaluation results, the smoothness characteristics of the hand tremor movement are analyzed according to the time series distribution law, and the gait speed, stride length, and hand tremor frequency are arranged in chronological order in combination with the timestamp to generate a time series data set; Based on the smoothness analysis results of various parameters in the time series, the time series data are further refined. First, the time series change data of gait speed and stride length are extracted, and the maximum, minimum and change amplitude of gait speed and stride length are calculated. Combined with the tremor motion smoothness index, the smooth interval and non-smooth interval in the time series are distinguished. Then, for the non-smooth interval, the corresponding data points in the time series are analyzed one by one, and the changes in gait speed and stride length are compared with the fluctuation trend of tremor frequency. The data points with excessively large or discontinuous changes in tremor frequency are further screened out. At the same time, the corresponding abnormal timestamps are marked in the time series, and the distribution law of the abnormal point data is statistically analyzed to sort out the distribution dense intervals of abnormal data. Finally, the gait speed, stride length, tremor frequency and timestamps are reintegrated in chronological order to form a complete multidimensional time series data set with abnormal points marked, providing a basis for subsequent analysis.

[0026] See also Figure 3 , the steps to obtain time segment data are: S211: according to the timestamp in the time series data set, the time series is divided by the set time window length, all data points in the time window are extracted according to the start time and end time of each time window, the gait speed, stride length and hand tremor frequency parameter data are screened, and the data are classified according to the time window to generate a parameter classification data set in the time window; According to the timestamp parameters in the time series data set, the time series is divided according to the set time window length. First, the timestamp of each data point in the time series data set is extracted, and the timestamps are sorted in chronological order to ensure that there is no duplication or jump in the timestamps. Then, according to the start time and end time range of the preset time window, the timestamp data is divided into multiple time windows. The division of the time window must ensure that the time range of each time window is continuous and covers all data points. Then, the values ​​of the three parameters of gait speed, stride length and hand tremor frequency are extracted in each time window, and the timestamps of the data points in each time window are matched one by one to ensure that all data points in each time window correspond to complete timestamps and the number of data points meets the basic requirements in the time window. Finally, check whether there are invalid data points or data loss in the divided time window, remove or mark the abnormal data in the time window, and rearrange the data points in the time window in chronological order. Finally, a set of classified data including gait speed, stride length and hand tremor frequency in the time window is generated to ensure that all data are reasonable and complete according to the time window division.

[0027] S212: Based on the parameter classification data set in the time window, for each data point in the time window, respectively calculate the mean value and standard deviation of the gait speed, stride length and hand tremor frequency data points, and correspond the statistical characteristics to the time window to establish a time window statistical characteristic set; The parameter classification data set within the time window is called, and the data points in each time window are further processed. First, the numerical data of gait speed, stride length and hand tremor frequency in each time window are extracted, and the validity of each data point is checked one by one. The abnormal data points beyond the reasonable range are eliminated. At the same time, the number of valid data points of gait speed, stride length and hand tremor frequency in each time window is counted to ensure that the statistical analysis uses complete and non-abnormal data. Then, the valid data points in each time window are numerically calculated one by one. The specific operation is as follows: for each parameter (gait speed, stride length length and hand tremor frequency), calculate their average value, that is, add up the values ​​of all valid data points and divide them by the number of data points, then calculate the standard deviation of the values, that is, sum the squares of the differences between the values ​​of each data point and the average value and divide them by the number of data points, and take the square root of the result to generate the average value and standard deviation data of each parameter in the time window. Finally, match the statistical characteristic data with the time window according to the parameter classification to form a statistical characteristic set of the time window, covering the average value and standard deviation of gait speed, stride length and hand tremor frequency, providing a complete statistical characteristic data basis for subsequent integration processing.

