Automatic Analysis and Classification System for Lung Function Test Data Based on Big Data

Abstract

This invention relates to the field of medical data mining technology, specifically to an automatic analysis and classification system for pulmonary function test data based on big data. First, it calculates a single-channel weighted evolution index based on the variation deviation of multi-channel data time-series curves, capturing the dynamic evolution characteristics of each dimension within a continuous observation window. Then, it combines the variation trend deviation between multiple channels to accurately assess the feature response sensitivity of each dimension in response to changes in pathological states. Finally, by comparing the differences between current data and historical lesion characteristics, it quantifies the discriminative contribution of each dimension when considering historical data for each lesion type. This allows the weighted discriminative signal intensity obtained based on feature response sensitivity and discriminative contribution to more comprehensively evaluate the sample contribution of each pulmonary function test dimension. This results in higher accuracy in training the classification model using the weighted discriminative signal intensity as sample weights, thereby improving the accuracy of pulmonary function test data classification.

Description

Technical Field

[0001] This invention relates to the field of medical data mining technology, specifically to an automatic analysis and classification system for lung function test data based on big data. Background Technology

[0002] Lung function tests are one of the important means of assessing the health status of the respiratory system. When combining big data to analyze and classify lung function test data, existing technologies usually train classification models based on data from various dimensions of lung function tests, and then automatically classify lung function test data based on the trained models.

[0003] Existing technologies typically train classification models based on data from various pulmonary function test dimensions using an equal-weighted input method, assigning the same sample weight to each dimension. However, this approach ignores the importance of each dimension in pulmonary function tests and exhibits inconsistent sensitivity to different lesion types, such as obstructive and restrictive pulmonary diseases. Consequently, the classification models trained using this equal-weighted input method have low accuracy, resulting in poor accuracy in classifying pulmonary function test data based on these models. Summary of the Invention

[0004] To address the problem that existing technologies using equal-weighted input methods to train classification models across all dimensions of pulmonary function tests result in low accuracy, leading to poor classification of pulmonary function test data based on these models, this application aims to provide an automatic analysis and classification system for pulmonary function test data based on big data. The specific technical solution adopted is as follows:

[0005] The first aspect of this application provides an automatic analysis and classification system for lung function test data based on big data, including:

[0006] The data acquisition and preprocessing module is used to acquire time-series curves of data for each dimension of lung function examination at each observation period;

[0007] The first determining module is used to determine the corresponding single-channel weighted evolution index based on the deviation of the data value time series curve of each lung function test dimension in each observation period compared with the data time series curves of all other data; and to determine the corresponding feature response sensitivity in the current observation period based on the deviation of the data change trend between the data time series curve of each lung function test dimension and the data time series curves of other lung function test dimensions, as well as the corresponding single-channel weighted evolution index.

[0008] The second determination module is used to determine the discriminant contribution of each lung function test dimension in the current observation period based on the data feature deviation between the time-series curve of each lung function test dimension and the historical data of each lesion feature.

[0009] The classification module is used to determine the weighted discriminant signal intensity of each lung function test dimension in the current observation period based on the feature response sensitivity and the discriminant contribution; and to classify the lung function test data based on the weighted discriminant signal intensity and the data time series curve of the current observation period.

[0010] Furthermore, the process of obtaining the single-channel weighted evolution index includes:

[0011] Based on the overall size and temporal variation disorder of the data values ​​at all sampling moments in the time-series curve of each lung function test dimension in each observation period, the corresponding temporal transient complexity is determined.

[0012] For each lung function test dimension, the complexity change value for each observation period is determined based on the difference between the reference temporal transient complexity of each observation period and the reference temporal transient complexity of the previous observation period; the overall change value is determined based on the mean of the complexity change values ​​for each lung function test dimension across all observation periods.

[0013] Based on the temporal transient complexity deviation between each lung function test dimension and other lung function test dimensions in each observation period, the corresponding overall deviation value is determined;

[0014] The product between the overall change value and the overall deviation value is normalized to determine the corresponding single-channel weighted evolution index.

[0015] Furthermore, the process of obtaining the temporal transient complexity includes:

[0016] In the time-series data curve of each lung function test dimension in each observation period, the mean of all sampled data values ​​is normalized to determine the physiological intensity value of each lung function test dimension in each observation period. Based on the information entropy of the tangent slope value of all sampled data values ​​in the time-series data curve of each lung function test dimension in each observation period, the corresponding time-series data complexity is determined. Based on the product between the physiological intensity value and the time-series data complexity, the time-series transient complexity of each lung function test dimension in each observation period is determined.

