A rapid extraction method for metabolic features based on the extraction of full waveform recognition

Abstract

The present invention provides a rapid extraction method for metabolic features based on the recognition and extraction of complete waveforms. By using a sliding window filtering method based on the number of breaths, filtering and noise reduction are carried out during measurement, the fluctuating curve is smoothed, abnormal fluctuations caused by external factor disturbances are filtered out, the wave peaks are identified, the complete waveforms are extracted, the mean differences of the complete waves are judged and the metabolic feature values are calculated. It is judged when to end the measurement according to the number of waveforms. After the measurement ends, the values of metabolic parameters including the respiratory quotient RQ and the resting energy expenditure REE are calculated based on all data points within two complete waveforms; the measurement of the metabolic parameters is based on indirect calorimetry, and the indirect calorimetry calculates the metabolic parameters by collecting the oxygen consumption VO2 and the carbon dioxide VCO2. The present invention can be used for the evaluation of human resting energy expenditure and the evaluation of the proportion of human glucose and lipid metabolism. It can personalize the measurement time based on the characteristics of human metabolic oscillations, and has the advantages of accurate measurement and short time consumption.

Description

Technical Field

[0001] The present invention belongs to the field of metabolic feature detection, and particularly relates to a method for rapidly extracting metabolic features based on the recognition and extraction of complete waveforms. Background Art

[0002] According to the Report on Nutrition and Chronic Diseases of Chinese Residents (2020), in 2019, the deaths caused by chronic diseases in China accounted for 88.5% of the total deaths, the prevalence of adult hypertension was 25.2%, the number of diabetes patients reached 116 million, the overweight rate of adults was 30.1%, and the obesity rate was 11.9%. The Scientific Research Report on Chinese Residents' Dietary Guidelines (2021) compiled by the Chinese Nutrition Society pointed out that "obesity has become the primary risk factor threatening the health of Chinese residents", and proposed to "aim at chronic disease prevention and lead a healthy lifestyle in all aspects". Obesity, as a form of malnutrition caused by overnutrition, is a risk factor for chronic diseases such as hypertension, type 2 diabetes, hyperlipidemia, stroke, and coronary heart disease. Malnutrition in chronic disease patients and critically ill patients can damage important tissues and organs of the human body, increasing the mortality rate and the incidence of complications of patients.

[0003] Since the human body converts nutrients into energy through metabolism, by extracting metabolic features such as resting energy metabolic rate, and using this to evaluate the metabolic state, determine nutritional needs, and formulate nutritional plans, chronic diseases can be effectively prevented and treated, guiding the public to lose weight reasonably and healthily, and improving the survival rate of critically ill patients.

[0004] Indirect calorimetry (IC) is a method for calculating human metabolic features by collecting respiratory gases and based on the fixed ratio formula of energy consumption. In the field of disease prevention and treatment, indirect calorimetry has been used as the gold standard method in the European and American critical care nutrition guidelines and recommended for clinical application. Accurately evaluating resting energy expenditure can effectively reduce the mortality rate of critically ill patients; at the same time, for chronic diseases such as diabetes, indirect calorimetry has also been proven to reveal changes in metabolic features, which precede the development and changes of diseases and have important value for the early warning and prevention of diseases. In the field of public health, indirect calorimetry can measure the daily energy consumption of the human body more accurately than the calculation by empirical formulas, and can assist in the formulation of daily nutritional plans, providing accurate nutritional goals and effective nutritional status monitoring methods for adjusting nutritional structure and precise nutritional support.

[0005] The principle of indirect calorimetry is to measure the oxygen consumption and carbon dioxide production of the human body during breathing, and use the Weir formula to calculate metabolic characteristics such as resting energy expenditure (REE) and respiratory quotient (RQ). The standard experimental process lasts for one hour. First, the subject needs to lie flat and rest for 30 minutes. Then, the subject wears a gas collection device such as a mask or a hood to continuously collect respiratory gases. Each breath is calculated once to generate a set of metabolic characteristic data. Because of the state change when lying down and the instability of the human body state when just wearing the collection device, the data of the first 5 minutes need to be discarded, and the average value of the data of the last 25 minutes is used to obtain the evaluation value. However, due to the influence of various factors such as sleep, breathing, and mood on metabolism, and due to compliance problems, most subjects cannot reach the ideal standard state for a long time during the actual test process, resulting in large data variability and low repeatability, which affects the accuracy of the test, and it is difficult to apply the test results well in clinical and public nutrition support. Summary of the Invention