[0028] S213: calling the time window statistical feature set, combining the average value and standard deviation of gait speed, stride length and hand tremor frequency with the timestamp range of each time window, integrating all time window data in time order, and forming time segment data; Based on the statistical characteristic set of each time window, the mean and standard deviation data of gait speed, stride length and hand tremor frequency in each time window are first extracted and matched with the time range of the time window. The statistical characteristic data of all time windows are arranged in the order of the start time of the time window to ensure that the order of the time windows is consistent with the timestamp order of the original time series. Then, the statistical characteristic data of each time window are checked for completeness to ensure that the mean and standard deviation data of gait speed, stride length and hand tremor frequency exist. Entries with incomplete parameters or abnormal data in the time window are eliminated. At the same time, it is checked whether the time range of the time window overlaps or jumps, and problematic time windows are eliminated or corrected. Finally, the statistical characteristic data of all time windows are integrated in chronological order to generate time segment data containing the statistical characteristics of the time window and the corresponding time range, where each time segment data includes the statistical characteristics of gait speed, stride length and hand tremor frequency in the time window and the corresponding time range, providing an accurate and clear data basis for subsequent time series analysis.

[0029] See also Figure 4 ,The steps for extracting key behavioral features are: S311: importing the time segment data into the neural network, fine-tuning it through LoRA, synchronously pairing the gait speed and stride length in combination with the start and end time of the time window, performing peak and trough detection on the tremor frequency data in each time window, and generating paired data of the gait speed and stride length in each time window; The time segment data is imported into the neural network. First, LoRA is used to reduce the training parameters and display usage through low-rank adaptation. Then, the gait speed, stride length and hand tremor frequency data of each time window are extracted from the time segment data. All data points are sorted according to the start and end time of the time window to ensure that the data points in the time window are arranged continuously in chronological order. Then, the two sets of gait speed and stride length data are paired point by point with the timestamp as the reference to form the paired data of gait speed and stride length. At the same time, the hand tremor frequency data is processed to mark the peak and trough data points in each time window, and the frequency and time position of the peak and trough are recorded to generate the paired data of gait speed and stride length in each time window and the peak and trough characteristic data of tremor frequency, which provide a basis for subsequent rhythm analysis and frequency characteristic analysis.

[0030] S312: Based on the paired data of gait speed and stride length in each time window, the changing trend of gait speed and stride length is analyzed using the formula: ; Calculate the gait rhythm change value in each time window , generate gait rhythm analysis results within the time window, where, and Indicates and Gait speed at each time point, and Indicates and The stride length at each time point, Represents the total number of data points in the time window; A time window contains 5 data points, and their gait speeds are , , , , , and the stride lengths are , , , , , the calculation process is as follows: Calculate the sum of squares of the change in gait speed:

[0031] Calculate the sum of squared changes in stride length:

[0032] Add the two changes and divide by the total number of data points:

[0033] The results show that the gait rhythm change value in the current time window is , which reflects the comprehensive variation characteristics of gait speed and stride length within the time window, and provides a quantitative basis for subsequent gait rhythm analysis.

[0034] S313: based on the gait rhythm analysis results within the time window, extract the typical physiological and behavioral characteristics of Parkinson's disease in each time window, combine the abnormal amplitude of the gait rhythm, the degree of change of the tremor frequency and the fluctuation of the tremor amplitude, summarize the key behavioral characteristics of each time window, and generate the key behavioral characteristics extraction results; Based on the abnormal detection results of gait rhythm and tremor frequency changes in the time window, firstly, the gait rhythm change value and the peak and valley characteristic data of tremor frequency in the time window are called, and the gait rhythm change and tremor frequency change in each time window are analyzed. The gait rhythm change value is associated with the peak and valley of tremor frequency to detect whether there is significant abnormal fluctuation. Then, the time window of abnormal fluctuation is marked according to the detection result of each time window, and the amplitude and time position of tremor frequency change are extracted. The typical physiological behavioral characteristics of Parkinson's disease are classified and sorted. Finally, the key behavioral characteristics of each time window are summarized and the related parameters are recorded to generate the key behavioral feature extraction results in the time window.

