CN114847926B

CN114847926B - Pulmonary function respiration detection method based on self-adaptive threshold

Info

Publication number: CN114847926B
Application number: CN202210491067.4A
Authority: CN
Inventors: 何国强; 徐文龙
Original assignee: China Jiliang University
Current assignee: China Jiliang University
Priority date: 2022-05-07
Filing date: 2022-05-07
Publication date: 2025-05-20
Anticipated expiration: 2042-05-07
Also published as: CN114847926A

Abstract

The invention discloses a lung function breath detection method based on a self-adaptive threshold. The existing fixed threshold method has limitations in improving the accuracy and robustness of breath detection. The method comprises the steps of carrying out convolution smoothing filtering on respiratory flow data to filter respiratory signal noise, carrying out zero-crossing detection on the smoothed respiratory flow data to obtain potential respiratory starting points and respiratory ending points, solving and standardizing first derivative peaks of the smoothed respiratory flow data, analyzing respiratory modes of the first derivative peaks according to time intervals, classifying the respiratory modes into 3 types of rapid respiration, normal respiration and slow respiration, screening the potential respiratory starting points and the potential respiratory ending points according to different self-adaptive adjustment time interval threshold Tn of each respiratory mode, and obtaining respiratory starting points and ending points to finish respiratory detection. According to the invention, the breathing threshold value can be adaptively changed according to the breathing characteristics and the breathing state of the patient, so that the accuracy and the robustness of the lung function breathing detection are improved.

Description

Pulmonary function respiration detection method based on self-adaptive threshold

Technical Field

The invention belongs to the technical field of breath detection, and particularly relates to a lung function breath detection method based on a self-adaptive threshold value.

Background

Breath detection refers to the detection of the time position of the beginning of expiration and the beginning of inspiration from within the respiratory flow signal in the lung function examination, which is a key step in the computer's automatic detection of lung function parameters, and is the starting point for calculating various clinical significance indicators, performing regression analysis or making predictions. Breath detection is a relatively easy task for a professional in the examination of lung function, but is still a challenge for automatic detection in a computer. Because of the variability of each person's breathing rate and breathing pattern, there is a high degree of variability between individuals, and the respiratory flow signal is highly noisy, not conforming to the assumptions made by most conventional automated digital signal processing analysis. Thus, a reliable method for automatic breath detection for use in a pulmonary function machine is achieved that requires constant verification and updating to accommodate the special features that may be present in a single respiratory flow signal.

The current respiration detection method used in most pulmonary function instruments is a fixed threshold method. In the fixed threshold method, first, all zero crossing points in the respiratory flow are identified, the flow peak value of the exhalation or inhalation is obtained from the two zero crossing points, and finally whether the condition is met between the two zero crossing points is analyzed through a preset fixed threshold value, and the respiratory cycle is judged to be qualified. However, a fixed threshold is affected by variability between individuals, and it is not possible to specify a single threshold for each case, and the subject can only choose to change the threshold until a satisfactory result is obtained. If the threshold is too low, false breaths may be detected, and if the threshold is too high, true breaths may be missed. Therefore, although of practical significance, fixed threshold breath detection methods have certain limitations.

Disclosure of Invention

Aiming at the defects of the prior art, the invention provides a lung function breath detection method based on an adaptive threshold, and the breath detection method is designed to adaptively change the breath mode threshold according to the breath characteristics and the breath state of a subject based on the difference of the breath modes of the subject, so as to form the breath detection with robustness and universality.

The invention discloses a lung function breath detection method based on a self-adaptive threshold value, which comprises the following steps of:

and step 1, carrying out breath detection.

And step 2, loading the respiratory flow data in a time flow mode at intervals t, and carrying out convolution smoothing on the respiratory flow data to obtain a smooth respiratory flow signal y (n), wherein n is a sampling point.

And 3, performing zero crossing detection on y (n) to obtain a zero crossing point time position set A.

Step 4, first derivative is obtained on y (n) to obtain first derivative X' (k), k=1, 2.

And 5, carrying out peak extraction on X' (k) by using a local extremum method to obtain a peak point set B, and normalizing elements in the peak point set B.

And 6, dividing y (n) into 3 breathing modes, namely a fast breathing mode, a normal breathing mode and a slow breathing mode according to the number of B elements of the peak point set.

And 7, adjusting the current time interval threshold Tn according to the judged breathing mode.

And 8, screening the zero-crossing point time position set A according to the current time interval threshold Tn, and determining the positions of the starting point and the ending point of respiration.