[0006] To overcome the bottleneck of the existing technology, the present invention proposes a rapid extraction method of metabolic characteristics based on complete waveform recognition and extraction, aiming to solve the problems of low accuracy in short-term measurement and difficulty for subjects to maintain the standard state in long-term measurement of the current method. Compared with other rapid extraction methods that shorten the test duration, it improves the accuracy, practicability, and convenience of indirect calorimetry. It can be used for the assessment of human resting energy consumption and the assessment of the proportion of human carbohydrate and lipid metabolism. It can measure the time individually according to the metabolic oscillation characteristics of the human body, and has the advantages of accurate measurement and short time consumption.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] A rapid extraction method of metabolic characteristics based on complete waveform recognition and extraction, which filters and reduces noise during measurement through a sliding window filtering method based on the number of breaths, smooths the fluctuation curve, filters out abnormal fluctuations caused by external factor disturbances, and identifies the wave peaks through the slope, extracts the complete waveform, judges the mean difference of the complete waves and calculates the metabolic characteristic values, determines when to end the measurement according to the number of waveforms, and calculates the values of metabolic parameters including respiratory quotient (RQ) and resting energy expenditure (REE) based on all data points within two complete waveforms after the measurement ends; the measurement of the metabolic parameters is based on indirect calorimetry, and the indirect calorimetry calculates the metabolic parameters by collecting oxygen consumption (VO2) and carbon dioxide (VCO2).

[0009] Further, the sliding window filtering method based on the number of breaths includes regarding each breath as a sampling point, and taking a current sampling point as the center to determine the window size of 2k, which altogether includes 2k + 1 data points. The metabolic parameter value of the current sampling point is equal to the average of the points within the window, that is , where represents the i-th metabolic parameter value. According to the window size, when the number of data points is greater than 2k + 1, filtering starts from the (k + 1)-th point, and the first k data points are not filtered; when the filtering window slides to the end and the number of remaining points is less than the window size, filtering is based on the remaining points, that is, the filtered value x of the last point last is equal to the average of the first k points, or a weighted average is implemented within the window where represents the weighting coefficient, the point is 1, and for other points, the farther away from the point the smaller the weighting coefficient, that is, the smaller the influence on the value of the current point.

[0010] Furthermore, the method for identifying the complete waveform sets a minimum interval between wave peaks, such that the minimum step length between two identified wave peaks is 7 sampling points, that is, 21 to 35 seconds; all data points between two wave peaks are counted as one complete waveform, and the wavelength Wavelength[i] is recorded. When two complete waveforms are collected, the measurement ends, and the data of several sampling points included in the complete waveform are averaged, and the average value of one complete waveform , where n represents the number of sampling points included in one waveform, represents the data of the i-th sampling point, and then the evaluation value is calculated by averaging according to the averages calculated from the two waveforms where, when the means of the two complete waveforms are similar, when the mean difference between the two complete waveforms is greater than 10%, the weighting coefficient is determined by the wavelength: , where Wavelength[1] is the wavelength of the first complete waveform and Wavelength[2] is the wavelength of the second complete waveform.

[0011] The present invention has the following advantages and effects:

[0012] (1) Based on the indirect calorimetry method, the present invention uses the sliding window method for noise reduction and uses wave peak recognition to capture the complete waveform, and real-time monitors the metabolic characteristic values. Filtering and noise reduction are performed for each breath. When two complete waveforms are captured, the measurement immediately ends, and the metabolic characteristic parameters such as the resting energy expenditure REE and the respiratory quotient RQ are calculated and output. This method can personalize the measurement time according to the differences in the oscillation frequencies of the human body. Since the tidal volume oscillation frequency is between about 0.003 and 0.03 Hz, the duration of one complete waveform is between 0.5 minutes and 5.5 minutes, the total measurement time is between 6 and 16 minutes, and the average time is about 10 minutes. Compared with the traditional 30-minute fixed-time method, it has the advantages of short time consumption and full consideration of individual differences.

[0013] (2) Different psychological and physiological states during actual measurement can lead to significant differences in breathing frequency. Since the breathing frequencies of different people at different stages are different, it is difficult to determine the size of the time window. For the filtering method based on the time window, when the filtering window is too small, it cannot cope with the noise generated by external disturbances, and when the window is too large, it is more likely to lose the changing trend. The filtering method based on the number of breathing points can still effectively filter and retain the changing trend when the breathing frequency changes greatly. The window filtering method based on the number of breathing points can better consider the fluctuations and differences in the human breathing frequency than the window filtering method based on time, and has better stability.