[0035] See also Figure 5 , the steps to obtain the physiological behavior characteristics analysis results are: S321: calling the key behavior feature extraction result, sorting the key behavior feature's expression form and corresponding value according to the time window sequence for the key behavior feature extracted in each time window, and recording the key behavior feature type and intensity of each time window, and generating a key behavior feature expression form and intensity record; The key behavior feature extraction results are called. First, the basic information of the key behavior features is extracted from the time window data, including the gait rhythm abnormality amplitude, tremor frequency change amplitude, and tremor fluctuation position data. The feature data is sorted in the order of the start time and end time of the time window to ensure that the feature data in the time window can completely correspond to each time period. During the extraction process, the key behavior features of each time window are checked to confirm whether there are missing data or abnormal values ​​in the values ​​of the gait rhythm abnormality amplitude and the tremor frequency change amplitude. For the missing data points, linear interpolation is performed according to the change trend of the adjacent points in the time window. At the same time, the abnormal values ​​that are obviously beyond the normal range are marked to ensure the data integrity of each time window. Subsequently, the extracted feature data are classified, and the gait rhythm abnormality amplitude is recorded separately as an independent type of feature data, and the tremor frequency change amplitude and the tremor fluctuation position are sorted into a group in time order to form a classification record table of behavior features. Finally, the feature expression forms and intensity information in all time windows are integrated in time series to generate a complete record containing the gait rhythm abnormality amplitude, tremor frequency change amplitude and its fluctuation position, providing basic data support for the subsequent trend analysis of key behavior features.

[0036] S322: Based on the key behavior feature expression form and strength record, perform time series trend analysis on the strength of each key behavior feature in time window order, analyze the change trend of the feature strength over time, and correspond it with the behavior feature type to generate the trend analysis result of the key behavior feature; Based on the manifestation and intensity records of key behavioral features, we first extracted feature data such as the amplitude of gait rhythm abnormality, the amplitude of tremor frequency change, and the position of tremor fluctuation, and arranged the feature intensity values ​​into a time series in the order of time windows, and ensured that the intensity value of each feature corresponded to the start and end time of the corresponding time window. For each key behavioral feature, the feature intensity change within the time window was analyzed time by time window, and the intensity difference between each adjacent time window was calculated to measure the amplitude of change within the time window. At the same time, the amplitude of change was recorded in correspondence with the time series position to form the time series information of the feature intensity change. Subsequently, according to the amplitude of the feature intensity change within the time window, the change trend value of the entire time series was calculated, and the overall change trend was quantified by accumulating the amplitude of change time by time window and dividing it by the total number of time windows. At the same time, the positive and negative direction information of the change trend was recorded to distinguish whether the feature intensity showed an increasing or decreasing trend. Finally, the change trend information was sorted and corresponded with each behavioral feature type to generate a complete trend analysis result including the amplitude of gait rhythm abnormality and the amplitude of tremor frequency change, laying the foundation for subsequent abnormal feature screening.

[0037] S323: combining the trend analysis results of key behavioral features, screening the feature types and intensities in the time window, extracting key behavioral features that exceed the normal threshold range, and marking the time position and abnormal indicators accordingly, generating physiological behavioral feature analysis results; Combined with the change trend information of the key behavioral features within the time window, the change trend value and intensity value of each feature are first screened, and the feature types with change trend values ​​beyond the normal range are extracted. At the same time, the key behavioral features with intensity values ​​outside the normal range are screened to ensure that the screened features are all abnormal features. Then the time window position of the screened abnormal features is marked, and the start and end time of the time window is recorded corresponding to the abnormal intensity and change trend of the feature. For the screened abnormal features, they are further classified and sorted according to the feature type, and the time window range, intensity value and change trend information of each type of abnormal feature are recorded separately to ensure the integrity of the abnormal behavior feature record. Finally, all the screened and classified abnormal features are integrated into a key behavior indicator list, and the list is sorted according to the time window order to generate a complete key behavior indicator marking result, ensuring that the time window range, behavior feature type, intensity value and change trend information are included, providing accurate and detailed data support for generating complete physiological behavior feature analysis results.