[0017] Furthermore, the process of obtaining the overall deviation value includes:

[0018] Within each observation period, the corresponding complexity deviation value is determined based on the difference between the temporal transient complexity of each lung function test dimension and the temporal transient complexity of each other lung function test dimension; the corresponding overall deviation value is determined based on the mean of the complexity deviation values ​​of each lung function test dimension and all other lung function test dimensions.

[0019] Furthermore, the process of obtaining the feature response sensitivity includes:

[0020] During the current observation period, the sampling time corresponding to the maximum value of the tangent slope of all sampling times of the time-series curve of each lung function test dimension is taken as the corresponding data mutation time.

[0021] Based on the temporal deviation of the data abrupt change time between each lung function test dimension and other lung function test dimensions, the corresponding perturbation response value is determined;

[0022] The corresponding feature response sensitivity is determined by multiplying the perturbation response value of each lung function test dimension in the current observation period with the single-channel weighted evolution index.

[0023] Furthermore, the process of obtaining the disturbance response value includes:

[0024] Each lung function test dimension is used as the target dimension; other lung function test dimensions besides the target dimensions are used as reference dimensions.

[0025] Within the current observation period, the difference between the index value of the data mutation moment in time sequence for each reference dimension and the index value of the data mutation moment in time sequence for the target dimension is normalized to determine the mutation time sequence deviation value for each reference dimension; based on the mean of the mutation time sequence deviation values ​​for all reference dimensions, the disturbance response value for the target dimension is determined.

[0026] Furthermore, the process of obtaining the contribution degree includes:

[0027] Acquire all historical sample data for each prior lesion type group under each lung function test dimension; within the current observation period, determine the corresponding overall category deviation value based on the overall magnitude of the data value of the time series curve of each lung function test dimension and the overall deviation between the historical sample data of each prior lesion type group;

[0028] The standard deviation of all historical sample data for each prior lesion type group under each lung function test dimension is used to determine the corresponding sample dispersion; the mean of the sample dispersion of all prior lesion type groups under each lung function test dimension is used to determine the corresponding overall dispersion.

[0029] The product of the negative correlation mapping value of the overall dispersion and the overall category deviation value is normalized to determine the discriminative contribution of each lung function test dimension in the current observation period.

[0030] Furthermore, the process of obtaining the overall category deviation value includes:

[0031] Calculate the average of all historical sample data for each prior lesion type group under each pulmonary function test dimension to determine the corresponding historical sample feature value; within the current observation period, determine the corresponding current observation feature value based on the average of the data values ​​at all sampling times in the data time series curve of each pulmonary function test dimension; determine the historical category deviation value for each pulmonary function test dimension under each prior lesion type group based on the difference between the current observation feature value and the historical sample feature value; determine the corresponding overall category deviation value for each pulmonary function test dimension under all prior lesion type groups based on the average of the historical category deviation values ​​for each pulmonary function test dimension under all prior lesion type groups.

[0032] Furthermore, the process of obtaining the weighted discrimination signal strength includes:

[0033] By positively mapping the mean values ​​of the feature response sensitivity and the discrimination contribution, the weighted discrimination signal intensity of each lung function examination dimension under the current observation period is determined.

[0034] Furthermore, the process of classifying lung function test data based on the weighted discrimination signal intensity and the time-series data curve of the current observation period includes:

[0035] Within the current observation period, the data values ​​of the time-series curves for each lung function examination dimension at each sampling time are arranged in chronological order to determine the corresponding time-series data sequence. The weighted discrimination signal intensity of each lung function examination dimension within the current observation period is used as the sample weight of the corresponding time-series data sequence to train the classification model, thus determining the trained lung function examination data classification model. The lung function examination data is then classified according to the lung function examination data classification model.

[0036] Secondly, this application provides a computer device including a memory and a processor. The memory is used to store computer program code, and the processor is used to call and run the computer program code from the memory to execute a system as described in the first aspect of this application or any embodiment of the first aspect.

[0037] Thirdly, this application provides a computer program product, which includes computer program code that, when executed, performs a system as described in the first aspect of this application or any embodiment thereof.

[0038] Fourthly, this application provides a computer-readable storage medium that stores computer program code, which, when executed, performs a system as described in the first aspect of this application or any embodiment thereof.

[0039] This application has the following beneficial effects:

[0040] First, a single-channel weighted evolution index is calculated based on the variation deviation of the time-series curves of multi-channel data to capture the dynamic evolution characteristics of each dimension within a continuous observation window. Then, combined with the variation deviation of data change trends between multiple channels, the feature response sensitivity of each dimension in response to changes in pathological state is accurately evaluated. Finally, by comparing the differences between current data and historical lesion characteristics, the discriminative contribution of each dimension is quantified when considering historical data of various lesion types. This allows the weighted discriminative signal intensity obtained based on feature response sensitivity and discriminative contribution to more comprehensively evaluate the sample contribution of each pulmonary function test dimension. This results in higher accuracy in training classification models using weighted discriminative signal intensity as sample weights, thereby improving the accuracy of pulmonary function test data classification. Attached Figure Description

[0041] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a structural diagram of an automatic analysis and classification system for lung function test data based on big data, provided in one embodiment of the present invention.