[0014] (3) For the weighted average method of two complete waveforms based on wavelength, in actual measurement, due to the requirements of real-time performance and accuracy, and the problems of individual state differences and external interference factors, the weighted average method of two complete waveforms based on wavelength can effectively average and offset the deviation caused by accidental disturbances in a short time. Due to the uncertainty of the oscillation period, the wavelengths of the extracted complete waveforms are different. Due to the changes in the environmental disturbances and the states of the subjects, the extracted wave peaks will also change accordingly. Therefore, the greater the difference in the means of the two complete waveforms, the greater the volatility, the greater the difference in the weighted weights, and the complete waveform with a longer wavelength has better stability, a higher weight value in calculation, and stronger anti-interference ability against disturbances. Description of the Drawings

[0015] Figure 1 is a schematic diagram of a rapid extraction method for metabolic characteristics based on the recognition and extraction of complete waveforms of the present invention;

[0016] Figure 2 is a flowchart of a rapid extraction method for metabolic characteristics based on the recognition and extraction of complete waveforms of the present invention. Detailed Embodiments

[0017] To make the above features, advantages, and implementation schemes of the present invention more obvious and understandable, the following further describes the implementation steps of the present invention in combination with the drawings and specific implementation schemes.

[0018] The present invention filters and reduces noise during measurement through a sliding window filtering method based on the number of breathing points, smooths the fluctuation curve, filters out abnormal fluctuations caused by external factor disturbances, identifies wave peaks through slopes, extracts complete waveforms, judges the mean differences of the complete waves and calculates metabolic characteristic values, determines when to end the measurement according to the number of waveforms, and calculates the values of metabolic parameters including respiratory quotient RQ and resting energy expenditure REE based on all data points within the two complete waveforms after the measurement ends; the measurement of the metabolic parameters is based on indirect calorimetry, and the indirect calorimetry calculates the metabolic parameters by collecting oxygen consumption VO2 and carbon dioxide VCO2.

[0019] Such as Figure 1As shown in the figure, in the rapid extraction method of metabolic characteristics based on complete waveform recognition and extraction of the present invention, the oxygen consumption VO2 and carbon dioxide production VCO2 are collected for each breath, and the metabolic characteristic value 1 is calculated through a fixed ratio formula. Taking the energy consumption as an example for the metabolic characteristic value 1 in the figure, other metabolic characteristic values such as the respiratory quotient RQ also conform to the same law and can be calculated and evaluated by the same method. Since the human body state is unstable when just wearing the metabolic measurement device, the data of the first five minutes will be discarded. The noise is filtered by a sliding window filtering method based on the number of breaths with a window size of 10, the peak 2 is found through the slope, and the false peaks 3 are filtered by conditions such as the minimum step length of the peak interval and the slope change range. All the points between two consecutive peaks are regarded as a complete waveform 4. When two complete waveforms are collected, the measurement ends. The average value of all the data within one complete waveform is obtained as the evaluation value Avg[i] of the metabolic characteristics, and the final evaluation value of the metabolic characteristics is equal to the weighted average of the evaluation value Avg[i] (metabolic characteristic value = C1 The mean value of each point within the first complete waveform + C2 The mean value of each point within the second complete waveform, where C1 and C2 are weighting coefficients), and the weighting coefficients depend on the difference between the two evaluation values and the wavelength of the complete waveform.

[0020] As Figure 2 shown, the rapid extraction method of metabolic characteristics based on complete waveform recognition and extraction of the present invention specifically includes the following implementation steps:

[0021] Step 1: The oxygen consumption VO2 and carbon dioxide production VCO2 are collected once for each breath, and the metabolic characteristic values of the human body are calculated through a fixed ratio formula, including resting energy expenditure (REE), respiratory quotient (RQ), etc. And due to the unstable state, the data of the first five minutes are discarded. Among them, the resting energy expenditure is the energy used by the human body to maintain the normal functions of body cells and organs and the waking state of the human body, and can be calculated by the weir formula from the oxygen consumption VO2 and carbon dioxide production VCO2:

[0022]

[0023]

[0024] The respiratory quotient RQ represents the proportion of different energy-providing substances. For example, for the three major energy-providing substances in the human body, sugar, fat, and protein, when metabolizing sugar, RQ = 1, and when metabolizing fat, RQ = 0.7.

[0025] The oxygen consumption VO2 and carbon dioxide production VCO2 are collected in real time, and various metabolic characteristic values are calculated once for each breath.

[0026] Step 2: Filter and denoise the metabolic feature values using a sliding window filtering method based on the number of breaths. Starting from the first data point after five minutes, the first 5 data points are not filtered. Filtering begins from the 6th data point. The filtered value is equal to the weighted average of the previous and next k points, with a geometric sequence of weighted coefficients of 0.8. The farther away from the center point, the smaller the weighted coefficient and the smaller the impact on the mean of the center point. The default window size is 10 (k = 5), that is, the average value of the current point is affected by a total of ten data points before and after. Through actual data verification, when the sliding window size is 10, the filtering effect is the best. In practical applications, the filtering window size can be adjusted as needed. While performing smoothing filtering, try to retain the waveform change characteristics as much as possible. In actual measurements, the filtering window size can also be adjusted according to the differences in equipment.

