A method, system and related device for electrocardiogram segment offset evaluation and output control

Abstract

The application discloses a method, system and related device for electrocardiogram segment offset evaluation and output control. The method comprises: signal quality screening, normality determination and state classification on historical electrocardiogram segments of a user, and construction of a multi-state individual historical baseline library; then, state features of a current segment are extracted, and a candidate baseline item is searched. A state matching confidence is generated based on feature distance, a baseline credibility score is generated based on timeliness, sample size, representativeness and historical segment backtracking detection information, an original offset score is generated based on rhythm offset features, and the original offset score is constrained by the baseline credibility score to obtain an output score. Finally, according to the baseline credibility score and the output score, shunt control is performed among a first type, a second type, a third type of output signal and an optional credibility limited identifier. The scheme can reduce the interference of state mismatch and low-quality baseline on segment offset output, and improve the output stability and reference consistency in a continuous acquisition scene.

Description

Technical Field

[0001] This invention belongs to the field of electrocardiogram signal processing and data evaluation technology, specifically relating to an electrocardiogram segment deviation evaluation and output control method, system and related device. Background Technology

[0002] With the development of wearable ECG acquisition devices, mobile terminals, and remote data processing platforms, long-term acquisition of continuous ECG segments has been widely used in scenarios such as home acquisition, postoperative follow-up, and long-term monitoring. In these scenarios, the system typically needs to quickly compare and filter the continuously acquired ECG segments to determine whether to continue acquisition, supplement acquisition, or trigger subsequent processing procedures.

[0003] In existing technologies, one type of scheme directly classifies or identifies based on the current ECG segment. This type of scheme focuses on extracting features from the current segment or directly outputting results using a model, but it is insufficient in adapting to the natural fluctuations within an individual for the same user under different states. Another type of scheme is based on population statistical models for evaluation. This type of scheme has certain advantages in large-scale processing, but because it mainly relies on population distribution as a reference, it does not adequately utilize the long-term stable patterns and natural variation range of the same user. Yet another type of scheme constructs an individualized baseline using the user's historical segments and compares the current segment with the historical baseline. This type of scheme has introduced individualized reference, but two prominent problems still exist in continuous acquisition scenarios.

[0004] First, there's the issue of state mismatch. The rhythmic behavior of the same user varies significantly across resting, sleeping, and activity recovery states. If the current segment to be evaluated is directly compared to a historical baseline in a different state, the difference in normal states may be mistakenly interpreted as an offset, leading to output bias.

[0005] Second, there is the issue of reference distortion. Even after selecting candidate baselines, the selected historical baseline entries may still suffer from insufficient sample size, reduced timeliness, insufficient representativeness, or the inclusion of potentially unsuitable samples within historical fragments. If the system defaults to allowing a selected baseline entry to directly participate in directional output, it can easily generate results with large amplitudes but insufficient basis when the reference object itself is unreliable.

[0006] It should be noted that even though existing technologies disclose individual baseline comparison, template maintenance, multi-parameter confidence concepts, state-related information, or anomaly detection methods, these technologies are mainly used for baseline acquisition timing selection, template updating, single scoring, or independent detection. They have not yet formed the following closed loop: first quantifying the state matching degree between the current segment and the candidate baseline entry, then quantifying the reference confidence of the candidate baseline entry itself, and then using the reference confidence to constrain the offset output and control the splitting between different output signals. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention provides an ECG segment offset assessment and output control method based on multi-state individual baseline credibility. When comparing segments using an individual's historical baseline, it not only considers the matching degree between the current segment and the candidate baseline entry, but also quantifies whether the candidate baseline entry itself is sufficiently credible. When the baseline credibility is insufficient, it constrains, compresses, or pulls back the output results to reduce misleading directional output.

