Opioid detection using EEG signals

Abstract

A system (1) for detecting the presence of a medicament in a human body (2) comprises an EEG sensor (10) and an analysis unit (11) having a processor arrangement (110) and a memory (111) storing instructions that, when executed by the processor arrangement (110), cause the analysis unit (11) to perform the steps of: receiving a signal (300-307) comprising a time series of values (310-313) using the EEG sensor (10); analyzing the values (310-313) of the signals (300-307) to convert the signals (300-307) into a sequence of symbols (Si); generating a plurality of words (500-504), each word (500-504) comprising a plurality of consecutive symbols (Si); and providing the plurality of words (500-504) and / or one or more variables derived from the plurality of words (500-504) to a classifier (600), the classifier (600) being configured to indicate the presence of a medicament in the human body (2) based on the plurality of words (500-504).

Description

[0001] This invention relates to systems and methods for detecting the presence of drugs in the human body, as well as to computer-readable storage media.

[0002] Electroencephalography (EEG) is an electrogram that records the spontaneous electrical activity of the human brain. EEG is used not only in cases of epilepsy but also, for example, to assess the depth of anesthesia in patients.

[0003] PL Purdon et al. described “Electroencephalogram signatures of loss and recovery of consciousness from propofol” in Proc. Natl. Acad. Sci., 110(12), E1142-E1151 in 2013. The authors recorded high-density EEG data from individuals during gradual propofol induction and awakening from unconsciousness. During deep unconsciousness, the alpha amplitude was found to be maximum at a low-frequency peak, while during entry into and exit from unconsciousness, the alpha amplitude was maximum at a low-frequency minimum.

[0004] EEG has several advantages over other techniques such as MRT: hardware costs can be lower and the required hardware can have a small shape factor. In addition, EEG is relatively tolerant of subject movement, is silent, and is non-invasive.

[0005] Therefore, it is expected that EEG will be used for other purposes.

[0006] EP 2 906 112 B1 describes systems and methods for monitoring and controlling the state of a patient, and more specifically, systems and methods for monitoring and controlling the state of a patient receiving a dose of an anesthetic compound (or, more commonly, a dose of anesthesia). Thresholding is used to analyze EEG datasets to generate binary time series from the EEG datasets, and a burst inhibition probability algorithm implementing a state-space model is used to generate a time series of burst inhibition probabilities from the binary time series, wherein the burst inhibition probability is correlated with the instantaneous probability that the patient's brain is in an inhibited state. Furthermore, the time series of burst inhibition probabilities is used to determine a characteristic spectrum consistent with the administration of at least one drug with anesthetic properties.

[0007] Burst inhibition is an EEG pattern characterized by alternating periods of high-voltage electrical activity with periods of inactivity or minimal brain activity. This pattern is seen in patients in a state of brain inactivation, such as under general anesthesia or in a coma.

[0008] For example, the effects of propofol can be monitored using the method described in EP 2 906 112 B1. However, opioids such as remifentanil only cause burst inhibition at extremely high doses, far exceeding the doses normally administered to patients. Therefore, known methods are limited to certain drugs and dose rates.

[0009] The purpose of this invention is to expand the application scope of EEG.

[0010] This objective is achieved by a system having the features of claim 1.

[0011] Therefore, a system for detecting the presence of a drug in the human body includes an EEG (electroencephalography) sensor and an analysis unit. The analysis unit includes a processor arrangement and a memory. The memory stores instructions that, when executed by the processor arrangement, cause the analysis unit to perform the following steps: receiving a time-series signal including values ​​using the EEG sensor; analyzing the values ​​of the signal to convert the signal into a sequence of symbols. The instructions, when executed by the processor arrangement, also cause the analysis unit to generate multiple words, where each word includes multiple (e.g., consecutive) symbols; and to provide (e.g., a trained) classifier to the generated multiple words and / or one or more variables derived from the multiple words (500-504), the classifier being configured to indicate the presence of a drug in the human body based on the multiple words.

[0012] This is based on the finding that the presence of remifentanil or other opioids in human blood is highly significantly correlated with the occurrence of a specific word (e.g., among several possible words). Therefore, the classifier can perform a simple determination of the proportion or percentage of one or more predetermined words, or the classifier can include a trained machine learning module. With the aid of the described system, the application of EEG can be greatly extended to the monitoring of opioids and other drugs. One or more variables derived from multiple words can indicate the distance between consecutive symbols of the corresponding words and / or the integral of the signal between two consecutive symbols. The symbols can indicate zero crossings (alternatively or additionally, the direction of crossing zero or another value).