[0038] See also Figure 6 , the sorting steps of key behavioral features are: S411: Based on the results of the physiological and behavioral characteristics analysis, the data are compared with the known standards of Parkinson's disease severity item by item, the difference between the actual value and the standard value of the key behavioral characteristics is calculated, and the difference size and direction of each characteristic are recorded to generate the key behavioral characteristics comparison difference results; The analysis results of physiological and behavioral characteristics are called. First, the data content such as the form, intensity and change trend of each key behavioral feature are extracted, and these data are sorted in the order of time windows to ensure that the key behavioral features in each time window can accurately correspond to their time range and behavioral content. In the process of sorting, the form and intensity of the key behavioral features are classified to ensure that the characteristic information such as gait rhythm, tremor frequency, and tremor amplitude are classified into separate data sets. At the same time, the change trend of the features is extracted and recorded in association with the form and intensity of each type of feature. Subsequently, the sorted key behavioral feature data is compared item by item with the known standards of the severity of Parkinson's disease, and the difference between the actual value and the standard value of the key behavioral feature is extracted, and the size and direction of the difference are recorded. In the comparison process, the integrity of the difference calculation is checked to ensure that each feature type contains complete form, intensity and trend information. At the same time, the positive and negative of the difference are marked to clarify whether the deviation exceeds or falls below the standard range. For feature data with large deviations, their feature categories and related differences are recorded, and all calculation results are checked to ultimately generate comparative difference results for each key behavioral feature, providing basic data support for subsequent comprehensive calculations and sorting.

[0039] S412: Based on the comparison difference results of key behavioral characteristics, the formula is used: ; Calculate the comprehensive impact of key behavioral characteristics on the condition , generate comprehensive impact assessment results of key behavioral characteristics, among which, is the characteristic intensity difference, is the characteristic frequency difference, is the characteristic duration difference; The actual values ​​of the intensity, frequency, and duration of a key behavior feature are , , , and the corresponding known standard values ​​are , , , then the calculation of each difference is: Calculate feature intensity difference :

[0040] Calculate the characteristic frequency difference :

[0041] Calculate feature duration difference :

[0042] Substitute into the formula to calculate the comprehensive impact value:

[0043] The results show that the comprehensive impact value of the current key behavioral characteristics , indicating that the feature has deviations in the difference comparisons in the three aspects, which can be used as the basis for subsequent sorting and classification to measure the comprehensive impact of the feature on the disease.

[0044] S413: Arrange the key behavior characteristics in descending order according to the comprehensive impact assessment results, record the sorting results and the corresponding comprehensive impact values, sort all the characteristics in order according to the degree of impact of the disease, and generate the key behavior characteristics sorting results; According to the comprehensive impact value of each key behavioral feature, they are arranged in descending order. First, the feature with the highest comprehensive impact value is extracted, and its corresponding feature type, intensity and expression form are recorded in association with the impact value to ensure that the impact degree of each feature is associated with its value. Then all features are sorted in order from large to small in terms of impact value, and the sorting position is marked for each type of feature. During the sorting process, for features with the same comprehensive impact value, their intensity and duration are given priority, and features with greater intensity are placed in the front row, and their specific sorting logic is marked. After the sorting is completed, the expression form, comprehensive impact value, intensity and sorting results of each feature are integrated into a feature sorting list to ensure that each feature has complete record information in the list. Finally, the key behavioral feature results sorted by the degree of disease impact are generated, providing comprehensive and detailed reference data for subsequent classification and disease stage marking.

[0045] See also Figure 7 , the steps to obtain the Parkinson's disease severity assessment results are: S421: According to the ranking results of the key behavior features, the manifestation form, intensity, frequency and duration information of the features are sorted item by item in combination with the ranking positions, and the corresponding impact level is marked for each type of feature to generate a key behavior feature impact record; Call the sorting results of key behavioral features, extract the core data content such as the comprehensive impact value, sorting position and expression form corresponding to each feature, sort them in sequence according to the sorting position, and ensure data integrity. First, extract the expression form, intensity, frequency and duration information of the features in sequence according to the comprehensive impact value, check each feature one by one, confirm whether the intensity and frequency are consistent within the data range, and mark the priority of each feature. For features with higher rankings, focus on marking the changing relationship between the feature range and the actual performance during the sorting process, especially accurately record the duration of high-impact features to ensure that no key details are missed in subsequent analysis. During the data sorting process, all key behavioral features are grouped in order to ensure that the intensity, frequency and duration corresponding to each group of data can correspond to its sorting logic, and finally generate key behavioral feature impact records sorted by priority, laying an accurate data foundation for subsequent feature analysis.

