A Heart Rate Intelligent Prediction Method

Abstract

This invention discloses an intelligent heart rate prediction method. Relating to the field of heart rate monitoring technology, the method involves continuously collecting heart rate data from patients using a heart rate monitoring instrument, constructing short-term heart rate sequences and corresponding waveforms, segmenting the waveforms into intervals, and extracting typical heart rate features for each interval, including the interval's average heart rate and standard deviation. These features are combined into a feature sequence, and a heart rate feature library is established. In the prediction phase, using the current time as a benchmark, the method acquires the patient's heart rate data from the past traceability period, generating the latest traceability short-term heart rate waveform and its feature sequence. By searching for matching historical feature sequences in the heart rate feature library, the difference rate is calculated to determine the optimal match. Finally, based on the optimal heart rate waveform and feature sequence, the method outputs a prediction result for the future heart rate. By leveraging historical data patterns and associating them with the patient's own characteristics, the method achieves intelligent prediction, improving the accuracy and timeliness of heart rate monitoring and providing support for clinical decision-making.

Description

Technical Field

[0001] This invention belongs to the field of heart rate monitoring technology, specifically, it relates to an intelligent heart rate prediction method. Background Technology

[0002] As a core vital sign, even slight fluctuations in heart rate can be the earliest sign of a worsening condition.

[0003] Current technologies in heart rate prediction typically rely on traditional statistical methods or simple machine learning models. These methods often perform simplistic analysis based on overall heart rate data, lacking in-depth capture of the dynamic changes in heart rate waveforms. They cannot effectively handle the complex fluctuations and nonlinear characteristics of heart rates in critically ill patients because subtle changes in heart rate data in the short term, such as arrhythmias or sudden fluctuations, are easily masked by the overall trend, leading to predictions that deviate from reality. Furthermore, current technologies usually do not perform fine-grained interval segmentation and feature extraction of the heart rate waveform, but rather treat the entire time period as a whole, ignoring the heart rate within different time intervals. Local patterns reduce the accuracy and robustness of predictions. In terms of real-time performance, existing technologies use fixed thresholds or rules for early warning, which are slow to respond to sudden changes in heart rate in critically ill patients and are prone to false alarms or missed alarms. In addition, existing technologies lack deep learning-driven feature learning and similarity retrieval mechanisms, resulting in low efficiency when processing large amounts of historical data and an inability to quickly identify the waveform pattern most similar to the current state. Due to the deficiencies of existing technologies in feature refinement, real-time matching, and complex pattern processing, heart rate prediction is prone to deviation and delay, failing to meet the high accuracy and timeliness requirements of critical care.

[0004] To address the aforementioned problems, this invention proposes an intelligent heart rate prediction method. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent heart rate prediction method that solves the problems of low heart rate prediction accuracy and poor real-time performance in existing technologies.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A method for intelligent heart rate prediction, the method comprising:

[0008] Step 1: Collect heart rate data from critically ill patients using a heart rate monitoring device, determine the heart rate data associated with the critically ill patients within the collection period, construct several short-term heart rate sequences, and determine several short-term heart rate waveforms associated with the critically ill patients.

[0009] Step 2: Divide the short-term heart rate waveforms into segments, determine the heart rate waveform intervals associated with each short-term heart rate waveform, construct the typical heart rate features associated with each heart rate waveform interval, combine them into a typical heart rate feature sequence associated with the corresponding short-term heart rate waveform, and construct a heart rate feature library by combining the corresponding short-term heart rate waveforms.

[0010] Step 3: Using the current time as the end time of the traceability period, obtain the heart rate data of the ICU patient in the past traceability period, and perform interval segmentation to determine the heart rate waveform interval and typical heart rate characteristics associated with the latest traceability short-term heart rate waveform. Search and match in the heart rate feature database to determine the optimal heart rate waveform and optimal typical heart rate feature sequence, which are used as the heart rate prediction output associated with the ICU patient.

[0011] As a further aspect of the present invention, in step one, the specific method for collecting heart rate data of critically ill patients using a heart rate monitoring instrument and determining the associated heart rate data of critically ill patients within the collection period is as follows:

[0012] Get any critically ill patient in the intensive care unit, denoted as Q;

[0013] Determine the current time, denoted as t0;

[0014] Obtain the preset collection period and determine the duration of the collection period as T1;

[0015] Using a pre-built heart rate monitoring instrument, the heart rate data associated with critical care patient Q at each time point is continuously collected within a preset acquisition period T1, resulting in a heart rate sequence R1, R2, ..., Rj, where R1 to Rj represent the heart rate from the first time point to the j-th time point within the acquisition period T1, and j represents the total number of time points within the acquisition period T1.

[0016] As a further aspect of the present invention, in step one, the specific method for constructing several short-term heart rate sequences and determining several short-term heart rate waveforms associated with intensive care patients is as follows:

[0017] Obtain the preset short-term and duration T2, where T2 < T1;

[0018] The acquisition period T1 is segmented according to the timeline using the short-term T2 to obtain several short-term sequences. The total number is recorded as m. These sequences are then sorted in chronological order to obtain the short-term sequences ST1, ST2, ..., STm.

[0019] Take any short-term STn and its associated heart rate data in the heart rate sequence R1, R2, ..., Rj, arrange them in timeline to obtain the short-term heart rate sequence r1, r2, ..., ro associated with the short-term STn, where ro is the total number of moments in duration T2, n is the counting index, and 1 ≤ n ≤ m;

[0020] Construct a two-dimensional coordinate system with time line as the horizontal axis and heart rate data as the vertical axis. Plot the o heart rate data points in the short-term heart rate sequence r1, r2, ..., ro associated with short-term STn as data points in the two-dimensional coordinate system in the order of time line, and obtain o data points. Fit the curve to obtain a complete curve, which is denoted as the short-term heart rate waveform Wn associated with short-term STn.

[0021] Repeat the above steps to determine the short-term heart rate sequence and short-term heart rate waveform associated with each short-term sequence in the short-term sequences ST1, ST2, ..., STm.

