Search

CN-122006118-A - Electrocardiogram pacing pulse detection method under low sampling rate condition

CN122006118ACN 122006118 ACN122006118 ACN 122006118ACN-122006118-A

Abstract

The invention discloses an electrocardiogram pacing pulse detection method under a low sampling rate condition, which comprises the steps of obtaining multi-lead electrocardiosignals, extracting energy concentration characteristics and symbol consistency characteristics, constructing multi-scale analysis windows, carrying out space vector concentration analysis on the multi-lead signals in each analysis window to obtain multi-scale vector concentration characteristics, screening pacing pulses with high confidence from full signals to serve as anchor points according to the energy concentration characteristics, the symbol consistency characteristics and the multi-scale vector concentration characteristics, screening out pacing pulses with high confidence from the full signals according to self-adaptive threshold values, searching at a prediction position by a relaxed threshold value when at least two anchor points are detected, carrying out forward and reverse parallel searching under a plurality of preset pacing interval hypotheses and carrying out morphological similarity matching and multi-feature verification when only a single anchor point is detected, and outputting a final pacing pulse detection list by integrating two detection results. The invention realizes complete and accurate detection of pacing pulse under the condition of low sampling rate.

Inventors

  • LV JUN
  • YU ZHENFENG
  • LIN JUNFENG
  • DU FANGYING

Assignees

  • 广东工业大学

Dates

Publication Date
20260512
Application Date
20260330

Claims (10)

  1. 1. A method for detecting an electrocardiogram pacing pulse under a low sampling rate condition, comprising the steps of: acquiring multi-lead electrocardiosignals and preprocessing; extracting energy concentration characteristics and symbol consistency characteristics according to the preprocessed multi-lead electrocardiosignals; Constructing multi-scale analysis windows, and carrying out space vector concentration analysis on the multi-lead signals in each analysis window to obtain multi-scale vector concentration characteristics; According to the energy concentration degree characteristic, the symbol consistency characteristic and the multi-scale vector concentration degree characteristic, combining with a self-adaptive threshold value, and screening pacing pulses with high confidence degree from the full signal to serve as anchor points; when at least two anchor points are detected, searching is carried out at the predicted position according to the time interval rule between the anchor points by using a relaxed threshold value to obtain a first remedy detection result; When only a single anchor point is detected, forward and reverse parallel searching is carried out under the assumption of a plurality of preset pacing intervals by taking the anchor point as a reference, and a second remedy detection result is obtained through morphological similarity matching and multi-feature verification; And integrating the first remedy detection result and the second remedy detection result, and outputting a final pacing pulse detection list.
  2. 2. The method according to claim 1, characterized in that extracting energy concentration features, in particular, comprises: calculating a first order differential signal for each lead; Calculating, for each lead, a sum of absolute values of differential signals over a short time window and a sum of absolute values of differential signals over a long time window, respectively, at each point in time; Taking the ratio of the sum in the short time window to the sum in the long time window as the energy concentration degree of the corresponding lead at the corresponding time point; and carrying out weighted average on the energy concentration of all the leads to obtain the multi-lead energy concentration characteristic.
  3. 3. The method according to claim 1, wherein extracting the symbol consistency feature comprises: For each time point, counting the polarity direction of the differential signal of each lead in the local window, wherein the polarity direction comprises positive direction and negative direction; Counting the number of leads with the polarity direction being the main positive direction and the number of leads with the polarity direction being the main negative direction in the local window; and taking the ratio of the maximum value of the number of the leads in the main positive direction and the number of the leads in the main negative direction to the total number of the leads as a symbol consistency index of the corresponding time point.
  4. 4. The method according to claim 1, wherein constructing a multi-scale analysis window, and performing spatial vector concentration analysis on the multi-lead signals in each analysis window, to obtain multi-scale vector concentration characteristics, specifically comprises: respectively constructing two different multi-lead signal matrixes with a short window radius and a long window radius; singular value decomposition is carried out on each multi-lead signal matrix to obtain a group of singular values; Calculating the ratio of the first singular value to the sum of all singular values of each multi-lead signal matrix as a vector concentration index under the corresponding window scale; and for each time point, selecting the maximum value in the vector concentration index under different window scales as the multi-scale vector concentration characteristic of the corresponding time point.
  5. 5. The method according to claim 1, wherein, according to the energy concentration characteristic, the symbol consistency characteristic and the multi-scale vector concentration characteristic, in combination with an adaptive threshold, pacing pulses with high confidence are selected from the full signal as anchor points, specifically comprising: constructing a comprehensive score according to the number of activated leads, the weighted average concentration of multiple leads and the sum of absolute values of all lead differential signals; calculating a self-adaptive detection threshold according to the global statistic and the local statistic of the comprehensive score; and taking a time point with the comprehensive score exceeding the self-adaptive detection threshold and meeting the preset characteristics of the number of activated leads, the symbol consistency index and the multi-scale vector concentration as a pacing pulse anchor point with high confidence.
  6. 6. The method of claim 1, wherein searching at the predicted location with a relaxed threshold value based on a time interval law between anchor points results in a first remedial detection, comprising: calculating the time interval between detected anchor points and judging the stability; if the time interval is stable and within the preset range, searching for a local maximum value of the comprehensive score in a search window before and after the predicted position by taking the time interval as a prediction interval; and verifying the candidate points corresponding to the searched local maximum values by applying a relaxed verification condition, and taking the verified candidate points as a first remedy detection result.
  7. 7. The method of claim 1, further comprising the step of R-wave detection and rejection, comprising: identifying the R wave crest value position according to the sum of the first-order difference absolute values of the multi-lead electrocardiosignals; a protection window is arranged around each R peak value, the area in the protection window is marked as an R area, and the rest areas are marked as quiet areas; In the case of remedial detection, the detection is performed only in the quiet zone.
  8. 8. The method of claim 7, further comprising performing an anomaly scan based on spatial vector concentration characteristics within the quiet zone, comprising: Calculating background baselines and background fluctuation of the multi-scale vector concentration characteristic through a sliding window in the quiet zone; calculating the overrun of the multi-scale vector concentration characteristic relative to the background baseline; taking a time point which satisfies the condition that the multi-scale vector concentration characteristic exceeds a first threshold value and the overrun exceeds a second threshold value in the quiet zone as a primary screening candidate point; And carrying out multidimensional verification on the primary screening candidate points by combining the symbol consistency characteristics, the distance between the primary screening candidate points and the R wave and whether the primary screening candidate points are local peaks, and supplementing the verified candidate points serving as pacing pulses to a detection list.
  9. 9. The method according to claim 1, wherein when only one anchor point is detected, performing multi-interval parallel search based on the anchor point, specifically comprising: setting a plurality of candidate pacing intervals; For each candidate pacing interval, predicting forward and backward from the anchor point position according to the candidate pacing interval to obtain a predicted position; determining candidate points through morphological similarity matching in search windows before and after each prediction position; performing de-duplication treatment on candidate points obtained in all candidate intervals; And checking the duplicate-removed candidate points by applying a preset verification path, and taking the checked candidate points as a second remedy detection result.
  10. 10. The method according to claim 9, wherein the determining the candidate point by morphological similarity matching specifically comprises: According to the detected anchor points, waveform fragments are intercepted on the original multi-lead electrocardiosignals, trend removal processing is carried out on each fragment, and a waveform template is constructed; In the search window, carrying out trending treatment on the original signal fragments at each time point, and calculating normalized cross-correlation coefficients of the original signal fragments and the waveform template; And taking the time point with the highest normalized cross-correlation coefficient as a candidate point.

Description

Electrocardiogram pacing pulse detection method under low sampling rate condition Technical Field The invention belongs to the technical field of biomedical engineering and digital signal processing intersection, and particularly relates to an electrocardiogram pacing pulse detection method under a low sampling rate condition. Background Cardiac pacemakers maintain or regulate heart rhythm by applying short electrical pulses to the heart muscle, which typically appear as sharp waveforms of very short duration and high amplitude in a body surface electrocardiogram, known as pacing pulses. The accurate detection of the position and number of pacing pulses is significant for assessing the working state of a pacemaker, analyzing the pacing-capturing relationship and judging abnormal heart rhythm. However, in dynamic electrocardiographic monitoring or long Cheng Chuandai-type devices, to reduce power consumption and storage pressure, the sampling rate is often limited to 1kHz or even lower, and accurate capture of pacing pulses under such low sampling rate conditions presents a significant challenge. Existing pacing pulse detection methods are mainly divided into two categories. The first type is a detection method based on high-frequency time domain characteristics, and pulse positioning is realized by amplifying steep changes by calculating first-order or high-order differences of signals by utilizing the characteristic that rising edges and falling edges of pacing pulses are extremely fast. For example Ricke et al construct microsecond time windows with extremely high sampling rates of 75kHz to accurately calculate derivatives, polpetta et al suppress noise-induced false slopes by nonlinear operators, haq et al convert 12-lead electrocardiograms to vector electrocardiograms and calculate slope abruptness. The second type is a detection method based on frequency domain energy characteristics, which utilizes the characteristic that pacing pulses have extremely wide frequency bands to extract energy characteristics in a specific frequency band or a transform domain. For example, jekova et al map a single-lead electrocardiogram to the time-frequency domain using an S-transform and quantify the energy concentration features by shannon' S energy, nallathambi et al extract the wideband energy features at a low sampling rate of 125Hz by combining analog filtering with digital decisions. However, the prior art still has the disadvantage that, first, the high frequency characteristics are severely distorted at low sampling rates. At 1kHz sampling rate, the pulse occupies only 1-2 sampling points, and phase jitter causes random attenuation or even disappearance of the differential signal, resulting in missed detection. Second, spatial directionality and signal characteristics are underutilized. The existing energy analysis method suffers from serious spectrum aliasing at low frequency, is difficult to distinguish pulse and broadband myoelectric noise, and mostly aims at single-lead or simple channel summation, the real pacing pulse is ignored to have highly consistent directivity on the multi-lead space electric vector, and the noise directivity is usually disordered. Third, there is a lack of adaptive remedial mechanisms for the intrinsic regularity of pacing. In the prior art, a point-by-point detection mode is mostly adopted, and when signals are extremely weak and only individual pulses can be detected, the inherent interval stability logic of a pacemaker cannot be utilized to automatically predict, search and remedy the residual pulses covered by noise, so that the detection integrity is poor. In view of the above drawbacks, it is desirable to provide an electrocardiographic pacing pulse detection method under low sampling rate conditions. Disclosure of Invention Aiming at the defects, the invention provides an electrocardiogram pacing pulse detection method under the condition of low sampling rate, which overcomes three key limitations of the existing pacing pulse detection method, namely (1) the differential slope characteristic is seriously distorted under the condition of low sampling rate to cause missed detection, (2) the utilization of multi-lead space directivity and symbol consistency is lacking to cause that the spectrum aliasing and myoelectric noise interference cannot be effectively caused, and (3) the multi-interval parallel search mechanism based on single-point anchoring is lacking, so that the masked or weak pacing signals cannot be self-adaptively remedied. The invention provides an electrocardiogram pacing pulse detection method under the condition of low sampling rate, which comprises the following steps: acquiring multi-lead electrocardiosignals and preprocessing; extracting energy concentration characteristics and symbol consistency characteristics according to the preprocessed multi-lead electrocardiosignals; Constructing multi-scale analysis windows, and carrying out space vector concentrati