[0043] Figure 2 This is a schematic diagram of a computer device structure provided in one embodiment of the present invention. Detailed Implementation

[0044] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an automatic analysis and classification system for lung function test data based on big data proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment, and specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0046] The following description, in conjunction with the accompanying drawings, details a specific solution for an automatic analysis and classification system for lung function test data based on big data, provided by this invention.

[0047] This application provides an automatic analysis and classification system for lung function test data based on big data. Please refer to [link to relevant documentation]. Figure 1 The diagram illustrates a structure of an automatic analysis and classification system for lung function test data based on big data, according to an embodiment of the present invention. The system includes: a data acquisition and preprocessing module 101, a first determination module 102, a second determination module 103, and a classification module 104.

[0048] The data acquisition and preprocessing module 101 is used to acquire the time-series curves of data for each lung function test dimension in each observation period.

[0049] In a medical database, the overall time-series curves for each historical patient across all pulmonary function test dimensions are collected. The time range of the overall time-series curve is divided into at least two observation periods. The local curve corresponding to each observation period on the overall time-series curve is then used as the data time-series curve for each pulmonary function test dimension within each observation period. In a specific implementation of this invention, the pulmonary function test dimensions include: flow rate, volume, heart rate, blood oxygen saturation, and transdiaphragmatic pressure. These dimensions can be adjusted according to the specific implementation environment, such as by introducing respiratory cycles and respiratory rates. The observation period is set to 5 minutes, the overall time-series curve is set to 30 minutes, and the sampling frequency corresponding to each sampling moment in the data time-series curve is set to 100Hz. These parameters can be adjusted according to the specific implementation environment and will not be elaborated further here. By setting the observation period, the temporal variation characteristics of the data corresponding to the pulmonary function test dimensions can be captured, improving the accuracy of subsequent analysis.

[0050] The first determining module 102 is used to determine the corresponding single-channel weighted evolution index based on the deviation of the data value time series curve of each lung function test dimension in each observation period compared with the data time series curves of all other data; and to determine the corresponding feature response sensitivity in the current observation period based on the deviation of the data change trend between the data time series curve of each lung function test dimension and the data time series curves of other lung function test dimensions, as well as the corresponding single-channel weighted evolution index.

[0051] By performing dimensional coupling analysis on the data generated during pulmonary function testing, the dynamic correlation patterns and coupling strength distributions between various dimensions can be revealed at the large-scale data level. This helps the system reduce the weight of dimensions corresponding to redundant parameters before model training, avoiding problems such as feature collinearity and gradient non-convergence in subsequent classification models. For each pulmonary function test dimension, the more significant its data change trend in time sequence, and the greater the deviation of its change trend from that of other pulmonary function test dimensions, the higher the independent discriminative value of the corresponding pulmonary function test indicator in time sequence. It is more likely to carry non-redundant and specific case physiological information, and therefore, the more attention should be paid to the corresponding pulmonary function test dimension. Therefore, based on the deviation of the data value time sequence curve of each pulmonary function test dimension in each observation period from the time sequence curves of all other data, the corresponding single-channel weighted evolution index is determined. The larger the single-channel weighted evolution index, the more attention should be paid to the corresponding pulmonary function test dimension.

[0052] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the single-channel weighted evolution index includes:

[0053] Based on the overall magnitude and temporal variation disorder of the data values ​​at all sampling moments in the data time-series curve of each lung function examination dimension in each observation period, the corresponding temporal transient complexity is determined. In a specific implementation of this invention, the process of obtaining the temporal transient complexity includes: normalizing the mean of the data values ​​at all sampling moments in the data time-series curve of each lung function examination dimension in each observation period to determine the physiological intensity value of each lung function examination dimension in each observation period; determining the corresponding temporal data complexity based on the information entropy of the tangent slope value at all sampling moments in the data time-series curve of each lung function examination dimension in each observation period; and determining the temporal transient complexity of each lung function examination dimension in each observation period based on the product between the physiological intensity value and the temporal data complexity.