[0022] As a further aspect of the present invention, in step two, the specific method for determining the several heart rate waveform intervals associated with each short-term heart rate waveform is as follows:

[0023] Obtain short-term sequences ST1, ST2, ..., STm and their corresponding short-term heart rate waveform sequences W1, W2, ..., Wm;

[0024] Obtain the preset total number of time points K and divide the short-term STn into intervals to obtain the short-term interval sequence DT1, DT2, ..., DTu associated with the short-term STn, where u is the total number of intervals in the short-term STn, K < 0, the start time of the short-term interval DT1 is the start time of the short-term STn, and the end time of the short-term interval DTu is the end time of the short-term STn.

[0025] Extract the portion of the short-term heart rate waveform Wn associated with each short-term interval in the short-term interval sequence DT1, DT2, ..., DTu from the short-term heart rate waveform Wn, and denot it as the heart rate waveform interval associated with each short-term interval. Summarize them as the heart rate waveform interval sequence w1, w2, ..., wu. Then, splice them in the order from w1 to wu to obtain the short-term heart rate waveform Wn.

[0026] As a further aspect of the present invention, the specific method for constructing the typical heart rate features associated with each heart rate waveform interval in step two is as follows:

[0027] Determine any short-term interval DTv in the short-term interval sequence DT1, DT2, ..., DTu, where v is the counting index and 1 ≤ v ≤ u;

[0028] Obtain the heart rate waveform interval wv associated with the short-term interval DTv;

[0029] Determine the heart rate data at each time point within the heart rate waveform interval wv, and take the average value, denoted as the interval average heart rate r_avg_v;

[0030] The standard deviation of all heart rate data in the heart rate waveform interval wv is determined based on the interval average heart rate r_avg_v, and is denoted as the interval heart rate standard deviation σ_v.

[0031] The combined calculated interval average heart rate r_avg_v and interval heart rate standard deviation σ_v are used as typical heart rate characteristics of the heart rate waveform interval wv associated with the short-term interval DTv, and are abbreviated as DXv.

[0032] Similarly, by determining the typical heart rate features associated with each heart rate waveform interval in the heart rate waveform interval sequence w1, w2, ..., wu, the typical heart rate feature sequence DX1, DX2, ..., DXu corresponding to the heart rate waveform interval sequence w1, w2, ..., wu is obtained.

[0033] As a further aspect of the present invention, the specific method for constructing the heart rate feature database in step two is as follows:

[0034] Extract the typical heart rate feature sequences DX1, DX2, ..., DXu and the heart rate waveform interval sequences w1, w2, ..., wu;

[0035] Any typical heart rate feature DXv is combined with the corresponding heart rate waveform interval wv to form a set of heart rate feature records, which are stored in a preset heart rate feature library.

[0036] Similarly, u sets of heart rate feature records are determined and stored in the heart rate feature database. Each set of heart rate feature records uses the interval average heart rate and the interval heart rate standard deviation as query indexes.

[0037] As a further aspect of the present invention, in step three, the specific method for obtaining the heart rate data of the intensive care patient within the past tracing cycle, using the current time as the end time of the tracing cycle, and performing interval segmentation, is as follows:

[0038] Get the current time, denoted as t0';

[0039] Obtain the traceability cycle and its duration T3 preset by the operator;

[0040] Take the current time t0' as the end time of a tracing period T3, and determine a tracing period T3 in the past;

[0041] Heart rate data of critically ill patient Q within this tracing period T3 are extracted and arranged in timeline to obtain the traced heart rate sequence R1', R2', ..., Rz', where R1' to Rz' represent the heart rate from the first time to the zth time within the tracing period T3, and z is the total number of times within the tracing period T3;

[0042] Using the same short-term T2 as in step one, segment the source tracing period T3 to obtain the source short-term sequence ST1',ST2',...,STe', where T2 < T3 and e is the total number of short-term sequences within the source tracing period T3;

[0043] Based on the source heart rate sequences R1', R2', ..., Rz', determine the source short-term heart rate sequences r1', r2', ..., ro' for any source short-term STf';

[0044] Using the same total number of time points K as in step two, perform interval segmentation on any one of the trace short-term sequences STf' in the trace short-term sequence ST1', ST2', ..., STe' to obtain the trace short-term interval sequence DT1', DT2', ..., DTu' associated with the trace short-term sequence STf', where f is the counting index and 1≤f≤e.

[0045] As a further aspect of the present invention, after determining the source short-term heart rate sequence r1', r2', ..., ro' of any source short-term STf' in step three, the source short-term heart rate waveform Wf' associated with the source short-term STf' is determined by using the method described in step one for determining several short-term heart rate waveforms associated with intensive care patients.

[0046] As a further aspect of the present invention, the specific method for determining the heart rate waveform interval and typical heart rate characteristics associated with the latest traceable short-term heart rate waveform in step three is as follows:

[0047] Based on the method of determining the traced short-term heart rate waveform Wf' associated with the traced short-term STf', the traced short-term heart rate waveform We' associated with the most recent traced short-term STf' of the critical care patient Q at the current time t0' is determined and used as the latest traced short-term heart rate waveform;

[0048] Based on the method of determining the source short-term interval sequence DT1',DT2',...,DTu' of the source short-term STf', the source short-term interval sequence associated with the source short-term STe' is determined. Then, the part of the latest source short-term heart rate waveform associated with each source short-term interval in the source short-term interval sequence associated with the source short-term STe' is extracted from the latest source short-term heart rate waveform. This part is denoted as the heart rate waveform interval associated with each source short-term interval, and the sum is denoted as the heart rate waveform interval sequence w1',w2',...,wu'.

[0049] Following the method described in step two, determine the interval average heart rate and interval standard deviation associated with each heart rate waveform interval in the heart rate waveform interval sequence w1', w2', ..., wu', and construct the corresponding typical heart rate feature sequence DX1', DX2', ..., DXu'.

[0050] As a further aspect of the present invention, in step three, the specific method for determining the optimal heart rate waveform and the optimal typical heart rate feature sequence as the heart rate prediction output associated with intensive care patients is as follows:

[0051] From the determined typical heart rate feature sequence DX1',DX2',...,DXu', continuously extract the g typical heart rate features closest to the current time t0', denoted as the first typical heart rate feature sequence, where g is a preset integer and 1≤g≤u;

[0052] Retrieve typical heart rate features from any set of g consecutive heart rate feature records from the heart rate feature database, and denote it as the second typical heart rate feature sequence;

[0053] Determine the difference rate between the corresponding heart rate typical features in the first heart rate typical feature sequence and the second heart rate typical feature sequence. If the average difference rate is less than the preset difference rate threshold, then the second heart rate typical feature sequence is selected as a candidate heart rate typical feature sequence.