And 9, judging whether the breath detection is continuous, and if so, returning to the step 2 for execution.

Preferably, the convolution smoothing processing in the step 2 is specifically:

And the window sequentially strokes through all the respiratory flow data to carry out polynomial fitting estimation until all the respiratory flow data are smoothed and a smooth respiratory flow signal y (n) is returned.

Preferably, the step 3 specifically includes:

The method comprises the steps of taking a window W1 to judge whether respiration flow values at the front two positions x (n-1) and x (n-2) of a center point position x (n) of the window W1 are negative, finding positions of a respiration starting point and an inspiration ending point according to an analysis strategy of judging whether respiration flow values at the rear two positions x (n+1) and x (n+2) of the center point position x (n) of the window W1 are positive, and taking the window W2 to judge whether respiration flow values at the front two positions x '(n-1) and x' (n-2) of the center point position x '(n) of the window W2 are positive or not and finding positions of an exhalation ending point and an inspiration starting point according to an analysis strategy of judging whether respiration flow values at the rear two positions x' (n+1) and x '(n+2) of the center point position x' (n) of the window W2 are negative or not. And alternately working the window W1 and the window W2, and sequentially marking and storing the positions of the central points of the window W1 and the positions of the central points of the window W2 which accord with the analysis strategy into a zero-crossing point time position set A.

Preferably, the step 5 specifically includes:

And 5.1, sequentially drawing a window adopting a local extreme point searching strategy through the first derivative X '(k), obtaining the position X (k) when the second derivative X' (k) is zero, and storing the position sequence labels conforming to the searching strategy into the peak point set B.

Step 5.2, standardizing elements in the peak point set B, wherein the standardized formula is as follows:

Wherein sigma is the standard deviation of x (k), X (k) is a normalized value of X (k) which is the mean value of X (k).

Preferably, the step 6 specifically includes:

If the number of elements in the peak point set B is 2 or 3, judging that the breathing mode is a normal breathing mode, if the number of elements in the peak point set B is more than 3, judging that the breathing mode is a fast breathing mode, and if the number of elements in the peak point set B is less than 2, judging that the breathing mode is a slow breathing mode.

Preferably, the step 7 specifically includes:

When the breathing mode is a rapid breathing mode and a normal breathing mode, taking the average value of time intervals between every two adjacent peak points in the peak point set B as T, if the breathing detection is carried out in the first time interval T, directly taking the T as a current time interval threshold Tn, otherwise taking the average value of the T and the time interval threshold calculated in the previous time interval T as the current time interval threshold Tn. When the breathing mode is a slow breathing mode, the smooth breathing flow signal y (n) is subjected to fast Fourier transform, and the minimum period is obtained by taking the reciprocal of the maximum frequency value f T' is taken as the current time interval threshold Tn.

Preferably, the step 8 specifically includes:

Judging the time interval between every two adjacent zero crossing points in the zero crossing point time position set according to the current time interval threshold Tn, if the time interval is smaller than the current time interval threshold Tn, considering that the latter zero crossing point is not a respiration starting point or an ending point, otherwise, considering that the latter zero crossing point is a respiration starting point or an ending point, and the exhalation starting point and the inhalation starting point are alternately distributed.

The invention has the beneficial effects that:

1. According to the invention, the interval segmentation method is adopted, long-time respiratory flow data are disassembled into a plurality of short-time interval respiratory flows, and the respiratory starting point and the respiratory ending point are independently processed in each interval, so that the detection process can be performed in real time, and the operation efficiency is high. More importantly, the self-adaptive threshold updating mode adopted by the invention can obtain the threshold value suitable for each individual breathing mode, and the current time interval threshold value updated in real time is better suitable for the breathing detection of the individual at different moments and different living states, so that the influence of the fixed threshold value on the overall result is avoided.

2. The invention has simple detection process, reasonable design and convenient realization.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a schematic diagram of raw respiratory flow data;

FIG. 3 is a schematic illustration of the flow rate data after convolution smoothing;

FIG. 4 is a schematic diagram of a smoothed respiratory flow signal and its first derivative peaks;

FIG. 5 is a schematic diagram of determining a breathing pattern based on the number of peak point set elements;

fig. 6 is a schematic diagram of respiratory detection using an adaptive threshold (current time interval threshold).

Detailed Description

The invention is further described below with reference to the accompanying drawings.

As shown in fig. 1, a lung function breath detection method based on an adaptive threshold includes the following steps:

and step 1, carrying out breath detection.

