A medical data fusion method, system, device and storage medium

Abstract

The application discloses a medical data fusion method, system, device and storage medium, and relates to the technical field of data processing.In the method, the behavior data and physiological index data of a patient are acquired; time alignment processing is performed to generate a first time sequence and a second time sequence respectively; the numerical fluctuation of the first time sequence and the second time sequence is calculated respectively, and the corresponding confidence weight is determined based on the numerical fluctuation; the first time sequence is time-shifted according to each time offset respectively, and the correlation coefficient between the time-shifted first time sequence and the second time sequence is calculated; the best time offset corresponding to the highest correlation coefficient is determined, and the first time sequence shifted according to the best time offset is taken as a target time sequence; and the target time sequence and the second time sequence are weighted and merged according to the corresponding confidence weight to obtain the fusion medical data of the patient.The application improves the accuracy of medical decision.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a medical data fusion method, system, device, and storage medium. Background Technology

[0002] With the rapid development of medical informatization, the sources of medical data are constantly expanding. In addition to physiological indicator data recorded by traditional medical equipment, wearable devices and smart medical terminals can also continuously collect patient behavioral data. The combined analysis of these multi-source data provides a new approach to comprehensively understand patients' health status.

[0003] Currently, doctors typically need to comprehensively analyze the relationship between patients' behavioral data and physiological indicators. For example, for patients with hypertension, blood pressure fluctuations may be influenced by various behavioral factors such as exercise and sleep. A common data processing method is to perform simple statistical calculations on data from different sources over a fixed time period, such as calculating the average or median.

[0004] However, this simple statistical processing method is insufficient to handle the complexities of actual clinical situations. Physiological data are typically sampled continuously at a high frequency, while behavioral data are often discrete event records, and there may be unknown time lags between the two, affecting the accuracy of medical decisions. Summary of the Invention

[0005] This application provides a medical data fusion method, system, device, and storage medium, which can improve the accuracy of medical decision-making.

[0006] The first aspect of this application provides a medical data fusion method, specifically including:

[0007] Acquire patient behavioral and physiological data;

[0008] Based on a preset first time window, the behavioral data and the physiological indicator data are time-aligned to generate a first time series and a second time series, respectively.

[0009] Calculate the numerical volatility of the first time series and the second time series within a preset period, and determine the confidence weights of the first time series and the second time series based on the numerical volatility.

[0010] The first time series is shifted according to each time offset within a preset search range, and the correlation coefficient between the shifted first time series and the second time series is calculated. The preset search range includes multiple discrete time offsets.

[0011] Determine the optimal time offset corresponding to the highest correlation coefficient, and use the first time series shifted according to the optimal time offset as the target time series.

[0012] The target time series and the second time series are weighted and merged according to their corresponding confidence weights to obtain the patient's fused medical data.

[0013] The above technical solution first involves an electronic device acquiring patient behavioral and physiological data, and then performing time alignment processing based on a preset first time window to generate a first time series and a second time series, achieving preliminary unification of data from different sources. By calculating the numerical volatility of the two time series within a preset period and determining the corresponding reliability weights, the problem of reliability differences between different data sources is effectively solved. Based on this, by traversing different time offsets within a preset search range, the correlation coefficient between the shifted first and second time series is calculated, and the optimal time offset is determined. Finally, the first time series is shifted based on the optimal time offset to obtain the target time series, which is then weighted and merged with the second time series according to their respective reliability weights. This data fusion method based on time alignment and reliability weighting can effectively handle the complex relationship between high-frequency continuously sampled physiological indicator data and discrete event-recorded behavioral data, improving the accuracy of fused medical data.

[0014] Optionally, the step of calculating the numerical volatility of the first time series and the second time series within a preset period, and determining the confidence weights of the first time series and the second time series based on each of the numerical volatility values, specifically includes:

[0015] A first rate of change sequence corresponding to the first time series and a second rate of change sequence corresponding to the second time series are generated based on the rate of change of data of adjacent sampling points in the first time series and the second time series. The first rate of change sequence includes multiple time periods corresponding to the first time series, and the second rate of change sequence includes multiple time periods corresponding to the second time series.

[0016] The ratio of the mean to the standard deviation of the first rate of change sequence is calculated as a first rate of change threshold, and the time period in the first rate of change sequence that exceeds the first rate of change threshold is marked as a first high rate of change segment. The ratio of the mean to the standard deviation of the second rate of change sequence is calculated as a second rate of change threshold, and the time period in the second rate of change sequence that exceeds the second rate of change threshold is marked as a second high rate of change segment.

[0017] The sum of the rate of change of the data corresponding to each of the first high rate of change segments is calculated to obtain the first numerical volatility of the first time series; the sum of the rate of change of the data corresponding to each of the second high rate of change segments is calculated to obtain the second numerical volatility of the second time series.

[0018] Based on the first numerical volatility and the second numerical volatility, the confidence weights corresponding to the first time series and the second time series are determined.

[0019] The above technical solution generates corresponding rate of change sequences based on the rate of change of data from adjacent sampling points of the first and second time series, accurately capturing the dynamic change characteristics of data over time. By calculating the ratio of the mean to the standard deviation of each rate of change sequence as a rate of change threshold, and using this to identify high rate of change segments, the problem of considering only the absolute value of data and failing to reflect the trend of change is effectively solved, achieving accurate identification of abnormal data fluctuation intervals. Furthermore, by calculating the sum of the rate of change corresponding to each high rate of change segment, numerical volatility is obtained, and the reliability weight of the time series is determined accordingly, establishing a correspondence between data dynamic characteristics and reliability. This weight determination method based on rate of change analysis can effectively distinguish between normal and abnormal data fluctuations, improving the accuracy of weight allocation in subsequent data fusion processes.

[0020] Optionally, determining the confidence weights corresponding to the first time series and the second time series based on the first numerical volatility and the second numerical volatility specifically includes:

[0021] The first numerical volatility and the second numerical volatility are normalized respectively.

[0022] Determine the maximum value of the first and second numerical variability after normalization, and take the logarithm of the maximum value as the benchmark attenuation factor;

[0023] Using the benchmark decay factor as the base and the normalized first numerical volatility as the exponent, the calculated exponential function value is used as the confidence weight of the first time series. Using the normalized second numerical volatility as the exponent, the calculated exponential function value is used as the confidence weight of the second time series.

[0024] The above technical solution first normalizes the first and second numerical variability, achieving a unified representation of data at different scales. By determining the maximum value after normalization and taking its logarithm as the benchmark attenuation factor, a nonlinear mapping relationship between numerical variability and confidence weights is established. Based on this, using the benchmark attenuation factor as the base and the normalized numerical variability as the exponent, an exponential function value is calculated as the confidence weight, ensuring that data with greater variability receives a smaller weight. This exponential function-based weight calculation method can reasonably reflect the impact of data variability on confidence.