[0054] Similarly, all candidate heart rate canonical feature sequences are determined, and the candidate heart rate canonical feature sequence with the smallest average difference rate is selected as the optimal heart rate canonical feature sequence.

[0055] Extract the heart rate feature records associated with each typical heart rate feature in the optimal heart rate typical feature sequence, extract the heart rate waveform interval, and reverse the source to determine the heart rate waveform diagram, which is then marked as the target heart rate waveform diagram;

[0056] Extract the target heart rate waveform from the g source short periods described above, and combine them to form the optimal heart rate waveform. Use the typical heart rate features from the optimal heart rate waveform as the associated heart rate prediction output for critically ill patients and display it. The calculation methods for the difference rate and the average difference rate are as follows:

[0057] Obtain any one typical heart rate feature from the first typical heart rate feature sequence and decompose it into the interval average heart rate r_avg1 and the interval heart rate standard deviation σ1;

[0058] Extract the heart rate typical feature corresponding to any heart rate typical feature in the first heart rate typical feature sequence from the second heart rate typical feature sequence, and decompose it into the interval average heart rate r_avg2 and the interval heart rate standard deviation σ2;

[0059] The interval average heart rate difference rate θ1 is calculated using |r_avg1-r_avg2| / [(r_avg1+r_avg2) / 2]=θ1;

[0060] The rate of difference in standard deviation of heart rate across intervals, θ2, is calculated using |σ1-σ2| / [(σ1+σ2) / 2]=θ2.

[0061] The difference rate θ between any heart rate typical feature in the first heart rate typical feature sequence and the corresponding heart rate typical feature in the second heart rate typical feature sequence is determined by θ = ω1*θ1 + ω2*θ2, where ω1 and ω2 are both calculation weights, and ω1 + ω2 = 1.

[0062] Similarly, determine the difference rate between all heart rate typical features in the first heart rate typical feature sequence and the corresponding heart rate typical features in the second heart rate typical feature sequence, and take the average value, which is denoted as the average difference rate θ_avg.

[0063] The beneficial effects of this invention are:

[0064] (1) This invention achieves efficient and personalized heart rate prediction by collecting heart rate data of critical care patients and constructing short-term heart rate sequences and waveforms. The method constructs a heart rate feature library through interval segmentation and typical feature extraction, thereby quickly locking the optimal prediction output during retrieval and matching, which significantly improves the accuracy and real-time performance of the prediction. Its advantage is that it can combine patient historical data for dynamic analysis and early warning, which helps clinicians to intervene in a timely manner and reduce medical risks. At the same time, it optimizes feature matching through deep learning models, which enhances the reliability and adaptability of the prediction, provides intelligent decision support for critical care, and ultimately improves the quality of patient care and medical efficiency.

[0065] (2) This invention utilizes a heart rate monitoring instrument to continuously collect heart rate data from critically ill patients within a preset collection period, forming a complete heart rate sequence, thus ensuring the comprehensiveness and real-time nature of data collection; secondly, by dividing the collection period into short segments and constructing short-term heart rate sequences and waveforms, it achieves refined analysis of heart rate changes, making it easier to capture abnormal fluctuations or trends in the short term, thereby improving the sensitivity and response speed to the patient's condition. The visualization of the waveforms enables medical staff to intuitively and quickly identify heart rate patterns and potential risks, promoting timely intervention and decision-making; the entire process is systematic and automated, improving monitoring efficiency and reducing human error, which is particularly beneficial in the intensive care environment and helps to optimize patient management.

[0066] (3) This invention constructs a systematic heart rate feature library by dividing the short-term heart rate waveform into multiple intervals and extracting the average heart rate and standard deviation of each interval as typical features. Its core advantage is that it realizes the refined description and efficient management of heart rate data. It can not only capture the dynamic fluctuation characteristics of heart rate in a short time more meticulously, improving the granularity and accuracy of data analysis, but also significantly enhance the retrieval and comparability of heart rate information by constructing the feature library with the interval average heart rate and standard deviation as standardized indexes. This provides stable and efficient data support for the rapid identification of ECG abnormalities, pattern matching and big data analysis.

[0067] (4) This invention achieves real-time monitoring and prediction of heart rate in critical care patients by dynamically tracing heart rate data from the past period based on the current time and performing refined interval segmentation and waveform construction. Its advantages are that it can efficiently process a large amount of heart rate data, accurately capture short-term fluctuations and long-term trends of heart rate through short-term sequence analysis and feature extraction, thereby improving the accuracy and reliability of prediction; it can intelligently match the optimal sequence in historical data and automatically generate personalized heart rate prediction output, significantly enhancing clinical decision support, reducing manual intervention, and improving monitoring efficiency and emergency response capabilities. Attached Figure Description

[0068] The invention will now be further described with reference to the accompanying drawings.

[0069] Figure 1 This is a flowchart illustrating the method described in this invention;

[0070] Figure 2 This is a flowchart illustrating the method described in Embodiment 2 of the present invention;

[0071] Figure 3 This is a flowchart illustrating the method described in Embodiment 3 of the present invention. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Example 1

[0074] A heart rate intelligent prediction method, such as Figure 1 As shown, this method includes the following:

[0075] This method is an intelligent heart rate prediction method mainly used for heart rate monitoring and prediction of critically ill patients in the intensive care unit. It combines signal processing, feature engineering and pattern recognition technologies to improve the timeliness and accuracy of clinical decision-making. The overall process achieves prediction by constructing a heart rate feature database and real-time retrieval and matching.

[0076] First, heart rate data of critically ill patients is collected using heart rate monitoring equipment. This data is then used to determine the associated heart rate data within the collection period, constructing several short-term heart rate sequences and identifying several associated short-term heart rate waveforms. Specifically:

[0077] The heart rate monitoring instrument is a medical-grade heart rate monitoring instrument, such as a continuous electrocardiogram (ECG) device or a photoplethysmography (PPG) sensor.

