Anesthesia state evaluation system and method suitable for patients of different ages
Technical Field
The invention relates to the field of neural signal analysis, in particular to an anesthesia state evaluation system and method suitable for patients of different ages.
Background
The problems of over-deep anesthesia in the operation, cognitive disorder after the operation and the like caused by difficult evaluation of the anesthesia state of the old are always difficult problems in the field of neuroscience. The electroencephalogram double-frequency index widely used in hospitals can accurately monitor the anesthesia state of adults, but problems frequently occur on the elderly over 60 years old, so that how to accurately evaluate the anesthesia state of the elderly has very important medical and scientific significance.
Studies have shown that the higher the age, the higher the sensitivity to anesthetic drugs, the half of the anesthetic dose required to achieve the same anesthetic state in older patients may be that required for younger patients. The lower anaesthetic needs of elderly patients are due to the decline of physical skills such as cardiovascular and cerebrovascular, respiratory, liver and kidney caused by aging. Although these factors certainly work, the main site of anesthesia is the central nervous system. At this stage, the determination of the anesthesia status is usually performed by analyzing the variation of the thermodynamic index, such as ordering entropy, approximate entropy, complexity, etc., for the analysis of the nervous system and the monitoring of the brain status. However, these methods only consider the signal sequence in the time sequence, and ignore the amplitude of the signal at each point, and lose a large amount of effective neural signal information, so that the calculation result is inaccurate, and there is a deviation from the real information of the neural signal.
The above analysis shows that the existing method cannot accurately evaluate the anesthesia status of people of different ages by using the same anesthesia index, and therefore, the application is provided.
Disclosure of Invention
The invention aims to provide an anesthesia state evaluation system and method suitable for patients of different ages. The invention carries out weighting calculation from two scales of time sequence and amplitude, can obtain real information in nerve signals, can accurately extract the anesthesia characteristic indexes of people, and can more accurately judge the anesthesia states of patients of different ages by adopting a three-classification algorithm.
In order to achieve the above purpose of the present invention, the following technical solutions are adopted:
an anesthesia state assessment system for patients of different ages, comprising: the system comprises an electroencephalogram signal acquisition module, a front-end signal processing module, a parameter calculation module and an anesthesia depth evaluation module;
the electroencephalogram signal acquisition module: the device is used for collecting forehead two-channel electroencephalogram signals of patient samples of different ages, and respectively intercepting the electroencephalogram signals of each patient in three periods of a waking period, an anesthesia period and a recovery period;
the front-end signal processing module: the electroencephalogram data acquisition and processing device is used for carrying out band-pass filtering on the electroencephalogram signals, carrying out segmentation processing on the electroencephalogram signals, calculating an overall electroencephalogram mean value and a mean value of each data segment, comparing the mean value of each data segment with the overall electroencephalogram mean value, and deleting the data segment if the deviation exceeds 30%; then comparing the average value between each data segment and the data segment adjacent to the data segment, and deleting the data segment if the deviation exceeds 50%;
the parameter calculation module: the device is used for calculating the double-scale entropy of each data segment and averaging to obtain the characteristic values of each patient in three states of a waking period, an anesthesia period and a recovery period;
the anesthesia depth evaluation module: the method is used for performing pattern recognition SVM three-classification by taking the characteristic values of each patient in three states of a waking period, an anesthesia period and a recovery period as input values and taking RBF as a kernel function, and the obtained classification result is the detection result of the anesthesia depth.
Further, in the preferred embodiment of the present invention, the electroencephalogram signal of each patient in the waking period, the anesthesia period and the recovery period is intercepted in the electroencephalogram signal acquisition module, and the length is 2min respectively.
Further, in a preferred embodiment of the present invention, in the front-end signal processing module, the performing a segmentation process on the electroencephalogram signal includes: the length of 12s is taken as one section, the overlapping rate of each data section and the previous data section is 75%, and the electroencephalogram signal data with the duration of 2min is divided into 13 sections.
Further, in a preferred embodiment of the present invention, in the front-end signal processing module, the electroencephalogram signal is subjected to band-pass filtering, a notch is used to remove 50Hz power frequency interference, filtering is performed to obtain a 0.1-45Hz signal, and finally data is down-sampled from 128Hz to 100 Hz.
Further, in a preferred embodiment of the present invention, in the front-end signal processing module, the mean value of each data segment is compared with the overall electroencephalogram mean value, if the deviation exceeds 30%, the data segment is deleted, and the step of performing segmentation processing on the electroencephalogram signal is returned again; if the deviation is less than 30%, the next step is carried out.
Further, in a preferred embodiment of the present invention, in the front-end signal processing module, the average value between each data segment and the data segment adjacent to the data segment is compared, and if the deviation exceeds 50%, the data segment is deleted, and the step of performing the segmentation processing on the electroencephalogram signal is resumed; if the deviation is less than 50%, the next step is carried out.