[0013] Analyzing the values ​​of a signal may include comparing the signal values ​​with predefined symbols to convert the signal into a sequence of symbols (e.g., a time series of symbols). The sequence of symbols may include the same number (or less) of the values ​​included in the signal.

[0014] An EEG sensor can be a frontal EEG sensor used to sense frontal cortical activity. It has been found that improved accuracy can be obtained by measuring frontal cortical activity, for example, using an EEG sensor configured to be adhesively or otherwise mounted to the forehead. An EEG sensor has multiple electrodes. For example, an EEG sensor has at least or exactly three electrodes.

[0015] The signal can be the original signal or include the original signal. This allows for simple analysis. Alternatively or additionally, the signal can be a preprocessed signal or include the preprocessed signal. The preprocessed signal can be obtained by preprocessing the original signal. This allows for improved accuracy of the results.

[0016] Preprocessing the raw signal may include bandpass filtering to obtain one or more frequency bands. For example, the frequency bands may be α, β1, β2, θ, δ, and γ bands, or one or more selected from them. The preprocessed signal may include one or more frequency bands, particularly the α, β1, β2, θ, δ, and / or γ bands. In this way, optimized results for a specific drug can be obtained.

[0017] Preprocessing the raw signal may include determining at least one signal envelope (e.g., of the raw signal or of one or more frequency bands (e.g., all frequency bands)). The preprocessed signal may be or include at least one signal envelope. It has been found that, for certain drugs, such as remifentanil, using an envelope can improve the accuracy of measurements.

[0018] Alternatively or additionally, preprocessing the original signal may include determining at least one phase signal.

[0019] Each of the predefined symbols can be assigned to a corresponding range of signal values. This reduces complexity and increases processing time. Furthermore, it allows for meaningful results to be obtained with a relatively small number of samples from the available data.

[0020] The width of at least one of the ranges can be equal to one standard deviation (or multiple standard deviations) of the corresponding signal. This allows for the reduction of analytical complexity while maintaining the informative content of the data.

[0021] For example, at least two, three, four, or even more different symbols are predefined. That is, in a very simple example, only two different symbols are defined, and each value of the signal is assigned to one of the two symbols. For example, the two symbols could be "high" for values ​​above zero and "low" for values ​​below zero.

[0022] In the example, each of the multiple words includes the same number of consecutive symbols, for example, two, three, four or more consecutive symbols. This allows for simple analysis that produces significant results.

[0023] The classifier can be configured to determine the frequency of word occurrences. It has been found that the occurrence of specific words is associated with the presence of certain drugs, particularly remifentanil, in humans.

[0024] Alternatively or additionally, the classifier can be configured to specifically compute... Entropy is used to determine the entropy of multiple words. It was also found that the entropy of multiple words correlates with the presence of certain drugs, particularly remifentanil, in the human body, and allows for particularly significant results.

[0025] The classifier may include a classifier module trained using signals obtained from drug-free individuals as a first dataset, and signals obtained from individuals administered opioids, particularly remifentanil, as a second dataset. Alternatively or additionally, the classifier module may be trained using signals obtained from drug-free individuals as a first dataset, and signals obtained from individuals administered anesthetics, particularly propofol, as a second dataset. This allows for the determination of high significance.

[0026] According to one aspect, a method for detecting the presence of a drug in a human body is provided. The method can use a system based on any aspect or embodiment described herein. The method includes the following steps performed, for example, by an analysis unit communicatively coupled to an EEG sensor (e.g., the system of any of the examples above): receiving a signal comprising a time series of values ​​using the EEG sensor; analyzing the values ​​of the signal to convert the signal into a sequence of symbols (e.g., comparing the values ​​of the signal with predefined symbols to convert the signal into a time series of symbols); generating a plurality of words, each word comprising a plurality of (e.g., consecutive) symbols; and providing the plurality of words (and / or one or more variables derived from the plurality of words) to a classifier configured to indicate the presence of a drug in a human body based on the plurality of words. Regarding the advantages of this method, refer to the above description of the system.

[0027] The classifier may include a classifier module trained using signals obtained from a drug-free human body as a first dataset, and signals obtained from a human body administered an opioid, particularly remifentanil, as a second dataset. Alternatively or additionally, the classifier module may be trained using signals obtained from a drug-free human body as a first dataset, and signals obtained from a human body administered an anesthetic, particularly propofol, as a second dataset. The method may include training the classifier module before detecting the presence of the drug in the human body.

