Construction method of rehabilitation evaluation model for department of cardiology

Abstract

The invention relates to the technical field of physiological parameter monitoring, in particular to a construction method of a cardiology rehabilitation evaluation model, which comprises the following steps: monitoring a systolic pressure peak value, a diastolic pressure valley value and duration time of a postoperative patient, constructing a waveform parameter set, calculating a peak sequence amplitude change rate through a moving average method, and calculating a peak sequence amplitude change rate. Marking a form abnormal period and a structure drift region; dividing an active phase and a passive phase to extract wave crest characteristics; analyzing time sequence difference to generate a three-dimensional rhythm vector; constructing a regression model; according to the method, heart pressure change is continuously monitored, a waveform structure is identified, compression period change is dynamically captured, a structure drift area is determined and positioned by combining an amplitude change rate and a difference value, multi-dimensional heart rhythm characteristics are extracted, day and night differences are analyzed, and multi-day deviation values are integrated to generate a scoring system; the abnormal recognition, rhythm quantification and trend prediction capabilities are improved, and a high-adaptability quantification basis is provided for rehabilitation evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of physiological parameter monitoring, and in particular to a method for constructing a cardiology rehabilitation assessment model. Background Art

[0002] The field of physiological parameter monitoring technology includes technologies that assess health status and disease risks by measuring and recording multiple physiological indicators of the human body in real time. This field focuses on how to accurately monitor physiological parameters such as heart rate, blood pressure, body temperature, respiratory rate, and blood oxygen saturation, and analyze the data to conduct health assessments, disease predictions, or adjust treatment plans. Physiological parameter monitoring technology is widely used in hospitals, medical equipment, health management, and rehabilitation treatment, especially for the monitoring and assessment of patients with chronic diseases at risk, such as cardiovascular diseases and respiratory diseases. The core content includes physiological signal acquisition technology, data transmission and storage technology, data analysis and processing technology, etc. The technology can obtain and analyze physiological data under dynamic and real-time conditions, and provide a scientific and healthy status assessment basis.

[0003] Among them, the method for constructing a cardiology rehabilitation assessment model refers to the process of establishing a model for rehabilitation assessment of cardiology patients. The key issue is how to use physiological parameter monitoring data, combined with the rehabilitation status of cardiovascular patients, to build a scientific and systematic assessment model. The method for constructing an assessment model involves the use of multiple physiological parameters, including the analysis of electrocardiogram, blood pressure, heart rate, blood oxygen and other data, combined with clinical rehabilitation indicators to construct a personalized assessment model for cardiology patients. In addition, the method also improves the rehabilitation assessment system for cardiology patients by comparing and integrating physiological parameter data with real-time monitoring data, combining statistics and medical theories. The key to the method lies in how to effectively reflect the changing patterns of multiple physiological indicators of patients during the rehabilitation process through matching data integration and modeling technology, thereby providing doctors with a scientific and accurate basis for rehabilitation assessment.

[0004] Existing physiological parameter monitoring technologies often rely on single measurements at discrete time points, lacking the ability to capture continuous dynamic changes. This can easily lead to overlooking transient abnormal signals or periodic fluctuations. This can include brief structural drifts in the ECG waveform that go unrecorded due to long sampling intervals, impacting the multidimensional nature of risk assessment. Traditional analysis methods employ fixed thresholds or single statistical indicators for abnormality assessment, making it difficult to account for individual differences and the fluctuating characteristics of complex physiological rhythms. For example, circadian blood pressure rhythm assessment relies solely on mean comparisons, ignoring the coordinated changes in peak spacing and duration, leading to misjudgment of compensatory regulation as pathological conditions. Existing assessment models are often based on static parameters and fail to integrate trend characteristics from multi-day continuous data. This makes it impossible to effectively distinguish short-term fluctuations from long-term pathological evolution. This can easily lead to overlooking cases in which a patient in the postoperative recovery period may have normal single-day data but a worsening trend over three days. Furthermore, existing technologies fail to differentiate the functions of the active and passive phases of the heart, resulting in a lack of analytical insight into rhythm analysis at the physiological level, making it difficult to guide the development of targeted interventions. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a method for constructing a cardiology rehabilitation assessment model.

[0006] In order to achieve the above-mentioned object, the present invention adopts the following technical solution: a method for constructing a cardiology rehabilitation assessment model, comprising the following steps:

[0007] S1: Through continuous monitoring, the patient's continuous systolic blood pressure peak and diastolic blood pressure trough values ​​are obtained in each cycle after surgery, and the corresponding duration and cardiac waveform vertex sequence are detected to identify the cardiac compression band cycle and construct a waveform structure parameter set;

[0008] S2: using a moving average method to calculate the average amplitude change rate of the cardiac waveform vertex sequence in the waveform structure parameter set, comparing the rate value of the previous cycle and calculating the amplitude difference, determining whether it exceeds the amplitude threshold, marking the morphological abnormality cycle and calculating the difference between the maximum amplitude and the mean, and if the difference continuously increases, marking the structural drift area;

[0009] S3: Based on the structural drift region, the active phase and passive phase of the heart are divided, the number of heart peaks, peak spacing and duration are extracted, and the difference in the number of vertices, the difference in average spacing and the difference in duration ratio are calculated using time series difference analysis to generate a three-dimensional circadian rhythm difference vector of the heart;

[0010] S4: Based on the three-dimensional cardiac circadian rhythm difference vector, the deviation amount, the average daily change amount, and the number of times the deviation falls within the daytime tolerance range for three consecutive days are input into a multiple linear regression model, and the model coefficients are obtained by the minimum residual fitting method to generate a cardiac rhythm regulation score;

[0011] S5: Based on the cardiac rhythm regulation score, a trend analysis algorithm is used to fit the three-day score sequence, the rhythm slope value is calculated and the direction of change is determined. If the slope is greater than zero, it is marked as an enhanced state; if the slope is equal to zero, it is marked as a stable state; if the slope is less than zero, it is marked as an unbalanced state, and the state classification result is optimized.