[0046] S422: According to the key behavioral feature impact records, the feature manifestation is matched with the corresponding disease state, and the change trend of each feature in the time dimension is summarized according to the significance of frequency and duration. Combined with the representativeness of the feature for Parkinson's disease, the severity classification results of the key behavioral features are generated; Based on the sorted records of the impact of key behavioral features, the correlation between the intensity, frequency and duration of each feature is analyzed item by item. First, the significant feature parameters in the expression form are extracted, and the feature frequency and duration are sorted in segments according to their range of change. For significant fluctuations in duration, the intensity parameters are compared with the change trend of duration, and the change characteristics corresponding to the key time points are extracted. At the same time, for the frequency changes of the features, the frequency fluctuation range is classified and summarized, and combined with the intensity change trend, the significant performance of the key features in frequency is extracted. The sorted frequency and duration data are matched with the feature intensity item by item to ensure that the data can fully reflect the feature performance. During the analysis process, the expression form and parameter range of each feature are recorded in detail, and the significant changes in the expression form are associated with the severity, providing accurate data support for subsequent classification, and generating a preliminary basis for the severity classification.

[0047] S423: Based on the severity classification results of key behavioral characteristics, the corresponding disease stage is marked according to the comprehensive manifestation and classification results of each type of key behavioral characteristics to generate a Parkinson's disease severity assessment result; Call the severity classification basis, and group each type of feature according to the sorted key behavioral feature data and correspond it to different stages of the disease. First, extract the feature severity classification results, and group and sort the features in order of mild, moderate, and severe to ensure that the feature range in each group can cover the main manifestations of the disease stage. For the mild classification group, extract the features with smaller changes in frequency and duration, and record them in association with the corresponding manifestations. At the same time, check the features of the moderate classification group and the severe classification group one by one to confirm the correspondence between the range of changes in intensity and duration and the disease stage. After completing the grouping, for each group of classification features, record its disease stage information and associate parameters such as manifestation, intensity, frequency and duration to generate the final classification record. By sorting each group of data, a Parkinson's disease severity assessment result containing complete data on the disease stage, manifestation and features is formed, ensuring that subsequent doctors can conduct accurate analysis based on specific classification information when diagnosing.

[0048] A Parkinson's intelligent detection system based on data recognition, comprising: The data acquisition module collects the user's physiological and motion data through smart watches, wristbands and mobile gait detectors, records the timestamp of each data point, calculates the smoothness index of hand tremor movement, and counts the continuous change amplitude of tremor frequency. Combined with the timestamp, it is arranged in chronological order to generate a time series data set; The time partitioning module divides the time series according to the timestamps in the time series data set by the set time window length, and classifies them according to the time windows, calculates the mean and standard deviation of each time window and combines them with the timestamp range of each time window to form time segment data; The behavior analysis module imports the time segment data into the neural network to synchronously pair the gait speed and stride length, calculates the gait rhythm change value in each time window, extracts the typical physiological behavior characteristics of Parkinson's disease in each time window, records the key behavior characteristic type and intensity of each time window, analyzes the change trend of the characteristic intensity over time and corresponds it with the behavior characteristic type, extracts the key behavior characteristics that exceed the normal threshold range, and marks the time position and abnormal indicators accordingly, and generates the physiological behavior characteristic analysis results; The assessment generation module is based on the results of physiological and behavioral characteristics analysis, compares the data with the known standards of Parkinson's disease severity item by item, calculates the comprehensive impact value of key behavioral characteristics on the disease according to the difference and arranges them in descending order, and organizes the manifestation, intensity, frequency and duration information of the characteristics item by item based on the sorting position. At the same time, the corresponding impact level is marked for each type of feature, and the key behavioral characteristics are classified according to the representativeness of Parkinson's disease in terms of severity. The corresponding disease stage is simultaneously marked to generate the Parkinson's disease severity assessment results.