[0054] Based on the definitions of information entropy and slope, the complexity of time-series data can characterize the local dynamic uncertainty structure of the corresponding time-series curve. When the slope distribution is wider, i.e., the curve has frequent acceleration and deceleration and other local abrupt changes, it corresponds to a higher information entropy, which corresponds to the complex characteristics of local dynamic changes. The physiological intensity value obtained based on the mean of the data values ​​characterizes the overall amplitude within the corresponding observation period, which belongs to the characterization of physiological intensity or respiratory driving force. When analyzing time-series data complexity that characterizes slope information entropy alone, it is impossible to distinguish between strong noise features and weak physiological signal features. When analyzing time-series data complexity alone, it is impossible to distinguish between structured transient signal change features and white noise features. Therefore, this embodiment of the invention combines physiological intensity value and time-series data complexity through a product, so that the obtained reference time-series transient complexity couples the signal structure and capability, which can more accurately reflect the sequential changes and potential abnormal features of lung function data. However, the reference time-series transient complexity is limited to the single-channel level and cannot effectively characterize the macroscopic evolution law of lung function data in time. Therefore, further analysis is needed based on the time-series transient complexity.

[0055] In one specific implementation of this invention, the process of obtaining the temporal transient complexity is expressed by the following formula: ;in, Dimensions of lung function tests During the observation period The time transient complexity; Dimensions of lung function tests During the observation period The information entropy of the tangent slope values ​​at all sampling times in the data time series curve, which is also the corresponding time series data complexity; Dimensions of lung function tests During the observation period Number of sampling times; Dimensions of lung function tests During the observation period The Middle Data values ​​at each sampling time; It is a minimum-maximum normalization function; Dimensions of lung function tests During the observation period The physiological strength value.

[0056] For each pulmonary function test dimension, the complexity change value for each observation period is determined based on the difference between the reference temporal transient complexity of each observation period and the reference temporal transient complexity of the previous observation period. The overall change value is determined based on the average of the complexity change values ​​for each pulmonary function test dimension across all observation periods. The calculation process of the overall change value shows that a larger overall change value indicates a more drastic jump in transient complexity between adjacent observation periods, indicating that its local dynamic characteristics change continuously over time rather than fluctuating randomly. The more obvious its dynamic temporal evolution trend, the more likely it is to represent that specific pathological features were captured in the corresponding pulmonary function test dimension within the corresponding observation period, and therefore, the more attention should be paid to the data for the corresponding pulmonary function test dimension within the corresponding observation period. It should be noted that the difference in this embodiment represents the absolute value of the difference, which will not be further elaborated here.

[0057] Further, based on the temporal transient complexity deviation between each lung function examination dimension and other lung function examination dimensions in each observation period, the corresponding overall deviation value is determined. In a specific implementation of this invention, the process of obtaining the overall deviation value includes: in each observation period, determining the corresponding complexity deviation value based on the difference between the temporal transient complexity of each lung function examination dimension and the temporal transient complexity corresponding to each other lung function examination dimension; and determining the corresponding overall deviation value based on the average of the complexity deviation values ​​between each lung function examination dimension and all other lung function examination dimensions.

[0058] The overall deviation value represents the deviation between different pulmonary function test dimensions within the same observation period. The larger the overall deviation value, the more significant the specificity of the corresponding pulmonary function test dimension relative to the overall pulmonary function test dimensions. The information it represents is more different from the overall signal pattern of all pulmonary function test dimensions, and it is more likely to correspond to physiological abnormalities or key discriminative signals. Therefore, it is more important to pay attention to the data of the corresponding pulmonary function test dimension within the corresponding observation period.

[0059] Therefore, this embodiment of the invention further normalizes the product between the overall change value and the overall deviation value to determine the corresponding single-channel weighted evolution index; so that when the single-channel weighted evolution index is larger, the time evolution trend shown by the lung function examination dimension in the corresponding observation period is more significant, the specificity of the corresponding lung function examination dimension is more significant, and the more attention needs to be paid to the data of the corresponding lung function examination dimension in the corresponding observation period.

[0060] In one specific implementation of this invention, the process of obtaining the single-channel weighted evolution index is expressed by the formula: ;in, Dimensions of lung function tests During the observation period The single-channel weighted evolution index; Dimensions of lung function tests The average of the complexity changes over all observation periods. The complexity change for the first observation period of all lung function test dimensions is set to 0 by default, but can be adjusted according to the specific implementation environment. Dimensions of lung function tests The number of other lung function test dimensions besides; Observation period Dimensions of Lung Function Test Its temporal transient complexity is different from other... The difference between the temporal transient complexity of each lung function test dimension, i.e., the corresponding complexity deviation value; Observation period Dimensions of Lung Function Test The overall deviation value. It should be noted that, in the calculation of all parameters in the formulas of this invention, the units of each parameter are not taken into account, and the formula calculation is based solely on the numerical value. Further details will not be provided here.

