A method and system for predicting the risk of copd complicated with chronic pulmonary heart disease

Abstract

This application discloses a method and system for predicting the risk of COPD complicated by chronic pulmonary heart disease, relating to the field of medical data analysis technology. The method includes the following steps: collecting physiological waveform data from patients based on multiple physically dispersed medical data sources with heterogeneous data formats; performing signal decomposition processing on the physiological waveform data to obtain multiple signal components used to reveal the time-frequency domain details of the physiological waveform data; extracting physiological signal features indicating the progression of COPD to chronic pulmonary heart disease from the signal components; generating a risk assessment of COPD complicated by chronic pulmonary heart disease based on the physiological signal features; generating explanatory information for the risk assessment, and presenting the explanatory information through a query interface. The query interface is configured to simultaneously display the physiological signal features and the waveform segments from the physiological waveform data from which the physiological signal features are derived in response to the user's query operation for the explanatory information. This application can improve the utilization rate and interpretability of available patient data.

Description

Technical Field

[0001] This application relates to the field of medical data analysis technology, and in particular to a method and system for predicting the risk of COPD complicated with chronic pulmonary heart disease. Background Technology

[0002] In medical practice, early and accurate detection of the risk of chronic obstructive pulmonary disease (COPD) patients developing chronic cor pulmonale is crucial. Traditional predictive systems primarily rely on processing pre-formatted electronic medical record data, typically stored in centralized databases, including basic patient information, diagnostic records, medication details, and standardized test results. This design assumes all relevant information can be integrated into a standardized and easily accessible storage environment to ensure efficient data retrieval and processing.

[0003] However, with the development of medical technology, new data sources are constantly emerging, such as high-resolution CT images, raw lung function waveform data, and physiological indicators from home monitoring devices. Due to their specialized nature, unique formats, and massive volume, these data cannot be smoothly integrated into existing centralized electronic medical record databases. For example, CT images are managed by independent PACS systems, raw lung function waveform data is stored in device-specific formats on diagnostic workstations, and home monitoring data flows into cloud platforms or independent research databases. This critical information is physically dispersed, heterogeneous in format, and follows different data management rules. Existing data query programs, designed for processing fixed-format electronic medical record data, cannot effectively access and understand this fragmented and diverse data landscape. They lack the complex natural language understanding capabilities required for in-depth analysis of free-text descriptions and cannot interpret raw physiological waveform data to extract meaningful features.

[0004] This architectural difference prevents the system from fully utilizing all available patient data. This inadequate data integration directly leads to a significant decrease in the accuracy of predicting the risk of COPD-related chronic pulmonary heart disease, especially when dealing with patients with complex or atypical conditions. Because the system cannot access these key, fragmented data points, it fails to incorporate nuanced indicators, and its risk assessment results often differ significantly from the judgments of experienced clinicians. Furthermore, the system's interpretability is impaired; its interpretive reports typically only cite fixed-format data and fail to provide insights into how crucial but fragmented information, such as CT reports or raw pulmonary function data, affects the results, leading to decreased trust from healthcare professionals. Summary of the Invention

[0005] This application proposes a method and system for predicting the risk of COPD complicated with chronic pulmonary heart disease, aiming to solve the technical problem that the physical dispersion of medical data sources and the heterogeneity of data formats lead to low utilization of available patient data, which in turn affects the accuracy and interpretability of prediction.

[0006] Firstly, this application provides a method for predicting the risk of COPD complicated with chronic pulmonary heart disease, comprising the following steps:

[0007] Physiological waveform data of patients are collected from multiple medical data sources that are physically dispersed and have heterogeneous data formats.

[0008] The physiological waveform data is subjected to signal decomposition processing to obtain multiple signal components used to reveal the time-frequency domain detailed features of the physiological waveform data;

[0009] From the signal components, physiological signal features indicating the progression of COPD to chronic cor pulmonale were extracted;

[0010] Based on the aforementioned physiological signal characteristics, a risk assessment for COPD complicated by chronic pulmonary heart disease is generated.

[0011] The risk assessment explanation information is generated and presented through a query interface. The query interface is configured to: respond to the user's query operation on the explanation information, synchronously display the physiological signal characteristics and the waveform segments from the physiological waveform data from which the physiological signal characteristics are derived.

[0012] In some embodiments of this application, the step of collecting patient physiological waveform data based on multiple physically distributed medical data sources with heterogeneous data formats includes:

[0013] Acquire raw data from multiple physically dispersed medical data sources with heterogeneous data formats. The raw data includes physiological monitoring data generated by medical monitoring devices and clinical context data from clinical interaction systems. The clinical context data includes at least one of structured electronic medical record data that records patient symptoms, activities, or medication events, and logs of cough and dyspnea attacks reported by patients.

[0014] The physiological monitoring data and the clinical scenario data are respectively converted into first intermediate data and second intermediate data with uniform format;

[0015] Based on the first intermediate data and the second intermediate data, the patient's physiological waveform data is generated.

[0016] In some embodiments of this application, the step of generating the patient's physiological waveform data based on the first intermediate data and the second intermediate data includes:

[0017] The first intermediate data is denoised and baseline drift corrected to obtain preliminary waveform data, and a first physiological waveform is generated based on the preliminary waveform data.

[0018] Based on the second intermediate data, exogenous physiological disturbance factors are identified and quantified, and corresponding disturbance waveform models are generated.

[0019] The second physiological waveform is generated by subtracting the perturbation component represented by the perturbation waveform model from the first physiological waveform.

[0020] The first physiological waveform and the second physiological waveform are used together as the patient's physiological waveform data for subsequent waveform selection steps.

[0021] In some embodiments of this application, generating explanatory information for the risk assessment and presenting the explanatory information through a query interface, wherein the query interface is configured to: in response to a user's query operation for the explanatory information, synchronously display the physiological signal characteristics and the waveform segments from the physiological waveform data from which the physiological signal characteristics are derived, the steps include:

[0022] Based on the risk assessment and the physiological signal characteristics on which the risk assessment is based, explanatory information is generated to illustrate the contribution of each physiological signal characteristic to the risk assessment.

[0023] The explanatory information is presented through the interactive query interface;

[0024] The query interface is configured to perform the following operations in response to a user's query for the explanatory information:

[0025] The analysis conclusions and data directly corresponding to the queried physiological signal features in the explanatory information are displayed;

[0026] The system displays the first or second physiological waveform derived from the queried physiological signal characteristics.

[0027] Highlight the waveform segments used to calculate the physiological signal characteristics being queried on the first or second physiological waveform displayed.

[0028] In some embodiments of this application, the query interface is further provided with a waveform switching control, which is used to respond to the user's switching command and switch between the first physiological waveform and the second physiological waveform.

[0029] In some embodiments of this application, the step of performing signal decomposition processing on the physiological waveform data to obtain multiple signal components for revealing the time-frequency domain detailed features of the physiological waveform data includes:

[0030] The type of the physiological waveform data is obtained, including pulmonary function oscillation waveforms and electrocardiogram waveforms;

[0031] For the physiological waveform data of type pulmonary function oscillation waveform, the continuous wavelet transform algorithm is used to decompose the physiological waveform data to obtain the wavelet coefficients of the physiological waveform data under different scales and time parameters, and the wavelet coefficients are used as multiple signal components to reveal the time-frequency domain detailed features of the physiological waveform data;

[0032] For the physiological waveform data of type ECG waveform, the empirical mode decomposition algorithm is used to decompose the physiological waveform data to obtain the intrinsic mode functions of the physiological waveform data under different scales and time parameters, and the intrinsic mode functions are used as the signal components.