Further, in a preferred embodiment of the present invention, in the parameter calculation module, the method for calculating the two-scale entropy of each data segment includes:
let x (n) be xiI is 1, … …, and N is data length;
x is to bei(j ═ 1, 2 …, N) toc categories, labeled 1 to c;
mapping x to y ═ y using Normal Cumulative Distribution (NCDF)
1,y
2,…,y
NIn the interval of [0,1 ]](ii) a Re-use linear algorithm for each y
jAn integer from 0 to c is assigned,
represents the jth classification time series;
creating an embedded vector with m-dimension and d-time delays
Each one of which is
Are all mapped to an ordered pattern
Wherein
The signal has m members, and each member is one of integers 1 to c, which can be assigned to each time series
C is the number of possible sort patterns
m;
For cmFor each possible ranking mode, the relative frequency is calculated as follows:
the two-scale entropy is calculated as follows:
further, in a preferred embodiment of the present invention, in the anesthesia depth evaluation module, a three-classification model OVO-SVMs is constructed according to an SVM two-classification model, the three-classification problem is divided into three SVM-based two-classification problems, two-by-two division is performed on three states to obtain three cases { C1(a, b), C2(a, C), and C3(b, C) }, and in the test, the result is voted, where a is 0;
c1, if a win, a ═ a + 1; further, b ═ b + 1;
c2, if a win, a ═ a + 1; further, c ═ c + 1;
c3, if b win, b ═ b + 1; further, c ═ c + 1;
the result is the maximum of a, b and c.
An anesthesia state evaluation method suitable for patients of different ages is based on the anesthesia state evaluation system and comprises the following steps:
the electroencephalogram signal acquisition module is used for acquiring electroencephalogram signals of a patient under general anesthesia in different anesthesia states;
segmenting the electroencephalogram signal by using the front-end signal processing module, and preprocessing to remove interference;
calculating the double-scale entropy of each data segment by using the parameter calculation module to obtain the characteristic values of each patient in three states of a waking period, an anesthesia period and a recovery period;
and by utilizing the anesthesia depth evaluation module, the characteristic values of each patient in three states of a waking period, an anesthesia period and a recovery period are used as input values, RBF is used as a kernel function, three classifications of the pattern recognition SVM are carried out, and the obtained classification result is the detection result of the anesthesia depth.
The invention has the following effects:
1. in data extraction, noise with abnormal amplitude is eliminated by data overlapping, calculating and comparing the mean values of all sections, the EEG noise is removed more reasonably and conveniently, long data are subdivided into short data, and transverse and longitudinal mean values are compared, so that the whole numerical value is more reasonable.
2. On the basis of the ordering entropy, the amplitude is added into the calculation of the ordering entropy, and the weighting calculation is carried out from two scales of the time sequence and the amplitude, so that the index is changed into the time-amplitude dual-scale signal entropy, the defect that the amplitude information in the electroencephalogram signal is lost by the conventional entropy algorithm is avoided, the real information in the nerve signal can be obtained, the anesthesia characteristic index of the person can be accurately extracted, and the intraoperative anesthesia states of the persons of different ages can be more comprehensively analyzed.
3. The invention adopts a three-classification algorithm, increases the recovery period as the third classification state on the basis of two classifications of anesthesia and wakefulness, and can more accurately judge the anesthesia states of patients of different ages. Therefore, the method has important significance in classifying the anesthesia states of patients of different ages.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention;
FIG. 2 is a diagram illustrating the training results of an SVM in accordance with an embodiment of the present invention;
FIG. 3 is a diagram of the effect of matching conventional ordering entropy to age;
FIG. 4 is a diagram illustrating the effect of fitting the dual-scale entropy to age according to an embodiment of the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to examples, but it will be understood by those skilled in the art that the following examples are only illustrative of the present invention and should not be construed as limiting the scope of the present invention. The examples, in which specific conditions are not specified, were conducted under conventional conditions or conditions recommended by the manufacturer. The reagents or instruments used are not indicated by the manufacturer, and are all conventional products available commercially.
An anesthesia state assessment system for patients of different ages, comprising: the system comprises an electroencephalogram signal acquisition module, a front-end signal processing module, a parameter calculation module and an anesthesia depth evaluation module;
the electroencephalogram signal acquisition module: the device is used for collecting forehead two-channel electroencephalogram signals of patient samples of different ages, and respectively intercepting the electroencephalogram signals of each patient in three periods of a waking period, an anesthesia period and a recovery period;
the front end signal processing module: the electroencephalogram data acquisition system is used for carrying out band-pass filtering on electroencephalogram signals, carrying out segmentation processing on the electroencephalogram signals, calculating an overall electroencephalogram mean value and a mean value of each data segment, comparing the mean value of each data segment with the overall electroencephalogram mean value, and deleting the data segment if the deviation exceeds 30%; then comparing the average value between each data segment and the data segment adjacent to the data segment, and deleting the data segment if the deviation exceeds 50%;
a parameter calculation module: the device is used for calculating the double-scale entropy of each data segment and averaging to obtain the characteristic values of each patient in three states of a waking period, an anesthesia period and a recovery period;
an anesthesia depth evaluation module: the method is used for performing pattern recognition SVM three-classification by taking the characteristic values of each patient in three states of a waking period, an anesthesia period and a recovery period as input values and taking RBF as a kernel function, and the obtained classification result is the detection result of the anesthesia depth.