[0061] In the current observation period, the sampling moment with the steepest slope on the time-series curve corresponding to each lung function test dimension usually corresponds to a significant local mutation. If the sampling moment with the steepest slope for a certain lung function test dimension is relatively earlier than the sampling moments with the steepest slope for other lung function test dimensions, it indicates that the lung function test dimension responds to the overall system disturbance earlier. From the perspective of disturbance response, the channel information generation rate of the corresponding lung function test dimension is high and the response is rapid. The sensitivity of the feature is higher, and the corresponding lung function test dimension needs more attention in the current observation period. Therefore, in this embodiment of the invention, in the current observation period, the corresponding feature response sensitivity is determined based on the deviation of the data change trend between the time-series curve of each lung function test dimension and the data time-series curve of other lung function test dimensions, as well as the corresponding single-channel weighted evolution index. The higher the feature response sensitivity, the more attention needs to be paid to the corresponding lung function test dimension.

[0062] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the feature response sensitivity includes:

[0063] During the current observation period, the sampling time corresponding to the maximum value of the tangent slope of all sampling times of the data time-series curve for each lung function test dimension is taken as the corresponding data mutation time. The smaller the time-series index value of the corresponding lung function test dimension's data mutation time compared to the data mutation times of other lung function test dimensions, the earlier the lung function test dimension responds to the overall system disturbance. Therefore, this embodiment of the invention further determines the corresponding disturbance response value based on the time-series deviation of the data mutation times between each lung function test dimension and other lung function test dimensions.

[0064] In one specific implementation of this invention, the process of obtaining the disturbance response value includes: sequentially using each lung function examination dimension as the target dimension; using other lung function examination dimensions besides the target dimension as reference dimensions; normalizing the difference between the index value of the data mutation moment in time sequence of each reference dimension and the index value of the data mutation moment in time sequence of the target dimension within the current observation period to determine the mutation time sequence deviation value of each reference dimension; and determining the disturbance response value of the target dimension based on the mean of the mutation time sequence deviation values ​​of all reference dimensions. Here, the index value in time sequence corresponds to the index value of the sampling moment corresponding to the data mutation moment within the current observation period. Therefore, the smaller the mutation time sequence deviation value, the earlier the overall system disturbance response of the target dimension appears compared to the corresponding reference dimension. Thus, the larger the disturbance response value, the higher the channel information generation rate and the faster the response of the corresponding lung function examination dimension from the perspective of disturbance response, and the higher the sensitivity in terms of features, the more attention needs to be paid to the corresponding lung function examination dimension within the current observation period.

[0065] Furthermore, by combining the perturbation response value and the single-channel weighted evolution index to characterize the lung function examination dimensions, the corresponding feature response sensitivity is determined based on the product between the perturbation response value and the single-channel weighted evolution index of each lung function examination dimension in the current observation period. This ensures that the greater the feature response sensitivity, the more attention needs to be paid to the corresponding lung function examination dimension in the current observation period.

[0066] In one specific implementation of this invention, the process of obtaining the feature response sensitivity is expressed by the following formula: ;in, Dimensions of lung function tests During the current observation period The characteristic response sensitivity under; Dimensions of lung function tests During the current observation period The single-channel weighted evolution index under the given conditions; Dimensions of lung function tests The number of reference dimensions; The current observation period Lower pulmonary function test dimensions The corresponding number The index value of the data mutation moment in each reference dimension in chronological order and the lung function test dimension The difference between the index values ​​in time sequence at the moment of data mutation; The current observation period Lower pulmonary function test dimensions The mutation timing deviation value.

[0067] The second determining module 103 is used to determine the discriminant contribution of each lung function examination dimension based on the data feature deviation between the time-series curve of each lung function examination dimension and the historical data of each lesion feature during the current observation period.

[0068] While feature response sensitivity can effectively characterize the immediate response speed of each pulmonary function test dimension to physiological disturbances, it is essentially an unsupervised, low-level signal feature and is not directly related to the final classification label. Therefore, it is necessary to further incorporate historical data of each pathological feature. Within the current observation period, based on the data feature deviation between the time-series curve of each pulmonary function test dimension and the historical data of each pathological feature, the discriminative contribution of each pulmonary function test dimension can be determined. Analyzing the contribution of feature parameters to classification discrimination can identify the true influence of different parameters on the classification structure in the multidimensional feature space, thereby revealing the nonlinear dependencies and collinear structural features of the feature space, and thus strengthening the weights of features with strong classification sensitivity.