[0025] Optionally, calculating the correlation coefficient between the shifted first time series and the second time series specifically includes:

[0026] The translated first time series and the second time series are divided into multiple consecutive second time windows according to a preset time length;

[0027] Within each of the second time windows, the Pearson correlation coefficient and Spearman correlation coefficient between the shifted first time series and the second time series are calculated respectively.

[0028] The correlation coefficient is obtained by weighted summing of the Pearson correlation coefficient and the Spearman correlation coefficient within each second time window.

[0029] By adopting the above technical solution, firstly, the segmented calculation of correlation is realized by dividing the second time window, which can capture the correlation characteristics within a local time period. Secondly, both Pearson correlation coefficient and Spearman correlation coefficient are used simultaneously to measure both linear correlation and monotonic correlation. Finally, a comprehensive correlation coefficient is obtained by weighted summation, making the correlation assessment more comprehensive and reliable.

[0030] Optionally, the step of weighting and merging the target time series and the second time series according to corresponding confidence weights to obtain the patient's fused medical data specifically includes:

[0031] Based on the highest correlation coefficient and each of the aforementioned confidence weights, a normalized fusion factor is constructed corresponding to the target time series and the second time series.

[0032] The target time series and the second time series are weighted and summed according to the corresponding normalized fusion factors to obtain the initial fusion sequence;

[0033] The initial fusion sequence is processed by an exponential moving average to obtain the patient's fusion medical data.

[0034] By adopting the above technical solution, a normalized fusion factor is first constructed by combining the correlation coefficient and the confidence weight, so that the fusion process can simultaneously consider the correlation and reliability of the data. Secondly, the initial fusion sequence is obtained by weighted summation, which ensures that the contribution of different time series to the fusion result matches its importance. Finally, the fusion result is smoothed by exponential moving average processing, which effectively suppresses the impact of short-term fluctuations on the fused medical data.

[0035] Optionally, the step of constructing the normalized fusion factor corresponding to the target time series and the second time series based on the highest correlation coefficient and each of the confidence weights specifically includes:

[0036] Multiply the highest correlation coefficient by a preset first weighting factor to obtain the relationship strength component;

[0037] The confidence weights corresponding to the first time series and the second time series are multiplied by a preset second weight factor to obtain the corresponding intrinsic confidence components.

[0038] The relationship strength component is added to each of the intrinsic credibility components to obtain the initial fusion factor corresponding to the target time series and the second time series;

[0039] The sum of all initial fusion factors is used as the normalization base, and each initial fusion factor is divided by the normalization base to obtain the normalized fusion factors corresponding to the target time series and the second time series.

[0040] By adopting the above technical solution, the relationship strength component is obtained by first adjusting the influence of the correlation coefficient through the first weight factor, and the intrinsic credibility component is obtained by adjusting the influence of the credibility weight through the second weight factor. Then, the two components are added together to obtain the initial fusion factor. Finally, normalization is performed to ensure that the sum of the fusion factors is 1. This makes the fusion process consider both the correlation strength between time series and the reliability of each series itself, thus achieving a more reasonable data fusion.

[0041] Optionally, the step of performing time alignment processing on the behavioral data and the physiological indicator data based on a preset first time window to generate a first time series and a second time series respectively includes:

[0042] Establish a unified timeline within the preset first time window;

[0043] The behavioral data is filtered according to the time interval determined by a preset nonlinear function to generate corresponding values ​​at each time point on the unified time axis, thereby obtaining a first time series aligned with the unified time axis.

[0044] The physiological index data are calculated using piecewise polynomial interpolation to generate corresponding values ​​at each time point on the unified time axis, resulting in a second time series aligned with the unified time axis.

[0045] By adopting the above technical solution, firstly, a unified time axis is established to provide a benchmark for data alignment. Secondly, the behavioral data is filtered using time intervals determined by nonlinear functions, which can better preserve the temporal information of behavioral characteristics. Finally, the physiological index data is resampled using a piecewise polynomial interpolation method, which not only ensures the continuity of the data but also avoids overfitting, thereby achieving high-quality time alignment of heterogeneous data.

[0046] A second aspect of this application provides a medical data fusion system, comprising:

[0047] The data acquisition module is used to acquire patients' behavioral and physiological data.

[0048] The time alignment module is used to perform time alignment processing on the behavioral data and the physiological indicator data based on a preset first time window, and generate a first time series and a second time series respectively.

[0049] The credibility weight determination module is used to calculate the numerical volatility of the first time series and the second time series within a preset period, and determine the credibility weight of the first time series and the second time series based on the numerical volatility.

[0050] The correlation analysis module is used to perform time shifting on the first time series according to each time offset within a preset search range, and to calculate the correlation coefficient between the shifted first time series and the second time series. The preset search range includes multiple discrete time offsets.

[0051] The optimal offset correction module is used to determine the optimal time offset corresponding to the highest correlation coefficient, and to take the first time series after being shifted according to the optimal time offset as the target time series.

[0052] The data fusion module is used to perform weighted merging of the target time series and the second time series according to the corresponding confidence weights to obtain the patient's fused medical data.

[0053] The above technical solution first involves an electronic device acquiring patient behavioral and physiological data, and then performing time alignment processing based on a preset first time window to generate a first time series and a second time series, achieving preliminary unification of data from different sources. By calculating the numerical volatility of the two time series within a preset period and determining the corresponding reliability weights, the problem of reliability differences between different data sources is effectively solved. Based on this, by traversing different time offsets within a preset search range, the correlation coefficient between the shifted first and second time series is calculated, and the optimal time offset is determined. Finally, the first time series is shifted based on the optimal time offset to obtain the target time series, which is then weighted and merged with the second time series according to their respective reliability weights. This data fusion method based on time alignment and reliability weighting can effectively handle the complex relationship between high-frequency continuously sampled physiological indicator data and discrete event-recorded behavioral data, improving the accuracy of fused medical data.

[0054] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any of the foregoing.

[0055] A fourth aspect of this application provides a computer-readable storage medium storing instructions that, when executed, perform the method described in any of the preceding descriptions. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the architecture of a medical data fusion system disclosed in an embodiment of this application;

[0057] Figure 2 This is a flowchart illustrating a medical data fusion method disclosed in an embodiment of this application;

[0058] Figure 3 yes Figure 2 A flowchart illustrating a sub-step of step S102;

[0059] Figure 4 This is a flowchart illustrating the steps of a numerical volatility calculation method.

[0060] Figure 5 yes Figure 4 A flowchart illustrating a sub-step of step S304;

[0061] Figure 6 This is a flowchart illustrating the steps involved in determining a correlation coefficient.