[0012] As a further solution of the present invention, the waveform structure parameter set includes the peak systolic pressure, the diastolic pressure trough, the duration, and the waveform vertex sequence; the structural drift area is specifically the area of ​​continuously increasing difference and the period of morphological abnormality; the three-dimensional circadian rhythm difference vector of the heart includes the difference in the number of vertices, the difference in average spacing, and the difference in duration proportion; the cardiac rhythm regulation score is specifically the deviation amount, the average daily change amount, and the number of hits in the daytime tolerance interval; the state classification results include enhanced state, stable state, and unbalanced state.

[0013] As a further solution of the present invention, the specific steps of continuously monitoring and obtaining the continuous systolic pressure peak and diastolic pressure trough values ​​of each cycle of the postoperative patient, detecting the corresponding duration and the cardiac waveform vertex sequence, identifying the cardiac compression band cycle and constructing the waveform structure parameter set are as follows:

[0014] S101: Continuously monitoring the patient's heart to collect peak systolic blood pressure and trough diastolic blood pressure values ​​during each cycle after surgery, using adjacent systolic and diastolic blood pressure time points to divide the complete cardiac cycle and generate a cardiac cycle data set;

[0015] S102: Calling the cardiac cycle data set, detecting the duration of each cycle, extracting all vertices in the waveform within the cycle that are higher than the diastolic pressure trough value, calculating the time interval and amplitude difference between adjacent vertices, recording the number of cardiac waveform vertices, vertex amplitudes and corresponding time series, and generating a cardiac waveform feature set;

[0016] S103: Based on the cardiac waveform feature set, filter out cycles whose duration is lower than the cardiac compression time threshold, extract the number of vertices, vertex amplitudes and timestamp data in the corresponding cycle, and generate a waveform structure parameter set.

[0017] As a further solution of the present invention, a moving average method is used to calculate the average amplitude change rate of the cardiac waveform vertex sequence in the waveform structure parameter set, and the rate value of the previous cycle is compared and the amplitude difference is calculated to determine whether it exceeds the amplitude threshold. The abnormal morphology period is marked and the difference between the maximum amplitude and the mean is calculated. If the difference continuously increases, the structural drift area is marked. The specific steps are as follows:

[0018] S201: Acquire a cardiac waveform vertex sequence in the waveform structure parameter set, calculate the average amplitude change rate of the waveform vertex in each cycle using a moving average method, and calculate the difference between the change rate and the change rate of the previous cycle to obtain an amplitude change rate difference;

[0019] S202: The amplitude change rate difference is called. If the rate difference of a single cycle exceeds the set amplitude threshold, it is determined to be a morphological abnormality cycle. At the same time, the maximum amplitude value and the average amplitude value of each vertex in the cycle are collected, and the difference between the maximum amplitude and the average is calculated. It is determined whether the difference is continuously increasing, and a structural drift determination result is generated.

[0020] The amplitude threshold is set by taking the standard deviation of the average signal amplitude plus or minus a multiple of 1.5 to 2.5 as the recognition boundary;

[0021] S203: calling the structural drift determination result and the amplitude difference, extracting the periodic sequence marked as abnormal and judging the time continuity, and marking the structural drift area by associating the time index mapping interval with the position segment in the ECG waveform.

[0022] As a further solution of the present invention, the active phase and passive phase of the heart are divided based on the structural drift region, the number of heart peaks, peak spacing, and duration are extracted, and the vertex number difference, average spacing difference, and duration ratio difference are calculated using time series difference analysis to generate the heart's three-dimensional circadian rhythm difference vector. The specific steps are as follows:

[0023] S301: Dividing the cardiac waveform sequence into active phases and passive phases based on the structural drift region, extracting phase distribution segments of continuous peaks within the region, marking phase boundaries according to the timing positions of the peaks within the segments, extracting the number of peaks within the phase, the spacing between adjacent peaks, and the duration of each phase segment, performing Z-score normalization processing, and generating phase timing features;

[0024] S302: calling the phase time series features, comparing the difference in the number of peaks, the average difference in the peak spacing, and the difference in the duration ratio between the two phases in the same time segment, and obtaining a time series difference factor group;

[0025] S303: According to the three types of difference parameters in the timing difference factor group, the difference in the number of vertices, the difference in the peak spacing and the difference in the time proportion are integrated, and a three-dimensional vector structure is established according to the arrangement order of the active phase and the passive phase. After encoding to form a sequence of difference items and unifying the dimensions, a three-dimensional circadian rhythm difference vector of the heart is generated.

[0026] As a further embodiment of the present invention, based on the three-dimensional cardiac circadian rhythm difference vector, the deviation amount, the average daily change amount, and the number of times the deviation falls within the daytime tolerance range for three consecutive days are input into a multiple linear regression model, and the model coefficients are obtained by the minimum residual fitting method to generate a cardiac rhythm regulation score. The specific steps are as follows:

[0027] S401: extracting the three-day deviation of the cardiac three-dimensional circadian rhythm difference vector, collecting the difference amplitude change value between two adjacent days and calculating the change amount within three days, and simultaneously counting the cumulative number of times the deviation value falls within the set daytime tolerance interval within three days to generate a rhythm difference daily series indicator group;

[0028] The quantitative standard of the daytime tolerance interval is set by ±1.5 times the standard deviation of the previous mean of the multiple difference vector components in the resting state;

[0029] S402: Calling the rhythm difference daily series indicator group to input the multiple linear regression model, setting the regression response variable corresponding to the change of the cardiac rhythm state, and constructing the functional relationship between the deviation input and the rhythm response through the minimum residual fitting process to obtain the rhythm control function result;

[0030] The multiple linear regression model includes a target output term, an input variable term, a constant term, a coefficient term, and a residual term;

[0031] S403: The rhythm control function result is processed in a unified output format, the result type is marked and the value range is standardized, and the stable output item in the function result is extracted as the rhythm state evaluation value to generate a cardiac rhythm control score.