[0028] According to one aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions that, when executed by a processor of an analysis unit (e.g., the system described above), cause the analysis unit to perform any example of the methods described above. Regarding advantages, refer again to the description of the system above.

[0029] The basic idea of ​​the invention will then be described in more detail with reference to the embodiments shown in the accompanying drawings, in which:

[0030] Figure 1 A system for detecting the presence of drugs in the human body is shown;

[0031] Figure 2 The original signals obtained using an EEG sensor and various preprocessed signals with different frequency bands are shown.

[0032] Figure 3 It shows Figure 2 One of the preprocessed signals and the envelope of the preprocessed signal;

[0033] Figure 4 It shows Figure 3 A portion of the preprocessed signal, wherein each of its values ​​is assigned to one of a plurality of predetermined symbols;

[0034] Figure 5 The training of the classifier and its application to test data are illustrated.

[0035] Figure 6 Training is shown Figure 6 The steps of classifying a classifier;

[0036] Figure 7 The steps of the optimization process for analyzing EEG signals are shown;

[0037] Figure 8A This shows portions of the pretreatment signals obtained from human subjects who had received and had not received the drug;

[0038] Figure 8B It shows Figure 5 A portion of the preprocessed signal of A, wherein, additionally, the random number signal is shown; and

[0039] Figure 9 Examples of 64 possible words are shown, consisting of three symbols that can have four values.

[0040] The system for detecting the presence of a drug in the human body will then be described in detail. The embodiments described herein should not be construed as limiting the scope of the invention.

[0041] Figure 1 A system 1 configured to detect the presence of a drug in a human body 2 is shown. The system 1 includes an electroencephalogram (EEG) sensor 10 and an analysis unit 11 coupled to the EEG sensor 10.

[0042] The EEG sensor 10 in this example is configured to sense frontal cortical activity. For this purpose, the EEG sensor 10 is formed to be attached to the forehead 20 of the human body 2. The EEG sensor 10 includes a plurality of electrodes 100A, 100B, 100C, i.e., at least two electrodes 100A-100C. Figure 1 In the example, the EEG sensor 10 includes three electrodes 100A-100C. Each of the electrodes 100A-100C has an adhesive layer and can be bonded to the forehead 20 of the human body 2. The EEG sensor 10 is elongated. The electrodes 100A-100C are arranged in a row. The EEG sensor 10 is made of a flexible material. Here, each of the electrodes 100A-100C of the EEG sensor 10 is mounted on a corresponding pad. The pad of the first electrode 100A of the electrodes 100A-100C is connected to the pad of the third electrode 100C of the electrodes 100A-100C via the pad of the second electrode 100B of the electrodes 100A-100C. Figure 1 In this example, the first electrode 100A is configured to be attached to the central portion of the forehead 20 of the human body 2. The third electrode 100C forms the end of the EEG sensor 10 and is configured to be attached to one side of the forehead 20 of the human body 2. The second sensor 100B is configured to be attached to the forehead 20 at a location between the two other electrodes 100A and 100B.

[0043] EEG sensor 10 is operatively connected to analysis unit 11. Here, each of electrodes 100A-100C is electrically connected to analysis unit 11. Analysis unit 11 is configured to receive EEG signals from EEG sensor 10, for example, as a voltage signal between two electrodes of electrodes 100A-100C. Alternatively, EEG sensor 10 can be connected to analysis unit 11 in any other wired or wireless manner.

[0044] The analysis unit 11 (also referred to as the analyzer) includes readout electronics 113, which is electrically connected to the EEG sensor 10, and more specifically, to electrodes 100A-100C. Readout electronics 113 includes a voltmeter for measuring the voltage between the two electrodes 100A-100B. Furthermore, readout electronics 113 includes an analog-to-digital converter (ADC). The ADC receives an analog signal from the voltmeter and outputs a corresponding digital signal.

[0045] The analysis unit 11 also includes a processor arrangement 110 and a memory 111. The processor arrangement 110 includes one or more processors. The memory 111 is a computer-readable storage medium. The memory 111 stores instructions for execution by the processor arrangement 110.