[0008] To achieve the above objectives, firstly, a method for assessing and controlling electrocardiogram segment deviation includes the following steps: Signal quality screening, normality determination, and status label classification and storage of electrocardiogram segments are performed to establish a multi-state individual historical baseline database; Obtain the current ECG segment to be evaluated and extract the state feature vector. The system retrieves the rhythm feature vector and searches the multi-state individual historical baseline library for candidate baseline entries that are consistent with the state label of the current ECG segment to be evaluated or have the smallest distance from the state feature vector. Based on the state feature vector The state feature vector of the candidate baseline entry The feature distance d between them generates the confidence score of the state matching. ; Calculate the timeliness factor of each candidate baseline entry. Sample size factor Representative factors and historical fragment retrospective factors This leads to the baseline validity score. And combined with the state matching confidence level Generate baseline credibility score ; An original offset score is generated based on the rhythm offset characteristics of the current ECG segment to be evaluated relative to the candidate baseline entry. And score based on baseline credibility. The original offset score Perform constraint processing to optimize the output score. Baseline credibility score As the value decreases, it gradually approaches the preset non-directional default value. ; Based on baseline reliability score and output rating It performs current splitting control among the first type of output signal, the second type of output signal, and the third type of output signal.

[0009] Furthermore, the status label includes at least two of the following: resting state, sleep state, and activity recovery state. When there is no baseline entry in the baseline library that is completely consistent with the status label of the current ECG segment to be evaluated, the baseline entry with the smallest state feature distance is selected as the candidate baseline entry.

[0010] Furthermore, the feature distance is a normalized Euclidean distance or a Mahalanobis distance, and the state matching confidence is... Obtained using a monotonically decreasing mapping function.

[0011] Furthermore, the timeliness factor and sample size factor Calculate them separately as follows: When the time span Δt between the latest sample in the candidate baseline entry and the current time is not greater than the preset effective window hour, ; When Δt is greater than the preset effective window hour, Decrease according to the decay function; When the number of samples N for candidate baseline entries is not less than the preset minimum number of samples hour, When N is less than hour, ; The representative factor The degree of coverage of the normal variation range in the corresponding state is determined based on the candidate baseline entries: Suppose that the candidate baseline entries contain M rhythm feature dimensions, and the coefficient of variation of the i-th dimension is... The minimum coverage threshold is Then when Time Coverage Indicator Select 1, otherwise select 0, and have ; The historical segment retrospective factor This was obtained by re-performing normality checks on historical segments in the candidate baseline entries: Let the total number of historical segments in the candidate baseline entries be . The number of segments that did not meet the baseline requirements after retrospective testing was: ,but =1-( / ); The baseline validity score Confidence of matching the state The baseline credibility score is generated using a weighted geometric mean. ; , where 0 < γ < 1.

[0012] Furthermore, the original offset score The output score is obtained by normalizing and weighting at least three of the following five rhythm deviation features: mean heart rate difference, RR interval standard deviation deviation ratio, RMSSD deviation ratio, HF / LF deviation, and QT interval deviation; and the output score is obtained by weighting and summing these features. Obtain it in the following ways: ,in This is the default value for non-directional properties.

[0013] Furthermore, the current shunting control among the first type of output signal, the second type of output signal, and the third type of output signal includes: when and ≥ When, output the first type of output signal; when and When, output the second type of output signal; when At that time, a third type of output signal is output, triggering a resampling, supplementary acquisition, delayed judgment, or acquisition quality check process; among which, ; when When this happens, a score output signal with a limited credibility indicator is output to indicate that the reference value of the current output score is limited.

[0014] Furthermore, it also includes: output scoring of multiple consecutive ECG segments currently being evaluated. Perform sliding window statistics, and output scores for K consecutive segments. All exceeded the preset offset threshold At that time, an upgrade output signal is generated.