[0032] As a further solution of the present invention, the change in three days is calculated using the formula:

[0033]

[0034] Where H is the trend parameter, Δd1 represents the amplitude change of the heart three-dimensional circadian rhythm difference vector between the first and second days, Δd2 represents the amplitude change of the heart three-dimensional circadian rhythm difference vector between the second and third days, ω1 represents the ratio parameter of the standard deviation of the three-day deviation under the square root to the mean, ω2 represents the ratio parameter of the range of the three-day deviation to the median, and η adj Represents the logarithm of the difference between adjacent days, and η adj =N-1, where N is the total number of days.

[0035] As a further embodiment of the present invention, based on the cardiac rhythm control score, a trend analysis algorithm is used to fit the three-day score sequence, calculate the rhythm slope value, and determine the direction of change. If the slope is greater than zero, it is marked as an enhanced state; if the slope is equal to zero, it is marked as a stable state; if the slope is less than zero, it is marked as an unbalanced state. The specific steps for optimizing the state classification result are as follows:

[0036] S501: Based on the cardiac rhythm control score, extract daily score values ​​in chronological order and construct a sequence, fit the score change trend through a trend analysis algorithm, perform linear trend modeling with time as the independent variable and score as the dependent variable, and generate a rhythm trend slope;

[0037] S502: calling the rhythm trend slope, judging the direction of change based on the positive and negative relationship of the slope, determining if the slope is greater than 0.05 as enhancement and marking it as an enhancement state, if the slope is zero, marking it as a stable state, and if the slope is less than zero, marking it as an imbalance state, and obtaining a rhythm state determination value;

[0038] S503: Based on the rhythm state determination value, the consistency between the current state mark and the previous state label is compared in combination with the change in the score trend in the past three days. If there is inconsistency, the current state mark is corrected, the result is updated and a structured output field is generated to optimize the state classification result.

[0039] As a further solution of the present invention, the trend analysis algorithm is used to fit the score change trend, using the formula:

[0040]

[0041] Where K is the trend correction coefficient, λ j Represents the time decay weight factor of day j, taking λ j =1 / (1+ln(j+1)), where j represents the time series index, ln represents the natural logarithm function, j+1 represents the index offset, and ρ j represents the cardiac rhythm control score value on day j, t j Represents the time series coded value of day j.

[0042] Compared with the prior art, the advantages and positive effects of the present invention are:

[0043] In the present invention, by continuously monitoring the dynamic changes of the peak value of cardiac systolic pressure and the valley value of diastolic pressure, combined with the duration and period identification of the waveform vertex sequence, a waveform structure parameter set is constructed to dynamically capture the period of the cardiac compression band, breaking through the limitations of traditional static parameter analysis. The moving average method calculates the average amplitude change rate of the waveform vertex sequence, combined with the dynamic comparison of the amplitude difference and the amplitude threshold, to enhance the sensitivity to the abnormal morphological period, and through the judgment mechanism of continuously increasing the difference, locates the structural drift area, and improves the early warning ability of abnormal signals. Based on the division of active phase and passive phase, multi-dimensional features such as the number of peaks, spacing and duration are extracted, and the circadian rhythm difference vector is generated in combination with the time series difference analysis. The circadian fluctuation characteristics of the cardiac rhythm are synchronously quantified from the spatial distribution and time dimensions, making up for the shortcomings of traditional single-dimensional rhythm evaluation. The multivariate linear regression model integrates the deviation, average daily change, and number of hits in the daily tolerance range for three consecutive days, generates a rhythm regulation score through coefficient weighting, and establishes a dynamic and multidimensional evaluation system to improve the individualized adaptability of rehabilitation assessment and the reliability of long-term trend prediction, providing a more scientific quantitative basis for clinical decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the main steps of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0046] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0047] See also Figure 1 The present invention provides a technical solution: a method for constructing a cardiology rehabilitation assessment model, comprising the following steps:

[0048] S1: Through continuous monitoring, the patient's continuous systolic blood pressure peak and diastolic blood pressure trough values ​​are obtained in each cycle after surgery, and the corresponding duration and cardiac waveform vertex sequence are detected to identify the cardiac compression band cycle and construct a waveform structure parameter set;

[0049] S2: Use the moving average method to calculate the average amplitude change rate of the cardiac waveform vertex sequence in the waveform structure parameter set, compare it with the rate value of the previous cycle and calculate the amplitude difference, determine whether it exceeds the amplitude threshold, mark the morphological abnormality cycle and calculate the difference between the maximum amplitude and the mean, and if the difference continues to increase, mark the structural drift area;

[0050] S3: Based on the structural drift area, the active phase and passive phase of the heart are divided, the number of heart peaks, peak spacing and duration are extracted, and the difference in the number of vertices, the average spacing difference and the difference in duration ratio are calculated using time series difference analysis to generate the three-dimensional circadian rhythm difference vector of the heart;

[0051] S4: Based on the three-dimensional circadian rhythm difference vector of the heart, the deviation amount, daily average change amount and the number of times the deviation falls into the daytime tolerance range for three consecutive days are input into the multivariate linear regression model. The model coefficients are obtained by the minimum residual fitting method to generate the cardiac rhythm regulation score;

[0052] S5: Based on the cardiac rhythm regulation score, a trend analysis algorithm is used to fit the three-day score sequence, calculate the rhythm slope value, and determine the direction of change. If the slope is greater than zero, it is marked as an enhanced state; if the slope is equal to zero, it is marked as a stable state; if the slope is less than zero, it is marked as an unbalanced state, and the state classification results are optimized.

[0053] The waveform structure parameter set includes the peak systolic pressure, diastolic pressure trough, duration, and waveform vertex sequence. The structural drift area is specifically the area of ​​continuously increasing difference and the period of morphological abnormality. The three-dimensional circadian rhythm difference vector of the heart includes the difference in the number of vertices, the difference in average spacing, and the difference in duration ratio. The cardiac rhythm regulation score is specifically the deviation amount, the average daily change, and the number of hits in the daytime tolerance interval. The state classification results include enhanced state, stable state, and unbalanced state.