[0046] Processor arrangement 110 is configured to process signals received from EEG sensor 10. In this example, processor arrangement 110 is configured to receive digital EEG signals. Analysis unit 11 also includes display 112. Processor arrangement 110 can be configured to display EEG signals on display 112. Processor arrangement 110 can also be configured to display on display 112 the result of detecting the presence of a drug in human body 2, which will be described in further detail below.

[0047] Figure 2 Graphs of various signals 300-306 examples obtained using EEG sensor 10 are shown. Each graph shows the amplitude of the signal on the y-axis versus the number of samples on the x-axis. The ADC is configured to sample the analog signal at a given sampling frequency. In this example, the sampling frequency is 256 Hz. The number of samples divided by the sampling frequency equals the time span, so the plotted window corresponds to approximately 5 seconds. The amplitude can be in any unit. In this example, the amplitude is shown in microvolts (μV).

[0048] Figure 2 The topmost figure (labeled "EEG") shows the EEG signal 300, encompassing the entire EEG frequency band, from 0.5 Hz to 45 Hz in this example (where the frequency resolution is 0.5 Hz). Signal 300 may have undergone artifact and noise removal. Signal 300 contains all the EEG information and is therefore referred to as the original signal 300 below for ease of reference.

[0049] Figure 2 Other figures in the diagram illustrate the bandpass filtered portion of the original signal 300. Therefore, these other figures illustrate the various preprocessed signals 301-306 obtained from the original signal 300.

[0050] The second figure from the top shows the preprocessed signal 301 corresponding to the delta band from 0.5 Hz to 4 Hz.

[0051] The third figure from the top shows the preprocessed signal 302 corresponding to the θ band from 4.5 Hz to 8 Hz.

[0052] The fourth figure from the top shows the preprocessed signal 303 corresponding to the α band from 8.5 Hz to 13 Hz.

[0053] The fifth figure from the top shows the preprocessed signal 304 corresponding to the β1 band from 13.5 Hz to 21 Hz.

[0054] The sixth figure from the top shows the preprocessed signal 305 corresponding to the β2 frequency band from 21.5 Hz to 30 Hz.

[0055] The seventh figure from the top shows the preprocessed signal 306 corresponding to the γ band from 30.5 Hz to 45 Hz.

[0056] When the instructions stored in memory 111 are executed by processor arrangement 110, the instructions cause analysis unit 11 to obtain raw signal 300 by reading out EEG sensor 10 using readout electronics 113 (and optionally performing artifact and noise removal and / or filtering to remove frequencies outside the α, β1, β2, θ, δ, and γ bands). Furthermore, when the instructions stored in memory 111 are executed by processor arrangement 110, the instructions cause analysis unit 11 to perform bandpass filtering on raw signal 300 to obtain various preprocessed signals 301-306 according to different EEG bands α, β1, β2, θ, δ, and γ.

[0057] Analysis unit 11 is configured to determine the signal envelope of signals 300-306 to obtain another preprocessed signal 307, for example, Figure 3 As shown.

[0058] Figure 3 A time-based graph on the x-axis is shown, illustrating a portion of the β1 band signal 304 and its envelope. Another preprocessed signal 307 includes this envelope. The envelope encloses signal 304. A maximum value of the envelope encloses multiple maximum values ​​of signal 304. The envelope encloses signal 304 on the positive side.

[0059] The envelope can be computed using the Hilbert transform. For example, analysis unit 11 (by executing corresponding instructions in memory) is configured to compute the modulus of a known Hilbert transform, which is defined using the Cauchy principal value (denoted by pv):

[0060]

[0061] Analysis unit 11 can be configured (e.g., by calculating the corresponding Hilbert transform) to calculate the envelopes of the original signal 300 and / or one or more (e.g., all) bandpass filter bands α, β1, β2, θ, δ, γ. According to the example, further processing of all seven signals (the original signal plus each band) and their seven envelopes, i.e., a total of 14 inputs to be further analyzed, is performed as described. Alternatively, only one or some of these inputs (e.g., predefined inputs or a set of inputs for a given drug) may be used for further analysis.

[0062] Each of the signals includes a time series of values. Therefore, when the instruction is executed by the processor arrangement 110, the analysis unit 11 uses the EEG sensor 10 to receive signals 300-307, which include a time series of values ​​310-313.

[0063] Figure 4 A portion of signal 304 (samples 10 to 24) is shown. Signal 304 includes a time series of values ​​311-313.