The present embodiment is a method based on the anesthesia status evaluation system, and referring to fig. 1, the method includes the following steps:
the method comprises the following steps of firstly, acquiring electroencephalogram signals of a patient under general anesthesia in different anesthesia states by using an electroencephalogram signal acquisition module, and specifically comprising the following steps:
(1) 124 forehead double-channel discrete electroencephalogram signals of patients of all ages are collected through Bispectral, 50 of the signals between 30 and 60 are named as adult groups, and 74 of the signals between 60 and 85 are named as old groups.
(2) And respectively intercepting the electroencephalogram signals of each patient in three periods of a waking period, an anesthesia period and a recovery period, wherein the duration of each period of the electroencephalogram signals is 2 min.
Note that the awake period refers to the state of the patient before administration; the anesthesia period refers to a period of time after administration until the patient is stopped; the recovery period refers to a period of time after the patient is completely awake after stopping taking the drug.
Segmenting the electroencephalogram signal by using a front-end signal processing module, and preprocessing to remove interference, wherein the method specifically comprises the following steps:
(1) performing band-pass filtering on the electroencephalogram signals, removing 50Hz power frequency interference by using trapped waves, then filtering to obtain 0.1-45Hz signals, and finally down-sampling the data from 128Hz to 100 Hz.
(2) The electroencephalogram signal is processed in a segmented mode, the length of 12s is taken as one segment, the overlapping rate of each data segment and the previous segment of data is 75%, the electroencephalogram signal data with the duration of 2min are divided into 13 segments, and the integral electroencephalogram mean value and the mean value of each data segment are calculated.
(3) Comparing the mean value of each data segment with the whole electroencephalogram mean value, deleting the segment if the deviation exceeds 30%, and returning to the step (2) again; if the deviation is less than 30%, the next step is carried out.
(4) Comparing the mean value between each data segment and the data segment adjacent to the data segment, if the deviation exceeds 50%, deleting the data segment, and returning to the step (2) again; if the deviation is less than 50%, the next step is carried out.
Calculating the double-scale entropy of each data segment by using a parameter calculation module to obtain the characteristic values of each patient in three states of a waking period, an anesthesia period and a recovery period;
the method for calculating the double-scale entropy of each data segment in the step comprises the following steps:
let x (n) be xiI is 1, … …, and N is data length;
x is to bei(j ═ 1, 2 …, N) maps to c classes, labeled 1 through c;
mapping x to y ═ y using Normal Cumulative Distribution (NCDF)
1,y
2,…,y
NIn the interval of [0,1 ]](ii) a Re-use linear algorithm for each y
jAn integer from 0 to c is assigned,
represents the jth classification time series;
creating an embedded vector with m-dimension and d-time delays
Each one of which is
Are all mapped to an ordered pattern
Wherein
The signal has m members, and each member is one of integers 1 to c, which can be assigned to each time series
C is the number of possible sort patterns
m;
For cmFor each possible ranking mode, the relative frequency is calculated as follows:
the two-scale entropy is calculated as follows:
and fourthly, using an anesthesia depth evaluation module to take the characteristic values of each patient in three states of a waking period, an anesthesia period and a recovery period as input values, and taking RBF as a kernel function to perform SVM three-classification, wherein the obtained classification result is the detection result of the anesthesia depth.
In the step, the obtained characteristic values of three states of 124 patients are used as input states, wherein 80 persons are used as training samples, 44 persons are used as detection samples, SVM three-classification is carried out by taking RBF as a kernel function, and according to the training result, an anesthesia state monitoring index with obvious dependence on age is obtained.
And constructing a three-classification model OVO-SVMs according to the SVM two-classification model, dividing the three-classification problem into three SVM-based two-classification problems, and dividing the three states pairwise to obtain three conditions { C1(a, b), C2(a, C) and C3(b, C) }. When in testing, voting is carried out on the result, wherein a is 0;
c1, if a win, a ═ a + 1; further, b ═ b + 1;
c2, if a win, a ═ a + 1; further, c ═ c + 1;
c3, if b win, b ═ b + 1; further, c ═ c + 1;
the result is the maximum of a, b and c.
As shown in FIG. 2, OVO three classification models were constructed from LIBSVM using RBF as kernel function, where 80 was human training set and 44 was human testing set, and the training results showed an accuracy of 97.561.
The obtained electroencephalogram data of the patient are respectively calculated into the double-scale entropy and the ordering entropy, linear fitting is carried out by taking the age as the abscissa, the results are shown in fig. 3 and fig. 4, and the fitting results show that the R, F and the p value of the double-scale entropy age fitting function are superior to the ordering entropy due to the consideration of the amplitude scale, so that the evaluation method for calculating the double-scale entropy provided by the invention is more suitable for age dependency analysis, and can be used for more comprehensively analyzing the intraoperative anesthesia states of people of different ages.
The above-mentioned embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements made to the technical solution of the present invention by those skilled in the art without departing from the spirit of the present invention shall fall within the protection scope defined by the claims of the present invention.