[0015] In a second aspect, the present invention provides an electrocardiogram segment deviation assessment and output control system, including a baseline construction module, a state matching module, a reliability assessment module, a deviation scoring module, and an output control module; The baseline construction module is used to screen signal quality, determine normality, and classify and store status labels of historical ECG segments in order to establish a multi-state individual historical baseline library. The state matching module is used to obtain the current ECG segment to be evaluated and extract the state feature vector. The system retrieves the rhythm feature vector and searches the multi-state individual historical baseline database for candidate baseline entries that match the state label of the current ECG segment to be evaluated or have the smallest distance from the state feature vector; based on the state feature vector... The state feature vector of the candidate baseline entry The feature distance d between them generates the confidence score of the state matching. ; The credibility assessment module is used to calculate the timeliness factor of the candidate baseline entries respectively. Sample size factor Representative factors and historical fragment retrospective factors This leads to the baseline validity score. And combined with the state matching confidence level Generate baseline credibility score ; The offset scoring module is used to generate an original offset score based on the rhythm offset characteristics of the current ECG segment to be evaluated relative to the candidate baseline entry. And score based on baseline credibility. The original offset score Perform constraint processing to optimize the output score. Baseline credibility score As the value decreases, it gradually approaches the preset non-directional default value. ; The output control module is used to score based on baseline confidence. and output rating Outputs a first-class output signal, a second-class output signal, a third-class output signal, and an optional confidence-limited flag; The system is deployed on at least one of a wearable ECG acquisition device, a mobile terminal that is communicatively connected to the wearable ECG acquisition device, an edge computing node, or a server.

[0016] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the above-described method for evaluating and controlling the electrocardiogram segment deviation is implemented.

[0017] Meanwhile, an electronic device is also provided, including a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the above-mentioned ECG segment deviation assessment and output control method.

[0018] Compared with the prior art, the present invention has at least the following beneficial effects: By quantifying the degree of state matching before segment comparison, the interference of normal differences in different states on the offset evaluation is reduced, and the offset calculation error caused by state mismatch is reduced. By quantifying the reference confidence of candidate baseline items separately, all selected baseline items are no longer assumed to have the same reference value, reducing misleading output under low-quality baseline conditions. The output score is made to approach the non-directional default value when the baseline confidence decreases, and the output signal is combined with different output signals for diversion control, improving the output stability and reference consistency in continuous acquisition scenarios. The core functional links of this invention include state matching confidence, baseline confidence score, restricted offset output, and output signal diversion in sequence. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall process of the method of the present invention.

[0020] Figure 2 This is a schematic diagram illustrating the construction and lifecycle management process of a multi-state individual historical baseline database.

[0021] Figure 3 A schematic diagram illustrating the process of generating state matching confidence and baseline confidence scores.

[0022] Figure 4 This is a schematic diagram of the output score generation and output signal splitting process.

[0023] Figure 5 A schematic diagram of the system structure for implementing the method of the present invention. Detailed Implementation

[0024] 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, not all, of the embodiments of the present invention. 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.

[0025] The objective of this invention is to reduce systematic bias errors caused by state mismatch when comparing ECG segments based on individual historical baselines by quantifying the degree of state matching between the current segment to be evaluated and candidate baseline entries. Simultaneously, it addresses how to quantify the reference reliability of candidate baseline entries when they may have insufficient sample size, decreased timeliness, insufficient representativeness, or potentially unsuitable samples included in historical segments, and uses this reliability to constrain the bias output, thereby avoiding results that are overly directional but lack sufficient basis.

[0026] The present invention will be further described below with reference to embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0027] This embodiment provides a method for assessing and controlling ECG segment deviation based on multi-state individual baseline reliability. The overall process is as follows: Figure 1 As shown, it includes the following steps.

[0028] Step S1: Establish a multi-state individual historical baseline database Historical ECG segments from the same user are processed segment by segment. Each historical ECG segment undergoes at least the following processing: signal quality screening, normality determination, status label determination, and storage according to status label classification.

[0029] The signal quality screening can be based on one or more of the following indicators: signal-to-noise ratio (SNR), baseline drift, lead contact stability, and artifact ratio. Specifically, after denoising preprocessing and R-peak localization for each historical ECG segment, the SNR, baseline drift amplitude (B), lead contact stability index (C), and artifact ratio (A) are calculated respectively; when all conditions are met... , , and At that time, the segment is determined to have passed the signal quality screening; among which , , , This refers to the preset threshold for the corresponding indicator. In another implementation, the above indicators can be normalized and then weighted and summed according to preset weights to form a quality score Q. If the segment passes, it is considered to pass; otherwise, the segment is discarded or marked as needing resampling.