[0033] In some embodiments of this application, the step of extracting physiological signal features indicating the progression of COPD to chronic cor pulmonale from the signal components includes:

[0034] For the wavelet coefficients, the wavelet energy spectral density of the physiological waveform data of the type of pulmonary function oscillation waveform in the preset frequency band is calculated, and the relative change of the wavelet energy spectral density between the inspiratory and expiratory phases of the respiratory cycle is extracted to obtain the physiological signal features that indicate the progression of COPD to chronic cor pulmonale.

[0035] For the intrinsic mode function, the duration, peak amplitude, and slope of the P wave and T wave in the physiological waveform data of the type of electrocardiogram waveform are extracted and compared with the pre-stored patient baseline data to obtain the physiological signal characteristics.

[0036] In some embodiments of this application, the step of generating a risk assessment for COPD complicated by chronic pulmonary heart disease based on the physiological signal characteristics includes:

[0037] The currently extracted physiological signal features, as well as the historical physiological signal features of the same patient, are obtained to form time-series data;

[0038] The time series data is analyzed within a preset time window to obtain a dynamic risk assessment threshold.

[0039] By comparing the values ​​of the currently extracted physiological signal features at continuous time points with the dynamic risk determination threshold, a continuous comparison result is obtained;

[0040] Based on the continuous comparison results, determine the changing trend of the currently extracted physiological signal features;

[0041] Based on the aforementioned trends, a risk assessment for COPD complicated by chronic pulmonary heart disease is generated.

[0042] In some embodiments of this application, the step of analyzing the time-series data within a preset time window to obtain a dynamic risk assessment threshold includes:

[0043] Calculate the mean and standard deviation of the time series data within a preset time window;

[0044] The dynamic risk assessment threshold is obtained by adding or subtracting the product of the preset multiple and the standard deviation from the average value.

[0045] Secondly, this application provides a risk prediction system for COPD complicated with chronic pulmonary heart disease, including:

[0046] The data acquisition module is used to collect patients' physiological waveform data from multiple physically dispersed medical data sources with heterogeneous data formats.

[0047] The data processing module is used to perform signal decomposition processing on the physiological waveform data to obtain multiple signal components that reveal the time-frequency domain details of the physiological waveform data;

[0048] The feature extraction module is used to extract physiological signal features that indicate the progression of COPD to chronic cor pulmonale.

[0049] The feature conversion module is used to generate a risk assessment for COPD complicated by chronic pulmonary heart disease based on the physiological signal features.

[0050] The information interaction module is used to generate explanatory information for the risk assessment and present the explanatory information through a query interaction interface. The query interaction interface is configured to: respond to the user's query operation on the explanatory information, synchronously display the physiological signal characteristics and the waveform segments from the physiological waveform data from which the physiological signal characteristics are derived.

[0051] The technical solution according to the embodiments of this application has at least the following beneficial effects:

[0052] This application, through the collaborative work of multi-source data acquisition, deep signal processing, intelligent feature extraction, risk assessment, and interpretable presentation, has formed an efficient, accurate, and transparent risk prediction system for COPD complicated by chronic cor pulmonale. The various technical features work closely together to improve the utilization and interpretability of available patient data, thereby enhancing the accuracy of COPD risk prediction for chronic cor pulmonale.

[0053] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0054] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0055] Figure 1 This is a flowchart illustrating a method for predicting the risk of COPD complicated with chronic pulmonary heart disease, provided in an embodiment of this application.

[0056] Figure 2 This is a schematic diagram of the architecture of a COPD-associated chronic pulmonary heart disease risk prediction system provided in an embodiment of this application. Detailed Implementation

[0057] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0058] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0059] like Figure 1 As shown, this application discloses a method for predicting the risk of COPD complicated with chronic pulmonary heart disease, including the following steps:

[0060] S110 collects patients' physiological waveform data based on multiple physically distributed medical data sources with heterogeneous data formats.

[0061] S120, perform signal decomposition processing on the physiological waveform data to obtain multiple signal components used to reveal the time-frequency domain detailed features of the physiological waveform data;

[0062] S130, extract physiological signal features indicating the progression of COPD to chronic cor pulmonale from the signal components;

[0063] S140, Based on the physiological signal characteristics, generate a risk assessment for COPD complicated by chronic pulmonary heart disease;

[0064] S150, generate the explanatory information of the risk assessment, and present the explanatory information through a query interaction interface. The query interaction interface is configured to: in response to the user's query operation on the explanatory information, synchronously display the physiological signal characteristics and the waveform segments in the physiological waveform data from which the physiological signal characteristics are derived.

[0065] To better understand the technical solution of this application, some key terms and implementation environments involved will be explained first.

[0066] "Physiological waveform data" refers to continuous data collected through medical monitoring equipment that reflects changes in a patient's physiological activities, such as electrocardiogram (ECG), pulmonary function oscillation waveforms, and respiratory waveforms. This data is usually presented in time series format and can intuitively reflect the patient's physiological state and pathological changes.

[0067] "Signal decomposition processing" refers to the process of breaking down complex physiological waveform data into multiple simpler, more physically meaningful signal components. Through signal decomposition, hidden time-frequency domain details in the original waveform data can be revealed, such as the energy distribution of different frequency components and instantaneous frequency changes.

[0068] "Time-frequency domain details" refer to the features obtained by analyzing a signal in both time and frequency dimensions. For example, in the time domain, one can focus on the signal's duration, amplitude, and slope; in the frequency domain, one can focus on the signal's frequency components and energy distribution. These details are crucial for understanding the intrinsic patterns and pathological significance of physiological waveform data.

[0069] "Physiological signal characteristics" refer to specific biomarkers or patterns extracted from signal components that can indicate the progression of COPD to chronic cor pulmonale. These characteristics can be quantified values ​​or specific waveform patterns, and they can reflect key information about the patient's disease progression.

[0070] "Risk assessment" refers to the quantitative or qualitative judgment of the likelihood of a patient with COPD developing chronic pulmonary heart disease based on extracted physiological signal characteristics. Risk assessment can be presented in the form of risk scores, risk levels, or risk probabilities.

[0071] "Explanatory information" refers to information that explains and clarifies the risk assessment results, aiming to help users understand the basis and reasons for the risk assessment. Explanatory information may include the contribution of various physiological signal characteristics to the risk assessment, relevant waveform segments, etc.

[0072] The “query interaction interface” refers to the graphical interface through which users interact with the system. Through this interface, users can query risk assessment results, explanatory information, and related physiological waveform data.

[0073] Several methods can be used to collect patients' physiological waveform data. For example, data can be manually exported from different medical data sources (such as hospital PACS systems, pulmonary function diagnostic workstations, and home monitoring cloud platforms), and then manually converted and integrated. While feasible, this method is inefficient and prone to human error. Another approach is to develop adapters or connectors for different data sources. These adapters can automatically identify and connect to specific medical data sources, such as acquiring data from Hospital Information Systems (HIS) and Picture Archiving and Communication Systems (PACS) via standard protocol interfaces like HL7 (Health Level Seven) or DICOM (Digital Imaging and Communications in Medicine). For non-standard format data sources, customized parsers can be developed to extract the required raw physiological waveform data. For example, for proprietary format waveform data stored in pulmonary function diagnostic workstations, specific scripts or programs can be written to parse their file structure and extract the raw pulmonary function oscillation waveforms. Data uploaded from home monitoring devices to the cloud platform can be obtained by calling the cloud platform's API interface.