[0064] Such as by means of Figure 4 As shown, when executed by processor arrangement 110, the instructions cause analysis unit 11 to analyze the values ​​310-313 of signal 304 to convert signal 304 into a sequence of symbols, or more specifically, to compare the values ​​310-313 of signal 304 with predefined symbols Si to convert signal 304 into a time sequence 400 of symbols.

[0065] exist Figure 4 In the example shown, signal 304 is illustrated, but any of the original or preprocessed signals 300-307 that include any one (e.g., all) of the envelope signal 307 of signals 300-306 can be processed accordingly.

[0066] Each of the predefined symbols Si is assigned a signal value (here: amplitude) within a corresponding range. In a specific example, the width of some ranges is equal to the standard deviation of the corresponding signal 304. That is, the symbol Si is defined based on the standard deviation of signal 304.

[0067] exist Figure 4In the example, four symbols Si are predefined (other numbers, such as two, three, five, or more, can also be conceived). Here, these symbols Si are simply labeled "1", "2", "3", and "4". Two symbols (symbols 2 and 3) are defined to cover the corresponding range from zero (amplitude or envelope amplitude) to + / - "a" multiplied by the standard deviation (e.g., one standard deviation). The other two symbols are defined to cover the corresponding range above / below + / - "a" multiplied by the standard deviation (e.g., one standard deviation).

[0068] Therefore, although the first three values ​​311-313 are different from each other, they all fall within the range of symbol 1. Thus, the first three values ​​311-313 are each assigned symbol 1, and the time series 400 of the symbol begins with three times the value of symbol 1. The next value falls within the range of symbol 3, and this also applies to the following two values, which are followed by the values ​​corresponding to symbol 2, and so on.

[0069] For further processing, the time series 400 with symbols is used.

[0070] Next, when the instructions are executed by the processor arrangement 110, the analysis unit 11 generates multiple words 500-504, each word 500-504 including multiple consecutive symbols Si.

[0071] exist Figure 4 Below the time series 400 in the diagram and symbols, some words 500-504 are shown. Words 500-504 have predefined lengths. In this example, the length is three, but other lengths are possible, such as two, four, five, or more.

[0072] Words 500-504 are defined by overlapping symbols Si. Here, each subsequent word 500-504 is defined as starting with one symbol difference. Each word 500-504 begins (and ends) one symbol further than the immediately preceding word 500-504. That is, the first word 500 begins with the first symbol Si in the symbol time sequence 400 (and contains the first, second, and third symbols Si), the second word 501 begins with the second symbol Si in the symbol time sequence 400 (and contains the second, third, and fourth symbols Si), and so on. Therefore, the number of words 500-504 corresponds to the number of symbols Si in the symbol time sequence 400 (ignoring the end of the time sequence). Alternatively, words can be defined as non-overlapping but containing adjacent groups of symbols Si. In this case, the number of words will be equal to the number of symbols Si divided by the word length. As another example, words can be shifted differently relative to each other; for example, words can overlap by only one symbol Si instead of two.

[0073] Thus, based on a predetermined number N of different symbols Si, the length L of the words 500 - 504, and the quantization a of the standard deviation, the value of the signal is analyzed and converted into a sequence of symbols Si.

[0074] When executed by the processor arrangement 110, the instructions may also cause the analysis unit 11 to determine one or more descriptors.

[0075] An example of such a descriptor is the occurrence count of the various different possible words 500 - 504. The word count for each possible word 500 - 504 can be a descriptor.

[0076] Another example of such a descriptor is the entropy of the generated plurality of words 500 - 504. To determine the entropy, the Shannon entropy of the words 500 - 504 can be calculated:

[0077]

[0078] Here, X is a random variable with a value range {x1, x2, …, xn}, and P is the occurrence probability of a given symbol.

[0079] Alternatively or additionally, the entropy of the words 500 - 504 can be calculated: Entropy:

[0080]

[0081] where q is the order of the entropy, and 0 < q < infinity, where q is not equal to 1. When q approaches 1, the limit of the entropy is the Shannon entropy.