[0078] The heart rate monitoring instrument is used to collect the heart rate of critically ill patients in real time and continuously until a collection cycle is completed. Then, based on the heart rate data within the collection cycle, a heart rate data sequence is formed in chronological order. The collection cycle is then initially divided into short periods by a preset short period, and the heart rate data within each short period is determined. In this way, several short-term heart rate sequences can be constructed.

[0079] Each short-term heart rate sequence is visualized as a short-term heart rate waveform, where the horizontal axis represents time (such as seconds or minutes) and the vertical axis represents heart rate data. The short-term heart rate waveform can intuitively show the fluctuations, trends and abnormalities of heart rate, such as arrhythmias or tachycardia. These short-term heart rate waveforms will serve as the basis for subsequent analysis.

[0080] It needs to be explained that heart rate changes in critically ill patients are often related to the deterioration of their condition. Short-term heart rate sequences capture the instantaneous dynamics of heart rate, avoiding the masking of long-term trends and helping to detect abnormalities early.

[0081] Secondly, short-term heart rate waveforms provide a visual representation of time series data, facilitating rapid assessment by healthcare professionals. They also provide structured data for computational analysis, and features such as heart rate variability can be extracted from short-term heart rate waveforms.

[0082] Then, the determined short-term heart rate waveforms are segmented into intervals to determine the heart rate waveform intervals associated with each short-term heart rate waveform. Typical heart rate features associated with each heart rate waveform interval are constructed and combined into a typical heart rate feature sequence associated with the corresponding short-term heart rate waveform. Finally, a heart rate feature library is constructed by combining the corresponding short-term heart rate waveforms. Specifically:

[0083] First, the operator uses a preset time to segment several short-term heart rate waveforms into intervals, thus obtaining several heart rate waveform intervals associated with each short-term heart rate waveform.

[0084] For each heart rate waveform interval, a set of typical heart rate features is calculated, including the interval average heart rate and the interval heart rate standard deviation. Then, the features of each interval are combined in chronological order to form a typical heart rate feature sequence for the corresponding short-term heart rate waveform.

[0085] Finally, all short-term heart rate waveforms and their corresponding feature sequences are stored in a database, denoted as the heart rate feature library.

[0086] It should be explained that segmenting a continuous waveform into smaller units makes it easier to capture local fluctuations and transition modes, and reduce the impact of noise. This is similar to piecewise linear approximation or windowing analysis in signal denoising and related processing.

[0087] Secondly, the purpose of identifying typical heart rate characteristics is to further extract meaningful numerical indicators from redundant heart rate data.

[0088] Finally, the heart rate feature library is equivalent to a historical knowledge base, which accumulates pattern data of patients or several patients. By storing feature sequences, it supports fast similarity search and pattern matching, providing reference templates for prediction, rather than simulating through pure models or pure AI, which lacks data support.

[0089] After the above operations, the current time is used as the end time of the traceability period. Heart rate data of ICU patients within the past traceability period is obtained and segmented into intervals. The heart rate waveform interval and typical heart rate characteristics associated with the latest traceability short-term heart rate waveform are determined. A search and matching process is performed in the heart rate feature database to determine the optimal heart rate waveform and optimal typical heart rate feature sequence, which are then used as the heart rate prediction output associated with the ICU patient. Specifically:

[0090] The source tracing cycle, which uses the current time as the end time of the source tracing cycle to review the past, can be understood as a retrospective time window (e.g., the past 24 hours, 48 ​​hours, or one week) to obtain the heart rate data of critical care patients within the source tracing cycle, ensuring that the data covers possible historical changes in the heart rate status of critical care patients.

[0091] Next, the same interval segmentation and feature extraction are performed on the heart rate data within the traceability period to obtain the heart rate waveform interval and typical heart rate feature sequence associated with the latest traceability short-term heart rate waveform. In the heart rate feature library, a similarity measurement algorithm is used to retrieve the historical feature sequence that best matches the latest feature sequence. The retrieved best match (i.e., the historical short-term heart rate waveform and feature sequence with the highest similarity) is used as the basis for prediction, and the predicted heart rate data is output.

[0092] Example 2

[0093] This embodiment improves upon some of the content described in Embodiment 1, and further discloses a method for determining several short-term heart rate waveforms associated with intensive care patients, such as... Figure 2 As shown, it specifically includes the following:

[0094] First, we need to select any one critically ill patient from all the critically ill patients and denote it as Q;

[0095] After identifying the critically ill patient Q, the current time is immediately obtained and recorded as t0;

[0096] Then, obtain the collection cycle preset by the operator based on the actual situation, and extract the duration of the collection cycle, denoted as T1;

[0097] Then, the heart rate monitoring instrument described in Example 1 is used to monitor the heart rate of intensive care patient Q. Heart rate data of intensive care patient Q is collected within a collection period T1 with the current time t0 as the start time of the collection period T1, including heart rate data at each moment within the collection period T1. All heart rate data are sorted in chronological order to obtain the heart rate sequence R1, R2, ..., Rj, where R1 to Rj represent the heart rate data from the first moment to the jth moment within the collection period T1, and j represents the total number of moments within the collection period T1.

[0098] Then obtain the short-term and short-term duration preset by the operator based on the actual situation, denoted as T2. Note that T2 < T1.

[0099] Obtain the first moment of the acquisition period T1 as described above and use it as the first moment of a short-term T2. Then, extract a short-term T2 from the acquisition period T1 and continue extracting until the acquisition period T1 is completely extracted. This completes the segmentation of the acquisition period T1 according to the timeline using short-term T2. Summarize all the extracted short-terms and record the total number as m. Sort the m short-terms in chronological order to obtain the short-term sequence ST1, ST2, ..., STm.

[0100] The principle is that a heart rate sequence with a length of T1 may be affected by a variety of factors such as diurnal rhythm, activity, treatment, and changes in the condition, making direct analysis extremely difficult. After dividing it into several short segments of T2, the physiological state within each segment is relatively stable, making it easier to capture specific patterns. In subsequent operations, each short segment STn not only becomes an independent basic analysis unit, but is also interconnected with each other.