[0074] Different signal decomposition algorithms can be selected based on the type and characteristics of physiological waveform data. For example, for electrocardiogram (ECG) waveform data, Fourier transform can be used to decompose it into a superposition of different frequency components, thereby obtaining the spectral information of the ECG signal. For pulmonary function oscillation waveform data, time-frequency analysis methods such as short-time Fourier transform (STFT) or wavelet transform (WT) can be used to obtain the energy distribution of the signal at different times and frequencies, thereby revealing its detailed features in the time and frequency domains. For example, wavelet decomposition can be performed on the acquired pulmonary function oscillation waveform data to obtain a series of wavelet coefficients, which represent the detailed information of the original waveform at different scales (frequency) and time positions.

[0075] Based on the signal components obtained after signal decomposition and processing, physiological signal features related to the progression of COPD to chronic cor pulmonale can be extracted. For example, for the spectral information of electrocardiogram signals, heart rate variability (HRV) indicators such as SDNN (standard deviation of all normal RR intervals) and RMSSD (root mean square of the difference between adjacent RR intervals) can be extracted. These indicators reflect the activity of the cardiac autonomic nervous system and are closely related to cardiopulmonary function. For the time-frequency domain details of pulmonary function oscillation waveforms, parameters such as respiratory rate, tidal volume, peak expiratory flow rate, and energy distribution changes in specific frequency ranges can be extracted. These parameters reflect pulmonary ventilation function and airway resistance. For example, the energy in a specific frequency range (e.g., 0.1-0.5 Hz) can be calculated from the wavelet coefficients of pulmonary function oscillation waveforms, and its relative changes between the inspiratory and expiratory phases can be analyzed to assess the degree of airway obstruction.

[0076] Various machine learning or statistical models can be used to generate risk assessments. For example, extracted physiological signal features can be used as input, and classification algorithms such as logistic regression models, support vector machines (SVM), decision trees, or neural networks can be employed to predict the risk level (e.g., low, medium, high risk) of a patient with COPD complicated by chronic pulmonary heart disease. Survival analysis methods such as the Cox proportional hazards model can also be used to predict the probability of a patient developing chronic pulmonary heart disease in the future. For instance, a multivariate logistic regression model can be constructed, using heart rate variability indicators and pulmonary function oscillation waveform characteristics as independent variables, and COPD complicated by chronic pulmonary heart disease as the dependent variable, to train the model and output the risk probability.

[0077] Explanatory artificial intelligence (XAI) techniques can be used to generate explanatory information. For example, methods such as LIME (Local Interpretable Model-agnostic Explanations) or SHAP (SHapley Additive explanations) can be used to analyze the contribution of each physiological signal feature to the risk assessment results and present them in an intuitive way (such as bar charts, heatmaps, etc.). The query interface can be designed to include risk assessment results, an explanatory information display area, a list of physiological signal features, and a waveform display area. When a user clicks or selects a physiological signal feature in the explanatory information display area, the query interface can automatically highlight the feature in the list of physiological signal features and simultaneously load and highlight the corresponding original physiological waveform data segment in the waveform display area. For example, when a user queries the explanatory information of "reduced heart rate variability," the query interface will not only show the contribution of this feature to the high-risk assessment but also simultaneously display the specific value of the heart rate variability index and highlight the RR interval sequence used to calculate the index on the electrocardiogram waveform.

[0078] The method for predicting the risk of COPD complicated with chronic pulmonary heart disease in this application works by constructing a complete closed loop from the collection of multi-source heterogeneous data to interpretable risk assessment.

[0079] First, through the data acquisition step, the system can overcome the limitations of traditional centralized databases and comprehensively acquire patients' physiological waveform data from multiple physically dispersed medical data sources with heterogeneous data formats. This step forms the basis for subsequent analysis, ensuring the integrity and diversity of the data, thereby avoiding prediction bias caused by missing data.

[0080] Secondly, the acquired raw physiological waveform data undergoes signal decomposition processing. This step aims to break down the complex raw waveform into multiple signal components with clearly defined physical meaning in the time and frequency domains. For example, through wavelet transform or empirical mode decomposition, the raw waveform can be decomposed into components of different frequencies and time scales, thereby revealing the hidden pathological information within the waveform. This deep signal processing provides more refined and richer input for subsequent feature extraction.

[0081] Next, the system extracts physiological signal features from these signal components that indicate the progression of COPD to chronic cor pulmonale. These features are key indicators screened using specialized knowledge and algorithms; they can sensitively reflect subtle changes in a patient's cardiopulmonary function, thus serving as biomarkers for disease progression. For example, specific frequency energy changes in pulmonary function oscillation waveforms or morphological changes in P and T waves in electrocardiogram waveforms can become important physiological signal features.

[0082] Subsequently, based on these extracted physiological signal features, the system generates a risk assessment for COPD complicated by chronic pulmonary heart disease. This step typically utilizes machine learning models or statistical analysis methods to integrate multiple physiological signal features and quantify or classify the patient's risk. By integrating multi-dimensional features, the accuracy and reliability of the risk assessment are significantly improved.

[0083] Finally, to enhance the credibility and clinical application value of the risk assessment, this application also generates explanatory information for the risk assessment and presents it through a query interface. When a user queries the explanatory information, the query interface simultaneously displays the relevant physiological signal characteristics, as well as waveform segments in the original physiological waveform data derived from these characteristics. This interactive explanation mechanism allows medical personnel not only to see the risk assessment results but also to gain a deeper understanding of the underlying physiological basis, thereby increasing their trust in the system's predictions and supporting clinical decision-making.

[0084] In summary, this application, through the collaborative work of multi-source data acquisition, deep signal processing, intelligent feature extraction, risk assessment, and interpretability presentation, has developed an efficient, accurate, and transparent risk prediction system for COPD-related chronic pulmonary heart disease. The close cooperation of these technical features enhances the utilization and interpretability of available patient data, thereby improving the accuracy of COPD-related chronic pulmonary heart disease risk prediction.

[0085] In some embodiments of this application, the step of collecting the patient's physiological waveform data preferably includes:

[0086] Acquire raw data from multiple physically dispersed medical data sources with heterogeneous data formats. The raw data includes physiological monitoring data generated by medical monitoring devices and clinical context data from clinical interaction systems. The clinical context data includes at least one of structured electronic medical record data that records patient symptoms, activities, or medication events, and logs of cough and dyspnea attacks reported by patients.

[0087] The physiological monitoring data and the clinical scenario data are respectively converted into first intermediate data and second intermediate data with uniform format;

[0088] Based on the first intermediate data and the second intermediate data, the patient's physiological waveform data is generated.

[0089] Acquiring raw data refers to collecting various types of data related to a patient's COPD and chronic pulmonary heart disease risk from multiple sources, including Hospital Information System (HIS), Electronic Medical Record System (EMR), Laboratory Information System (LIS), Picture Archiving and Communication System (PACS), and various wearable devices or remote monitoring platforms. Physiological monitoring data can be understood as continuous data such as electrocardiogram signals, respiratory waveforms, and blood oxygen saturation generated in real-time or periodically by medical monitoring equipment such as electrocardiographs, pulmonary function analyzers, and pulse oximeters. Clinical contextual data refers to information reflecting a patient's disease state, treatment process, and daily activities. For example, structured electronic medical record data may include the patient's diagnostic records, medication regimens, past medical history, and physical examination results, while patient-reported cough and dyspnea event logs provide detailed time points and intensity information about the patient's subjective feelings and symptom onset.