[0062] Figure 7 yes Figure 2 A flowchart illustrating a sub-step of step S106;

[0063] Figure 8 yes Figure 7 A flowchart illustrating a sub-step of step S601;

[0064] Figure 9 This is a schematic diagram of a medical data fusion system provided in an embodiment of this application;

[0065] Figure 10 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application.

[0066] Figure labeling: 21, Data acquisition module; 22, Time alignment module; 23, Credibility weight determination module; 24, Correlation analysis module; 25, Optimal offset correction module; 26, Data fusion module; 901, Processor; 902, Communication bus; 903, User interface; 904, Network interface; 905, Memory. Detailed Implementation

[0067] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification 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.

[0068] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0069] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0070] Figure 1An exemplary system architecture for a medical data fusion system is shown.

[0071] like Figure 1 As shown, the system architecture may include electronic device 11, network 12, and medical acquisition instrument 13. Network 12 serves as the medium for providing a communication link between electronic device 11 and medical acquisition instrument 13. Network 12 may include various connection types, such as wired, wireless communication links, or fiber optic cables.

[0072] Users can use electronic device 11 to interact with medical acquisition instrument 13 via network 12 to initiate data acquisition and transmit the acquired raw data to electronic device 11. Electronic device 11 may be equipped with a medical data fusion processing application that performs fusion processing of behavioral and physiological indicator data.

[0073] Electronic device 11 is hardware and can be various electronic devices with a display screen, including but not limited to smartphones, tablets, laptops, and desktop computers, which integrate a processing unit that performs data fusion methods.

[0074] The medical data acquisition instrument 13 can be one or more hardware devices used to acquire patient behavioral data and physiological indicator data, such as wearable sensors, multi-parameter monitors, etc. It is responsible for the real-time acquisition of raw data and provides this data to the electronic device 11 for analysis, processing and fusion.

[0075] The following detailed explanation uses the electronic device side as an example.

[0076] This embodiment discloses a medical data fusion method. Figure 2 This is a flowchart illustrating a medical data fusion method disclosed in an embodiment of this application, as shown below. Figure 2 As shown, the method includes steps S101 to S106:

[0077] S101: Obtain patient behavioral and physiological data.

[0078] In this embodiment of the application, behavioral data refers to a set of data obtained by continuously monitoring the patient's movement status, activity intensity, and positional changes through various medical acquisition instruments. For example, the number of steps, walking speed, and posture changes recorded by wearable devices, or information such as body movement frequency and lying-up status collected by smart hospital beds. In order to comprehensively assess the patient's health status, it is also necessary to obtain their physiological index data, which are used to represent various quantitative measurement results reflecting the patient's vital signs and bodily functions, such as real-time measurement values ​​of physiological parameters such as heart rate, blood pressure, body temperature, and blood oxygen.

[0079] Specifically, the electronic device establishes a data connection with the patient's associated medical data acquisition instruments, continuously receiving raw data collected by these devices. During data acquisition, the sampling frequency for behavioral data is adaptively adjusted according to the patient's activity level, including three preset levels: resting state sampling frequency, mild activity sampling frequency, and vigorous exercise sampling frequency. Physiological indicator data is collected using a preset fixed sampling frequency, with different physiological indicators having different fixed sampling periods. After receiving this heterogeneous data, the electronic device first verifies the data's validity, including checking data integrity, timestamp continuity, and the reasonableness of numerical ranges. Then, the verified data is categorized and stored according to patient identification and time information, preparing for subsequent data processing. The entire data acquisition process is real-time and continuous, ensuring timely monitoring and analysis of the patient's condition.

[0080] For example, electronic devices establish data connections with smart bracelets, smartphones, and bedside monitors worn by patients via Bluetooth, WiFi, or 5G networks. For behavioral data sampling, a low-frequency sampling rate of once per minute is used when the patient is detected as resting; this increases to a medium-frequency rate of 1-2 times per second when the patient gets up or walks; and a high-frequency rate of 5-10 times per second is used during strenuous exercise. For physiological indicator data sampling, heart rate is sampled at a fixed frequency of once per second, blood pressure at a fixed frequency of once every 15 minutes, and body temperature at a fixed frequency of once every 30 minutes. When verifying the received data, the electronic device checks whether the heart rate is within the reasonable range of 40-200 beats per minute, whether the blood pressure data includes both systolic and diastolic pressure values, and whether there are duplicate or discontinuous timestamps.

[0081] S102: Based on the preset first time window, perform time alignment processing on behavioral data and physiological indicator data to generate a first time series and a second time series respectively.

[0082] In this embodiment of the application, the first time series corresponds to the ordered arrangement of behavioral data on the time axis, and the second time series corresponds to the ordered arrangement of physiological indicator data on the time axis.

[0083] Specifically, the electronic device first sets the length of a preset first time window as the baseline unit for time alignment, and then processes behavioral data and physiological indicator data separately. For behavioral data with variable sampling frequencies, the electronic device aggregates multiple sampling points within each time window to generate the first time series data points corresponding to that window. For physiological indicator data with fixed sampling frequencies, the electronic device remaps the original sampling points to the time scale of the time window to generate the second time series data points. In this way, the two types of heterogeneous data achieve a unified discretized representation in the time dimension, laying the foundation for subsequent data analysis.

[0084] For example, electronic devices are set to a preset first time window of 1 minute. For behavioral data, if multiple sampling points exist within a certain time window, such as acceleration data collected 5 times per second during a patient's strenuous exercise, these data points are aggregated into a representative value within that minute. For physiological indicator data, the heart rate sampling data per second is resampled into 1 point based on 60 points per minute, and the blood pressure data collected every 15 minutes is interpolated to be expanded into one point per minute. The resulting first and second time series have the same time interval, and each series corresponds to one data point within the 1-minute time window.

[0085] Reference Figure 3 , Figure 3 This is provided by the embodiments of this application. Figure 2 A flowchart illustrating a sub-step of step S102, including steps S201 to S203, is as follows:

[0086] S201: Establish a unified timeline within the scope of the preset first time window.

[0087] In this embodiment of the application, a unified timeline refers to a standardized time reference coordinate system established within a preset first time window, used to uniformly represent the time point positions of different data sources.

[0088] Specifically, the electronic device first determines the length of a preset first time window, using it as the basic unit of measurement for constructing a unified time axis. Then, using the preset first time window as a step size, the electronic device creates a uniformly distributed sequence of time points throughout the entire data processing interval, with the interval between each time point strictly equal to the length of the preset first time window. This method constructs a unified time axis with a fixed time scale, providing a basic framework for subsequently mapping data from different sources and with different sampling frequencies to the same time reference system.