[0054] See also Figure 1 The present invention provides a technical solution: a method for constructing a cardiology rehabilitation assessment model, comprising the following steps:

[0055] S101: Continuously monitoring the patient's heart to collect peak systolic blood pressure and trough diastolic blood pressure values ​​during each cycle after surgery, using adjacent systolic and diastolic blood pressure time points to divide the complete cardiac cycle and generate a cardiac cycle data set;

[0056] Postoperative recovery data were collected using a non-invasive blood pressure monitor. The sensor was attached to the patient's upper arm, and the sampling frequency was set to 100 Hz. 24-hour blood pressure waveforms were recorded continuously. When the rising slope of the systolic pressure waveform exceeded 5 mmHg / ms, it was marked as a peak point. The corresponding diastolic pressure trough point was determined by the lowest pressure value in the adjacent trough. For three consecutive cycles, the patient's systolic pressure peaks were 122 mmHg, 119 mmHg, and 125 mmHg, respectively, and the corresponding diastolic pressure troughs were 78 mmHg, 74 mmHg, and 80 mmHg. The cycles were divided by adjacent peak-trough time points. When the cycle start time was detected as 10:05:23.450 and the end time was 10:05:24.310, the cycle duration was calculated to be 860 ms, forming a cardiac cycle data set containing three columns of data: timestamp, systolic pressure, and diastolic pressure, as shown in Table 1.

[0057] Table 1 Example of cardiac cycle data table

[0058]

[0059]

[0060] As shown in Table 1, the system records the timestamp starting point and physiological parameters of each cycle. When the systolic blood pressure peak exceeds 140 mmHg five times consecutively within a cycle, an abnormal flag is triggered. Each entry in the dataset includes 12 metadata fields, with the timestamp accuracy reaching the millisecond level. The blood pressure value is quantized into a 16-bit digital signal through an AD converter.

[0061] S102: Calling the cardiac cycle data set, detecting the duration of each cycle, extracting all vertices in the waveform within the cycle that are higher than the diastolic pressure trough value, calculating the time interval and amplitude difference between adjacent vertices, recording the number of cardiac waveform vertices, vertex amplitudes and corresponding time series, and generating a cardiac waveform feature set;

[0062] The data set in Table 1 was called, and the sliding window method was used to calculate the cycle duration. The window width was set to 3 cycles. When the standard deviation of the cycle duration within the window was detected to be more than 50ms, an abnormality was marked. When extracting the waveform vertices, the amplitude threshold was set to the diastolic pressure trough value + 5mmHg, including when the cycle diastolic pressure trough value was 80mmHg, and the peak points with an amplitude higher than 85mmHg were screened. When three vertices were detected, namely 89mmHg@10:05:23.500, 92mmHg@10:05:23.750, and 88mmHg@10:05:24.100, the time intervals between adjacent vertices were calculated to be 250ms and 350ms, respectively, and the amplitude differences were +3mmHg and -4mmHg, respectively. The number of vertices was recorded as 3, forming a cardiac waveform feature set including time series and amplitude series.

[0063] S103: Based on the cardiac waveform feature set, screening cycles whose duration is less than the cardiac compression time threshold, extracting the number of vertices, vertex amplitudes, and timestamp data within the corresponding cycle, and generating a waveform structure parameter set;

[0064] The cardiac compression time threshold was set to 800 ms, and the feature set was traversed to filter entries with a cycle length ≤ 800 ms. When the cycle length was 790 ms, four vertices were extracted, the amplitude sequence was [91, 95, 93, 90] mmHg, and the timestamp sequence was [0 ms, 210 ms, 430 ms, 650 ms]. When generating the waveform structure parameter set, the vertex density was calculated to be 4 / 0.79s = 5.06 / second, the amplitude coefficient of variation was (standard deviation 2.06) / (mean 92.25) = 2.23%, and the time interval mean was 650 ms / 3 = 216.67 ms. When the vertex density exceeded 6 / second for three consecutive cycles, it was marked as a high-frequency abnormal mode. The parameter set storage format includes 12 fields, including the cycle number, the filter flag, and the vertex tuple array. The timestamp data is converted to the offset of the cycle starting point for storage.

[0065] See also Figure 1 The present invention provides a technical solution: a method for constructing a cardiology rehabilitation assessment model, comprising the following steps:

[0066] S201: Obtaining a cardiac waveform vertex sequence in a waveform structure parameter set, calculating an average amplitude change rate of the waveform vertex in each cycle using a moving average method, and performing a difference calculation based on the change rate and the change rate of the previous cycle to obtain an amplitude change rate difference;

[0067] Based on the vertex sequence of the waveform structure parameter set, the moving average method with a window width of 5 is used, where k represents the current calculation point index, Δp k；1 Indicates the amplitude difference of the previous vertex (such as +3mmHg), Δp k is the current amplitude difference (such as +4mmHg), Δp k:1 For subsequent amplitude differences (e.g. -2 mmHg), calculate the average rate of change When k = 2, the value is R2 = (3 + 4 - 2) / 3 = 1.67 mmHg / cycle, compared with the rate value of the previous cycle R prev =1.33mmHg / cycle, calculate the difference value ΔR=|R k -R prev |=0.34mmHg / cycle. When the difference value of three consecutive cycles exceeds 0.3, an abnormal flag is generated. Data storage uses floating point precision of 0.01mmHg / cycle.

[0068] Table 2 Amplitude change rate difference table

[0069]

[0070]

[0071] As shown in Table 2, the field "rate difference" is obtained through ΔR calculation. When the C0241 cycle difference of 0.34 exceeds the threshold of 0.3, the abnormal code 0x01 is marked in the database.