Step 2, loading the respiratory flow data in a time flow form (and visualizing by using PyQt 5) at intervals t, simulating a real lung function checking flow, wherein t is preferably 2 seconds, and carrying out convolution smoothing on the respiratory flow data to obtain a smooth respiratory flow signal y (n), wherein n is a sampling point, and figures 2 and 3 respectively illustrate a respiratory flow data example and the respiratory flow signal after convolution smoothing of the respiratory flow data. Because the sampling rate of the respiratory signals collected by the pulmonary function instrument is 100Hz, and the useful frequency band of the real respiratory signals is concentrated between 0.5 Hz and 1.5Hz, a great amount of respiratory noise can influence respiratory detection, and therefore, the respiratory signals are processed by convolution smoothing.

Step 3, performing zero-crossing detection on y (n) to obtain a zero-crossing time position set a, wherein the zero-crossing time position set a is regarded as a group of potential respiration start and end point sets.

Step 4, first derivative is obtained for y (n), resulting in first derivative X '(k), k=1, 2,..m, wherein, m=m-1, M being the total number of sampling points M, X' (k) in the smoothed respiratory flow signal y (n) reflecting the change in respiratory rate during breathing.

And 5, carrying out peak extraction on X' (k) by using a local extremum method to obtain a peak point set B, and normalizing elements in the peak point set B. Fig. 4 illustrates the smoothed respiratory flow signal of fig. 3 and peak points extracted from the signal.

And 6, dividing y (n) into 3 breathing modes, namely a fast breathing mode, a normal breathing mode and a slow breathing mode according to the number of B elements of the peak point set. Fig. 5 illustrates a breathing pattern that is a partial signal division of the smoothed respiratory flow signal of fig. 3.

And 8, screening the zero-crossing point time position set A according to the current time interval threshold Tn, and determining the positions of the starting point and the ending point of respiration. Fig. 6 illustrates respiratory start and end point positions (black filled circle positions) determined in the local signal of the smoothed respiratory flow signal in fig. 3.

Preferably, the convolution smoothing processing in step 2 is specifically:

Taking a window with the window length of 0.55 seconds, returning the window to an L-degree polynomial by using a least square method, taking L to be a value in 3, 5, 7 or 9, taking L=5 in the embodiment, then using polynomial fitting to estimate the respiratory flow value at the central point of the window, and sequentially dividing the window through all respiratory flow data to carry out polynomial fitting estimation until all respiratory flow data are smoothed and a smooth respiratory flow signal y (n) is returned.

Preferably, the step 3 specifically comprises:

Taking window W1 and window W2 with window length of 0.05 seconds, judging whether the respiratory flow values at the front two positions x (n-1) and x (n-2) of the central point position x (n) of the window W1 are negative or not by the window W1, searching the positions of a respiratory start point and an inhalation end point by an analysis strategy of whether the respiratory flow values at the rear two positions x (n+1) and x (n+2) of the central point position x (n) of the window W1 are positive or not, judging whether the respiratory flow values at the front two positions x '(n-1) and x' (n-2) of the central point position x '(n) of the window W2 are positive or not by the window W2, and searching the positions of the respiratory end point and the inhalation start point by an analysis strategy of whether the respiratory flow values at the rear two positions x' (n+1) and x '(n+2) of the central point position x' (n) of the window W2 are negative or not. The window W1 and the window W2 work alternately, all window W1 center point positions and window W2 center point positions conforming to the analysis strategy are marked and stored in a zero-crossing point time position set A in sequence, and each element in the zero-crossing point time position set A is a potential respiration starting or ending point.

Preferably, the step 5 specifically comprises:

and 5.1, sequentially drawing a window with the window length of 0.5 seconds through the first derivative X ' (k), adopting a local extreme point searching strategy for the window, solving the position X (k) when the second derivative X ' ' (k) is zero, and storing position sequence labels conforming to the searching strategy into the peak point set B.

Wherein sigma is the standard deviation of x (k), Is the mean value of x (k).

Preferably, the step 6 specifically comprises:

If the number of elements in the peak point set B is 2 or 3, the respiration mode is judged to be a normal respiration mode (Normal Breathing Phase, NBP), if the number of elements in the peak point set B is more than 3, the respiration mode is judged to be a fast respiration mode (Fast Breathing Phase, FBP) if the number of elements in the peak point set B is more than 3, and if the number of elements in the peak point set B is less than 2, the respiration mode is judged to be a slow respiration mode (Slow Breathing Phase, SBP) if the number of elements in the peak point set B is less than 2, the respiration mode is judged to be a normal level.