[0089] For example, if an electronic device sets a preset first time window of 1 minute, for a 24-hour data processing interval, a unified timeline containing 1440 evenly distributed time points will be created, with each time point spaced 1 minute apart. If data processing begins at 00:00 AM, the time points on the unified timeline will be 00:00, 00:01, 00:02, and so on until 23:59, forming a complete daily time scale sequence.

[0090] S202: The behavioral data is filtered according to the time interval determined by the preset nonlinear function to generate corresponding values ​​at each time point on the unified time axis, thereby obtaining a first time series aligned with the unified time axis.

[0091] In this embodiment, the preset nonlinear function refers to the data filtering rules set according to the characteristics of behavioral data, which are used to determine the selection method and weight allocation of data at different time points.

[0092] Specifically, the electronic device first uses a unified timeline as a reference system, and then calculates the time interval range for each time point according to a preset nonlinear function. For each time point, the electronic device filters behavioral data within this time interval range, generating the corresponding value for that time point by performing a weighted average or feature extraction on the filtered data. Due to the characteristics of the nonlinear function, the time interval range may differ at different time points, allowing the data filtering process to better adapt to the dynamic characteristics of the behavioral data. Finally, the electronic device arranges the generated values ​​in chronological order according to the unified timeline, forming the first time series.

[0093] For example, an electronic device uses a unified timeline as a reference system and a Gaussian function as a preset nonlinear function. Based on the nonlinear function, the electronic device calculates the time interval range at a given time point t to be within 30 seconds before and after it. Then, the electronic device filters behavioral data within this time interval range and determines the weight of each data point based on its distance from time point t (the closer the distance, the greater the weight), generating the value corresponding to time point t through a weighted average. The electronic device repeats this process for each point on the unified timeline, ultimately arranging the generated values ​​in chronological order to form the first time series.

[0094] S203: Physiological index data are calculated using piecewise polynomial interpolation to generate corresponding values ​​at each time point on a unified time axis, thereby obtaining a second time series aligned with the unified time axis.

[0095] In the embodiments of this application, piecewise polynomial interpolation refers to a method of interpolating physiological index data using different polynomial functions in different time intervals to generate continuous and smooth values ​​between the original sampling points.

[0096] Specifically, the electronic device first arranges the physiological indicator data in chronological order to determine the time intervals between adjacent sampling points. For each time interval, the electronic device selects a polynomial function of appropriate order based on the data characteristics within and outside the interval, and calculates the coefficients of the polynomial function. Then, the electronic device uses the obtained piecewise polynomial function to calculate the interpolation result for each time point on a unified time axis. During the interpolation calculation, the electronic device must ensure the continuity of function values ​​and derivatives between adjacent intervals to guarantee the smoothness of the generated data. Finally, the electronic device arranges the interpolated values ​​in chronological order according to the unified time axis to form a second time series.

[0097] For example, for blood pressure data sampled every 15 minutes, the electronic device interpolates between adjacent sampling points using a cubic polynomial function. If the blood pressure measurement at a certain moment is 120 mmHg, and the measurement at the next sampling point is 125 mmHg, the electronic device calculates the cubic polynomial function for those 15 minutes and uses this function to generate the corresponding blood pressure value for each one-minute time point on a unified time axis. The generated second time series not only achieves alignment with the unified time axis but also maintains the continuity and smoothness of the physiological indicator data.

[0098] S103: Calculate the numerical volatility of the first time series and the second time series within a preset period, and determine the confidence weights of the first time series and the second time series based on the numerical volatility.

[0099] In the embodiments of this application, numerical volatility refers to the degree of change of the values ​​of a time series within a preset period, reflecting the stability and reliability of the data; confidence weight refers to the weight coefficient assigned to the time series based on numerical volatility, used to represent the confidence level of the data in subsequent analysis.

[0100] Specifically, the electronic device first sets a preset period as the time range for calculating numerical volatility. For the first and second time series, it calculates the statistical characteristics of the values ​​within each preset period, including indicators such as mean, standard deviation, and range. The combination of these statistical indicators is used to represent numerical volatility. Subsequently, the electronic device converts the calculated numerical volatility into corresponding confidence weights according to a pre-set mapping rule. Generally, the smaller the numerical volatility, the more stable the data, and the greater the corresponding confidence weight; the greater the numerical volatility, the more volatile the data, and the smaller the corresponding confidence weight.

[0101] For example, if an electronic device is set to a preset period of 30 minutes, for the patient's first time series, the standard deviation of the data within these 30 minutes is calculated as σ1, and the mean is μ1. The numerical volatility can then be expressed as σ1 / μ1. Similarly, for the patient's second time series, the standard deviation σ2 and the mean μ2 are calculated, and the numerical volatility is expressed as σ2 / μ2. If the numerical volatility of the first time series is 0.2 and the numerical volatility of the second time series is 0.1, then the first time series is assigned a confidence weight of 0.6, and the second time series is assigned a confidence weight of 0.8.

[0102] Reference Figure 4 , Figure 4 This application provides a flowchart illustrating the steps of a numerical volatility calculation method, including steps S301 to S304, as follows:

[0103] S301: Generate a first rate of change sequence corresponding to the first time series and a second rate of change sequence corresponding to the second time series based on the rate of change of data of adjacent sampling points in the first time series and the second rate of change sequence. The first rate of change sequence includes multiple time periods corresponding to the first time series, and the second rate of change sequence includes multiple time periods corresponding to the second time series.

[0104] In this embodiment of the application, the rate of change sequence refers to a sequence composed of the data change rates in each time period in chronological order.

[0105] Specifically, for the first and second time series, the electronic device sequentially takes two adjacent sampling points as the start and end points of a time period. It calculates the difference between the end point value and the start point value to obtain the numerical difference for that time period. Then, it divides this numerical difference by the time interval of that period to obtain the rate of change of the data within that time period. After repeating the above calculation for all time periods, the calculated rates of change of the data for each time period of the first and second time series are organized chronologically to obtain a first rate of change sequence corresponding to the first time series and a second rate of change sequence corresponding to the second time series. The first rate of change sequence reflects the dynamic change characteristics of the data in the first time series, while the second rate of change sequence describes the fluctuation trend of the data in the second time series.

[0106] S302: Calculate the ratio of the mean to the standard deviation of the first rate of change sequence as the first rate of change threshold, and mark the time period in the first rate of change sequence that exceeds the first rate of change threshold as the first high rate of change segment. Calculate the ratio of the mean to the standard deviation of the second rate of change sequence as the second rate of change threshold, and mark the time period in the second rate of change sequence that exceeds the second rate of change threshold as the second high rate of change segment.