[0049] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A Parkinson's intelligent detection method based on data recognition, characterized in that: The following steps are involved: S1: Collect user physiological and motion data through smart watches, wristbands and mobile gait detectors, record gait speed, stride length and hand tremor frequency, analyze the smoothness of hand movement, record the timestamp of each data point, organize the data in chronological order, and obtain a time series data set; S2: dividing the time series data set into time windows, counting the continuous data points in each time window, calculating the average and standard deviation of gait speed, stride length and hand tremor frequency in each time window, analyzing the statistical characteristics of each time window, and obtaining time segment data; S3: Import the time segment data into the neural network, perform fine-tuning through LoRA, identify the changes in gait rhythm and hand tremor frequency and amplitude in each time window, compare with the normal pattern, extract key behavioral features from each time window for typical physiological behaviors of Parkinson's disease, analyze and record the manifestation and intensity of each key behavioral feature based on the key behavioral feature extraction results, analyze the change trend, mark the key behavioral indicators, and form the physiological behavioral feature analysis results; S4: Based on the analysis results of the physiological and behavioral characteristics, each key behavioral characteristic is compared and analyzed with the known standards of Parkinson's disease severity. The key behavioral characteristics are ranked according to the degree of impact of each key behavioral characteristic on the disease. According to the ranking results, the duration, frequency and intensity of the symptoms are comprehensively analyzed. The key behavioral characteristics are classified according to the severity of the disease, and the disease stage corresponding to each behavioral characteristic is recorded to generate a Parkinson's disease severity assessment result.

2. The Parkinson's disease intelligent detection method based on data recognition according to claim 1 is characterized in that: The steps for obtaining the time series data set are: S111: Collect user physiological and motion data through smart watches, wristbands and mobile gait detectors, extract gait speed, stride length and hand tremor frequency, record the timestamp of each data point, and perform preliminary integration of the data to obtain a multidimensional data set; S112: Based on the multidimensional data set, gait speed, stride length and hand tremor frequency are arranged in time order according to timestamps, using the formula: ； Calculating smoothness index of hand tremor movement , and at the same time, the continuous change amplitude of the tremor frequency is counted to generate a preliminary evaluation result of smoothness, where Indicates hand tremor frequency data points, Indicates hand tremor frequency data points, Indicates the total number of data points; S113: In combination with the preliminary smoothness evaluation result, the smoothness characteristics of the hand tremor movement are analyzed according to the time series distribution law, and the gait speed, stride length, and hand tremor frequency are arranged in chronological order in combination with the timestamp to generate a time series data set.

3. The Parkinson's disease intelligent detection method based on data recognition according to claim 2 is characterized in that: The steps for obtaining the time segment data are as follows: S211: dividing the time series by a set time window length according to the timestamp in the time series data set, extracting all data points in the time window according to the start time and end time of each time window, screening the gait speed, stride length and hand tremor frequency parameter data, and classifying them according to the time window to generate a parameter classification data set in the time window; S212: Based on the parameter classification data set within the time window, for each data point within the time window, respectively calculate the mean value and standard deviation of the gait speed, stride length and hand tremor frequency data points, and correspond the statistical characteristics to the time window to establish a time window statistical characteristic set; S213: calling the time window statistical feature set, combining the mean and standard deviation of gait speed, stride length and hand tremor frequency with the timestamp range of each time window, integrating all time window data in chronological order to form time segment data.

4. The Parkinson's disease intelligent detection method based on data recognition according to claim 3 is characterized in that: The steps of extracting the key behavior features are: S311: importing the time segment data into a neural network, synchronously pairing the gait speed and stride length in combination with the start and end time of the time window, performing peak and trough detection on the tremor frequency data in each time window, and generating paired data of the gait speed and stride length in each time window; S312: Based on the paired data of gait speed and stride length in each time window, the changing trend of gait speed and stride length is analyzed using the formula: ； Calculate the gait rhythm change value in each time window , generate gait rhythm analysis results within the time window, where, and Indicates and Gait speed at each time point, and Indicates and The stride length at each time point, Represents the total number of data points in the time window; S313: Based on the gait rhythm analysis results within the time window, typical physiological behavioral characteristics of Parkinson's disease in each time window are extracted, and the key behavioral characteristics of each time window are summarized in combination with the abnormal amplitude of the gait rhythm, the degree of change of the tremor frequency and the fluctuation of the tremor amplitude, to generate the key behavioral characteristics extraction results.