[0030] The normality determination can be based on: RR interval regularity, P wave presence, PR interval range, and / or intra-segment rhythm stability. Specifically, the RR interval sequence is calculated based on the R-peak sequence, and P wave detection, PR interval estimation, and intra-segment rhythm stability analysis are performed on the segment; when the proportion of RR intervals exceeding the preset normal fluctuation range does not exceed the preset regularity threshold... The target QRS complex is preceded by an identifiable P wave, and its detection rate is not lower than a preset P wave threshold. The average PR interval within the fragment falls within the preset PR interval range. , The intra-segment rhythm stability index is not lower than the preset stability threshold. If the condition is met, the segment is deemed to meet the baseline requirements; otherwise, it is not included in the baseline database. The aforementioned preset range can be a fixed interval, a user-specific baseline interval, or a combination of both. Historical segments that fail the normality assessment are not included in the baseline database.

[0031] The status label includes at least two of the following: resting state, sleep state, and activity recovery state. The status label can be determined based on at least one of the following: heart rate level, heart rate variability characteristics, frequency domain energy distribution, activity sensor information, or user context information. Specifically, the average heart rate (HR), heart rate variability indices RMSSD and SDNN, frequency domain low-frequency / high-frequency energy ratio (HF / LF), and activity intensity are extracted from the segment. In addition to state characteristics such as the collection period T, the state characteristics are compared with preset rules or the corresponding state baseline center: when When the heart rate (HR) is below the individual's resting mean by a preset range and the RMSSD is above the individual's upper quartile resting value, it is marked as a sleep state; when the activity intensity of consecutive segments... When the HR changes from high to low and shows a downward trend, it is marked as an activity recovery state; when When the conditions for sleep state and activity recovery state are not met, it is marked as resting state; when it cannot be determined, it can be marked as transitional state or classified into candidate state according to the principle of minimum state feature distance. The state features can also be combined with the collection time period or user context information for auxiliary discrimination. When evaluating the current ECG segment to be evaluated, the same state label determination rule as historical ECG segments is used to synchronously generate the state label of the current ECG segment to be evaluated, so as to ensure that the semantics of the current state are consistent with those of historical segments.

[0032] The baseline repository stores corresponding historical segments and their statistics according to status labels. The baseline repository also includes a lifecycle management mechanism, which is used to perform at least one of the following operations: new sample inclusion, statistic update, baseline freezing, and outlier segment removal.

[0033] Step S2: Retrieve candidate baseline entries and generate state matching confidence scores. After obtaining the current ECG segment to be evaluated, it is preprocessed to extract the state feature vector. and rhythmic feature vectors.

[0034] In the multi-state individual historical baseline database, baseline entries that match the state label of the current ECG segment to be evaluated are preferentially retrieved; if no baseline entry with completely matching state labels exists, the baseline entry with the smallest state feature distance is selected as the candidate baseline entry.

[0035] Let the state feature vector of the candidate baseline entry be... Then the state matching confidence level Based on state feature vectors and The feature distance d between them is obtained. The feature distance d can be a normalized Euclidean distance or a Mahalanobis distance. Preferably, the state matching confidence is... .

[0036] Step S3: Generate baseline credibility score After retrieving candidate baseline entries, the validity of the candidate baseline entries is quantified to generate a baseline credibility score. .

[0037] In a preferred embodiment, the timeliness factor of each candidate baseline is calculated. Sample size factor Representative factors and historical fragment retrospective factors .

[0038] Time factor This is used to reflect the time span Δt between the latest sample in the candidate baseline entry and the current time, when Δt is not greater than a preset effective window. hour, Take 1; when Δt is greater than hour, Decrease according to the decay function. Specifically, let... The preset effective time window, λ is a positive time decay coefficient, and F_min is an optional lower limit of the time proximity factor: when When, the time proximity factor is 1; when At that time, the time proximity factor is... Decreasing. Based on this exponential decay function, the longer a candidate baseline entry has not been updated, the lower its reference weight for offset evaluation; when configured... In this case, the time proximity factor has a non-zero lower bound to prevent historically valid but temporarily unupdated baseline entries from being completely masked. In another implementation, the time proximity factor can also take a linear decay form. Implementation, in which This is the preset upper limit time window.