[0107] Specifically, the electronic device processes the first rate of change sequence and the second rate of change sequence separately. First, it calculates the arithmetic mean of all rate of change values ​​in both sequences to obtain the mean. Then, it calculates the square root of the average of the sum of squares of deviations of each rate of change value from the mean to obtain the standard deviation. Dividing the mean by the standard deviation yields the corresponding rate of change thresholds, i.e., the first rate of change threshold for the first rate of change sequence and the second rate of change threshold for the second rate of change sequence. Next, it compares the rate of change value for each time period in both sequences with the corresponding rate of change threshold. When the rate of change value in the first rate of change sequence is greater than the first rate of change threshold, that time period is marked as the first high rate of change segment, indicating that the first time series changes drastically within that time period. When the rate of change value in the second rate of change sequence is greater than the second rate of change threshold, that time period is marked as the second high rate of change segment, indicating that the second time series changes significantly within that time period.

[0108] S303: Calculate the sum of the data change rates corresponding to each of the first high rate of change segments to obtain the first numerical volatility of the first time series; calculate the sum of the data change rates corresponding to each of the second high rate of change segments to obtain the second numerical volatility of the second time series.

[0109] Specifically, the electronic device processes the high-rate-of-change segments of the first and second time series respectively. For the first time series, all time periods marked as the first high-rate-of-change segment are identified, the corresponding data rate-of-change values ​​are extracted, and these rate-of-change values ​​are summed to obtain the first numerical volatility, which is used to characterize the overall volatility of the first time series. For the second time series, all time periods marked as the second high-rate-of-change segment are identified, the corresponding data rate-of-change values ​​are extracted, and these rate-of-change values ​​are summed to obtain the second numerical volatility, which is used to characterize the overall volatility of the second time series.

[0110] S304: Determine the confidence weights corresponding to the first time series and the second time series based on the first numerical volatility and the second numerical volatility.

[0111] Specifically, the electronic device matches the first and second numerical fluctuations with a preset confidence weight table. In the confidence weight table, different ranges of numerical fluctuation correspond to different confidence weight values. The electronic device searches for the corresponding weight value in the confidence weight table based on the first numerical fluctuation to determine the confidence weight of the first time series, and searches for the corresponding weight value in the confidence weight table based on the second numerical fluctuation to determine the confidence weight of the second time series.

[0112] Reference Figure 5 , Figure 5 yes Figure 4 A flowchart illustrating a sub-step of step S304 includes steps S401 to S403, as follows:

[0113] S401: Normalize the first numerical volatility and the second numerical volatility respectively.

[0114] Specifically, the electronic device normalizes the first and second numerical volatility respectively. It obtains the first maximum and first minimum values ​​corresponding to the volatility values ​​of the first time series, subtracts the first minimum value from the first numerical volatility, and divides the result by the difference between the first maximum and the first minimum value to obtain the first normalized numerical volatility. Similarly, it obtains the second maximum and second minimum values ​​corresponding to the volatility values ​​of the second time series, subtracts the second minimum value from the second numerical volatility, and divides the result by the difference between the second maximum and the second minimum value to obtain the second normalized numerical volatility.

[0115] S402: Determine the maximum value of the first and second numerical volatility after normalization, and take the logarithm of the maximum value as the benchmark attenuation factor.

[0116] Specifically, the electronic device compares the fluctuations of the first and second normalized values ​​after normalization and selects the maximum value. Then, it performs a logarithmic operation on this maximum value and uses the calculated logarithmic value as the benchmark attenuation factor.

[0117] S403: Using the benchmark decay factor as the base, and the normalized first numerical volatility as the exponent, the calculated exponential function value is used as the confidence weight of the first time series. Using the normalized second numerical volatility as the exponent, the calculated exponential function value is used as the confidence weight of the second time series.

[0118] Specifically, the electronic device calculates the confidence weights for the two time series separately. For the first time series, an exponential function value is calculated using the benchmark decay factor as the base and the first normalized numerical volatility as the exponent, and the calculation result is used as the confidence weight of the first time series; for the second time series, the electronic device calculates an exponential function value using the benchmark decay factor as the base and the second normalized numerical volatility as the exponent, and the calculation result is used as the confidence weight of the second time series.

[0119] S104: Shift the first time series according to each time offset within the preset search range, and calculate the correlation coefficient between the shifted first time series and the second time series. The preset search range includes multiple discrete time offsets.

[0120] Specifically, the electronic device first sets a preset search range containing multiple discrete time offsets. For each time offset within the preset search range, the electronic device shifts the first time series according to that time offset, resulting in multiple shifted first time series. Then, the electronic device calculates the correlation coefficient between each shifted first time series and the second time series using the Pearson correlation coefficient formula. The Pearson correlation coefficient formula is: the covariance of the data at corresponding times of the two series divided by the product of the standard deviations of the two series.

[0121] For example, the preset search range includes time offsets of -2 minutes, -1 minute, 0 minutes, 1 minute, and 2 minutes. When the time offset is 1 minute, the electronic device shifts the first time series forward by 1 minute, and the correlation coefficient between the shifted first time series and the second time series is calculated to be 0.82 using the Pearson correlation coefficient formula; when the time offset is 2 minutes, the electronic device shifts the first time series forward by 2 minutes, and the correlation coefficient is calculated to be 0.75.

[0122] Reference Figure 6 , Figure 6 This application provides a flowchart illustrating a method for determining correlation coefficients, including steps S501 to S503, as follows:

[0123] S501: Divide the shifted first time series and second time series into multiple consecutive second time windows according to a preset time length.

[0124] Specifically, the electronic device sets a preset time length as the basis for dividing the second time window. For each time offset corresponding to the shifted first and second time series, the electronic device divides the sequence sequentially from the start time according to the preset time length, resulting in multiple consecutive and non-overlapping second time windows. Each second time window contains the same time length.

[0125] For example, an electronic device sets a preset time length of 10 minutes and divides the first and second time series after shifting with a time offset of 1 minute. If the total duration of the two series is 60 minutes, then six second time windows can be divided, containing data for 0-10 minutes, 10-20 minutes, 20-30 minutes, 30-40 minutes, 40-50 minutes, and 50-60 minutes respectively.

[0126] S502: Within each second time window, calculate the Pearson correlation coefficient and Spearman correlation coefficient between the shifted first time series and the second time series, respectively.

[0127] Specifically, within each second time window, the electronic device calculates the correlation between the shifted first and second time series using both the Pearson correlation coefficient and Spearman correlation coefficient formulas. The Pearson correlation coefficient is obtained by dividing the covariance of the corresponding time data of the two series by the product of the standard deviations of the two series; the Spearman correlation coefficient is obtained by calculating the Pearson correlation coefficient between the sorted ranks of the corresponding time data of the two series.