[0090] The purpose of converting physiological monitoring data and clinical context data into first and second intermediate data with uniform formats is to eliminate data format differences between different data sources. For example, binary data, text data or XML data output from different devices are uniformly converted into standardized JSON or CSV formats, and preprocessing operations such as data cleaning, deduplication and missing value filling are performed to ensure data availability and consistency.

[0091] Based on the first and second intermediate data, the patient's physiological waveform data is generated. Specifically, this involves associating and integrating physiological monitoring data and clinical context data that have been standardized and preprocessed. For example, by aligning with timestamps, clinical events are matched with physiological waveforms, thereby constructing a complete patient physiological waveform dataset containing rich contextual information.

[0092] This application's approach first acquires raw data containing physiological monitoring data and clinical context data, ensuring the comprehensiveness of the data source and reflecting the patient's physiological state and disease progression from multiple dimensions. Subsequently, by converting these heterogeneous data into first and second intermediate data with unified formats, it effectively solves the data integration difficulties caused by incompatible data formats in traditional methods, laying the foundation for subsequent data processing. Finally, physiological waveform data is generated based on these unified-format intermediate data. This results in waveform data that not only contains pure physiological signals but also incorporates key information related to the patient's clinical context, thereby more accurately and comprehensively characterizing the patient's risk of COPD complicated by chronic pulmonary heart disease.

[0093] The following is a specific example to illustrate this.

[0094] Imagine a COPD patient using smart wearable devices at home for daily physiological monitoring. These devices record real-time physiological data such as heart rate, respiratory rate, and blood oxygen saturation, storing it in a specific binary format. Simultaneously, the patient visits the hospital regularly for follow-up appointments, and their diagnostic records, medication information, and pulmonary function test results are stored in the hospital's EMR system as structured electronic medical records. Furthermore, the patient proactively logs their cough and dyspnea episodes via a mobile application; these logs are stored in text format.

[0095] First, the data acquisition module retrieves raw data from the cloud platform of the smart wearable device, the hospital's EMR system, and the mobile application's server. The data from the smart wearable device is identified as physiological monitoring data, while the EMR data and mobile application logs are identified as clinical context data.

[0096] Next, the data processing module converts the format of this raw data. For example, it parses and converts the binary physiological monitoring data from smart wearable devices into a unified CSV format, forming the first intermediate data. Simultaneously, it converts the structured electronic medical record data from the EMR system and the text log data from the mobile application into a unified JSON format, forming the second intermediate data. During this process, data cleaning is also performed, such as removing duplicate records, filling in missing values ​​(e.g., through interpolation), and standardizing timestamps to ensure time alignment across different data sources.

[0097] Finally, based on these uniformly formatted first and second intermediate data, the system generates the patient's physiological waveform data. For example, it timestamps and associates CSV-formatted physiological monitoring data (such as ECG waveforms and respiratory waveforms) with JSON-formatted clinical context data (such as medication events and cough onset times) to construct a comprehensive dataset containing physiological waveforms and their corresponding clinical event annotations. This dataset not only contains continuous physiological waveforms but also clearly indicates which clinical events occurred within specific waveform segments, thus providing more contextually meaningful input for subsequent signal decomposition and feature extraction.

[0098] In a further embodiment of this application, the step of generating the patient's physiological waveform data preferably includes:

[0099] The first intermediate data is denoised and baseline drift corrected to obtain preliminary waveform data, and a first physiological waveform is generated based on the preliminary waveform data.

[0100] Based on the second intermediate data, exogenous physiological disturbance factors are identified and quantified, and corresponding disturbance waveform models are generated.

[0101] The second physiological waveform is generated by subtracting the perturbation component represented by the perturbation waveform model from the first physiological waveform.

[0102] The first physiological waveform and the second physiological waveform are used together as the patient's physiological waveform data for subsequent waveform selection steps.

[0103] The denoising and baseline drift correction of the first intermediate data refers to using signal processing techniques to eliminate random noise and systematic baseline drift in the original physiological monitoring data. Denoising can be understood as removing high-frequency or low-frequency interference unrelated to physiological activity from the signal through methods such as filtering and wavelet transform. Baseline drift correction refers to correcting slow changes in the signal baseline caused by poor electrode contact, changes in body position, etc., using algorithms such as polynomial fitting and moving average. Through these processes, cleaner preliminary waveform data can be obtained, and the first physiological waveform can be generated based on this data. The purpose is to provide a physiological signal view that has undergone basic purification.

[0104] Identifying and quantifying exogenous physiological disturbances based on second intermediate data involves using clinical contextual data (such as patient activity logs, medication records, and symptom feedback) to determine external interference events that may affect physiological waveform data and quantifying their degree of influence. For example, a patient's cough, change of position, eating, or the administration of specific medications can all produce transient or persistent non-pathophysiological changes in the physiological waveform. By analyzing the second intermediate data, the timing and type of these events can be accurately labeled, and corresponding disturbance waveform models can be generated based on pre-set models or empirical knowledge. The purpose is to provide a basis for subsequent removal of these disturbances.

[0105] Subtracting the perturbation component represented by the perturbation waveform model from the first physiological waveform means removing the effects of the identified and quantified exogenous physiological perturbations from the first physiological waveform. For example, if the perturbation waveform model indicates the presence of a respiratory waveform artifact caused by coughing during a certain time period, this artifact is subtracted from the first physiological waveform to generate the second physiological waveform. The second physiological waveform can be understood as a waveform that is closer to the patient's intrinsic physiological state after the removal of known exogenous interferences, with the aim of providing a purer view of physiological signals focused on the progression of the disease itself.

[0106] The above technical solutions significantly improve the purity and accuracy of patient physiological waveform data. Specifically, by denoising and correcting baseline drift in the first intermediate data, random noise and baseline drift in the signal are effectively suppressed, allowing the first physiological waveform to more realistically reflect the patient's physiological state. Furthermore, by combining clinical context data to identify and remove exogenous physiological disturbances, the generated second physiological waveform can eliminate interference from non-disease factors to the greatest extent, thereby more accurately revealing the potential physiological signal characteristics of COPD progression to chronic cor pulmonale. This hierarchical processing and multi-waveform output approach provides more reliable and discriminative input for subsequent signal decomposition and feature extraction, greatly improving the accuracy and clinical value of COPD risk assessment for chronic cor pulmonale complications.

[0107] In a further embodiment of this application, the step of generating explanatory information for risk assessment and presenting the explanatory information through a query interface, wherein the query interface is configured to: in response to a user's query operation for the explanatory information, synchronously display the physiological signal characteristics and the waveform segments from the physiological waveform data from which the physiological signal characteristics are derived, preferably includes:

[0108] Based on the risk assessment and the physiological signal characteristics on which the risk assessment is based, explanatory information is generated to illustrate the contribution of each physiological signal characteristic to the risk assessment.

[0109] The explanatory information is presented through the interactive query interface;

[0110] The query interface is configured to perform the following operations in response to a user's query for the explanatory information:

[0111] The analysis conclusions and data directly corresponding to the queried physiological signal features in the explanatory information are displayed;

[0112] The system displays the first or second physiological waveform derived from the queried physiological signal characteristics.

[0113] Highlight the waveform segments used to calculate the physiological signal characteristics being queried on the first or second physiological waveform displayed.