[0039] Sample size factor This is used to reflect whether the sample size of candidate baseline entries is sufficient. The sample size N of candidate baseline entries is not less than the preset minimum sample size. hour, Set to 1; when N is less than hour, It should be noted that the sample size N mentioned in this paper does not refer to the number of candidate baseline entries, but rather to the cumulative number of historical ECG segments within the same candidate baseline entry that have passed signal quality screening and normality assessment and been included in that state baseline; let N be the sample size N. To preset the minimum number of samples, when When the sample size factor is 1, Time sample size factor Linear growth.

[0040] Representative factor This is used to reflect the degree to which candidate baseline entries cover the normal variation range under the corresponding state. Suppose that each candidate baseline entry contains M rhythmic feature dimensions, and the coefficient of variation of the i-th dimension is... The minimum coverage threshold is .when Time Coverage Indicator Select 1, otherwise select 0, and have .

[0041] Historical Fragment Retrospective Factor This reflects whether there are historical segments in the candidate baseline entries that are no longer suitable as baselines after re-examination. Let the total number of historical segments in the candidate baseline entries be... The number of segments that did not meet the baseline requirements after re-testing was [number missing]. Then it is acceptable. As an example, when When the value is below the preset rejection threshold, the segments that do not meet the baseline requirements are removed from the candidate baseline entries, and the statistics are recalculated based on the remaining segments.

[0042] In a preferred embodiment, baseline validity score Obtain it as follows: ,in , , and The preset weights are 1.

[0043] Furthermore, baseline credibility score It can be obtained by the following formula: , Where 0 < γ < 1. Baseline reliability scores can also be obtained using a weighted arithmetic mean or by looking up a table. .

[0044] Step S4: Generate the original offset score and the output score Extract the rhythm offset features of the current ECG segment to be evaluated relative to the candidate baseline entry. The rhythm offset features are determined based on at least three of the following five items: mean heart rate difference, RR interval standard deviation offset ratio, RMSSD offset ratio, HF / LF offset, and QT interval offset.

[0045] In a preferred embodiment, the original offset score is obtained by normalizing and weighting the sum of the rhythm offset features. .

[0046] The key to this invention is not merely generating the original offset score. , but based on right Perform constraint processing to ensure that when the baseline confidence score is... As the value decreases, the output score gradually approaches the preset non-directional default value. Instead of maintaining the directional strength of the original offset score.

[0047] In a preferred embodiment, the output score is calculated using the following formula:

[0048] in It is preferable to set it to an intermediate value, which is used to express that when the reliability of the reference is insufficient, it is not advisable to give results that are too directional.

[0049] In continuous data acquisition scenarios, if the system does not output a score at all when the baseline confidence is insufficient, the downstream processing module will have difficulty distinguishing between "evaluation pause due to insufficient baseline confidence" and "output loss due to data acquisition interruption or system anomaly." This is detrimental to the distinguishability of system states and the correct triggering of subsequent processes. Therefore, a non-directional default value is adopted. It continues to output scores instead of stopping outputting them when the baseline credibility is insufficient.

[0050] In addition to the linear fusion method mentioned above, the original offset score is also evaluated. Constraint handling can also be achieved through interval compression, segmented mapping, or gated output logic.

[0051] Step S5: Output signal splitting Based on baseline reliability score With output score The combined result is split among the first type of output signal, the second type of output signal, and the third type of output signal.

[0052] In a preferred embodiment, when ≥ and When, output the first type of output signal; when and When, output the second type of output signal; when When, output a third type of output signal; when It can output a score signal with a credibility limitation indicator.