[0082] The memory 111 also stores a classifier 600 (to be described in further detail below). The instructions, when executed by the processor arrangement 110, cause the analysis unit 11 to provide the classifier 600 with a plurality of generated words 500 - 504, where the classifier 600 is configured to indicate the presence of a drug in the human body 2 based on the plurality of words 500 - 504. The words 500 - 504 contain information about the electrical activity of the brain of the human body 2 indicating the presence of a specific drug. The classifier 600 has been trained using training data obtained when a specific drug is present in the human body and training data when no drug is present in the human body. Figure 5 As

[0083] shown, a database containing a training set of EEG data and a test set of EEG data is provided. Figure 5

[0084] ​For this example, frontal lobe EEGs were recorded and analyzed from 168 patients: 81 were awake (ND, no medication) before administration of any medication, 59 received propofol only (PROP), and 28 received remifentanil only (REMI). There was no overlap between patients and groups. A subset of patients (n=84) was balanced to balance the number of patients in each group (n=28). A single window (10 seconds) for each patient in this subset was randomly selected as the training set. The test set consisted of 15,929 ND windows, 7,289 PROP windows, and 1,855 REMI windows from all patients (n=168).

[0085] As described above and as will be referred to below. Figure 6 Analyze the training set as described further (in Figure 5 The abbreviation for symbolic dynamics is "SD". Symbolic dynamics produces descriptors, such as one or more (e.g., all) descriptors mentioned above.

[0086] These descriptors are provided to one or more classifier modules 601, 602, and 603. Examples of classifier modules 601-603 perform linear descriptor analysis (LDA), quadratic descriptor analysis (QDA), and support vector machine (SVM). Classifier 600 may include one, two, or all of classifier modules 601-603. Additional or alternatively, other classifier modules may be conceived for application. For example, the classifier module of classifier 600 may be any linear or nonlinear classifier, or any algorithm used for classification (e.g., quadratic, decision tree, ANFIS model, etc.). The classification module of classifier 600 may be an artificial intelligence module. The classification module of classifier 600 may be configured for machine learning.

[0087] At least one or at least two inputs can be provided to classifier modules 601-603. In the example, the following inputs are provided to classifier modules 601-603: the entropy (amplitude signal or envelope signal or both) of one or more of the original EEG signal 300 or any frequency component signals 301-307, and / or the entropy calculated using a combination of parameters N (number of different symbols), L (word length), and a (standard deviation). As a further input, the following can be provided to classifier modules 601-603: the occurrence of one or more given words (e.g., under the same definition (for entropy, amplitude / envelope, original EEG / frequency component, N, L, combination)).

[0088] It is worth noting that, instead of words containing the continuous symbol Si, or words other than those containing the continuous symbol Si, one or more variables derived from multiple words can be provided to classifier 600. For example, a symbol can indicate a zero crossing (or crossing another predefined value), and one or more variables can indicate the distance between consecutive symbols or the area under the signal between consecutive symbols. Optionally, the descriptor as described above can be determined based on such variables and then provided to classifier modules 601-603. Alternatively or additionally, the variables can be provided directly to classifier modules 601-603.

[0089] During training, information about the categories of the training data is also provided to classifier modules 601-603, which are ND, REMI, or PROP.

[0090] Training can be performed using the following methods: using multiple words 500-504 generated from signals 300-307 obtained from a drug-free human body 2, and using multiple words 500-504 generated from signals 300-307 obtained from a human body 2 that has been given opioids, particularly remifentanil; and / or using multiple words 500-504 generated from signals 300-307 obtained from a drug-free human body 2, and using multiple words 500-504 generated from signals 300-307 obtained from a human body 2 that has been given anesthetics, particularly propofol; and / or using multiple words 500-504 generated from signals 300-307 obtained from a human body 2 that has been given opioids, particularly remifentanil, and using multiple words 500-504 generated from signals 300-307 obtained from a human body 2 that has been given anesthetics, particularly propofol.

[0091] Therefore, the analysis unit 11 is configured to perform a method including the following steps:

[0092] - Use EEG sensor 10 to receive signals 300-307, which include a time series of values ​​310-313;

[0093] - Compare the values ​​310-313 of signals 300-307 with predefined symbols Si to convert signals 300-307 into a time series of symbols 400;

[0094] - Generate multiple words 500-504, each word 500-504 including multiple consecutive symbols Si; and

[0095] - Provide multiple words 500-504 to classifier 600, which is configured to indicate the presence of a drug in the human body 2 based on multiple words 500-504.

[0096] After training, the test data can be applied to the trained classifier 600. Figure 5 The example shown is the classification module 601.