[0101] Extract any short-term sequence ST1, ST2, ..., STm from the determined short-term sequences, denoted as STn. Extract the heart rate data associated with short-term STn from the heart rate sequences R1, R2, ..., Rj, and sort them according to the timeline to obtain the short-term heart rate sequences r1, r2, ..., ro associated with short-term STn, where o is the total number of moments within duration T2, n is the counting index, 1≤n≤m, and o<j.

[0102] Next, a two-dimensional coordinate system is constructed with the time line as the horizontal axis and the heart rate data as the vertical axis. Then, o heart rate data points are extracted from the short-term heart rate sequence r1, r2, ..., ro associated with the short-term STn. These o heart rate data points are plotted in the constructed two-dimensional coordinate system in chronological order as data points to obtain o data points. Finally, the o data points are fitted with a curve to obtain a complete curve, which is recorded as the short-term heart rate waveform Wn associated with the short-term STn.

[0103] By repeating the above curse, the short-term heart rate sequences and short-term heart rate waveforms associated with each of the short-term sequences ST1, ST2, ..., STm can be determined.

[0104] Example 3

[0105] This embodiment, based on embodiment 2, further discloses a method for storing each heart rate waveform interval and its associated typical heart rate features using a preset heart rate feature database, such as... Figure 3 As shown, it specifically includes the following:

[0106] This embodiment is based on segmented feature extraction and indexed storage of heart rate time series data to achieve efficient heart rate pattern analysis and retrieval. In Embodiment 2, heart rate data was segmented to determine short-term heart rate sequences and short-term heart rate waveforms. Directly processing short-term heart rate sequences and waveforms is computationally intensive and makes it difficult to capture local features. Therefore, this embodiment decomposes the sequence into smaller intervals, extracts the statistical features of each interval, and constructs a feature library to support feature-based fast querying and pattern matching, as detailed below:

[0107] First, obtain m short-term heart rates ST1 to STm and their corresponding short-term heart rate waveforms W1 to Wm. Each short-term heart rate and short-term heart rate waveform represents heart rate data over a period of time, presented in both data and image formats.

[0108] Obtain the total number of time points K preset by the operator, and use the total number of time points K to segment any short-term STn in the short-term sequence ST1, ST2, ..., STm into intervals to obtain the short-term interval sequence DT1, DT2, ..., DTu associated with the short-term STn, where u is the total number of intervals in the short-term STn, and K < 0, the start time of the short-term interval DT1 is the start time of the short-term STn, and the end time of the short-term interval DTu is the end time of the short-term STn.

[0109] Next, in the short-term heart rate waveform Wn associated with the short-term STn, portions of the short-term heart rate waveform Wn associated with all short-term intervals in the short-term interval sequence DT1, DT2, ..., DTu are extracted to obtain the heart rate waveform interval associated with each short-term interval. These intervals are then sorted according to the timeline to obtain the heart rate waveform interval sequence w1, w2, ..., wu. If the heart rate waveform interval sequences w1, w2, ..., wu are spliced ​​together in the order from w1 to wu, the complete short-term heart rate waveform Wn can be obtained.

[0110] Based on the determined short-term interval sequence DT1, DT2, ..., DTu, extract any short-term interval, denoted as DTv, where v is the counting index, and 1 ≤ v ≤ u;

[0111] Extract the heart rate waveform interval wv associated with the short-term interval DTv from the determined heart rate waveform interval sequence w1, w2, ..., wu, then obtain the heart rate data corresponding to each moment in the heart rate waveform interval wv, and average the heart rate data corresponding to each extracted moment, which is denoted as the interval average heart rate r_avg_v.

[0112] Then, based on the determined interval average heart rate r_avg_v and the standard deviation of all heart rate data in the determined heart rate waveform interval wv, the standard deviation of the interval heart rate is denoted as σ_v.

[0113] Next, the calculated interval average heart rate r_avg_v and interval heart rate standard deviation σ_v are extracted and combined to obtain the typical heart rate characteristics of the heart rate waveform interval wv associated with the short-term interval DTv, which is abbreviated as DXv.

[0114] By repeating the above steps, the typical heart rate features associated with each heart rate waveform interval in the heart rate waveform interval sequence w1, w2, ..., wu can be determined, and the typical heart rate feature sequence corresponding to the heart rate waveform interval sequence w1, w2, ..., wu can be obtained, denoted as: DX1, DX2, ..., DXu.

[0115] Then, any typical heart rate feature DXv is combined with the corresponding heart rate waveform interval wv to form a set of heart rate feature records, which are stored in the heart rate feature library pre-built by the operator.

[0116] If the above approach is not feasible, u sets of heart rate feature records can be constructed based on the heart rate waveform interval sequence w1, w2, ..., wu and the typical heart rate feature sequence DX1, DX2, ..., DXu, and stored synchronously in a preset heart rate feature database. Each set of heart rate feature records uses the interval average heart rate and the interval heart rate standard deviation as query indexes.

[0117] Example 4

[0118] This embodiment further discloses a method for predicting the heart rate of critically ill patients, based on embodiment 3, specifically including the following:

[0119] First, the current time needs to be determined again and recorded as t0'. Then, the traceability period and duration T3 preset by the operator are obtained.

[0120] Next, the current time t0' is taken as the end time of a tracing period T3. Based on the end time, the process extends to the past and a tracing period T3 is determined.

[0121] Then, the heart rate data of critically ill patient Q within the determined traceability period T3 are obtained and sorted in chronological order to obtain the traceability heart rate sequence, represented as: R1',R2',...,Rz', where traceability heart rates R1' to Rz' represent the heart rates from the first moment to the zth moment within the traceability period T3, and z is the total number of moments within the traceability period T3.

[0122] Next, the same short-term T2 as in step one is used to perform segmentation processing on the tracing period T3, and finally the tracing short-term sequence associated with the tracing period T3 can be obtained, represented as: ST1',ST2',...,STe', where T2<T3 and e is the total number of short-term sequences within the tracing period T3.

[0123] Obtain the source heart rate sequences R1', R2', ..., Rz', and determine the source short heart rate sequence associated with each source short sequence ST1', ST2', ..., STe'. The source short heart rate sequence of any source short sequence STf' is represented as r1', r2', ..., ro'.