[0069] Preferably, in some possible implementations of the embodiments of the present invention, the process of determining the contribution level includes:

[0070] The process involves acquiring all historical sample data for each prior lesion type group under each pulmonary function test dimension; within the current observation period, determining the corresponding overall category deviation value based on the overall magnitude of the data values ​​in the time-series curve of each pulmonary function test dimension and the overall deviation between the historical sample data of each prior lesion type group; the process of obtaining the overall category deviation value includes: calculating the average value of all historical sample data for each prior lesion type group under each pulmonary function test dimension to determine the corresponding historical sample feature value; within the current observation period, determining the corresponding current observation feature value based on the average value of the data values ​​at all sampling times in the time-series curve of each pulmonary function test dimension; and determining the historical category deviation value for each pulmonary function test dimension under each prior lesion type group based on the difference between the current observation feature value and the historical sample feature value. In a specific implementation of this invention, the prior lesion type grouping includes obstructive, restrictive, and mixed types, which can be adjusted according to the specific implementation environment.

[0071] The larger the historical category deviation value, the more the data value characteristics of the corresponding pulmonary function test dimension deviate from the corresponding prior lesion type group. For each pulmonary function test dimension, if its data value characteristics have significant data value deviation characteristics in all prior lesion type groups, it indicates that the pulmonary function test dimension has strong separating characteristics between different prior lesion type groups in the data space, and does not belong to any known category. The corresponding pulmonary function test dimension has a higher potential discriminative ability for classification. Therefore, the overall category deviation value is further determined based on the mean of the historical category deviation values ​​of each pulmonary function test dimension under all prior lesion type groups. A larger overall category deviation value should correspond to a higher discriminative contribution.

[0072] Furthermore, the standard deviation of all historical sample data for each prior lesion type group under each pulmonary function test dimension is used to determine the corresponding sample dispersion. The mean of the sample dispersion for all prior lesion type groups under each pulmonary function test dimension is used to determine the corresponding overall dispersion. For each prior lesion type group, the smaller the sample dispersion, the higher the stability of the sample features in the corresponding prior lesion type group, and the more reliable the data. Therefore, the larger the overall dispersion, the higher the reliability of the overall class bias value in representing the discriminative contribution. Finally, the product between the negative correlation mapping value of the overall dispersion and the overall class bias value is normalized to determine the discriminative contribution of each pulmonary function test dimension in the current observation period. This ensures that the larger the discriminative contribution, the stronger the discriminative power of the corresponding pulmonary function test dimension in the sample data, and the more attention should be paid to the corresponding pulmonary function test dimension.

[0073] In one specific implementation of this invention, the process of determining contribution is expressed by the following formula: ;in, The current observation period Lower pulmonary function test dimensions The discriminant contribution; Dimensions of lung function tests The mean of the sample dispersion of all prior lesion types grouped together, that is, the overall dispersion; The number of groups for prior lesion types; Dimensions of lung function tests Next The average value of all historical sample data for each prior lesion type group, which is also the corresponding historical sample feature value; Dimensions of lung function tests The mean of all data values ​​at all sampling times in the time series curve during the current observation period, which is also the current observed characteristic value; it is an absolute value. Lung function test dimensions for the current observation period Next Historical category bias values ​​for prior lesion type groupings; Dimensions of lung function tests The overall category deviation value for the current observation period.

[0074] The classification module 104 is used to determine the weighted discriminant signal intensity of each lung function test dimension in the current observation period based on the feature response sensitivity and discriminant contribution; and to classify lung function test data based on the weighted discriminant signal intensity and the data time series curve of the current observation period.

[0075] Since both feature response sensitivity and discriminant contribution can characterize the degree of attention paid to each lung function examination dimension during the current observation period from different perspectives, this embodiment of the invention further determines the weighted discriminant signal intensity of each lung function examination dimension during the current observation period based on feature response sensitivity and discriminant contribution. This means that the greater the weighted discriminant signal intensity, the more attention is needed to the corresponding lung function examination dimension, the higher the importance of the sample data for the corresponding lung function examination dimension during the current observation period, and the greater the sample weight that needs to be given when training the classification model. Therefore, in a specific implementation of this embodiment, a positive correlation mapping is performed between the mean of feature response sensitivity and discriminant contribution to determine the weighted discriminant signal intensity of each lung function examination dimension during the current observation period. The greater the weighted discriminant signal intensity determined by feature response sensitivity and discriminant contribution, the more sensitive and discriminative the sample data for the corresponding lung function examination dimension during the current observation period, and the higher its importance. Therefore, the corresponding sample data needs to be paid more attention to during training, and the corresponding sample weight should be greater.

[0076] In one specific implementation of this invention, the process of obtaining the weighted discrimination signal strength is expressed by the following formula: ;in, The current observation period Lower pulmonary function test dimensions The weighted discrimination signal strength; The current observation period Lower pulmonary function test dimensions The characteristic response sensitivity; The current observation period Lower pulmonary function test dimensions The discriminant contribution; To preset the positive correlation mapping parameter, this embodiment of the invention sets it to 0.5, which is used to adjust the value range of the weighted discrimination signal intensity so as to adjust the sample weights when training the classification model later as needed. It can be adjusted according to the specific implementation environment.