[0097] Figure 6 The diagram illustrates how classification performance can be determined. Optionally, the training dataset is fed to the Hilbert transform module to determine the envelope. The selected input is fed to the symbolic dynamics analysis module, which also receives the number of symbols N, word length L, and quantization value (here, standard deviation a). Optimal parameters can be determined by iteratively performing analysis using variations in these parameters. The descriptors obtained through symbolic dynamics analysis are then optionally fed into a discriminant function after a logarithmic transformation. This can include selected classifier modules 601-603. The results are used to determine the true positive rate (TPR, i.e., sensitivity), the true negative rate (TNR, i.e., specificity), and / or the area under the curve (AUC).

[0098] according to Figure 7 The training set is used to determine the best descriptor and the best classification model, depending on the highest TPR, TNR, and / or AUC results. These may vary depending on the drug.

[0099] Once the optimal descriptor and classification model are determined, the test data are analyzed to obtain practical results. These results are then used to evaluate the quality of the overall analysis.

[0100] For comparison, Figure 8A Signals from a human without drug ND, a human using remifentanil (REMI), and a human using propofol (PROP) are shown.

[0101] When the number of symbols N is (where "high" means a value above 0 and "low" means a value below 0) and the word length is 2, only 4 different words can appear: high-high, high-low, low-high, and low-low.

[0102] By performing the analysis as described above, analysis unit 11 provides the following word counts:

[0103]

[0104] It can be seen that these differences are significant for the three scenarios.

[0105] In order to Figure 8B The word counts were determined to be 648, 641, 640, and 627 by comparison with the random number signal (RNG) shown in the figure.

[0106]

[0107] In each case, two parameters were used.

[0108] For example, for detecting remifentanil (or without the drug), the first parameter is of order 2. Entropy. For this purpose, the envelope ENV of the β1 band was used as input (symbol count 2, word length 2). Furthermore, the high-low occurrence counts of words from the β1 envelope were used, resulting in the sensitivity, specificity, and area under the curve indicated on the right. AMP represents amplitude, without envelope.

[0109] These results demonstrate that nonlinear parameters based on EEG symbol dynamics can identify the presence of remifentanil in the frontal lobe EEG of patients and distinguish it from the propofol effect in both the amplitude and envelope domains.

[0110] The classifier 600 can provide an output indicating the result of the classification, i.e., whether a drug (e.g., remifentanil or propofol) is present in the human body. This output can be provided on the display 112.

[0111] As an example, this allows for determining whether a patient is already in a brain state equivalent to that under opioid administration, or checking whether the effects of opioids still exist in the patient's body, before treatment begins. Furthermore, the described systems and methods can be used to indicate whether a patient has correctly declared (or has been declared) to be awake (e.g., unaffected by opioids such as remifentanil).

[0112] A word length of 2 and a symbol count of 2 result in 4 different words, while a word length of 3 and a symbol count of 4 result in 64 different words. Figure 9 The diagram shows all the different terms for this example. These parameters can produce improved results depending on the available statistics and medications. The optimal parameters are determined as described above.

[0113] The concept of the present invention is not limited to the above-described embodiments, but can be implemented in different ways.

[0114] List of reference numerals

[0115] 1 System

[0116] 10 EEG sensors

[0117] 100A-100C electrode

[0118] 11 Analysis Unit

[0119] 110 Processor Layout

[0120] 111 Memory

[0121] 112 monitors

[0122] 113 Reading out electronic devices

[0123] 2. Human body

[0124] 20 Forehead

[0125] 300 Original Signal

[0126] 301-307 Preprocessing Signals

[0127] Values ​​310-313

[0128] Time series of 400 symbols

[0129] 500-504 words

[0130] 600 classifier

[0131] 601-603 Classifier Module

[0132] Si symbol

[0133] α, β1, β2, θ, δ, γ frequency bands

Claims

1. A system (1) for detecting the presence of a drug in a human body (2), comprising: - EEG sensor (10); as well as - An analysis unit (11) having a processor arrangement (110) and a memory (111) storing instructions that, when executed by the processor arrangement (110), cause the analysis unit (11) to perform the following steps: The EEG sensor (10) is used to receive a time series signal (300-307) including values ​​(310-313). Analyze the values ​​(310-313) of the signal (300-307) to convert the signal (300-307) into a sequence of symbols (Si); Generate multiple words (500-504), each word (500-504) comprising multiple consecutive symbols (Si); and The plurality of words (500-504) and / or one or more variables derived from the plurality of words (500-504) are provided to a classifier (600), which is configured to indicate the presence of the drug in the human body (2) based on the plurality of words (500-504).