[0128] For example, within a second time window of 0-10 minutes, the electronic device calculates a Pearson correlation coefficient of 0.85 and a Spearman correlation coefficient of 0.82 between the shifted first and second time series; within a second time window of 10-20 minutes, the calculated Pearson correlation coefficient is 0.78 and the Spearman correlation coefficient is 0.75. In this way, both correlation coefficients are obtained for each second time window.

[0129] S503: The correlation coefficient is obtained by weighted summing of the Pearson correlation coefficient and the Spearman correlation coefficient within each second time window.

[0130] Specifically, the electronic device assigns weighting coefficients to the Pearson correlation coefficient and Spearman correlation coefficient calculated within each second time window. Then, the electronic device multiplies the Pearson correlation coefficient and the Spearman correlation coefficient within each second time window by their respective weighting coefficients, and sums all the weighted results to obtain the correlation coefficient.

[0131] For example, in an electronic device, the weight of the Pearson correlation coefficient is set to 0.6, and the weight of the Spearman correlation coefficient is set to 0.4. Within the second time window of 0-10 minutes, the Pearson correlation coefficient (0.85) multiplied by the weight 0.6 yields 0.51, and the Spearman correlation coefficient (0.82) multiplied by the weight 0.4 yields 0.328; adding them together gives 0.838. Within the second time window of 10-20 minutes, the same calculation yields 0.766. Finally, the results from all the second time windows are summed to obtain the final correlation coefficient.

[0132] S105: Determine the optimal time offset corresponding to the highest correlation coefficient, and use the first time series after shifting according to the optimal time offset as the target time series.

[0133] Specifically, after obtaining the correlation coefficients corresponding to each time offset, the electronic device compares the magnitudes of the correlation coefficients to determine the highest correlation coefficient. The electronic device then determines the time offset corresponding to the highest correlation coefficient as the optimal time offset and uses the first time series shifted at the optimal time offset as the target time series.

[0134] S106: The target time series and the second time series are weighted and merged according to their corresponding confidence weights to obtain the patient's fused medical data.

[0135] Specifically, the electronic device uses the confidence weights of the target time series and the second time series, which have been determined through volatility analysis. It multiplies the data point at each moment in the target time series by its corresponding confidence weight, and the corresponding data point in the second time series by its corresponding confidence weight. The weighted data are then merged to obtain the patient's fused medical data for that moment. This process is repeated for all moments to obtain the complete fused medical data sequence.

[0136] Reference Figure 7 , Figure 7 This is provided by the embodiments of this application. Figure 2 A flowchart illustrating a sub-step of step S106, including steps S601 to S603, is as follows:

[0137] S601: Based on the highest correlation coefficient and each confidence weight, construct the normalized fusion factor corresponding to the target time series and the second time series.

[0138] Specifically, the electronic device multiplies the highest correlation coefficient by the confidence weight of the target time series to obtain the fusion factor of the target time series, and multiplies the highest correlation coefficient by the confidence weight of the second time series to obtain the fusion factor of the second time series. Then, the electronic device normalizes the two fusion factors by dividing each by the sum of the fusion factors to obtain the normalized fusion factors corresponding to the target time series and the second time series.

[0139] For example, the highest correlation coefficient is 0.85, the confidence weight of the target time series is 0.7, and the confidence weight of the second time series is 0.3. The fusion factor of the target time series is 0.85 × 0.7 = 0.595, and the fusion factor of the second time series is 0.85 × 0.3 = 0.255. Dividing each fusion factor by the sum of 0.85, we obtain a normalized fusion factor of 0.7 for the target time series and 0.3 for the second time series.

[0140] Reference Figure 8 , Figure 8 This is provided by the embodiments of this application. Figure 7 A flowchart illustrating a sub-step of step S601, including steps S701 to S704, is as follows:

[0141] S701: Multiply the highest correlation coefficient by the preset first weighting factor to obtain the relationship strength component.

[0142] In the embodiments of this application, the relationship strength component refers to a quantitative indicator that characterizes the degree of correlation between the target time series and the second time series.

[0143] Specifically, the electronic device multiplies a preset first weighting factor with the highest correlation coefficient obtained through time alignment to calculate the relationship strength component.

[0144] For example, if the highest correlation coefficient is 0.85 and the preset first weighting factor is 0.6, then the relationship strength component is 0.85 × 0.6 = 0.51. This relationship strength component reflects the correlation strength between the two time series.

[0145] S702: Multiply the confidence weights corresponding to the first time series and the second time series by the preset second weight factor to obtain the corresponding intrinsic confidence components.

[0146] In the embodiments of this application, the intrinsic reliability component refers to a quantitative indicator that characterizes the reliability of the time series data itself.

[0147] Specifically, the electronic device multiplies the preset second weighting factor by the credibility weights of the first and second time series respectively to calculate the intrinsic credibility components corresponding to the two time series.

[0148] For example, if the preset second weighting factor is 0.4, the confidence weight of the first time series is 0.7, and the confidence weight of the second time series is 0.3, then the intrinsic confidence component of the first time series is 0.4 × 0.7 = 0.28, and the intrinsic confidence component of the second time series is 0.4 × 0.3 = 0.12.

[0149] S703: Add the relation strength component to each intrinsic credibility component to obtain the initial fusion factor corresponding to the target time series and the second time series.

[0150] Specifically, the electronic device adds the relationship strength component to the intrinsic credibility components corresponding to the target time series and the second time series, respectively, to obtain the initial fusion factor for the two time series. The initial fusion factor reflects the overall credibility of the time series.

[0151] For example, if the relation strength component is 0.51, the intrinsic credibility component of the target time series is 0.28, and the intrinsic credibility component of the second time series is 0.12, then the initial fusion factor of the target time series is 0.51 + 0.28 = 0.79, and the initial fusion factor of the second time series is 0.51 + 0.12 = 0.63. These initial fusion factors reflect the overall credibility of the time series.

[0152] S704: Use the sum of all initial fusion factors as the normalization base, and divide each initial fusion factor by the normalization base to obtain the normalized fusion factors corresponding to the target time series and the second time series.

[0153] Specifically, the electronic device adds the initial fusion factors of the target time series and the second time series to obtain a normalized base, and then divides each initial fusion factor by the normalized base to obtain the corresponding normalized fusion factor.

[0154] For example, the initial fusion factor of the target time series is 0.79, and the initial fusion factor of the second time series is 0.63. The normalized base is 0.79 + 0.63 = 1.42. The normalized fusion factor of the target time series is 0.79 ÷ 1.42 ≈ 0.556, and the normalized fusion factor of the second time series is 0.63 ÷ 1.42 ≈ 0.444. The sum of the two normalized fusion factors is 1.