[0072] S202: Calling the amplitude change rate difference, if the rate difference of a single cycle exceeds the set amplitude threshold, it is determined to be a morphological abnormality cycle, and at the same time collecting the maximum amplitude value and average amplitude value of each vertex in the cycle, calculating the difference between the maximum amplitude and the average value, and determining whether the difference is continuously increasing, to generate a structural drift determination result;

[0073] The amplitude threshold is set by taking the average amplitude of the signal plus or minus the standard deviation between 1.5 and 2.5 times as the recognition boundary;

[0074] The amplitude threshold T is calculated by T = μ + 1.5σ, where μ is the mean of the rate difference of the last 10 cycles (e.g., 0.28 mmHg / cycle) and σ is the standard deviation (e.g., 0.12 mmHg / cycle). Substituting into T = 0.28 + 1.5 × 0.12 = 0.46 mmHg / cycle, when the rate difference ΔR of cycle C0250 is detected to be 0.52, an abnormality is triggered because 0.52 > 0.46. The peak amplitude sequence of this cycle [89, 92, 95, 91] mmHg is extracted and the mean is calculated. Maximum amplitude p max =95mmHg, difference Δp max =95-91.75=3.25mmHg, three consecutive period difference sequence [2.1,2.8,3.25]mmHg, calculate the slope When s>0.3, the structure drift code 0xA1 is marked.

[0075] S203: Calling the structural drift determination result and the amplitude difference, extracting the periodic sequence marked as abnormal and judging the time continuity, and marking the structural drift area by associating the time index mapping interval with the position segment in the ECG waveform.

[0076] Extract the abnormal cycle number [C0245, C0246, C0247], the timestamp sequence [10:15:33.200, 10:15:33.950, 10:15:34.700], calculate the adjacent intervals Δt1 = 750ms, Δt2 = 750ms, and judge the continuity condition Δt ≤ 800ms. If it is met, generate the continuous event ID #20240522-003. When mapping to the ECG waveform R wave index, use the index calculation formula When the base time epoch = 10:00:00.000, the start time of C0245 corresponds to Index = 3320, the end index Index = 3470, and the storage drift area is recorded as [3320, 3470, 0xA3].

[0077] See also Figure 1 The present invention provides a technical solution: a method for constructing a cardiology rehabilitation assessment model, comprising the following steps:

[0078] S301: Divide the cardiac waveform sequence into active and passive phases based on the structural drift region, extract the phase distribution segments of continuous peaks within the region, mark the phase boundaries according to the temporal position of the peaks within the segment, extract the number of peaks within the phase, the spacing between adjacent peaks, and the duration of each phase segment, perform Z-score normalization processing, and generate phase timing features;

[0079] Based on the structural drift region index [3320,3470], extract all peak timestamp sequences in the segment

[0080] [3325,3340,3355,3370,3385]ms, calculate the adjacent peak intervals as 15ms, 15ms, 15ms, 15ms, determine that the active phase starts at 3325ms, ends at 3385ms, lasts for 60ms, and includes 5 peaks, and the passive phase starts at 3400ms, ends at 3450ms, lasts for 50ms, and includes 4 peaks, and perform Z-score standardization on the number of peaks. The formula is: The number of active phase peaks X1 = 5, the number of passive phase peaks X2 = 4, the mean μ of the entire data set = 4.2, the standard deviation σ = 0.84, and the calculated value Z1 = (5-4.2) / 0.84 = 0.95.

[0081] Z2=(4-4.2) / 0.84=-0.24. When the interval is standardized, the average interval of the active phase is 15ms, the mean of the entire data set is 14.5ms, and the standard deviation is 1.2ms. Z spacing =(15-14.5) / 1.2=0.42, and the generated phase timing characteristics include the normalized number of peaks, spacing value and duration, as shown in Table 3.

[0082] Table 3 Phase timing characteristics

[0083] Phase Type Normalized peak number Normalized spacing values Normalized duration Active phase 0.95 0.42 1.12 Passive phase -0.24 -0.17 0.89

[0084] As shown in Table 3, when calculating the standardized duration, the active phase duration of 60ms is compared with the mean of 53ms and the standard deviation of 6.2ms, and Z duration =(60-53) / 6.2=1.12. Data is stored in single-precision floating-point format with two decimal places.

[0085] S302: Calling the phase timing characteristics, comparing the difference in the number of peaks, the average difference in the peak spacing, and the difference in the duration ratio between the two phases in the same time segment, to obtain a timing difference factor group;

[0086] Comparing the two-phase data in Table 3, the peak number difference ΔN is calculated as 0.95-(-0.24)=1.19. The average peak spacing difference is the absolute difference ΔS=|0.42-(-0.17)|=0.59, which is the difference between the active phase 0.42 and the passive phase -0.17. The duration ratio difference is calculated as 60 / (60+50)=54.5% for the active phase and 45.5% for the passive phase, resulting in ΔT=54.5%-45.5%=9.0%. The timing difference factor group (1.19, 0.59, 9.0%) is generated. The percentage difference is converted to decimal form 0.09 for calculation. Fixed-point encoding is used for data conversion with an accuracy of 0.0001.

[0087] S303: Based on the three types of difference parameters in the timing difference factor group, the vertex number difference, the peak spacing difference, and the time proportion difference are integrated, and a three-dimensional vector structure is established according to the arrangement order of the active phase and the passive phase. After encoding the difference item sequence and unifying the dimensions, a three-dimensional cardiac circadian rhythm difference vector is generated;

[0088] Construct a three-dimensional vector by the difference factor group (1.19, 0.59, 0.09) in the order of active phase priority Normalize the vector and calculate the modulus Get the unit vector 8-bit quantization is used during encoding to map multiple components to the range of 0-255. The calculation formula is: This includes converting 0.895 to round(255×(0.895+1) / 2)=round(255×0.9475)=242, generating a difference term sequence [242,185,134]. When stored as a three-dimensional vector, the little-endian byte order is used, occupying 3 bytes of storage space.

[0089] See also Figure 1 The present invention provides a technical solution: a method for constructing a cardiology rehabilitation assessment model, comprising the following steps:

[0090] S401: extracting the three-day deviation of the heart's three-dimensional circadian rhythm difference vector, collecting the difference amplitude change value between two adjacent days and calculating the change amount within three days, and simultaneously counting the cumulative number of times the deviation value falls within the set daytime tolerance interval within three days to generate a rhythm difference daily series indicator group;

[0091] The quantitative standard of the daytime tolerance interval was set by ±1.5 times the standard deviation of the previous mean of the multiple difference vector components in the resting state;

[0092] Calculate the daily vector modulus based on the three-day difference vector sequence [242,185,134], [238,192,141], [235,189,138] Get the modulus length sequence [1.33, 1.34, 1.32], calculate the adjacent daily changes Δd1 = |1.34-1.33| = 0.01, Δd2 = |1.32-1.34| = 0.02, and count the number of times the three-day deviation value [0.01, 0.02] falls into the daily tolerance interval [0.005, 0.03] is 2. When calculating the trend parameter H, first find Where the standard deviation σ=0.0055, the mean μ=0.015, and we get ω1=0.0055 / 0.015≈0.367. Range = 0.02-0.01 = 0.01, median 0.015, ω2 = 0.01 / 0.015 ≈ 0.667, adjustment coefficient η adj =3-1=2, substitute into the formula The indicator group for the generated rhythm difference daily series includes H value 0.088 and tolerance hit number 2, as shown in Table 4.

[0093] Table 4 Daily sequence indicators of rhythm differences

[0094]

[0095]

[0096] As shown in Table 4, the H value is calculated using the formula. When H < 0.1, it is marked as a stable state. Data storage uses double-precision floating point type, and the timestamp is accurate to the daily level.

[0097] S402: Calling the rhythm difference daily series indicator group to input the multiple linear regression model, setting the regression response variable corresponding to the change of the cardiac rhythm state, and constructing the functional relationship between the deviation input and the rhythm response through the minimum residual fitting process to obtain the rhythm control function result;

[0098] The multiple linear regression model includes target output term, input variable term, constant term, coefficient term, and residual term;

[0099] A multiple linear regression model Y = β0 + β1X1 + β2X2 + ∈ was constructed, where the response variable Y was the clinical assessment score (ranging from 70 to 100), the input variable X1 was the trend parameter H, and X2 was the number of tolerance hits. The three-day data in Table 4 [(0.088, 2, 85), (0.095, 3, 88)] were used for fitting, and the regression coefficients β0 = 72.3, β1 = 15.6, and β2 = 4.2 were calculated. The residual ∈ = 0.7, and the functional relationship Y = 72.3 + 15.6X1 + 4.2X2 was obtained. When the new data X1 = 0.09 and X2 = 2 were substituted, the calculation result was Y = 72.3 + 15.6 × 0.09 + 4.2 × 2 = 72.3 + 1.404 + 8.4 = 82.104, which was rounded to 82 points. The model fit R 2 =0.92, the least squares method is used for parameter estimation, and the matrix operation accuracy is kept to 10 decimal places.

[0100] S403: Unifying the output format of the rhythm control function results, marking the result type and standardizing the value range, extracting the stable output items in the function results as rhythm state assessment values, and generating a cardiac rhythm control score;

[0101] The output value 82.104 of the rhythm control function was formatted, and the value range was set to [0,100]. When the calculated result exceeded 100, the truncation function Y′=min(max(Y,0),100) was used, including taking Y′=100 when Y=105.6 and taking Y′=0 when Y=-5.3. The result was rounded to 82 points. When marking the result type, the score interval was set to [0,60] for high risk, 60,80 for concern, and 80,100 for normal. When the score was 82, it was marked as normal. The stable term β0=72.3 in the regression model was extracted as the baseline assessment value. The difference between the score on the day and the baseline Δ=82-72.3=9.7 was calculated. The difference threshold is T = 10. When |Δ| ≥ T, an abnormal fluctuation mark is added. The data is stored in JSON structured format, including three fields: score, baseline value, and status mark, including {"score": 82, "baseline": 72.3, "status": "normal"}. At the same time, a CRC-8 checksum 0x5B is added to ensure data integrity. The field encoding rules are as follows: score occupies 1 byte (0-100 is mapped to 0x00-0x64), baseline occupies 2 bytes (floating point number × 100 rounded), status occupies 1 byte (0x00: high risk, 0x01: attention, 0x02: normal), and the checksum is calculated by the polynomial x 8 +x 2 +x+1 is calculated and generated, and the total length of the data packet is fixed at 5 bytes.

[0102] See also Figure 1The present invention provides a technical solution: a method for constructing a cardiology rehabilitation assessment model, comprising the following steps:

[0103] S501: Based on the cardiac rhythm regulation score, daily score values ​​are extracted in chronological order and a sequence is constructed. The score change trend is fitted using a trend analysis algorithm. A linear trend model is performed with time as the independent variable and the score as the dependent variable to generate a rhythm trend slope.

[0104] Extract the rhythm scores of 5 consecutive days [82,85,88,90,87] to construct a time series, and code the time t j Pick

[0105] [1,2,3,4,5], calculate the daily decay weight λ j =1 / (1+ln(j+1)), when j=1, λ1=

[0106] 1 / (1+ln2)≈0.590, when j=2, λ2=1 / (1+ln3)≈0.461, calculate the daily score change Including when j=2 When j=3 Numerator sum calculation: Denominator calculation: Where K is the trend correction coefficient, λ j Represents the time decay weight factor of the jth day, taking λ j =1 / (1+

[0107] ln(j+1)), where j represents the time series index, ln represents the natural logarithm function, j+1 represents the index offset, and ρ j represents the cardiac rhythm control score value on day j, t j Representing the time series code value of day j, the trend correction coefficient K = 6.333 / 5.477≈1.156 is obtained, generating the rhythm trend slope 1.156, as shown in Table 5.

[0108] Table 5 Trend analysis calculation parameters

[0109] Date Index score Time Code Decay Weight Square root of change Weighted calculation value 1 82 1 0.590 1.414 0.834 2 85 2 0.461 1.732 1.603 3 88 3 0.385 1.732 2.010 4 90 4 0.333 1.414 1.886

[0110] As shown in Table 5, the sum of the numerators is 6.333, the denominator is 5.477, the trend slope K=1.156, and the data is stored with three decimal places.

[0111] S502: Call the rhythm trend slope and determine the change direction based on the positive and negative relationship of the slope. If the slope is greater than 0.05, it is determined to be enhanced and marked as an enhanced state. If the slope is equal to zero, it is marked as a stable state. If the slope is less than zero, it is marked as an unbalanced state. The rhythm state determination value is obtained;

[0112] The slope threshold is set to 0.05. When K = 1.156 > 0.05, it is marked as an enhanced state. When K = -0.03 < 0, it is marked as an unbalanced state. When K = 0.02∈(-0.05, 0.05), it is marked as a stable state. The state coding rule is: enhanced state 0x01, stable state 0x00, unbalanced state 0xFF. When the data is updated, it is compared with the previous day's state mark. If the current mark 0x01 is inconsistent with the previous day's 0x00, the state correction flag is triggered and the number of corrections is recorded in the log file. The threshold of 0.05 is set according to clinical data statistics to cover 90% of the normal fluctuation range.

[0113] S503: Based on the rhythm state determination value and the score trend change over the past three days, the consistency of the current state mark and the previous state label is compared. If there is inconsistency, the current state mark is corrected, the result is updated, and a structured output field is generated to optimize the state classification result;

[0114] Extract the state sequence of the last three days [0x01, 0x01, 0xFF], detect that the state of the third day 0xFF is inconsistent with the state of the previous two days 0x01, and calculate the trend continuity index When C < 0.75, the correction process is initiated. The K values ​​for the previous three days are recalculated to [1.156, 1.043, 0.987], with an average slope of K = 1.062. Based on K > 0.05, the status for the third day is corrected to 0x01, generating the structured output fields {"trend": "↑stability": 0.667, "history": [1.156, 1.043, 0.987]}. The timestamp 20240522T0845Z is added for data persistence, and compressed storage is performed using the MsgPack binary format.

[0115] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for constructing a cardiology rehabilitation assessment model, characterized in that: The following steps are involved: S1: Through continuous monitoring, the patient's continuous systolic blood pressure peak and diastolic blood pressure trough values ​​are obtained in each cycle after surgery, and the corresponding duration and cardiac waveform vertex sequence are detected to identify the cardiac compression band cycle and construct a waveform structure parameter set; S2: using a moving average method to calculate the average amplitude change rate of the cardiac waveform vertex sequence in the waveform structure parameter set, comparing the rate value of the previous cycle and calculating the amplitude difference, determining whether it exceeds the amplitude threshold, marking the morphological abnormality cycle and calculating the difference between the maximum amplitude and the mean, and if the difference continuously increases, marking the structural drift area; S3: Based on the structural drift region, the active phase and passive phase of the heart are divided, the number of heart peaks, peak spacing and duration are extracted, and the difference in the number of vertices, the difference in average spacing and the difference in duration ratio are calculated using time series difference analysis to generate a three-dimensional circadian rhythm difference vector of the heart; S4: Based on the three-dimensional cardiac circadian rhythm difference vector, the deviation amount for three consecutive days, the average daily change amount and the number of times the deviation falls into the daytime tolerance range are input into the multivariate linear regression model, and the model coefficients are obtained by the minimum residual fitting method to generate a cardiac rhythm regulation score.

2. The method for constructing a cardiology rehabilitation assessment model according to claim 1, characterized in that: The waveform structure parameter set includes the peak systolic pressure, the diastolic pressure trough, the duration, and the waveform vertex sequence. The structural drift area is specifically the area of ​​continuously increasing difference and the period of morphological abnormality. The three-dimensional circadian rhythm difference vector of the heart includes the difference in the number of vertices, the difference in average spacing, and the difference in duration ratio. The cardiac rhythm regulation score is specifically the deviation amount, the average daily change amount, and the number of hits in the daytime tolerance interval.

3. The method for constructing a cardiology rehabilitation assessment model according to claim 1, characterized in that: The specific steps for obtaining the continuous systolic pressure peak and diastolic pressure trough values ​​of each cycle of the postoperative patient through continuous monitoring, detecting the corresponding duration and cardiac waveform vertex sequence, identifying the cardiac compression band cycle and constructing the waveform structure parameter set are as follows: S101: Continuously monitoring the patient's heart to collect peak systolic blood pressure and trough diastolic blood pressure values ​​during each cycle after surgery, using adjacent systolic and diastolic blood pressure time points to divide the complete cardiac cycle and generate a cardiac cycle data set; S102: Calling the cardiac cycle data set, detecting the duration of each cycle, extracting all vertices in the waveform within the cycle that are higher than the diastolic pressure trough value, calculating the time interval and amplitude difference between adjacent vertices, recording the number of cardiac waveform vertices, vertex amplitudes and corresponding time series, and generating a cardiac waveform feature set; S103: Based on the cardiac waveform feature set, filter out cycles whose duration is lower than the cardiac compression time threshold, extract the number of vertices, vertex amplitudes and timestamp data in the corresponding cycle, and generate a waveform structure parameter set.

4. The method for constructing a cardiology rehabilitation assessment model according to claim 1, characterized in that: The moving average method is used to calculate the average amplitude change rate of the cardiac waveform vertex sequence in the waveform structure parameter set, and the rate value of the previous cycle is compared and the amplitude difference is calculated to determine whether it exceeds the amplitude threshold. The abnormal morphology cycle is marked and the difference between the maximum amplitude and the mean is calculated. If the difference continuously increases, the structural drift area is marked. The specific steps are as follows: S201: Acquire a cardiac waveform vertex sequence in the waveform structure parameter set, calculate the average amplitude change rate of the waveform vertex in each cycle using a moving average method, and calculate the difference between the change rate and the change rate of the previous cycle to obtain an amplitude change rate difference; S202: The amplitude change rate difference is called. If the rate difference of a single cycle exceeds the set amplitude threshold, it is determined to be a morphological abnormality cycle. At the same time, the maximum amplitude value and the average amplitude value of each vertex in the cycle are collected, and the difference between the maximum amplitude and the average is calculated. It is determined whether the difference is continuously increasing, and a structural drift determination result is generated. The amplitude threshold is set by taking the standard deviation of the average signal amplitude plus or minus a multiple of 1.5 to 2.5 as the recognition boundary; S203: calling the structural drift determination result and the amplitude difference, extracting the periodic sequence marked as abnormal and judging the time continuity, and marking the structural drift area by associating the time index mapping interval with the position segment in the ECG waveform.

5. The method for constructing a cardiology rehabilitation assessment model according to claim 1, characterized in that: The steps for dividing the heart's active and passive phases based on the structural drift region, extracting the number of heart peaks, peak spacing, and duration, and calculating the vertex number difference, average spacing difference, and duration ratio difference using time series difference analysis to generate a three-dimensional heart circadian rhythm difference vector are as follows: S301: Dividing the cardiac waveform sequence into active phases and passive phases based on the structural drift region, extracting phase distribution segments of continuous peaks within the region, marking phase boundaries according to the timing positions of the peaks within the segments, extracting the number of peaks within the phase, the spacing between adjacent peaks, and the duration of each phase segment, performing Z-score normalization processing, and generating phase timing features; S302: calling the phase time series features, comparing the difference in the number of peaks, the average difference in the peak spacing, and the difference in the duration ratio between the two phases in the same time segment, and obtaining a time series difference factor group; S303: According to the three types of difference parameters in the timing difference factor group, the difference in the number of vertices, the difference in the peak spacing and the difference in the time proportion are integrated, and a three-dimensional vector structure is established according to the arrangement order of the active phase and the passive phase. After encoding to form a sequence of difference items and unifying the dimensions, a three-dimensional circadian rhythm difference vector of the heart is generated.

6. The method for constructing a cardiology rehabilitation assessment model according to claim 1, characterized in that: Based on the three-dimensional cardiac circadian rhythm difference vector, the deviation amount, daily average change amount, and the number of times the deviation falls into the daytime tolerance range for three consecutive days are input into a multiple linear regression model. The model coefficients are obtained by the minimum residual fitting method and the specific steps of generating the cardiac rhythm regulation score are as follows: S401: extracting the three-day deviation of the cardiac three-dimensional circadian rhythm difference vector, collecting the difference amplitude change value between two adjacent days and calculating the change amount within three days, and simultaneously counting the cumulative number of times the deviation value falls within the set daytime tolerance interval within three days to generate a rhythm difference daily series indicator group; The quantitative standard of the daytime tolerance interval is set by ±1.5 times the standard deviation of the previous mean of the multiple difference vector components in the resting state; S402: Calling the rhythm difference daily series indicator group to input the multiple linear regression model, setting the regression response variable corresponding to the change of the cardiac rhythm state, and constructing the functional relationship between the deviation input and the rhythm response through the minimum residual fitting process to obtain the rhythm control function result; The multiple linear regression model includes a target output term, an input variable term, a constant term, a coefficient term, and a residual term; S403: The rhythm control function result is processed in a unified output format, the result type is marked and the value range is standardized, and the stable output item in the function result is extracted as the rhythm state evaluation value to generate a cardiac rhythm control score.

7. The method for constructing a cardiology rehabilitation assessment model according to claim 6, characterized in that: The calculation of the change within three days uses the formula: Where H is the trend parameter, Δd1 represents the amplitude change of the heart three-dimensional circadian rhythm difference vector between the first and second days, Δd2 represents the amplitude change of the heart three-dimensional circadian rhythm difference vector between the second and third days, ω1 represents the ratio parameter of the standard deviation of the three-day deviation under the square root to the mean, ω2 represents the ratio parameter of the range of the three-day deviation to the median, and η adj Represents the logarithm of the difference between adjacent days, and η adj =N-1, where N is the total number of days.

8. The method for constructing a cardiology rehabilitation assessment model according to claim 1, characterized in that: The method further comprises: S5: Based on the cardiac rhythm control score, a trend analysis algorithm is used to fit the three-day score sequence, the rhythm slope value is calculated, and the direction of change is determined. If the slope is greater than zero, it is marked as an enhanced state; if the slope is equal to zero, it is marked as a stable state; if the slope is less than zero, it is marked as an unbalanced state, and the state classification result is optimized; The state classification results include enhanced state, stable state, and unbalanced state.

9. The method for constructing a cardiology rehabilitation assessment model according to claim 8, characterized in that: Based on the cardiac rhythm control score, a trend analysis algorithm is used to fit the three-day score sequence, calculate the rhythm slope value, and determine the direction of change. If the slope is greater than zero, it is marked as an enhanced state; if the slope is equal to zero, it is marked as a stable state; if the slope is less than zero, it is marked as an unbalanced state. The specific steps for optimizing the state classification results are as follows: S501: Based on the cardiac rhythm control score, extract daily score values ​​in chronological order and construct a sequence, fit the score change trend through a trend analysis algorithm, perform linear trend modeling with time as the independent variable and score as the dependent variable, and generate a rhythm trend slope; S502: calling the rhythm trend slope, judging the direction of change based on the positive and negative relationship of the slope, determining if the slope is greater than 0.05 as enhancement and marking it as an enhancement state, if the slope is zero, marking it as a stable state, and if the slope is less than zero, marking it as an imbalance state, and obtaining a rhythm state determination value; S503: Based on the rhythm state determination value, the consistency between the current state mark and the previous state label is compared in combination with the change in the score trend in the past three days. If there is inconsistency, the current state mark is corrected, the result is updated and a structured output field is generated to optimize the state classification result.

10. The method for constructing a cardiology rehabilitation assessment model according to claim 9, characterized in that: The trend analysis algorithm is used to fit the score change trend, using the formula: Where K is the trend correction coefficient, λ j Represents the time decay weight factor of the jth day, taking λ j =1 / (1+ln(j+1)), where j represents the time series index, ln represents the natural logarithm function, j+1 represents the index offset, and ρ j represents the cardiac rhythm control score value on day j, t j Represents the time series coded value of day j.

Images