2. The system (1) according to claim 1, characterized in that, Analyzing the values ​​(310-313) of the signal (300-307) includes comparing the values ​​(310-313) of the signal (300-307) with predefined symbols (Si) to convert the signal (300-307) into a sequence (400) of the symbols.

3. The system (1) according to claim 1 or 2, characterized in that, The EEG sensor (10) is a frontal EEG sensor (10) used to sense frontal cortical activity.

4. The system (1) according to any one of the preceding claims, characterized in that, The signals (300-307) are the original signal (300) or preprocessed signals (301-307) obtained by preprocessing the original signal (300).

5. The system (1) according to claim 4, characterized in that, Preprocessing the original signal (300) includes performing bandpass filtering to obtain multiple frequency bands (α, β1, β2, θ, δ, γ), wherein the preprocessed signal (301-307) includes one or more of the frequency bands (α, β1, β2, θ, δ, γ).

6. The system (1) according to claim 4 or 5, characterized in that, Preprocessing the original signal (300) includes: determining at least one signal envelope, wherein the preprocessed signal (307) includes the at least one signal envelope.

7. The system (1) according to any one of the preceding claims, characterized in that, Each of the predefined symbols (Si) is assigned a corresponding range of signal values.

8. The system (1) according to claim 7, characterized in that, The width of at least one of the ranges is equal to the standard deviation of the corresponding signal (300-307) or a multiple of the standard deviation.

9. The system (1) according to any one of the preceding claims, characterized in that, Predefine two, three, four or more distinct symbols (Si).

10. The system (1) according to any one of the preceding claims, characterized in that, Each of the plurality of words (500-504) includes the same number of two, three or more consecutive symbols (Si).

11. The system (1) according to any one of the preceding claims, characterized in that, The classifier (600) is configured to determine the number of occurrences of the words (500-504).

12. The system (1) according to any one of the preceding claims, characterized in that, The classifier (600) is specifically configured to compute... Entropy is used to determine the entropy of the plurality of words (500-504).

13. The system (1) according to any one of the preceding claims, characterized in that, The classifier (600) includes classifier modules (601-603) that have been trained using the following: - Multiple words (500-504) generated using signals (300-307) obtained from a drug-free human body (2), and multiple words (500-504) generated using signals (300-307) obtained from a human body (2) administered opioids, particularly remifentanil; and / or - Multiple words (500-504) generated using signals (300-307) obtained from a drug-free human body (2), and multiple words (500-504) generated using signals (300-307) obtained from a human body (2) administered with an anesthetic, particularly propofol; and / or - Multiple words (500-504) generated from signals (300-307) obtained from human bodies (2) that have been given opioids, particularly remifentanil, and multiple words (500-504) generated from signals (300-307) obtained from human bodies (2) that have been given anesthetics, particularly propofol.

14. A method for detecting the presence of a drug in a human body (2), comprising the steps of said method performed by an analysis unit (11) communicatively coupled to an EEG sensor (10): - The EEG sensor (10) is used to receive a time series signal (300-307) including values ​​(310-313). - Analyze the values ​​(310-313) of the signal (300-307) to convert the signal (300-307) into a sequence of symbols (Si); - Generate multiple words (500-504), each word (500-504) comprising multiple consecutive symbols (Si); and - Provide the plurality of words (500-504) and / or one or more variables derived from the plurality of words (500-504) to a classifier (600), the classifier (600) being configured to indicate the presence of the drug in the human body (2) based on the plurality of words (500-504).

15. The method according to claim 14, wherein, The classifier (600) includes classifier modules (601-603) that have been trained using the following: - Multiple words (500-504) generated using signals (300-307) obtained from a drug-free human body (2), and multiple words (500-504) generated using signals (300-307) obtained from a human body (2) administered opioids, particularly remifentanil; and / or - Multiple words (500-504) generated using signals (300-307) obtained from a drug-free human body (2), and multiple words (500-504) generated using signals (300-307) obtained from a human body (2) administered with an anesthetic, particularly propofol; and / or - Multiple words (500-504) generated from signals (300-307) obtained from human bodies (2) that have been given opioids, particularly remifentanil, and multiple words (500-504) generated from signals (300-307) obtained from human bodies (2) that have been given anesthetics, particularly propofol.

16. A computer-readable storage medium storing instructions that, when executed by a processor arrangement (110) of an analysis unit (11), cause the analysis unit (11) to perform the method according to claim 14 or 15.

Images