[0155] S602: The target time series and the second time series are weighted and summed according to the corresponding normalized fusion factors to obtain the initial fusion sequence.

[0156] Specifically, the electronic device multiplies each data point of the target time series by its corresponding normalized fusion factor, multiplies each data point of the second time series by its corresponding normalized fusion factor, and then adds the weighted results at the corresponding positions to obtain the initial fusion sequence.

[0157] For example, if the data point of the target time series at a certain moment is 100, and the corresponding data point of the second time series is 80, the normalized fusion factor of the target time series is 0.556, and the normalized fusion factor of the second time series is 0.444, then the data point of the initial fused sequence at that moment is: 100 × 0.556 + 80 × 0.444 = 55.6 + 35.52 = 91.12. By calculating the data point for each moment in this way, the complete initial fused sequence can be obtained.

[0158] S603: Perform exponential moving average processing on the initial fusion sequence to obtain the patient's fusion medical data.

[0159] Specifically, the electronic device processes the initial fused sequence using an exponential moving average algorithm. For each time t in the initial fused sequence, its smoothed value is equal to the sum of the initial value at the current time multiplied by the smoothing coefficient and the complement of the smoothed value at the previous time multiplied by the smoothing coefficient.

[0160] For example, if the electronic device sets a smoothing coefficient of 0.3, the initial fused sequence value at the current moment is 91.12, and the smoothed value at the previous moment is 88.50, then the smoothed value at the current moment equals 91.12 multiplied by 0.3 plus 88.50 multiplied by 0.7, i.e., 91.12 × 0.3 + 88.50 × 0.7 = 27.336 + 61.95 = 89.286. The electronic device obtains the final fused medical data sequence by performing this calculation on each data point in the initial fused sequence. This processing method can effectively smooth short-term fluctuations while preserving the long-term trend of the data.

[0161] This embodiment also discloses a medical data fusion system. Figure 9 This is a schematic diagram of a medical data fusion system disclosed in an embodiment of this application. The system includes:

[0162] Data acquisition module 21 is used to acquire patients' behavioral data and physiological indicator data;

[0163] The time alignment module 22 is used to perform time alignment processing on the behavioral data and the physiological indicator data based on a preset first time window, and generate a first time series and a second time series respectively.

[0164] The credibility weight determination module 23 is used to calculate the numerical volatility of the first time series and the second time series within a preset period, and determine the credibility weight of the first time series and the second time series based on the numerical volatility.

[0165] The correlation analysis module 24 is used to perform time shifting on the first time series according to each time offset within a preset search range, and calculate the correlation coefficient between the shifted first time series and the second time series. The preset search range includes multiple discrete time offsets.

[0166] The optimal offset correction module 25 determines the optimal time offset corresponding to the highest correlation coefficient and uses the first time series after being shifted according to the optimal time offset as the target time series.

[0167] The data fusion module 26 is used to perform weighted merging of the target time series and the second time series according to the corresponding confidence weights to obtain the patient's fused medical data.

[0168] Optionally, the confidence weight determination module 23 is further configured to generate a first rate of change sequence corresponding to the first time series and a second rate of change sequence corresponding to the second time series based on the rate of change of data of adjacent sampling points in the first time series and the second time series. The first rate of change sequence includes multiple time periods corresponding to the first time series, and the second rate of change sequence includes multiple time periods corresponding to the second time series. The module is further configured to: calculate the ratio of the mean to the standard deviation of the first rate of change sequence as a first rate of change threshold; mark time periods in the first rate of change sequence that exceed the first rate of change threshold as first high rate of change segments; calculate the ratio of the mean to the standard deviation of the second rate of change sequence as a second rate of change threshold; mark time periods in the second rate of change sequence that exceed the second rate of change threshold as second high rate of change segments; calculate the sum of the data rates of change corresponding to each of the first high rate of change segments to obtain a first numerical volatility of the first time series; calculate the sum of the data rates of change corresponding to each of the second high rate of change segments to obtain a second numerical volatility of the second time series; and determine the confidence weights corresponding to the first time series and the second time series based on the first numerical volatility and the second numerical volatility.

[0169] Optionally, the confidence weight determination module 23 is further configured to normalize the first numerical volatility and the second numerical volatility respectively; determine the maximum value of the first numerical volatility and the second numerical volatility after normalization, and take the logarithm of the maximum value as the benchmark attenuation factor; use the benchmark attenuation factor as the base, and use the first numerical volatility after normalization as the exponent to calculate the exponential function value as the confidence weight of the first time series, and use the second numerical volatility after normalization as the exponent to calculate the exponential function value as the confidence weight of the second time series.

[0170] Optionally, the correlation analysis module 24 is further configured to divide the shifted first time series and the second time series into multiple consecutive second time windows according to a preset time length; within each second time window, calculate the Pearson correlation coefficient and the Spearman correlation coefficient between the shifted first time series and the second time series respectively; and perform a weighted summation of the Pearson correlation coefficient and the Spearman correlation coefficient within each second time window to obtain the correlation coefficient.

[0171] Optionally, the data fusion module 26 is further configured to construct normalized fusion factors corresponding to the target time series and the second time series based on the highest correlation coefficient and each of the confidence weights; to perform weighted summation of the target time series and the second time series according to the corresponding normalized fusion factors to obtain an initial fusion sequence; and to perform exponential moving average processing on the initial fusion sequence to obtain the fused medical data of the patient.

[0172] Optionally, the data fusion module 26 is further configured to multiply the highest correlation coefficient by a preset first weighting factor to obtain a relationship strength component; multiply the confidence weights corresponding to the first time series and the second time series by a preset second weighting factor to obtain corresponding intrinsic confidence components; add the relationship strength component to each of the intrinsic confidence components to obtain initial fusion factors corresponding to the target time series and the second time series; use the sum of all initial fusion factors as a normalization base, and divide each of the initial fusion factors by the normalization base to obtain normalized fusion factors corresponding to the target time series and the second time series.

[0173] Optionally, the time alignment module 22 is further configured to establish a unified time axis within a preset first time window; filter the behavioral data according to a time interval determined by a preset nonlinear function to generate corresponding values ​​at each time point of the unified time axis, thereby obtaining a first time series aligned with the unified time axis; and calculate the physiological indicator data using piecewise polynomial interpolation to generate corresponding values ​​at each time point of the unified time axis, thereby obtaining a second time series aligned with the unified time axis.

[0174] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0175] This embodiment also discloses an electronic device, as shown in the reference. Figure 10 The electronic device may include: at least one processor 901, at least one communication bus 902, user interface 903, network interface 904, and at least one memory 905.

[0176] The communication bus 902 is used to enable communication between these components.