[0124] Next, using the same total number of times K as in Example 3, interval segmentation is performed on any one of the trace short-term sequences STf' in the trace short-term sequence ST1', ST2', ..., STe'. Finally, the trace short-term interval sequence DT1', DT2', ..., DTu' associated with the trace short-term sequence STf' can be obtained, where f is the counting index and 1≤f≤e.

[0125] Then, according to the method described in Example 2, the traced short heart rate waveform Wf' associated with the traced short heart rate sequence r1',r2',...,ro' associated with the traced short STf' is constructed.

[0126] Based on the method described above for determining the short-term heart rate waveform Wf' associated with the short-term STf', the short-term heart rate waveform associated with the most recent short-term STf' of the critical care patient Q at the current time t0' can be determined and denoted as We'. The short-term heart rate waveform We' is then marked as the latest short-term heart rate waveform.

[0127] Based on the method of determining the source short-term interval sequence DT1',DT2',...,DTu' associated with the source short-term STf', the source short-term interval sequence associated with the source short-term STe' is determined. Then, from the determined latest source short-term heart rate waveform, the part of the latest source short-term heart rate waveform associated with each source short-term interval in the source short-term interval sequence associated with the source short-term STe' is extracted and denoted as the heart rate waveform interval associated with each source short-term interval in the source short-term interval sequence DT1',DT2',...,DTu'. All heart rate waveform intervals are summarized and denoted as the heart rate waveform interval sequence, represented as: w1',w2',...,wu';

[0128] Based on the method described in Example 3, the mean heart rate and standard deviation of the heart rate associated with each heart rate waveform interval in the heart rate waveform interval sequence w1', w2', ..., wu' are determined and combined to construct the corresponding typical heart rate feature sequence DX1', DX2', ..., DXu' associated with the heart rate waveform interval.

[0129] Next, the determined typical heart rate feature sequence DX1',DX2',...,DXu' is extracted, and the g typical heart rate features closest to the current time t0' are extracted from the typical heart rate feature sequence DX1',DX2',...,DXu'. The g typical heart rate features are then sorted in chronological order and denoted as the first typical heart rate feature sequence, where g is a preset integer and 1≤g≤u.

[0130] Next, retrieve the typical heart rate features from any set of g consecutive heart rate feature records in the heart rate feature database, and sort them in chronological order, which is recorded as the second typical heart rate feature sequence.

[0131] Determine the difference rate of the time-sequential heart rate typical features between the first and second heart rate typical feature sequences. If the average difference rate is less than a preset difference rate threshold, then the second heart rate typical feature sequence is selected as a candidate heart rate typical feature sequence, as follows:

[0132] First, obtain any typical heart rate feature from the first typical heart rate feature sequence and decompose it into the interval average heart rate r_avg1 and the interval heart rate standard deviation σ1;

[0133] Then, extract a typical heart rate feature from the second typical heart rate feature sequence that corresponds to the typical heart rate feature with an interval average heart rate of r_avg1 and an interval heart rate standard deviation of σ1 determined in the first typical heart rate feature sequence, and decompose this typical heart rate feature extracted from the second typical heart rate feature sequence into an interval average heart rate of r_avg2 and an interval heart rate standard deviation of σ2.

[0134] The interval average heart rate difference rate θ1 is calculated using: |r_avg1-r_avg2| / [(r_avg1+r_avg2) / 2]=θ1;

[0135] Then, the difference rate θ2 of the standard deviation of heart rate in the interval is calculated by using: |σ1-σ2| / [(σ1+σ2) / 2]=θ2;

[0136] Then, the difference rate θ between any heart rate typical feature in the first heart rate typical feature sequence and the corresponding heart rate typical feature in the second heart rate typical feature sequence is determined by using θ=ω1*θ1+ω2*θ2, where ω1 and ω2 are both calculation weights, and ω1+ω2=1;

[0137] Finally, by repeating the above operations, the difference rate between all heart rate typical features in the first heart rate typical feature sequence and the corresponding heart rate typical features in the second heart rate typical feature sequence is determined, and the mean of all difference rates is calculated and marked as the average difference rate θ_avg.

[0138] For example, the first typical heart rate feature sequence contains 10 typical heart rate features, numbered sequentially as 1, 2, 3, ..., 10; the second typical heart rate feature sequence also contains 10 typical heart rate features, numbered sequentially as 11, 12, 13, ..., 20.

[0139] The average difference rate θ_avg is equal to the sum of the difference rates between typical heart rate feature 1 and typical heart rate feature 11, the difference rates between typical heart rate feature 2 and typical heart rate feature 12, ..., the difference rates between typical heart rate feature 10 and typical heart rate feature 20, and then divided by 10 to obtain the average difference rate θ_avg.

[0140] The average difference rate θ_avg is essentially an indicator of the overall similarity between two heart rate canonical feature sequences. It is calculated by taking the difference rate θ (i.e., the weighted average heart rate difference rate θ1 and the standard deviation difference rate θ2) for each pair of heart rate canonical features in the two heart rate canonical feature sequences in chronological order (each heart rate canonical feature consists of the interval average heart rate and the interval heart rate standard deviation) and then taking the arithmetic mean of the difference rates at all corresponding positions. The smaller the average difference rate θ_avg, the higher the overall similarity between the two heart rate canonical feature sequences.

[0141] Repeat the above steps to determine all candidate heart rate canonical feature sequences. Extract the unique candidate heart rate canonical feature sequence with the smallest average difference rate from all the determined candidate heart rate canonical feature sequences and mark it as the optimal heart rate canonical feature sequence. Then, extract the heart rate waveform intervals corresponding to the heart rate feature records that are commonly associated with each heart rate canonical feature in the optimal heart rate canonical feature sequence, and investigate the heart rate waveform diagrams associated with all heart rate waveform intervals, and use them as the target heart rate waveform diagram.