[0114] Specifically, generating explanatory information to illustrate the contribution of each physiological signal feature to the risk assessment refers to the system's further analysis of the role and weight of each physiological signal feature in forming the risk assessment after completing the risk assessment for COPD complicated by chronic pulmonary heart disease. For example, interpretable artificial intelligence (XAI) techniques, such as SHAP (SHapley Additive exPlanations) values ​​or LIME (Local Interpretable Model-agnostic Exlanations) methods, can be used to quantify the positive or negative contribution of each physiological signal feature to the final risk score, thereby generating easily understandable explanatory text or charts.

[0115] Presenting the explanatory information through a query interface means displaying the generated explanatory information to the user in a visual manner. For example, the interface can display each physiological signal feature and its corresponding contribution in the form of a list, bar chart, or pie chart, allowing users to clearly understand which features have the greatest impact on risk assessment.

[0116] The query interface is configured to perform the following operations in response to a user's query for the explanatory information. Specifically, when a user queries the explanatory information of a certain physiological signal feature presented on the interface (e.g., by clicking, hovering, or selecting), the system immediately executes a series of linked display operations. First, it displays the analysis conclusions and data directly corresponding to the queried physiological signal feature in the explanatory information. This means that the interface displays the detailed value of the specific physiological signal feature, its normal range, and the preliminary analysis and judgment based on the feature value (e.g., "significantly reduced heart rate variability, suggesting autonomic nervous system dysfunction"). Second, it displays the first or second physiological waveform derived from the queried physiological signal feature. That is, the system automatically jumps to or synchronously displays the original physiological waveform data from which the physiological signal feature originates on the current interface. This can be the unprocessed first physiological waveform or the second physiological waveform after denoising and perturbation correction, depending on the source of the feature extraction. Finally, the waveform segment used to calculate the queried physiological signal feature is highlighted on the first or second physiological waveform displayed. This means that on the linked physiological waveform diagram, the system will use eye-catching colors or marks to highlight which time period or local waveform the physiological signal feature was calculated or extracted from, so that users can intuitively see the original data basis of the feature.

[0117] This application's solution provides a highly interconnected and visualized interpretation mechanism, enabling users to directly trace from abstract risk assessment results and feature contribution levels to specific original physiological waveform data and their key segments. This layered presentation, moving from result to cause and from abstract to concrete, greatly enhances the transparency and interpretability of the risk assessment process. By displaying the analytical conclusions and data of the queried feature, users can obtain quantitative information and preliminary interpretations of that feature; by displaying the original physiological waveform in conjunction with the data, users can intuitively see the context of the feature; and by highlighting waveform segments, users can accurately locate the source of the feature, thereby gaining a more comprehensive understanding of the reliability of the risk assessment.

[0118] Through the aforementioned technical solution, this application effectively addresses the limitations of traditional risk prediction methods in interpreting information presentation, significantly enhancing users' understanding and trust in the risk assessment results for COPD-related chronic pulmonary heart disease. This intuitive and traceable interpretation mechanism enables clinicians to more confidently adopt prediction results and make comprehensive judgments in conjunction with patients' actual physiological data, thereby optimizing the clinical decision-making process and improving the accuracy and efficiency of patient management.

[0119] The following is a specific example to illustrate this.

[0120] Suppose the system assesses a patient's risk of COPD complicated by chronic cor pulmonale as "high risk," and the explanatory information indicates that "abnormal relative change between the inspiratory and expiratory phases of the respiratory cycle" is the main physiological signal characteristic leading to this high risk. When a clinician clicks or selects this explanatory information on the query interface, the system immediately displays the specific numerical value of this characteristic (e.g., relative change of X%) and provides an analytical conclusion (e.g., "This change significantly exceeds the normal range, suggesting abnormal pulmonary function oscillations"). Simultaneously, the query interface displays the patient's pulmonary function oscillation waveform (i.e., the first or second physiological waveform), highlighting the specific waveform segments of the inspiratory and expiratory phases on which the calculation of the "relative change between the inspiratory and expiratory phases of the respiratory cycle" is based. In this way, the doctor not only understands the abnormality of this characteristic but also intuitively sees the specific abnormal manifestations in the original waveform, thus gaining a clear and intuitive understanding of the physiological basis of the risk assessment results.

[0121] In a preferred embodiment of this application, the query interface is further provided with a waveform switching control, which is used to respond to the user's switching command and switch between the first physiological waveform and the second physiological waveform.

[0122] The waveform switching control can be understood as a user interface element, such as a button, drop-down menu, checkbox, or slider, configured to receive user input commands. The function of this control is to allow the query interface to instantly switch between displaying a first physiological waveform and a second physiological waveform when the user issues a switching command. The first physiological waveform typically refers to the raw or pre-processed physiological waveform data without correction for perturbation factors, while the second physiological waveform is obtained by subtracting exogenous physiological perturbation factors from the first physiological waveform. Through this switching function, users can easily compare the two waveforms to gain a more comprehensive understanding of the source of physiological signal characteristics and their performance at different processing stages. This design allows users to quickly compare physiological waveform data from different processing stages without leaving the current analysis interface or performing complex operations, thereby more intuitively assessing the effectiveness and reliability of physiological signal characteristics and gaining a deeper understanding of the basis for risk assessment.

[0123] In some embodiments of this application, the step of performing signal decomposition processing on physiological waveform data to obtain multiple signal components for revealing the time-frequency domain detailed features of the physiological waveform data preferably includes:

[0124] The type of the physiological waveform data is obtained, including pulmonary function oscillation waveforms and electrocardiogram waveforms;

[0125] For the physiological waveform data of type pulmonary function oscillation waveform, the continuous wavelet transform algorithm is used to decompose the physiological waveform data to obtain the wavelet coefficients of the physiological waveform data under different scales and time parameters, and the wavelet coefficients are used as multiple signal components to reveal the time-frequency domain detailed features of the physiological waveform data;

[0126] For the physiological waveform data of type ECG waveform, the empirical mode decomposition algorithm is used to decompose the physiological waveform data to obtain the intrinsic mode functions of the physiological waveform data under different scales and time parameters, and the intrinsic mode functions are used as the signal components.

[0127] The type of physiological waveform data acquired refers to the predefined type of the physiological waveform data to be processed, which is automatically identified by the system or specified by the user, such as pulmonary function oscillation waveform or electrocardiogram (ECG) waveform. This type identification is the basis for subsequently selecting an appropriate decomposition algorithm. Pulmonary function oscillation waveforms typically refer to the waveforms of respiratory airflow or volume changes acquired through pulmonary function testing equipment; they are periodic but may exhibit non-stationarity. ECG waveforms refer to the waveforms of cardiac electrical activity acquired through electrocardiogram (ECG) equipment, containing typical structures such as P waves, QRS complexes, and T waves, and may be subject to various noise interferences.

[0128] For physiological waveform data of the type of pulmonary function oscillation waveform, the continuous wavelet transform (CWT) algorithm is used to decompose it. Continuous wavelet transform (CWT) is a time-frequency analysis method that can decompose a signal into wavelet coefficients at different scales (frequency) and time positions. These wavelet coefficients can effectively reveal the local features of the signal at different frequency components and their changes over time, and are particularly suitable for analyzing non-stationary, transient, or multi-scale physiological signals, such as the high-frequency oscillation components in respiratory airflow. Through CWT, wavelet coefficients of physiological waveform data at different scales and time parameters can be obtained, and these wavelet coefficients are used as signal components to reveal the detailed time-frequency domain features of the physiological waveform data.