[0177] The user interface 903 may include a display screen and a camera. Optional user interfaces may also include standard wired interfaces and wireless interfaces.

[0178] The network interface 904 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0179] The processor 901 may include one or more processing cores. The processor connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory, and by calling data stored in memory. Optionally, the processor may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor.

[0180] The memory 905 may include random access memory (RAM) or read-only memory. Optionally, the memory may include a non-transitory computer-readable storage medium. The memory can be used to store instructions, programs, code, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor. As shown in the figure, the memory, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a medical data fusion method.

[0181] exist Figure 10In the electronic device shown, the user interface is mainly used to provide an input interface for the user and to obtain the user input data; while the processor can be used to call an application program that stores a medical data fusion method in the memory. When executed by one or more processors, the electronic device performs one or more methods as described in the above embodiments.

[0182] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0183] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0184] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some service interfaces; indirect couplings or communication connections between apparatuses or units may be electrical or other forms.

[0185] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0186] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0187] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0188] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Other embodiments of this disclosure will be readily apparent to those skilled in the art upon consideration of the disclosure in this specification. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A medical data fusion method, characterized in that, The method includes: Acquire patient behavioral and physiological data; Based on a preset first time window, the behavioral data and the physiological indicator data are time-aligned to generate a first time series and a second time series, respectively. A first rate of change sequence corresponding to the first time series and a second rate of change sequence corresponding to the second time series are generated based on the rate of change of data of adjacent sampling points in the first time series and the second time series. The first rate of change sequence includes multiple time periods corresponding to the first time series, and the second rate of change sequence includes multiple time periods corresponding to the second time series. The ratio of the mean to the standard deviation of the first rate of change sequence is calculated as a first rate of change threshold, and the time period in the first rate of change sequence that exceeds the first rate of change threshold is marked as a first high rate of change segment. The ratio of the mean to the standard deviation of the second rate of change sequence is calculated as a second rate of change threshold, and the time period in the second rate of change sequence that exceeds the second rate of change threshold is marked as a second high rate of change segment. The sum of the rate of change of the data corresponding to each of the first high rate of change segments is calculated to obtain the first numerical volatility of the first time series. The sum of the rate of change of the data corresponding to each of the second high rate of change segments is calculated to obtain the second numerical volatility of the second time series. Based on the first numerical volatility and the second numerical volatility, determine the confidence weights corresponding to the first time series and the second time series; The first time series is shifted according to each time offset within a preset search range, and the correlation coefficient between the shifted first time series and the second time series is calculated. The preset search range includes multiple discrete time offsets. Determine the optimal time offset corresponding to the highest correlation coefficient, and use the first time series shifted according to the optimal time offset as the target time series. The target time series and the second time series are weighted and merged according to their corresponding confidence weights to obtain the patient's fused medical data.

2. The medical data fusion method according to claim 1, characterized in that, The step of determining the confidence weights corresponding to the first time series and the second time series based on the first numerical volatility and the second numerical volatility specifically includes: The first numerical volatility and the second numerical volatility are normalized respectively. Determine the maximum value of the first and second numerical variability after normalization, and take the logarithm of the maximum value as the benchmark attenuation factor; Using the benchmark decay factor as the base and the normalized first numerical volatility as the exponent, the calculated exponential function value is used as the confidence weight of the first time series. Using the normalized second numerical volatility as the exponent, the calculated exponential function value is used as the confidence weight of the second time series.

3. The medical data fusion method according to claim 1, characterized in that, The calculation of the correlation coefficient between the shifted first time series and the second time series specifically includes: The translated first time series and the second time series are divided into multiple consecutive second time windows according to a preset time length; Within each of the second time windows, the Pearson correlation coefficient and Spearman correlation coefficient between the shifted first time series and the second time series are calculated respectively. The correlation coefficient is obtained by weighted summing of the Pearson correlation coefficient and the Spearman correlation coefficient within each second time window.

4. The medical data fusion method according to claim 1, characterized in that, The step of weighting and merging the target time series and the second time series according to corresponding confidence weights to obtain the patient's fused medical data specifically includes: Based on the highest correlation coefficient and each of the aforementioned confidence weights, a normalized fusion factor is constructed corresponding to the target time series and the second time series. The target time series and the second time series are weighted and summed according to the corresponding normalized fusion factors to obtain the initial fusion sequence; The initial fusion sequence is processed by an exponential moving average to obtain the patient's fusion medical data.

5. The medical data fusion method according to claim 4, characterized in that, The construction of normalized fusion factors corresponding to the target time series and the second time series based on the highest correlation coefficient and each of the aforementioned confidence weights specifically includes: Multiply the highest correlation coefficient by a preset first weighting factor to obtain the relationship strength component; The confidence weights corresponding to the first time series and the second time series are multiplied by a preset second weight factor to obtain the corresponding intrinsic confidence components. The relationship strength component is added to each of the intrinsic credibility components to obtain the initial fusion factor corresponding to the target time series and the second time series; The sum of all initial fusion factors is used as the normalization base, and each initial fusion factor is divided by the normalization base to obtain the normalized fusion factors corresponding to the target time series and the second time series.

6. The medical data fusion method according to claim 1, characterized in that, The step of performing time alignment processing on the behavioral data and the physiological indicator data based on a preset first time window to generate a first time series and a second time series, specifically includes: Establish a unified timeline within the preset first time window; The behavioral data is filtered according to the time interval determined by a preset nonlinear function to generate corresponding values ​​at each time point on the unified time axis, thereby obtaining a first time series aligned with the unified time axis. The physiological index data are calculated using piecewise polynomial interpolation to generate corresponding values ​​at each time point on the unified time axis, resulting in a second time series aligned with the unified time axis.

7. A medical data fusion system, characterized in that, The method for performing a medical data fusion as described in claim 1 specifically includes: The data acquisition module is used to acquire patients' behavioral and physiological data. The time alignment module is used to perform time alignment processing on the behavioral data and the physiological indicator data based on a preset first time window, and generate a first time series and a second time series respectively. The credibility weight determination module is used to calculate the numerical volatility of the first time series and the second time series within a preset period, and determine the credibility weight of the first time series and the second time series based on the numerical volatility. The correlation analysis module is used to perform time shifting on the first time series according to each time offset within a preset search range, and calculate the correlation coefficient between the shifted first time series and the second time series. The preset search range includes multiple discrete time offsets. The optimal offset correction module is used to determine the optimal time offset corresponding to the highest correlation coefficient, and to take the first time series after being shifted according to the optimal time offset as the target time series. The data fusion module is used to perform weighted merging of the target time series and the second time series according to the corresponding confidence weights to obtain the patient's fused medical data.

8. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions. The user interface and the network interface are both used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-6.

Images