[0142] Then, extract the next g heart rate typical features (g heart rate waveform intervals / short-term source intervals) in time relative to the g heart rate typical features (i.e., g heart rate waveform intervals / short-term source intervals) mentioned above in the target heart rate waveform (i.e., find the next g heart rate typical features after the segment corresponding to the current most recent g heart rate typical features in similar target heart rate waveforms matched in history). This is because if there exists a historical period in which the previous g heart rate typical features are highly similar to the current g heart rate typical features, then the next g heart rate typical features immediately following this historical period are considered the next g heart rate typical features. The g typical heart rate features following the historical pattern are called the "subsequent g typical heart rate features". They are used to predict future heart rate changes (that is, to infer the possible future heart rate changes of the current patient by using the evolution trend after the historical similar pattern). The heart rate waveform intervals of the subsequent g typical heart rate features are extracted and combined into the optimal heart rate waveform. Then, the typical heart rate features in the optimal heart rate waveform and the optimal heart rate waveform are used together as the heart rate prediction output associated with the intensive care patient, so as to display it to medical staff and operators.

[0143] All data in the formulas described above have been calculated with dimensions removed. Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0144] The above description is merely an example and illustration of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the invention or exceed the scope defined in the claims, they should all fall within the protection scope of the present invention.

[0145] It should be stated that all user data collected in this application was collected with the user's consent and authorization. Furthermore, the uses of user data are legal and compliant, and the use and processing of user data comply with the relevant laws, regulations, and standards of the relevant regions.

Claims

1. A method for intelligent heart rate prediction, characterized in that, The method includes: Step 1: Collect heart rate data from critically ill patients using a heart rate monitoring device, determine the heart rate data associated with the critically ill patients within the collection period, and construct several short-term heart rate sequences R1, R2, ..., Rj, where j represents the total number of moments within the preset collection period; Obtain the preset short-term and duration T2. ​​Use the short-term T2 to segment the acquisition period T1 in chronological order to obtain several short-term sequences. The total number is recorded as m. Determine the short-term heart rate sequences and short-term heart rate waveforms W1, W2, ..., Wm associated with each short-term sequence ST1, ST2, ..., STm associated with the intensive care patient. Step 2: Divide the short-term heart rate waveforms W1, W2, ..., Wm into segments to determine the heart rate waveform intervals associated with each short-term heart rate waveform, including: obtaining the short-term sequences ST1, ST2, ..., STm and their corresponding short-term heart rate waveform sequences W1, W2, ..., Wm; Based on the total number of time points K, the short-term STn is divided into intervals to obtain the short-term interval sequence DT1, DT2, ..., DTu, where K < 0, the start time of the short-term interval DT1 is the start time of the short-term STn, and the end time of the short-term interval DTu is the end time of the short-term STn. Extract the portion of the short-term heart rate waveform Wn associated with each short-term interval in the short-term interval sequence DT1, DT2, ..., DTu from the short-term heart rate waveform Wn, and denot it as the heart rate waveform interval associated with each short-term interval. Summarize them as the heart rate waveform interval sequence w1, w2, ..., wu. The short-term heart rate waveform Wn is obtained by splicing them in the order from w1 to wu. Construct typical heart rate features associated with each heart rate waveform interval, including: Determine the heart rate waveform interval wv for any short interval DTv in the short interval sequence, 1≤v≤u; Take the average heart rate data at each time point in wv, and denote it as the interval average heart rate r_avg_v; The standard deviation of all heart rate data in wv is determined based on r_avg_v and denoted as the interval heart rate standard deviation σ_v. The combination of r_avg_v and σ_v serves as the typical heart rate feature DXv of the heart rate waveform interval wv associated with the short-term interval DTv. Based on the method of determining the typical heart rate features DXv of the heart rate waveform interval wv associated with the short-term interval DTv, the typical heart rate features associated with each heart rate waveform interval in the sequence w1, w2, ..., wu are determined, and the typical heart rate feature sequence DX1, DX2, ..., DXu corresponding to the heart rate waveform interval sequence w1, w2, ..., wu is obtained. Determine the typical heart rate feature sequences associated with each short-term heart rate waveform, and construct a heart rate feature library by combining the corresponding short-term heart rate waveforms. Specifically: Extract the typical heart rate feature sequences DX1,DX2,...,DXu and the heart rate waveform interval sequences w1,w2,...,wu; Any typical heart rate feature DXv is combined with the corresponding heart rate waveform interval wv as a set of heart rate feature records and stored in a preset heart rate feature library. Similarly, u sets of heart rate feature records are determined and stored in the heart rate feature database. Each set of heart rate feature records uses the interval average heart rate and the interval heart rate standard deviation as query indexes. Step 3: Using the current time as the end time of the tracing cycle, obtain the heart rate data of the intensive care patient in the past tracing cycle, and divide the data into intervals to determine the short-term tracing interval sequence DT1', DT2', ..., DTu'. Then, determine the heart rate waveform interval and typical heart rate features associated with the latest short-term tracing heart rate waveform. The g most recent typical heart rate features are recorded as the first typical heart rate feature sequence, where g is a preset integer and 1≤g≤u; Retrieve typical heart rate features from any set of g consecutive heart rate feature records in the heart rate feature database, arrange them in chronological order of retrieval time, and denot them as the second typical heart rate feature sequence. Determine the difference rate between the corresponding heart rate typical features in the first heart rate typical feature sequence and the second heart rate typical feature sequence. If the average difference rate is less than the preset difference rate threshold, then the second heart rate typical feature sequence is selected as a candidate heart rate typical feature sequence. Similarly, all candidate heart rate canonical feature sequences are determined, and the candidate heart rate canonical feature sequence with the smallest average difference rate is selected as the optimal heart rate canonical feature sequence. Extract the heart rate feature records associated with each typical heart rate feature in the optimal heart rate typical feature sequence, extract the heart rate waveform interval, and reverse the source to determine the heart rate waveform diagram, which is then marked as the target heart rate waveform diagram; Extract the target heart rate waveform and the heart rate waveform corresponding to the g source short intervals of the timeline of the g heart rate typical features in the first heart rate typical feature sequence. Combine them and call them the optimal heart rate waveform. Use the heart rate typical features in the optimal heart rate waveform as the heart rate prediction output associated with the critical care patient and display it.

2. The intelligent heart rate prediction method according to claim 1, characterized in that, In step one, the specific method for collecting heart rate data from critically ill patients using a heart rate monitoring device and determining the associated heart rate data of critically ill patients within the collection period is as follows: Get any critically ill patient in the intensive care unit, denoted as Q; Determine the current time, denoted as t0; Obtain the collection period, and denote the duration of the collection period as T1; Using a pre-built heart rate monitoring instrument, heart rate data associated with critically ill patient Q at each time point are continuously collected within the acquisition period T1, resulting in heart rate sequences R1, R2, ..., Rj, where R1 to Rj represent the heart rate from the first time point to the j-th time point within the acquisition period T1.

3. The intelligent heart rate prediction method according to claim 2, characterized in that, In step one, the specific method for constructing several short-term heart rate sequences and determining several short-term heart rate waveforms associated with intensive care patients is as follows: Find the shortest-term T2, where T2 < T1; Extract the short-term sequences ST1, ST2, ..., STm; Take any short-term STn and its associated heart rate data in the heart rate sequence R1, R2, ..., Rj, arrange them in timeline to obtain the short-term heart rate sequence r1, r2, ..., ro associated with the short-term STn, where ro is the total number of moments in duration T2, n is the counting index, and 1 ≤ n ≤ m; Construct a two-dimensional coordinate system with time line as the horizontal axis and heart rate data as the vertical axis. Plot the o heart rate data points in the short-term heart rate sequence r1, r2, ..., ro associated with short-term STn as data points in the two-dimensional coordinate system in the order of time line, and obtain o data points. Fit the curve to obtain a complete curve, which is denoted as the short-term heart rate waveform Wn associated with short-term STn. Repeat the above steps to determine the short-term heart rate sequence and short-term heart rate waveform associated with each short-term sequence in the short-term sequences ST1, ST2, ..., STm.

4. The intelligent heart rate prediction method according to claim 3, characterized in that, In step three, the current time is used as the end time of the tracing cycle. The specific method for obtaining the heart rate data of critically ill patients in the past tracing cycle and performing interval segmentation is as follows: Get the current time, denoted as t0'; Obtain the traceability cycle and its duration T3 preset by the operator; Take the current time t0' as the end time of a tracing period T3, and determine a tracing period T3 in the past; Heart rate data of critically ill patient Q within this tracing period T3 are extracted and arranged in timeline to obtain the traced heart rate sequence R1', R2', ..., Rz', where R1' to Rz' represent the heart rate from the first time to the zth time within the tracing period T3, and z is the total number of times within the tracing period T3; Using the same short-term T2 as in step one, segment the source tracing period T3 to obtain the source short-term sequence ST1',ST2',...,STe', where T2 < T3 and e is the total number of short-term sequences within the source tracing period T3; Based on the source heart rate sequences R1', R2', ..., Rz', determine the source short-term heart rate sequences r1', r2', ..., ro' for any source short-term STf'; Using the same total number of time points K as in step two, perform interval segmentation on any one of the trace short-term sequences STf' in the trace short-term sequence ST1', ST2', ..., STe' to obtain the trace short-term interval sequence DT1', DT2', ..., DTu' associated with the trace short-term sequence STf', where f is the counting index and 1≤f≤e.

5. The intelligent heart rate prediction method according to claim 4, characterized in that, In step three, after determining the source short-term heart rate sequence r1', r2', ..., ro' of any source short-term STf', the source short-term heart rate waveform Wf' associated with the source short-term STf' is determined using the method described in step one for determining several short-term heart rate waveforms associated with intensive care patients.

6. The intelligent heart rate prediction method according to claim 5, characterized in that, In step three, the specific method for determining the heart rate waveform interval and typical heart rate characteristics associated with the latest traceable short-term heart rate waveform is as follows: Based on the method of determining the traced short-term heart rate waveform Wf' associated with the traced short-term STf', the traced short-term heart rate waveform We' associated with the most recent traced short-term STf' of the critical care patient Q at the current time t0' is determined and used as the latest traced short-term heart rate waveform; Based on the method of determining the source short-term interval sequence DT1',DT2',...,DTu' of the source short-term STf', the source short-term interval sequence associated with the source short-term STe' is determined. Then, the part of the latest source short-term heart rate waveform associated with each source short-term interval in the source short-term interval sequence associated with the source short-term STe' is extracted from the latest source short-term heart rate waveform. This part is denoted as the heart rate waveform interval associated with each source short-term interval, and the sum is denoted as the heart rate waveform interval sequence w1',w2',...,wu'. As described in step two, determine the interval average heart rate and interval standard deviation associated with each heart rate waveform interval in the heart rate waveform interval sequence w1', w2', ..., wu', and construct the corresponding typical heart rate feature sequence DX1', DX2', ..., DXu'.

7. The intelligent heart rate prediction method according to claim 6, characterized in that, In step three, the specific methods for calculating the difference rate and the average difference rate include: From the determined typical heart rate feature sequence DX1',DX2',...,DXu', continuously extract the g typical heart rate features closest to the current time t0', and denote them as the first typical heart rate feature sequence; Extract typical feature sequences of the second heart rate based on the heart rate feature library; Obtain any one typical heart rate feature from the first typical heart rate feature sequence and decompose it into the interval average heart rate r_avg1 and the interval heart rate standard deviation σ1; Extract the heart rate typical feature corresponding to any heart rate typical feature in the first heart rate typical feature sequence from the second heart rate typical feature sequence, and decompose it into the interval average heart rate r_avg2 and the interval heart rate standard deviation σ2; The interval average heart rate difference rate θ1 is calculated using |r_avg1-r_avg2| / [(r_avg1+r_avg2) / 2]=θ1; The rate of difference in standard deviation of heart rate across intervals, θ2, is calculated using |σ1-σ2| / [(σ1+σ2) / 2]=θ2. The difference rate θ between any heart rate typical feature in the first heart rate typical feature sequence and the corresponding heart rate typical feature in the second heart rate typical feature sequence is determined by θ = ω1*θ1 + ω2*θ2, where ω1 and ω2 are both calculation weights, and ω1 + ω2 = 1. Similarly, determine the difference rate between all heart rate typical features in the first heart rate typical feature sequence and the corresponding heart rate typical features in the second heart rate typical feature sequence, and take the average value, which is denoted as the average difference rate θ_avg.

Images