[0129] For physiological waveform data of the type electrocardiogram (ECG) waveform, Empirical Mode Decomposition (EMD) is used to decompose it. EMD is an adaptive time-frequency analysis method that can decompose complex nonlinear, non-stationary signals into a series of physically meaningful intrinsic mode functions (IMFs). Each IMF represents the oscillation mode of the signal at different time scales and has good locality and adaptability, requiring no pre-defined basis functions. EMD is particularly suitable for processing complex rhythm changes, arrhythmia events, and various noise components in ECG signals. Through EMD, the IMFs of physiological waveform data at different scales and time parameters can be obtained. These IMFs are used as signal components to reveal the detailed time-frequency domain features of the physiological waveform data.

[0130] Through the above technical solution, this application can significantly improve the accuracy and effectiveness of physiological waveform data signal decomposition. Because it employs the most suitable signal decomposition algorithms (such as continuous wavelet transform and empirical mode decomposition) for different types of physiological waveform data (such as pulmonary function oscillation waveforms and electrocardiogram waveforms), it can capture the time-frequency domain details contained in various waveforms more precisely and comprehensively. This not only avoids feature loss or noise interference that may be caused by a single general-purpose algorithm, but also provides higher-quality input for subsequent extraction of physiological signal features indicating the progression of COPD to chronic cor pulmonale, thereby ultimately improving the accuracy and reliability of COPD-associated chronic cor pulmonale risk assessment.

[0131] In a specific embodiment of this application, the step of extracting physiological signal features indicating the progression of COPD to chronic cor pulmonale preferably includes:

[0132] For the wavelet coefficients, the wavelet energy spectral density of the physiological waveform data of the type of pulmonary function oscillation waveform in the preset frequency band is calculated, and the relative change of the wavelet energy spectral density between the inspiratory and expiratory phases of the respiratory cycle is extracted to obtain the physiological signal features that indicate the progression of COPD to chronic cor pulmonale.

[0133] For the intrinsic mode function, the duration, peak amplitude, and slope of the P wave and T wave in the physiological waveform data of the type of electrocardiogram waveform are extracted and compared with the pre-stored patient baseline data to obtain the physiological signal characteristics.

[0134] Wavelet coefficients refer to a series of coefficients obtained at different scales and time parameters after decomposing the pulmonary function oscillation waveform using a continuous wavelet transform algorithm. These coefficients reflect the local characteristics of the signal in the time-frequency domain. The pulmonary function oscillation waveform can be understood as waveform data reflecting the mechanical characteristics of the patient's respiratory system, such as respiratory impedance waveforms or flow-volume loop waveforms. The wavelet energy spectral density of the preset frequency band refers to the energy distribution represented by the wavelet coefficients within a specific frequency range. This frequency band is usually set according to the clinical characteristics of respiratory mechanics changes in COPD patients, such as 5-20Hz. Extracting the relative change of the wavelet energy spectral density between the inspiratory and expiratory phases of the respiratory cycle aims to quantify the dynamic changes in parameters such as airway resistance and lung compliance during respiration. These changes are important indicators of COPD progression.

[0135] Intrinsic mode functions (IMFs) refer to a series of oscillating components with good locality and adaptability obtained after decomposing an electrocardiogram (ECG) waveform using an empirical mode decomposition algorithm. Each component represents the inherent oscillation mode of the signal at different time scales. ECG waveforms refer to the patient's cardiac electrical activity data acquired through ECG monitoring equipment. The duration, peak amplitude, and slope of the P wave and T wave are key parameters in an ECG, reflecting the electrical activity of the atria and ventricles, respectively. An excessively long P wave duration or abnormal amplitude may indicate increased atrial load, while changes in T wave morphology and amplitude may be related to abnormal ventricular repolarization or ischemia. Comparing these parameters with pre-stored patient baseline data aims to identify abnormal changes in the patient's individual physiological state, rather than relying solely on absolute values, thereby improving the specificity and sensitivity of the diagnosis. Pre-stored patient baseline data typically includes historical physiological signal characteristics of the patient in a healthy or stable state, used to establish individualized reference standards.

[0136] Through the aforementioned technical solution, this application can extract more specific and sensitive physiological signal features from decomposed physiological waveform data, thereby significantly improving the accuracy of predicting the risk of COPD complicated by chronic cor pulmonale. Specifically, by quantifying the relative changes in wavelet energy spectral density of pulmonary function oscillation waveforms within the respiratory cycle, the dynamic evolution of respiratory mechanics parameters can be assessed more precisely, providing crucial evidence for the early detection of COPD progression. Simultaneously, by conducting detailed analysis of the P and T waves in electrocardiogram waveforms and comparing them with individual baseline data, early cardiac dysfunction associated with chronic cor pulmonale can be detected in a timely manner, avoiding missed diagnoses or misdiagnoses that may result from relying solely on general indicators. This comprehensive analysis combining respiratory and cardiac physiological characteristics makes risk assessment more comprehensive and reliable, helping clinicians to intervene early and improve patient prognosis.

[0137] The following is a specific example to illustrate this.

[0138] Suppose that after continuous wavelet transform, the pulmonary function oscillation waveform of a COPD patient shows a sustained increasing trend in the relative change of wavelet energy spectral density between the expiratory and inspiratory phases within a preset frequency range of 5-20 Hz. This may indicate a progressive increase in airway resistance. Simultaneously, after empirical mode decomposition, the patient's electrocardiogram (ECG) waveform shows a 15% increase in P wave duration compared to pre-stored baseline data, and an increase in T wave peak amplitude in the right precordial leads. The combination of these physiological signal characteristics—deterioration of respiratory mechanics parameters and signs of increased right ventricular load on the ECG—will be extracted and used as evidence of an increased risk of COPD progression to chronic cor pulmonale. In this way, this application provides a multi-dimensional, high-precision risk prediction model.

[0139] In some embodiments of this application, the step of generating a risk assessment for COPD complicated with chronic pulmonary heart disease based on physiological signal characteristics preferably includes:

[0140] The currently extracted physiological signal features, as well as the historical physiological signal features of the same patient, are obtained to form time-series data;

[0141] The time series data is analyzed within a preset time window to obtain a dynamic risk assessment threshold.

[0142] By comparing the values ​​of the currently extracted physiological signal features at continuous time points with the dynamic risk determination threshold, a continuous comparison result is obtained;

[0143] Based on the continuous comparison results, determine the changing trend of the currently extracted physiological signal features;

[0144] Based on the aforementioned trends, a risk assessment for COPD complicated by chronic pulmonary heart disease is generated.

[0145] The process of acquiring currently extracted physiological signal features, along with historical physiological signal features from the same patient, to form time-series data means that the system not only focuses on the patient's current physiological signal features but also integrates all relevant physiological signal feature data recorded from the patient's past periods, arranging this data in chronological order to form a continuous time-series dataset. The purpose is to provide comprehensive historical background information for subsequent dynamic analysis.

[0146] The system analyzes the time-series data within a preset time window to obtain a dynamic risk assessment threshold. This can be understood as the system statistically analyzing the time-series data based on the patient's historical data over a specific time span (e.g., the past week, month, or longer) to calculate a personalized, time-varying risk assessment threshold. This threshold better adapts to the patient's own physiological fluctuations and disease progression patterns, rather than using a uniform, fixed threshold.

[0147] The values ​​of the currently extracted physiological signal features at consecutive time points are compared with the dynamic risk assessment threshold to obtain a continuous comparison result. For example, this could mean that the system continuously monitors the patient's physiological signal features and compares them with the dynamically updated risk assessment threshold in real time. This comparison is performed continuously to capture any physiological changes that exceed the normal fluctuation range.