5. The Parkinson's disease intelligent detection method based on data recognition according to claim 4 is characterized in that: The steps for obtaining the physiological behavior characteristic analysis result are: S321: calling the key behavior feature extraction result, sorting the key behavior feature expression forms and corresponding values ​​according to the time window sequence for the key behavior feature extracted in each time window, and recording the key behavior feature type and intensity of each time window to generate a key behavior feature expression form and intensity record; S322: Based on the key behavior feature expression form and strength record, perform time series trend analysis on the strength of each key behavior feature in time window order, analyze the change trend of the feature strength over time, and generate trend analysis results of the key behavior features in correspondence with the behavior feature type; S323: In combination with the trend analysis results of the key behavioral features, the feature types and intensities in the time window are screened, the key behavioral features that exceed the normal threshold range are extracted, and the time positions and abnormal indicators are marked accordingly to generate physiological behavioral feature analysis results.

6. The Parkinson's disease intelligent detection method based on data recognition according to claim 5 is characterized in that: The steps for sorting the key behavior features are: S411: Based on the physiological behavior feature analysis results, the data is compared with the known Parkinson's disease severity standards item by item, the difference between the actual value and the standard value of the key behavior feature is calculated, and the difference size and direction of each feature are recorded to generate the key behavior feature comparison difference result; S412: According to the comparison difference result of the key behavior feature, the formula is used: ； Calculate the comprehensive impact of key behavioral characteristics on the condition , generate comprehensive impact assessment results of key behavioral characteristics, among which, is the characteristic intensity difference, is the characteristic frequency difference, is the characteristic duration difference; S413: Arrange the key behavior characteristics in descending order according to the comprehensive impact assessment results, record the sorting results and the corresponding comprehensive impact values, sort all the characteristics in order according to the degree of impact of the disease, and generate the key behavior characteristics sorting results.

7. The Parkinson's disease intelligent detection method based on data recognition according to claim 6 is characterized in that: The steps for obtaining the Parkinson's disease severity assessment result are: S421: According to the ranking results of the key behavior features, the manifestation form, intensity, frequency and duration information of the features are sorted item by item in combination with the ranking positions, and the corresponding impact level is marked for each type of feature to generate a key behavior feature impact record; S422: According to the key behavior feature impact record, the feature expression form is matched with the corresponding disease state, and the change trend of each feature in the time dimension is summarized according to the significance of frequency and duration, and the key behavior feature severity classification result is generated in combination with the representativeness of the feature for Parkinson's disease; S423: Based on the severity classification results of the key behavioral characteristics, the corresponding disease stage is marked according to the comprehensive manifestation and classification results of each type of key behavioral characteristics to generate a Parkinson's disease severity assessment result.

8. A Parkinson's intelligent detection system based on data recognition, characterized in that: The system is used to execute the Parkinson's intelligent detection method based on data recognition according to any one of claims 1 to 7, comprising: The data acquisition module collects the user's physiological and motion data through smart watches, wristbands and mobile gait detectors, records the timestamp of each data point, calculates the smoothness index of hand tremor movement, and counts the continuous change amplitude of tremor frequency. Combined with the timestamp, it is arranged in chronological order to generate a time series data set; The time partitioning module partitions the time series by the set time window length according to the timestamps in the time series data set, and classifies according to the time windows, calculates the mean and standard deviation of each time window and combines them with the timestamp range of each time window to form time segment data; The behavior analysis module imports the time segment data into a neural network to synchronously pair the gait speed and stride length, calculates the gait rhythm change value in each time window, extracts the typical physiological behavior characteristics of Parkinson's disease in each time window, records the key behavior characteristic type and intensity of each time window, analyzes the change trend of the characteristic intensity over time and corresponds it with the behavior characteristic type, extracts the key behavior characteristics that exceed the normal threshold range, and marks the time position and abnormal indicators accordingly, and generates the physiological behavior characteristic analysis results; The assessment generation module compares the data with the known standards of Parkinson's disease severity item by item based on the physiological and behavioral characteristics analysis results, calculates the comprehensive impact value of the key behavioral characteristics on the disease according to the difference and arranges them in descending order, sorts out the manifestation, intensity, frequency and duration information of the characteristics item by item based on the sorting position, and marks the corresponding impact level of each type of feature. The key behavioral characteristics are classified according to the representativeness of Parkinson's disease based on the characteristics, and the corresponding disease stage is simultaneously marked to generate a Parkinson's disease severity assessment result.

Images