[0077] Furthermore, it is necessary to train a classification model based on the sample weights represented by the weighted discriminant signal intensity, thereby enabling more accurate classification of pulmonary function test data based on the classification model. Preferably, in some possible implementations of this invention, the process of classifying pulmonary function test data based on the weighted discriminant signal intensity combined with the data time-series curve of the current observation period includes:

[0078] Within the current observation period, the data values ​​of the time-series curves for each lung function examination dimension at each sampling time are arranged chronologically to determine the corresponding time-series data sequence. The weighted discriminant signal intensity of each lung function examination dimension within the current observation period is used as the sample weight of the corresponding time-series data sequence for training the classification model, thus determining the trained lung function examination data classification model. Lung function examination data is then classified according to the lung function examination data classification model. The weighted discriminant signal intensity is equivalent to the sample weight of the corresponding lung function examination dimension within the current observation period. A higher weighted discriminant signal intensity requires more attention to the corresponding sample data (i.e., the corresponding time-series data sequence) during classification model training, thereby strengthening the data features of high-attention lung function examination dimensions and weakening the data features of low-attention lung function examination dimensions. This results in higher accuracy in classifying lung function examination data based on the trained lung function examination data classification model. In one specific implementation of this invention, the classification model uses the XGBoost model. Training the XGBoost model is a technique well-known to those skilled in the art and will not be further limited or described here.

[0079] In summary, an automatic analysis and classification system for pulmonary function test data based on big data first calculates a single-channel weighted evolution index based on the variation deviation of the time-series curves of multi-channel data, capturing the dynamic evolution characteristics of each dimension within a continuous observation window. Then, by combining the variation trend deviation of data across multiple channels, it accurately assesses the feature response sensitivity of each dimension in response to changes in pathological states. Finally, by comparing the differences between current data and historical lesion characteristics, it quantifies the discriminative contribution of each dimension when considering historical data for various lesion types. This allows the weighted discriminative signal intensity obtained based on feature response sensitivity and discriminative contribution to more comprehensively evaluate the sample contribution of each pulmonary function test dimension. Furthermore, it improves the accuracy of classification model training when using weighted discriminative signal intensity as sample weights, thereby enhancing the accuracy of pulmonary function test data classification.

[0080] This application also provides a computer device; please refer to [link / reference]. Figure 2 The illustration shows a schematic diagram of a computer device structure according to an embodiment of the present invention. The computer device includes a memory 201, a processor 202, and a computer program 203 stored in the memory 201 and running on the processor 202. When the processor 202 executes the computer program 203, the computer device can execute any of the aforementioned automatic analysis and classification systems for lung function test data based on big data.

[0081] This application also provides a computer program product that, when run on a computer device, enables the computer device to execute any of the aforementioned big data-based automatic analysis and classification systems for lung function test data.

[0082] This application also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer device, the computer device can execute any of the aforementioned big data-based automatic analysis and classification systems for lung function test data.

[0083] In the embodiments provided in this application, it should be understood that the computer device, computer program product and computer-readable storage medium provided are all used to execute the corresponding system provided above, and therefore the beneficial effects they can achieve can be referred to the beneficial effects of the system provided above, which will not be repeated here.

[0084] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0085] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A big data based automatic analysis and classification system for pulmonary function test data, characterized in that, The system comprises: a data acquisition preprocessing module, configured to acquire a data time series curve of each lung function test dimension in each observation time period; a first determination module, configured to determine a corresponding single-channel weighted evolution index according to a data value time series change deviation of the data time series curve of each lung function test dimension in each observation time period compared with all other data time series curves; in the current observation time period, determine a corresponding feature response sensitivity according to a data change trend deviation between the data time series curve of each lung function test dimension and the data time series curves of other lung function test dimensions and the corresponding single-channel weighted evolution index; the acquisition process of the single-channel weighted evolution index comprises: determine a corresponding time series transient complexity according to an overall size of data values at all sampling moments in the data time series curve of each lung function test dimension in each observation time period and a time series change confusion situation; under each lung function test dimension, determine a complexity change value of each observation time period according to a difference between a reference time series transient complexity of each observation time period and a reference time series transient complexity of a previous observation time period; determine a corresponding overall change value according to a mean value of the complexity change values of each lung function test dimension in all observation time periods; determine a corresponding overall deviation value according to a time series transient complexity deviation between each lung function test dimension and other lung function test dimensions in each observation time period; normalize a product between the overall change value and the overall deviation value to determine the corresponding single-channel weighted evolution index; the acquisition process of the time series transient complexity comprises: in the data time series curve of each lung function test dimension in each observation time period, normalize a mean value of data values at all sampling moments to determine a physiological intensity value of each lung function test dimension in each observation time period; determine a corresponding time series data complexity according to information entropy of tangent slope values at all sampling moments in the data time series curve of each lung function test dimension in each observation time period; determine a time series transient complexity of each lung function test dimension in each observation time period according to a product between the physiological intensity value and the time series data complexity; the acquisition process of the feature response sensitivity comprises: in the current observation time period, take a sampling moment corresponding to a maximum value of tangent slopes of all sampling moments of the data time series curve of each lung function test dimension as a corresponding data mutation moment; determine a corresponding perturbation response value according to a time series deviation of data mutation moments between each lung function test dimension and other lung function test dimensions; determine a corresponding feature response sensitivity according to a product between the perturbation response value of each lung function test dimension in the current observation time period and the single-channel weighted evolution index; a second determination module, configured to determine a discriminant contribution degree of each lung function test dimension according to a data feature deviation between the data time series curve of each lung function test dimension and historical data of each lesion feature in the current observation time period; the acquisition process of the discriminant contribution degree comprises: Obtain all historical sample data of each prior lesion type grouping under each lung function test dimension; in the current observation period, determine the corresponding overall category deviation value according to the overall deviation between the overall size of the data value of the data time curve of each lung function test dimension and the historical sample data of each prior lesion type grouping; Determine the sample dispersion degree of each prior lesion type grouping under each lung function test dimension by the standard deviation of all historical sample data of each prior lesion type grouping under each lung function test dimension; determine the overall dispersion degree according to the mean value of the sample dispersion degree of all prior lesion type groupings under each lung function test dimension. Determine the discriminant contribution degree of each lung function test dimension in the current observation period by normalizing the product between the negative correlation mapping value of the overall dispersion degree and the overall category deviation value. The classification module is used to determine the weighted discriminant signal intensity of each lung function test dimension in the current observation period according to the feature response sensitivity and the discriminant contribution degree; and perform lung function test data classification by combining the weighted discriminant signal intensity with the data time curve in the current observation period.

2. The system as claimed in claim 1, wherein, The overall deviation value acquisition process includes: In each observation period, determine the complexity deviation value of each lung function test dimension according to the difference between the time sequence transient complexity of each lung function test dimension and the time sequence transient complexity of other lung function test dimensions; and determine the overall deviation value according to the mean value of the complexity deviation values between each lung function test dimension and all other lung function test dimensions.

3. The system as claimed in claim 1, wherein The perturbation response value acquisition process includes: In turn, take each lung function test dimension as the target dimension; and take other lung function test dimensions outside the target dimension as the reference dimension; In the current observation period, normalize the difference between the index value of the data mutation time of each reference dimension in time sequence and the index value of the data mutation time of the target dimension in time sequence to determine the mutation time sequence deviation value of each reference dimension; and determine the perturbation response value of the target dimension according to the mean value of the mutation time sequence deviation values of all reference dimensions.

4. The system as claimed in claim 1, wherein, The overall category deviation value acquisition process includes: Calculate the average value of all historical sample data of each prior lesion type grouping under each lung function test dimension to determine the corresponding historical sample feature value; in the current observation period, determine the corresponding current observation feature value according to the mean value of the data values of all sampling time points in the data time curve of each lung function test dimension; determine the historical category deviation value of each lung function test dimension under each prior lesion type grouping according to the difference between the current observation feature value and the historical sample feature value; and determine the corresponding overall category deviation value according to the mean value of the historical category deviation values of each lung function test dimension under all prior lesion type groupings.

5. The system as claimed in claim 1, wherein, The weighted discriminant signal intensity acquisition process includes: Determine the weighted discriminant signal intensity of each lung function test dimension in the current observation period by positively correlating the mean value between the feature response sensitivity and the discriminant contribution degree.

6. The big data based automatic analysis and classification system for pulmonary function test data of claim 1, wherein, The process of classifying the pulmonary function test data according to the weighted discriminant signal strength in combination with the data time sequence curve of the current observation time period comprises: In the current observation time period, the data values of the data time sequence curve of each pulmonary function test dimension at each sampling time are arranged in time sequence to determine the corresponding time sequence data sequence; the weighted discriminant signal strength of each pulmonary function test dimension in the current observation time period is used as the sample weight of the corresponding time sequence data sequence to train the classification model, and a trained pulmonary function test data classification model is determined; and the pulmonary function test data classification is performed according to the pulmonary function test data classification model.

Images