Preferably, the step 7 specifically comprises:

When the breathing mode is a rapid breathing mode and a normal breathing mode, taking the average value of time intervals between every two adjacent peak points in the peak point set B as T, if the breathing detection is carried out in the first time interval T, directly taking the T as a current time interval threshold Tn, otherwise taking the average value of the T and the time interval threshold calculated in the previous time interval T as the current time interval threshold Tn. When the breathing mode is the slow breathing mode, the smooth breathing flow signal y (n) is subjected to Fast Fourier Transform (FFT), and the minimum period is obtained by taking the reciprocal of the maximum frequency value f T' is taken as the current time interval threshold Tn.

Preferably, the step 8 specifically comprises:

Judging the time interval between every two adjacent zero crossing points in the zero crossing point time position set according to the current time interval threshold Tn, if the time interval is smaller than the current time interval threshold Tn, considering that the latter zero crossing point is not a respiration starting point or an ending point, otherwise, considering that the latter zero crossing point is a respiration starting point or an ending point, and the respiration starting point are alternately distributed, and the respiration ending point is an respiration starting point, and the respiration ending point do not need to be judged in addition.

Claims

1. A lung function breath detection method based on an adaptive threshold is characterized by comprising the following steps:

Step 1, respiratory detection is carried out;

step 2, loading the respiratory flow data in a time flow mode at intervals t, and carrying out convolution smoothing on the respiratory flow data to obtain a smooth respiratory flow signal y (n), wherein n is a sampling point;

Step 3, performing zero crossing detection on y (n) to obtain a zero crossing point time position set A;

Step 4, first derivative is calculated on y (n) to obtain first derivative X' (k), k=1, 2, & M, wherein m=m-1, and M is the total number of sampling points M in the smooth respiratory flow signal y (n);

Step 5, carrying out peak extraction on X' (k) by using a local extremum method to obtain a peak point set B, and normalizing elements in the peak point set B;

step 6, dividing y (n) into 3 breathing modes, namely a fast breathing mode, a normal breathing mode and a slow breathing mode according to the number of B elements of the peak point set;

step 7, adjusting a current time interval threshold Tn according to the judged breathing mode;

step 8, screening a zero-crossing point time position set A according to a current time interval threshold Tn, and determining respiration starting and ending point positions;

step 9, judging whether the breath detection is continuous, if so, returning to the step 2 for execution;

The step 7 specifically comprises the following steps:

When the breathing mode is a fast breathing mode and a normal breathing mode, taking the average value of time intervals between every two adjacent peak points in the peak point set B as T, if the breathing detection is carried out in the first time interval T, directly taking T as a current time interval threshold Tn, otherwise taking the average value of the T and the time interval threshold calculated in the previous time interval T as the current time interval threshold Tn, when the breathing mode is a slow breathing mode, carrying out fast Fourier transform on a smooth breathing flow signal y (n), taking the reciprocal of a maximum frequency value f to obtain the minimum period T' is taken as the current time interval threshold Tn.

2. The method for detecting lung function respiration based on the adaptive threshold according to claim 1, wherein the convolution smoothing process in the step 2 is specifically:

3. The method for detecting lung function respiration according to claim 1, wherein the step 3 is specifically:

Taking a window W1, judging whether the respiratory flow value at the front two positions x (n-1) and x (n-2) of the central point position x (n) of the window W1 is negative, finding the positions of a respiratory start point and an inhalation end point according to an analysis strategy of judging whether the respiratory flow value at the rear two positions x (n+1) and x (n+2) of the central point position x (n) of the window W1 is positive, taking the window W2, judging whether the respiratory flow value at the front two positions x '(n-1) and x' (n-2) of the central point position x '(n) of the window W2 is positive, finding the positions of an exhalation end point and an inhalation start point according to an analysis strategy of judging whether the respiratory flow value at the rear two positions x' (n+1) and x '(n+2) of the central point position x' (n) of the window W2 is negative, and alternately working all intersection point positions of the window W1 and the central point position of the window W2 according to an analysis strategy are sequentially marked and stored in a zero time position set A.

4. The method for detecting lung function respiration according to claim 1, wherein the step 5 is specifically:

Step 5.1, sequentially drawing a window adopting a local extreme point searching strategy through a first derivative X '(k), obtaining a position X (k) when a second derivative X' (k) is zero, and marking and storing the position sequence conforming to the searching strategy into a peak point set B;

5. The method for detecting lung function respiration according to claim 1, wherein the step 6 is specifically:

6. The method for detecting lung function respiration according to claim 1, wherein the step 8 is specifically: