Adaptive breathing phase estimation method
By preprocessing and detecting anomalies in respiratory data, combined with physiological feature prediction algorithms, the respiratory phase is accurately identified, solving the problem of misjudgment in non-rhythmic states of intelligent respiratory assistive equipment and achieving system stability and safety.
Patent Information
- Application Number
- CN202510986595.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-17
- Publication Date
- 2025-11-07
AI Technical Summary
Existing intelligent respiratory assist devices are unable to adapt to changes in patients' non-rhythmic breathing states, leading to misjudgments and human-machine asynchrony, which affects patient safety.
By acquiring and preprocessing respiratory data, using the extreme values of lung volume curves to segment the phase, and combining real-time anomaly detection and prediction algorithms based on physiological characteristics, non-rhythmic respiratory data is identified and eliminated, thus accurately identifying the respiratory phase.
It improves the accuracy of respiratory phase recognition, reduces recognition time, ensures the stability and safety of the respiratory assistance system, and avoids human-machine aggression.
Smart Images
Figure CN120899222A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of respiratory assistance, in particular to an adaptive respiratory phase estimation method. BACKGROUND
[0002] The number of patients with respiratory dysfunction worldwide is increasing year by year, and the shortage of hospital resources for a large number of patients with respiratory dysfunction makes home respiratory rehabilitation an increasingly urgent need. Intelligent respiratory assistance equipment can help patients to carry out targeted respiratory training, improve the strength and endurance of respiratory muscles, and thus improve respiratory function. However, when the current intelligent equipment carries out respiratory assistance, it is often difficult to adapt to the changes in respiratory rhythm during the rehabilitation process of the patient, especially when the respiratory is in a non-rhythmic state, such as speaking, coughing, deep breathing, etc. The occurrence of these non-rhythmic breaths is uncertain and can occur at any time point, and their respiratory patterns are quite different from rhythmic breathing. If these non-rhythmic breaths cannot be effectively processed in the system, it will cause the system to make a false judgment, causing the human-machine to be out of sync or even to be in conflict, thereby causing damage to the chest and abdominal tissues of the patient.
[0003] Therefore, how to accurately judge the respiratory intention, identify the respiratory phase, adapt to the changes of different individuals and different respiratory states, and complete the real-time monitoring of the respiratory parameters of the patient is the key to ensuring the human-machine adaptability of the respiratory assistance system. SUMMARY
[0004] In view of the deficiencies of the prior art, the present application provides an adaptive respiratory phase estimation method, which effectively solves the influence of the randomness and diversity of respiratory behavior in the respiratory state recognition process, and realizes accurate discrimination of respiratory state, providing a basis for control of the respiratory assistance system.
[0005] The technical solution adopted by the present application to achieve the above-mentioned purpose is:
[0006] An adaptive respiratory phase estimation method, comprising the following steps:
[0007] Obtaining respiratory data and pre-processing the same;
[0008] Obtaining a lung volume curve from the pre-processed respiratory data and performing extreme value processing thereon to obtain phase segmentation points and segment the respiratory data phase;
[0009] Eliminating non-rhythmic respiratory data through a real-time anomaly detection mechanism;
[0010] Using a prediction algorithm based on physiological characteristics to predict and identify the rhythmic respiratory data to obtain the final respiratory phase recognition result.
[0011] The pre-processing of the respiratory data specifically removes noise interference in the respiratory signal.
[0012] The respiratory data phase segmentation process, in particular:
[0013] By finding the extreme value of the lung volume curve, locating the time point of the extreme value point, finding the flow point at the corresponding time on the respiratory flow curve, taking it as a zero-crossing point, and taking the difference between the adjacent points and the value of the point, the positive value corresponds to the inspiration stage, and the negative value corresponds to the expiration stage. The zero-crossing point is used as a phase segmentation point for respiratory phase segmentation. At the same time, the minimum phase length is set to check the consistency of the signs of consecutive sampling points, and the segmentation result is verified.
[0014] The non-rhythmic respiratory data is removed through a real-time anomaly detection mechanism, in particular:
[0015] An elliptical region is set as the judgment boundary of the stable respiratory cycle, and the respiratory cycle in the region is defined as the effective auxiliary interval, and the outside of the boundary is regarded as abnormal and excluded.
[0016] The rhythmic respiratory data is predicted and identified using a prediction algorithm based on physiological characteristics, in particular:
[0017] The respiratory phase θ is calculated:
[0018]
[0019] Where V ex_cur is the current respiratory volume; V ex is the predicted total respiratory volume of the current cycle, and γ(t) is the adaptive ratio at time t:
[0020]
[0021] Where, is the kth expiration volume, is the kth inspiration volume, k∈[t-60,t] is the respiratory cycle number in the sliding window, and the window length is 60 seconds. W(t) is the respiratory data in the current sliding window.
[0022] An adaptive respiratory phase estimation system, comprising:
[0023] A respiratory data preprocessing module for acquiring and preprocessing respiratory data;
[0024] A phase segmentation module for detecting the sign change of adjacent sampling points in the respiratory data, obtaining a phase segmentation point, and segmenting the respiratory data phase;
[0025] An anomaly detection module for removing non-rhythmic respiratory data through a real-time anomaly detection mechanism;
[0026] The phase recognition module is used for predicting and recognizing the rhythm breathing data by using a prediction algorithm based on physiological characteristics, so as to obtain a final breathing phase recognition result.
[0027] The present application has the following advantages and benefits:
[0028] 1. The present application provides an adaptive breathing phase recognition method, which is different from the previous breathing phase two classification. The present application performs fine segmentation on the breathing phase on the basis of fast recognition, significantly improves the accuracy in the breathing phase fine recognition process, and greatly reduces the recognition time of the breathing phase estimation. Through the prediction algorithm based on physiological characteristics, the dynamic factor disturbance problem in the breathing phase estimation process is effectively solved, and the precise control of the breathing assistance system is ensured.
[0029] 2. The present application effectively solves the problem of breathing phase recognition under the condition of non-rhythmic breathing, avoids the man-machine confrontation problem caused by the misrecognition of breathing intention in the control process of the breathing assistance system, and ensures the stability in the breathing assistance process. BRIEF DESCRIPTION OF DRAWINGS
[0030] Figure 1 is a method flowchart of the present application;
[0031] Figure 2 is a breathing phase division schematic diagram of the present application;
[0032] Figure 3 is a breathing prediction result comparison diagram of the method of the present application and the prior art. DETAILED DESCRIPTION
[0033] The present application will be further described in detail below in combination with the drawings and embodiments.
[0034] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below in combination with the drawings. In the following description, a large number of specific details are set forth in order to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the spirit of the present application, so the present application is not limited by the specific embodiments disclosed below.
[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application.
[0036] As Figure 1 is a method flowchart of the present application;
[0037] An adaptive respiratory phase estimation method, comprising the following parts:
[0038] Firstly, the respiratory data of the subject is collected by a respiratory flow sensor, and the flow change of the respiratory process is recorded; secondly, the respiratory data is preprocessed to remove the noise interference in the respiratory signal; then, the extreme value of the lung volume curve is solved, and the potential phase segmentation point is identified, so as to preliminarily realize the division of the respiratory phase; further, through the real-time anomaly detection mechanism, the non-rhythm respiratory data is quickly eliminated, and only the rhythm respiratory data is reserved for estimating the respiratory phase; finally, the respiratory rhythm change of the human body is estimated by using the prediction algorithm based on the physiological characteristics, and the accurate identification of the respiratory phase is completed.
[0039] The respiratory data collection experiment paradigm is as follows: firstly, the subject adopts the supine position, maintains the natural rhythm breathing state, and opens the respiratory flow sensor to collect the respiratory data of the subject; the used sensor is FS6122 gas flow sensor, and the sampling frequency is 200Hz; and zero point calibration is carried out before use to ensure the accuracy of data collection. At the same time, a standard medical nasal mask is equipped for respiratory flow monitoring to ensure the standardization and consistency of the measurement process. In the collection, the subject alternately performs rhythm breathing and rest, and 30 minutes of rhythm breathing data is collected. Since the respiratory flow signal has quasi-periodic and non-stationary characteristics, and is often interfered by multiple sources of noise, the Kalman filter is used for preprocessing the respiratory signal in the present application.
[0040] As Figure 2 shown is a respiratory phase division schematic diagram of the present application;
[0041] The estimation of the respiratory phase mainly depends on the respiratory flow Q and the lung volume V collected by the respiratory flow meter. In the respiratory cycle, the respiratory flow Q (solid line) and the lung volume V (dashed line) both show significant phase characteristics: the zero point of Q marks the change of the respiratory phase, and the positive and negative values correspond to the inspiration and expiration stages respectively. By solving the extreme value point of the lung volume curve, the time when the maximum value and the minimum value appear is found, that is, the zero-crossing point of the respiratory flow curve. The difference between the adjacent points and the value of the point is calculated, and the positive value corresponds to the inspiration stage and the negative value corresponds to the expiration stage, so as to preliminarily decompose the inspiration and expiration stages. In order to improve the robustness of the method, the consistency check of the continuous sampling point symbol is adopted, and the minimum phase time constraint is combined to determine the accuracy of the respiratory phase segmentation. The minimum phase time constraint is set to 300ms.
[0042] The stable breathing data of different individuals presents an ellipse-like distribution in a two-dimensional space, and the rhythmic breathing cycle points are gathered in a specific ellipse area. When coughing, sneezing and other non-rhythmic breathings occur during the breathing process, the breathing data will deviate from the specific area. Therefore, the application designs a fast abnormality detection method, takes the ellipse area as the judgment boundary of the stable breathing cycle, defines the breathing cycle in the area as the effective auxiliary interval, and regards the outside of the boundary as abnormal and excluded, so as to obtain stable and reliable respiratory flow and lung volume data.
[0043] Since the measurement of the inhalation volume and the exhalation volume can be dynamically affected by various internal and external factors, such as physiological state, motion activity state, environmental factor and the like. Therefore, the breathing phase estimation is completed by means of algorithm prediction, and the prediction algorithm based on physiological characteristics is described as follows.
[0044] The breathing phase θ calculation formula can be described as:
[0045]
[0046] Among them, V ex_cur is the current respiratory volume; V ex is the total respiratory volume predicted in the current cycle, and γ(t) is the adaptive ratio at t time.
[0047]
[0048] Among them, is the kth exhalation volume; is the kth inhalation volume; r is time (second); γ(t) is the adaptive ratio at t time; k ∈ [t-60, t] is the breathing cycle number in the sliding window, and the window length is 60 seconds. In the application, the algorithm presets a standard ratio γ = 0.98 when the system is started, and then updates the ratio γ(t) every minute.
[0049] In order to further illustrate the beneficial effects of the method of the application, the proposed method is compared and verified with commonly used algorithms, including recursive least square (RLS) algorithm, extended Kalman filter (EKF) algorithm and long short-term memory (LSTM) algorithm.
[0050] Figure 3 The comparison chart of the breathing prediction results of the method of the application and the prior art
[0051] From the experimental results in the figure, among various prediction algorithms, the scatter points of the adaptive respiratory phase prediction method proposed in the application are distributed on the contour line in a concentrated manner, indicating that the deviation between the predicted value and the actual value is small. At the same time, from the normal distribution of errors, the adaptive respiratory phase prediction algorithm proposed in the application and the RLS algorithm are most concentrated in the low error part, indicating that the probability of large error of the algorithm is low, and the prediction accuracy and stability are higher.
[0052] Table 1 is the error measurement and prediction performance comparison of the method of the application and the existing method
[0053]
[0054]
[0055] From the data in the table, after introducing the respiratory abnormality detection, the MAE (mean absolute error) of each algorithm is significantly reduced, and the maximum error is also effectively controlled. For example, the MAE of the method before abnormality detection is 3.543, and after detection, it is reduced to 2.003; the MAE of the RLS algorithm before detection is 3.816, and after detection, it is 2.231; the MAE of the EKF algorithm before detection is 4.036, and after detection, it is 1.920; the MAE of the LSTM algorithm before detection is 4.064, and after detection, it is 2.169. This shows that the respiratory abnormality detection mechanism can effectively improve the accuracy of the algorithm for respiratory phase prediction.
[0056] In terms of average time, the average time of the method is the shortest, only 0.001 ms, the average time of the RLS and EKF algorithms is the same, both 0.027 ms, and the average time of the LSTM algorithm is 0.081 ms. Although the LSTM algorithm performs better in prediction accuracy, it has a relatively long time cost. In actual real-time control scenarios, the accuracy and time cost need to be weighed comprehensively.
[0057] In terms of replacement rate, the higher the replacement rate means that the algorithm is more disturbed by abnormal breathing and has lower stability. The replacement rate of the LSTM algorithm is as high as 89.19%, indicating that the proportion of correction in abnormal breathing detection is extremely high, and the stability is relatively poor. The replacement rate of the method is only 2.70%, which performs well in stability. Considering factors such as MAE, average time and replacement rate, the method proposed in the application can effectively reduce the error after introducing abnormality detection, ensure the prediction accuracy, and has extremely short average time and high stability, which is the best choice.
Claims
1. A method of adaptive respiratory phase estimation, characterized by, The method comprises the following steps: acquiring and preprocessing the respiratory data; obtaining a lung volume curve from the preprocessed respiratory data and performing extreme value processing to obtain phase segmentation points and segment the respiratory data phase; eliminating non-rhythmic respiratory data through a real-time anomaly detection mechanism; using a prediction algorithm based on physiological characteristics to predict and identify the rhythmic respiratory data to obtain a final respiratory phase identification result.
2. The adaptive breath phase estimation method of claim 1, wherein, The preprocessing of the respiratory data specifically comprises removing noise interference in the respiratory signal.
3. The adaptive breath phase estimation method of claim 1, wherein, The respiratory data phase segmentation process specifically comprises: locating the time point of the extreme value point by performing extreme value processing on the lung volume curve, finding the flow point at the corresponding time point on the respiratory flow curve, taking the flow point as a zero-crossing point, taking the difference between the values of the adjacent points of the zero-crossing point and the point, taking the positive value as the inspiration stage and the negative value as the expiration stage, taking the zero-crossing point as the phase segmentation point, segmenting the respiratory phase, and verifying the segmentation result by checking the consistency of the signs of the continuous sampling points using a set minimum phase length.
4. The adaptive breath phase estimation method of claim 1, wherein, The elimination of non-rhythmic respiratory data through a real-time anomaly detection mechanism specifically comprises: setting an elliptical region as the judgment boundary of a stable respiratory cycle, defining the respiratory cycle in the region as an effective auxiliary interval, and regarding the cycles outside the boundary as abnormal and excluding them.
5. The adaptive breath phase estimation method of claim 1, wherein, The prediction and identification of the rhythmic respiratory data using a prediction algorithm based on physiological characteristics specifically comprises: calculating the respiratory phase θ: Where, V ex_cur is the current breath volume; V ex is the total breath volume predicted for the current cycle, and γ(t) is the adaptive ratio at time t: wherein, V k is the volume of the kth expiration, V k is the volume of the kth inspiration, k ∈ [t-60, t] is the respiratory cycle number in the sliding window, the window length is 60 seconds, and W(t) is the respiratory data in the current sliding window.
6. An adaptive respiratory phase estimation system, characterized by, comprises: a respiratory data preprocessing module for acquiring and preprocessing the respiratory data; a phase segmentation module for detecting the sign change of adjacent sampling points in the respiratory data, obtaining phase segmentation points, and segmenting the respiratory data phase; an anomaly detection module for eliminating non-rhythmic respiratory data through a real-time anomaly detection mechanism; a phase identification module for using a prediction algorithm based on physiological characteristics to predict and identify the rhythmic respiratory data to obtain a final respiratory phase identification result.