[0148] Based on continuous comparison results, determining the changing trend of the currently extracted physiological signal features means that the system analyzes whether the physiological signal features show an upward, downward, stable, or increasingly volatile trend based on continuous comparison results. For example, if the physiological signal features continuously exceed the upper limit of the dynamic risk assessment threshold for multiple consecutive time points, it may indicate a deteriorating trend. Therefore, based on the changing trend, a risk assessment of COPD complicated by chronic pulmonary heart disease is generated. The purpose is to generate a more accurate and prospective risk assessment of COPD complicated by chronic pulmonary heart disease based on the dynamic changing trend of physiological signal features, rather than just the absolute value at a single time point. For example, a continuously deteriorating trend may lead to a higher risk assessment result.

[0149] Through the aforementioned technical solution, this application enables dynamic and personalized assessment of the risk of COPD complicated by chronic cor pulmonale. Compared to traditional methods that rely solely on data from a single time point, this application significantly improves the accuracy and foresight of risk assessment by integrating the patient's historical physiological signal characteristics and analyzing their changing trends. The introduction of this dynamic risk threshold makes the assessment results more adaptable to individual patient physiological fluctuations, reducing the possibility of false positives and false negatives. Furthermore, by judging the changing trends of physiological signal characteristics, doctors and patients can identify early warning signs of disease progression more quickly, providing a valuable window for timely intervention and adjustment of treatment plans, effectively delaying or preventing the progression of COPD to chronic cor pulmonale.

[0150] In a specific embodiment of this application, the step of analyzing the time-series data within a preset time window to obtain a dynamic risk assessment threshold preferably includes:

[0151] Calculate the mean and standard deviation of the time series data within a preset time window;

[0152] The dynamic risk assessment threshold is obtained by adding or subtracting the product of the preset multiple and the standard deviation from the average value.

[0153] The time-series data refers to time-series data composed of currently extracted physiological signal features and historical physiological signal features of the same patient, reflecting changes in the patient's physiological state over time. The preset time window can be understood as a continuous time range used to analyze the time-series data, such as data from the past 7 days, 30 days, or longer. The length of this time window can be flexibly configured according to the disease progression characteristics, data collection frequency, and clinical needs. Calculating the average value of the time-series data within the preset time window involves taking the arithmetic mean of the values ​​of all physiological signal features within that time window to reflect the central trend of the physiological signal features during that period. The standard deviation is used to quantify the dispersion or volatility of the physiological signal feature values ​​within the time window; a larger standard deviation indicates greater data volatility. The preset multiplier is an adjustable parameter whose value can be optimized based on clinical experience, statistical analysis, or machine learning models. For example, it can be set to 1, 1.5, 2, or 3, etc., to control the sensitivity of the dynamic risk assessment threshold. Adding or subtracting the product of the average value and the standard deviation from the preset multiple means that the dynamic risk assessment threshold is dynamically adjusted based on the patient's own historical data distribution, which can better adapt to individual differences and physiological fluctuations. For example, the product of the average value plus the preset multiple and the standard deviation can be used as the upper risk threshold, and the product of the average value minus the preset multiple and the standard deviation can be used as the lower risk threshold, or only one of them can be used as a one-way risk threshold.

[0154] The dynamic risk assessment threshold generated through the above technical solution can better adapt to the physiological fluctuations and dynamic progression of individual patients. Compared with methods using fixed thresholds or simple averages as thresholds, this solution can provide more personalized and robust risk assessment. Specifically, using the mean and standard deviation allows the threshold to be adjusted according to the patient's own historical data distribution, thereby improving the accuracy and sensitivity of risk assessment, especially in identifying early and subtle physiological changes, which has a significant advantage and helps to achieve early warning and intervention for COPD complicated by chronic pulmonary heart disease.

[0155] like Figure 2 As shown, this application also discloses a COPD complication chronic pulmonary heart disease risk prediction system 200, comprising:

[0156] The data acquisition module 210 is used to collect physiological waveform data of patients from multiple physically dispersed medical data sources with heterogeneous data formats.

[0157] Data processing module 220 is used to perform signal decomposition processing on the physiological waveform data to obtain multiple signal components for revealing the time-frequency domain detailed features of the physiological waveform data;

[0158] The feature extraction module 230 is used to extract physiological signal features that indicate the progression of COPD to chronic cor pulmonale;

[0159] The feature conversion module 240 is used to generate a risk assessment of COPD complicated with chronic pulmonary heart disease based on the physiological signal features.

[0160] The information interaction module 250 is used to generate explanatory information for the risk assessment and present the explanatory information through a query interaction interface. The query interaction interface is configured to: respond to the user's query operation on the explanatory information, synchronously display the physiological signal characteristics and the waveform segments from the physiological waveform data from which the physiological signal characteristics are derived.

[0161] The data acquisition module 210 can be configured to connect to and collect data from heterogeneous data sources in various ways. For example, this module can integrate multiple data interfaces, such as standardized medical data exchange protocols (e.g., HL7, DICOM), for acquiring data from hospital information systems (HIS), image archiving and communication systems (PACS), etc. For non-standard format data sources, the data acquisition module 210 can include a customized parser to identify and extract the required raw physiological waveform data. For example, a corresponding parsing submodule can be developed for the data format of a specific model of pulmonary function diagnostic workstation or home monitoring device. In addition, the data acquisition module 210 can also support a manual import function, allowing users to upload locally stored physiological waveform data files to meet data input needs in specific scenarios.

[0162] The data processing module 220 can be configured to include a library of various signal processing algorithms to adapt to different types of physiological waveform data. For example, for electrocardiogram (ECG) waveform data, the data processing module 220 can call algorithms such as Fourier transform, wavelet transform, or empirical mode decomposition (EMD) for processing. For pulmonary function oscillation waveform data, the data processing module 220 can employ time-frequency analysis methods such as short-time Fourier transform (STFT) or continuous wavelet transform (CWT) to obtain the energy distribution of the signal at different times and frequencies. The data processing module 220 can also automatically or manually select a suitable decomposition algorithm according to preset rules or user selection.

[0163] The feature extraction module 230 can be configured to include a series of predefined feature extraction algorithms and rules. For example, for the spectral information of electrocardiogram signals, the feature extraction module 230 can calculate heart rate variability (HRV) indices, such as SDNN and RMSSD. For the time-frequency domain details of pulmonary function oscillation waveforms, the feature extraction module 230 can calculate parameters such as respiratory rate, tidal volume, peak expiratory flow rate, and energy distribution changes in specific frequency bands. The feature extraction module 230 can also support custom feature extraction rules, allowing clinical experts to add or modify feature extraction logic based on new research findings or specific needs.

[0164] The feature transformation module 240 can be configured to integrate multiple machine learning models or statistical models. For example, the feature transformation module 240 can use classification algorithms such as logistic regression, support vector machines (SVM), decision trees, random forests, or neural networks, taking extracted physiological signal features as input, to predict the risk level of COPD patients developing chronic pulmonary heart disease. The feature transformation module 240 can also support ensemble learning methods, combining the prediction results of multiple models to improve the robustness and accuracy of the assessment. In addition, the feature transformation module 240 can provide model training and update functions, allowing the system to optimize the model based on new clinical data.

[0165] The information interaction module 250 can be configured to use interpretable artificial intelligence (XAI) technology to generate explanatory information, such as using methods like LIME or SHAP to analyze the contribution of each physiological signal feature to the risk assessment results. The query interface can be designed to include a risk assessment result display area, an explanatory information display area, a list of physiological signal features, and a waveform display area. This interface supports multiple interaction methods; for example, clicking, hovering, or selecting a physiological signal feature will automatically highlight that feature in the physiological signal feature list and simultaneously load and highlight the corresponding original physiological waveform data segment in the waveform display area. The information interaction module 250 can also provide a report generation function, exporting the risk assessment results and explanatory information into a printable report format.