[0027] Step 3: Identify the peaks by slope change. The characteristic of a peak is that the slope before the peak is positive and the slope after the peak is negative. Filter out misidentified peaks through threshold judgment. After experiments, the reasonable slope intervals are [-40, -10] and [10, 40], and the reasonable slope intervals include but are not limited to this. Since multiple peaks and valleys are collected at one oscillation, in order to extract the complete waveform, a minimum interval will be set between the peaks. Since the oscillation frequency is between 0.003 and 0.03 Hz, the oscillation period fluctuates between 30 seconds and 5 minutes, and the time for one breath is between 3 - 5 seconds, the default minimum step between two identified peaks is 7 sampling points (i.e., 21 - 35 seconds); all data points between two peaks are regarded as a complete waveform, and the measurement ends when two complete waveforms (i.e., three peaks) are extracted.

[0028] Step 4: Calculate the metabolic feature value, which is the average value of one complete waveform , where n represents the number of sampling points included in one waveform, represents the value of the i-th sampling point, and then determine the final evaluation value according to the average values calculated from the two waveforms. The average calculated evaluation value is . . When the difference in the means of two complete waveforms is less than 10%, we consider their means to be similar, that is, the change is less than 10%, -10% , and at this time the weight coefficient , that is, the final evaluation value is equal to the average of the means of the two complete waveforms; when the difference in the means of two complete waveforms is greater than 10%, determine the weighted coefficient according to the wavelength. The waveform with a longer wavelength has a greater weight. The longer the wavelength of the waveform, the smaller the probability of being disturbed by external factors, , where Wavelength[1] is the wavelength of the first complete waveform and Wavelength[2] is the wavelength of the second complete waveform.

[0029] Step Five: Save and output the final evaluation value.

[0030] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A rapid extraction method for metabolic features based on the recognition and extraction of complete waveforms, characterized in that: Filtering and noise reduction are performed during measurement through a sliding window filtering method based on the number of breaths, smoothing the fluctuation curve, filtering out abnormal fluctuations caused by external factor disturbances, identifying the peak through the slope, extracting the complete waveform, judging the mean difference of the complete wave and calculating the metabolic characteristic value, judging when to end the measurement according to the number of waveforms, and calculating the values of metabolic parameters including respiratory quotient RQ and resting energy expenditure REE based on all data points within two complete waveforms after the measurement ends; the measurement of the metabolic parameters is based on indirect calorimetry, and the indirect calorimetry calculates the metabolic parameters by collecting oxygen consumption VO2 and carbon dioxide VCO2; The sliding window filtering method based on the number of breaths includes regarding each breath as a sampling point, taking a current sampling point as the center, determining the window size 2k, which altogether contains 2k + 1 data points, and the metabolic parameter value of the current sampling point is equal to the average of each point within the window; The method for identifying the complete waveform sets a minimum interval between peaks, such that the minimum step length between two identified peaks is 7 sampling points, that is, 21 to 35 seconds; all data points between two peaks are counted as one complete waveform, and the wavelength Wavelength[i] is recorded, and when two complete waveforms are collected, the measurement ends.

2. The rapid extraction method of metabolic features based on complete waveform recognition and extraction according to claim 1, characterized in that: The metabolic parameter value of the current sampling point being equal to the average of each point within the window is: ， Among them represents the i-th metabolic parameter value. According to the window size, when the number of data points is greater than 2k + 1, filtering starts from the (k + 1)-th point, and the first k data points are not filtered; when the filtering window slides to the end and the number of remaining points is less than the window size, filtering is based on the remaining points, that is, the filtered value x last of the last point is equal to the average of the first k points, or a weighted average is implemented within the window , where , , , represent the weighting coefficients. The current sampling point is 1. For other points, 0 < weighting coefficient , and the farther away from the current sampling point, the smaller the weighting coefficient, that is, the smaller the influence on the metabolic parameter value of the current sampling point.

3. A rapid extraction method for metabolic features based on complete waveform recognition and extraction according to claim 1, characterized in that: When two complete waveforms are collected and the measurement ends, the data of several sampling points included in the complete waveform are averaged, and the average value of one complete waveform , where n represents the number of sampling points included in one waveform, represents the data of the i-th sampling point, and then the evaluation value is calculated by averaging the average values calculated from the two waveforms Among them, , when the means of the two complete waveforms are similar, When the mean difference between the two complete waveforms is greater than 10%, the weighting coefficient is determined by the wavelength: , where Wavelength[1] is the wavelength of the first complete waveform and Wavelength[2] is the wavelength of the second complete waveform.

Images