[0053] The first type of output signal is used to mark the degree of deviation of the current ECG segment to be evaluated as reaching a preset condition; the second type of output signal is used to mark the current ECG segment to be evaluated as being in a state of continued acquisition; the third type of output signal is used to trigger resampling, supplementary acquisition, delayed judgment, or acquisition quality check process.

[0054] Step S6: Example of continuous offset trend judgment and state recognition In another implementation, output scores are given for multiple consecutive ECG segments currently being evaluated. Perform sliding window statistics. Calculate the output score for K consecutive segments. All exceeded the preset offset threshold At that time, an upgrade output signal is generated to increase the priority of subsequent processing.

[0055] In one exemplary implementation, the status labels are resting state, sleep state, and activity recovery state. For the sleep state, it can be determined according to the following rules: the current segment average heart rate is lower than the individual's resting mean by a preset range, the RMSSD is higher than the individual's upper quartile value in the resting state, and the activity level is lower than a preset threshold.

[0056] For the activity recovery state, the following rules can be used to determine it: the average heart rate of the current segment is higher than the individual's resting average by a preset range, and the average heart rate of multiple consecutive segments shows a decreasing trend, indicating a transition from high to low activity levels. For the resting state, segments that do not meet the rules for the sleep state and activity recovery state and whose activity levels are below the resting activity threshold can be marked as resting states.

[0057] like Figure 5 As shown, the present invention also provides a system including a baseline construction module, a state matching module, a reliability assessment module, an offset scoring module, and an output control module. Each module can be deployed inside the same device or distributed in wearable ECG acquisition devices, mobile terminals, edge nodes, or servers.

[0058] Among them, the baseline construction module can run in the non-real-time stage and is responsible for batch processing of historical segments; it performs signal quality screening, normality determination and status label classification and storage of ECG segments, and establishes a multi-state individual historical baseline library; The state matching module is used to obtain the current ECG segment to be evaluated and extract the state feature vector. And rhythm feature vectors, and retrieve candidate baseline entries from the multi-state individual historical baseline database that have the same state label as the current ECG segment to be evaluated, generated synchronously using the same state label determination rules, or have the smallest distance between their state feature vectors; based on the state feature vectors The state feature vector of the candidate baseline entry The feature distance d between them generates the confidence score of the state matching. ; The offset scoring module is used to calculate the timeliness factor of the candidate baseline entries respectively. Sample size factor Representative factors and historical fragment retrospective factors This leads to the baseline validity score. And combined with the state matching confidence level Generate baseline credibility score A raw offset score is generated based on the rhythm offset characteristics of the current ECG segment to be evaluated relative to the candidate baseline entry. And score based on baseline credibility. The original offset score Perform constraint processing to optimize the output score. Baseline credibility score As the value decreases, it gradually approaches the preset non-directional default value. ; The output control module is used to score based on baseline confidence. and output rating It performs current splitting control among the first type of output signal, the second type of output signal, and the third type of output signal.

[0059] The state matching module, credibility assessment module, offset scoring module, and output control module can be executed sequentially in the real-time phase.

[0060] The present invention also provides an electronic device, including a processor and a memory. The memory stores a computer program, and when the processor executes the computer program, it implements the method steps of any of the above embodiments. The present invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above methods.

[0061] To further illustrate the technical principles of this invention, a comparative verification scheme can be used. This scheme can utilize either real collected data or simulated data that conforms to the distribution patterns of electrocardiogram parameters. The following verification results are generated based on a reproducible simulation script and are used only to illustrate the principles and technical effects of this invention; they do not constitute a limitation on practical application effects.

[0062] The following simulation verification scenarios illustrate the technical effects of the state matching mechanism and baseline credibility assessment mechanism of this invention. Each scenario provides simulation parameter settings, the processing methods for the control group and the experimental group, and the corresponding quantitative evaluation results.

[0063] In the first simulation verification, three normal state parameter distributions were set up: resting state, sleep state, and activity recovery state. The baseline sample size for each state was 50, and the test sample size was 200. The control group merged the samples from all three states into a single baseline, while the experimental group established baseline entries for each state and performed state matching. The offset judgment threshold was determined by adding 1.5 times the standard deviation to the mean of the offset score distribution of the state-matched groups, resulting in a threshold of 1.37. The verification results showed that: when comparing with a unified baseline, the average score was 3.48, and the false offset rate was 74.3%; after state matching, the average score was 0.83, and the false offset rate decreased to 8.5%. These results indicate that state matching helps reduce false offsets caused by normal state transitions.

[0064] In the second set of simulation verifications, the number of test samples in the fixed resting state was 300, and three baseline quality scenarios were set: high quality, medium quality, and low quality. The number of high-quality baseline samples was 50. The value is set to 0.92, and the medium-quality baseline sample size is 5. The value is set to 0.50, and the number of low-quality baseline samples is 2. We set the value to 0.18. A normalized offset score exceeding 0.65 was defined as a misleading high offset output. Validation results show that: in high-quality scenarios, the proportion of unconstrained misleading high offsets is 4.3%, decreasing to 3.7% after constraints; in medium-quality scenarios, the proportion is 5.0%, decreasing to 1.0% after constraints; in low-quality scenarios, the proportion is 60.0%, decreasing to 0.0% after constraints, and the proportion of output scores falling within the middle range [0.4, 0.6] is 100.0%. These results indicate that baseline credibility constraints help suppress misleading directional outputs under low-quality baseline conditions.

[0065] On the other hand, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the electrocardiogram segment deviation assessment and output control method described in the present invention.

[0066] The present invention can also provide a computer device, including a processor and a memory, wherein the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and the processor can implement the electrocardiogram segment deviation assessment and output control method described in the present invention when executing the computer executable program.

[0067] The computer device may be a laptop, tablet, desktop computer, or workstation.

[0068] The processor can be a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or an off-the-shelf programmable gate array (FPGA).

[0069] The memory described in this invention can be an internal storage unit of a laptop, tablet, desktop computer, or workstation, such as memory or hard disk; or it can be an external storage unit, such as a portable hard disk or flash memory card.

[0070] Computer-readable storage media can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media can include: read-only memory (ROM), random access memory (RAM), solid-state drives (SSDs), or optical discs, etc. Random access memory can include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

[0071] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for assessing and controlling electrocardiogram segment deviation, characterized in that, Includes the following steps: Signal quality screening, normality determination, and status label classification and storage of electrocardiogram segments are performed to establish a multi-state individual historical baseline database; Obtain the current ECG segment to be evaluated and extract the state feature vector. The system retrieves the rhythm feature vector and searches the multi-state individual historical baseline library for candidate baseline entries that are consistent with the state label of the current ECG segment to be evaluated or have the smallest distance from the state feature vector. Based on the state feature vector The state feature vector of the candidate baseline entry The feature distance d between them generates the confidence score of the state matching. ; Calculate the timeliness factor of each candidate baseline entry. Sample size factor Representative factors and historical fragment retrospective factors This leads to the baseline validity score. And combined with the state matching confidence level Generate baseline credibility score ; An original offset score is generated based on the rhythm offset characteristics of the current ECG segment to be evaluated relative to the candidate baseline entry. And score based on baseline credibility. The original offset score Perform constraint processing to optimize the output score. Baseline credibility score As the value decreases, it gradually approaches the preset non-directional default value. ; Based on baseline reliability score and output rating It performs current splitting control among the first type of output signal, the second type of output signal, and the third type of output signal.

2. The ECG segment deviation assessment and output control method according to claim 1, characterized in that, The status label includes at least two of the following: resting state, sleep state, and activity recovery state. When there is no baseline entry in the baseline library that is completely consistent with the status label of the current ECG segment to be evaluated, the baseline entry with the smallest state feature distance is selected as the candidate baseline entry.

3. The ECG segment deviation assessment and output control method according to claim 1, characterized in that, The feature distance is expressed as normalized Euclidean distance or Mahalanobis distance, and the state matching confidence is expressed as... Obtained using a monotonically decreasing mapping function.

4. The ECG segment deviation assessment and output control method according to claim 1, characterized in that, The timeliness factor and sample size factor Calculate them separately as follows: When the time span Δt between the latest sample in the candidate baseline entry and the current time is not greater than the preset effective window hour, ; When Δt is greater than the preset effective window hour, Decrease according to the decay function; When the number of samples N for candidate baseline entries is not less than the preset minimum number of samples hour, When N is less than hour, ; The representative factor The degree of coverage of the normal variation range in the corresponding state is determined based on the candidate baseline entries: Suppose that the candidate baseline entries contain M rhythm feature dimensions, and the coefficient of variation of the i-th dimension is... The minimum coverage threshold is Then when Time Coverage Indicator Select 1, otherwise select 0, and have ; The historical segment retrospective factor This was obtained by re-performing normality checks on historical segments in the candidate baseline entries: Let the total number of historical segments in the candidate baseline entries be . The number of segments that did not meet the baseline requirements after retrospective testing was: ,but =1-( / ); The baseline validity score Confidence of matching the state The baseline credibility score is generated using a weighted geometric mean. ; , where 0 < γ < 1.

5. The ECG segment deviation assessment and output control method according to claim 1, characterized in that, The original offset score The output score is obtained by normalizing and weighting at least three of the following five rhythm deviation features: mean heart rate difference, RR interval standard deviation deviation ratio, RMSSD deviation ratio, HF / LF deviation, and QT interval deviation; and the output score is obtained by weighting and summing these features. Obtain it in the following ways: ,in This is the default value for non-directional properties.

6. The ECG segment deviation assessment and output control method according to claim 1, characterized in that, Shunting control between the first type of output signal, the second type of output signal, and the third type of output signal includes: when and ≥ When, output the first type of output signal; when and When, output the second type of output signal; when At that time, a third type of output signal is output, triggering a resampling, supplementary acquisition, delayed judgment, or acquisition quality check process; among which, ; when When this happens, a score output signal with a limited credibility indicator is output to indicate that the reference value of the current output score is limited.

7. The ECG segment deviation assessment and output control method according to claim 1, characterized in that, Also includes: Output scoring of multiple consecutive ECG segments to be evaluated Perform sliding window statistics, and output scores for K consecutive segments. All exceeded the preset offset threshold At that time, an upgrade output signal is generated.

8. A control system for evaluating and outputting electrocardiogram segment deviation, characterized in that, It includes a baseline construction module, a state matching module, a credibility assessment module, an offset scoring module, and an output control module; The baseline construction module is used to screen signal quality, determine normality, and classify and store status labels for historical ECG segments in order to establish a multi-state individual historical baseline library. The state matching module is used to obtain the current ECG segment to be evaluated and extract the state feature vector. The system retrieves the state feature vector and the candidate baseline entry in the multi-state individual historical baseline database that matches the state label of the current ECG segment to be evaluated or has the smallest distance from the state feature vector; based on the state feature vector... The state feature vector of the candidate baseline entry The feature distance d between them generates the confidence score of the state matching. ; The credibility assessment module is used to calculate the timeliness factor of the candidate baseline entries respectively. Sample size factor Representative factors and historical fragment retrospective factors This leads to the baseline validity score. And combined with the state matching confidence level Generate baseline credibility score ; The offset scoring module is used to generate an original offset score based on the rhythm offset characteristics of the current ECG segment to be evaluated relative to the candidate baseline entry. And score based on baseline credibility. The original offset score Perform constraint processing to optimize the output score. Baseline credibility score As the value decreases, it gradually approaches the preset non-directional default value. ; The output control module is used to score based on baseline confidence. and output rating Outputs a first-class output signal, a second-class output signal, a third-class output signal, and an optional confidence-limited flag; The system is deployed on at least one of a wearable ECG acquisition device, a mobile terminal that is communicatively connected to the wearable ECG acquisition device, an edge computing node, or a server.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the ECG segment deviation assessment and output control method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the ECG segment deviation assessment and output control method according to any one of claims 1 to 7.

Images