[0166] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0167] The foregoing has provided a detailed description of the preferred embodiments of this application. However, this application is not limited to the above-described embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined in this application.

Claims

1. A method for predicting the risk of COPD complicated with chronic cor pulmonale, characterized in that, Includes the following steps: Physiological waveform data of patients are collected from multiple medical data sources that are physically dispersed and have heterogeneous data formats. The physiological waveform data is subjected to signal decomposition processing to obtain multiple signal components used to reveal the time-frequency domain detailed features of the physiological waveform data; From the signal components, physiological signal features indicating the progression of COPD to chronic cor pulmonale were extracted; Based on the aforementioned physiological signal characteristics, a risk assessment for COPD complicated by chronic pulmonary heart disease is generated. The risk assessment explanation information is generated and presented through a query interface. The query interface is configured to: respond to the user's query operation on the explanation information, synchronously display the physiological signal characteristics and the waveform segments from the physiological waveform data from which the physiological signal characteristics are derived.

2. The method for predicting the risk of COPD complicated with chronic cor pulmonale according to claim 1, characterized in that, The steps for collecting patients' physiological waveform data based on multiple physically dispersed and heterogeneous medical data sources include: Acquire raw data from multiple physically dispersed medical data sources with heterogeneous data formats. The raw data includes physiological monitoring data generated by medical monitoring devices and clinical context data from clinical interaction systems. The clinical context data includes at least one of structured electronic medical record data that records patient symptoms, activities, or medication events, and logs of cough and dyspnea attacks reported by patients. The physiological monitoring data and the clinical scenario data are respectively converted into first intermediate data and second intermediate data with uniform format; Based on the first intermediate data and the second intermediate data, the patient's physiological waveform data is generated.

3. The method for predicting the risk of COPD complicated with chronic cor pulmonale according to claim 2, characterized in that, The step of generating the patient's physiological waveform data based on the first intermediate data and the second intermediate data includes: The first intermediate data is denoised and baseline drift corrected to obtain preliminary waveform data, and a first physiological waveform is generated based on the preliminary waveform data. Based on the second intermediate data, exogenous physiological disturbance factors are identified and quantified, and corresponding disturbance waveform models are generated. The second physiological waveform is generated by subtracting the perturbation component represented by the perturbation waveform model from the first physiological waveform. The first physiological waveform and the second physiological waveform are used together as the patient's physiological waveform data for subsequent waveform selection steps.

4. The method for predicting the risk of COPD complicated with chronic cor pulmonale according to claim 3, characterized in that, The steps of generating explanatory information for the risk assessment and presenting the explanatory information through a query interface, wherein the query interface is configured to: respond to a user's query operation for the explanatory information and simultaneously display the physiological signal characteristics, and the waveform segments from the physiological waveform data from which the physiological signal characteristics are derived, include: Based on the risk assessment and the physiological signal characteristics on which the risk assessment is based, explanatory information is generated to illustrate the contribution of each physiological signal characteristic to the risk assessment. The explanatory information is presented through the interactive query interface; The query interface is configured to perform the following operations in response to a user's query for the explanatory information: The analysis conclusions and data directly corresponding to the queried physiological signal features in the explanatory information are displayed; The system displays the first or second physiological waveform derived from the queried physiological signal characteristics. Highlight the waveform segments used to calculate the physiological signal characteristics being queried on the first or second physiological waveform displayed.

5. The method for predicting the risk of COPD complicated with chronic cor pulmonale according to claim 4, characterized in that, The query interface also includes a waveform switching control, which is used to respond to the user's switching command and switch between the first physiological waveform and the second physiological waveform.

6. The method for predicting the risk of COPD complicated with chronic cor pulmonale according to claim 1, characterized in that, The step of performing signal decomposition processing on the physiological waveform data to obtain multiple signal components used to reveal the time-frequency domain detailed features of the physiological waveform data includes: The type of the physiological waveform data is obtained, including pulmonary function oscillation waveforms and electrocardiogram waveforms; For the physiological waveform data of type pulmonary function oscillation waveform, the continuous wavelet transform algorithm is used to decompose the physiological waveform data to obtain the wavelet coefficients of the physiological waveform data under different scales and time parameters, and the wavelet coefficients are used as multiple signal components to reveal the time-frequency domain detailed features of the physiological waveform data; For the physiological waveform data of type ECG waveform, the empirical mode decomposition algorithm is used to decompose the physiological waveform data to obtain the intrinsic mode functions of the physiological waveform data under different scales and time parameters, and the intrinsic mode functions are used as the signal components.

7. The method for predicting the risk of COPD complicated with chronic cor pulmonale according to claim 6, characterized in that, The step of extracting physiological signal features indicating the progression of COPD to chronic cor pulmonale from the signal components includes: For the wavelet coefficients, the wavelet energy spectral density of the physiological waveform data of the type of pulmonary function oscillation waveform in the preset frequency band is calculated, and the relative change of the wavelet energy spectral density between the inspiratory and expiratory phases of the respiratory cycle is extracted to obtain the physiological signal features that indicate the progression of COPD to chronic cor pulmonale. For the intrinsic mode function, the duration, peak amplitude, and slope of the P wave and T wave in the physiological waveform data of the type of electrocardiogram waveform are extracted and compared with the pre-stored patient baseline data to obtain the physiological signal characteristics.

8. The method for predicting the risk of COPD complicated with chronic cor pulmonale according to claim 1, characterized in that, The steps for generating a risk assessment for COPD complicated by chronic pulmonary heart disease based on the aforementioned physiological signal characteristics include: The currently extracted physiological signal features, as well as the historical physiological signal features of the same patient, are obtained to form time-series data; The time series data is analyzed within a preset time window to obtain a dynamic risk assessment threshold. By comparing the values ​​of the currently extracted physiological signal features at continuous time points with the dynamic risk determination threshold, a continuous comparison result is obtained; Based on the continuous comparison results, determine the changing trend of the currently extracted physiological signal features; Based on the aforementioned trends, a risk assessment for COPD complicated by chronic pulmonary heart disease is generated.

9. The method for predicting the risk of COPD complicated with chronic cor pulmonale according to claim 8, characterized in that, The step of analyzing the time-series data within a preset time window to obtain the dynamic risk assessment threshold includes: Calculate the mean and standard deviation of the time series data within a preset time window; The dynamic risk assessment threshold is obtained by adding or subtracting the product of the preset multiple and the standard deviation from the average value.

10. A risk prediction system for COPD complicated with chronic cor pulmonale, characterized in that, include: The data acquisition module is used to collect patients' physiological waveform data from multiple physically dispersed medical data sources with heterogeneous data formats. The data processing module is used to perform signal decomposition processing on the physiological waveform data to obtain multiple signal components that reveal the time-frequency domain details of the physiological waveform data; The feature extraction module is used to extract physiological signal features that indicate the progression of COPD to chronic cor pulmonale. The feature conversion module is used to generate a risk assessment for COPD complicated by chronic pulmonary heart disease based on the physiological signal features. The information interaction module is used to generate explanatory information for the risk assessment and present the explanatory information through a query interaction interface. The query interaction interface is configured to: respond to the user's query operation on the explanatory information, synchronously display the physiological signal characteristics and the waveform segments from the physiological waveform data from which the physiological signal characteristics are derived.

Images