RU2797878C1 - Method of differential diagnosis of essential tremor and the first stage parkinson's disease using spurge analysis on cross-wavelet spectrum of electromyographic signals of antagonist muscles - Google Patents

Abstract

FIELD: medicine.

SUBSTANCE: joint registration of electromyograms (EMG) is carried out on antagonist muscles — on radial flexors of the wrist and long radial extensor of the wrist. The study is carried out when the patient is relaxed in a sitting position, while the arms are bent at the elbow joints and lie freely on the armrests of the chair, the palms hang down relaxedly. The cross-wavelet spectrum (CWS) of the envelopes of the EMG signals of the antagonist muscles is calculated. The local maxima on the FAC and the selection of bursts on the FAC are calculated. The parameters of bursts on the FAC of the envelopes of the recorded EMG signals are estimated, namely, the central frequency, the maximum power spectral density, the duration at a height of 1/sqrt(2), the bandwidth at a height of 1/sqrt(2), and the instantaneous phase. The number of bursts per FAC per second are calculated. The value of the coefficient F according to the stated formula is calculated. The F factor is compared to a T threshold of -0.31026 for the patient's left hand and 0.52174 for the patient's right hand. If the value of the coefficient F is less than the threshold value T for the corresponding arm of the patient, the first stage of Parkinson's disease is diagnosed. If the value of the coefficient F is not less than the threshold value T, the essential tremor is diagnosed.

EFFECT: method provides differential diagnosis of essential tremor and the first stage of Parkinson's disease due to quantitative signs obtained in the study of bursts on the cross-wavelet spectrum of envelopes of EMG signals of the antagonist muscles of patients.

1 cl, 7 dwg, 2 tbl, 5 ex

Description

The invention relates to digital signal processing and can be used in medicine for rapid differential diagnosis of essential tremor (ET) and Parkinson's disease (PD).

Differential diagnosis of ET and PD is an urgent medical problem. The complexity of the differential diagnosis of ET and PD lies in the fact that in some cases these neurophysiological diseases are accompanied by similar clinical manifestations. In addition, in clinical practice, there are cases of transition of ET to PD. One of the most promising approaches to distinguish between ET and PD is considered to be a comparative analysis of electromyographic signals (EMG) of the antagonist muscles of the patient's extremities. Antagonist muscles are called flexor and extensor muscles corresponding to one limb. Comparison of the signals of antagonist muscles consists in studying the phase shift of the EMG signals of these muscles. Depending on the magnitude of the phase shift, synchronous and alternating types of tremor are distinguished. Synchronous tremor corresponds to a phase shift close to zero, alternating tremor corresponds to a phase shift close to π radians. Unfortunately, there is no one-to-one correspondence between the disease and the type of tremor. During the examination, the patient may experience both synchronous and alternating tremor.

For the differential diagnosis of ET and PD, various mathematical methods for analyzing EMG signals from antagonist muscles are used.

Cross-spectra EMG of antagonist muscles are analyzed [Boose A., Spieker S., Jentgens C., Dichgans J. Wrist tremor: investigation of agonist-antagonist interaction by means of long-term EMG recording and cross-spectral analysis // Electroencephalography and Clinical Neurophysiology/Electromyography and Motor Control. - 1996. - V. 101, N. 4. - PP. 355-363]. The disadvantage of this method of signal analysis is that the cross spectrum is calculated immediately for the entire EMG signal, including areas with alternating and synchronous tremor. As a result, one type of tremor may go unnoticed against another type of tremor. The method for differential diagnosis of PD and ET suffers from the same disadvantage (RU 2558176, FGBNU NTSN, 07/27/2015), based on the analysis of EMG coherence of antagonist muscles at a double frequency of pathological tremor. This method provides an accuracy of about 88% in distinguishing between PD and ET.

There is a method for analyzing EMG signals of antagonist muscles based on the study of statistical distributions of the phase difference of the envelopes of EMG signals [Sushkova O.S., Morozov A.A., Kershner I.A., Petrova N.G., Gabova A.V., Chigaleichik L.A., Karabanov A.V. Investigation of distribution laws of the phase difference of the envelopes of electromyograms of antagonist muscles in Parkinson's disease and essential tremor patients // RENSIT: Radioelectronics. Nanosystems. information technologies. - 2020. - V. 12, No. 3. - P. 415-428]. The disadvantage of this method of signal analysis is that statistical distributions are calculated immediately for the entire EMG signal, as a result of which a tremor of one type may go unnoticed against the background of a tremor of another type.

More advanced tools for analyzing biomedical signals are wavelet coherence and wavelet phase coherence. Wavelet coherence shows the degree of relatedness of two signals at different times and at different frequencies, in contrast to ordinary coherence, which depends only on frequency. This increases the amount of useful information extracted from the signals. A known method for the differential diagnosis of PD and Alzheimer's disease, based on wavelet coherence [Jeong D.H., Young-Do K., In-Uk S., Yong-An Ch., Jaeseung J. Wavelet energy and wavelet coherence as EEG biomarkers for the diagnosis of Parkinson's disease-related dementia and Alzheimer's disease // Entropy. - 2016. - V. 18, N. 1. - P. 8]. This method is not intended for the differential diagnosis of ET and PD. A method for extracting useful EMG characteristics is based on wavelet coherence (CN 109567799 A, Hangzhou Dianzi University, Hangzhou Electronic Science and Technology University, 04/05/2019). The possibility of using this method for the differential diagnosis of neurodegenerative diseases of ET and PD is unknown.

A known method for the differential diagnosis of PD and ET, based on the analysis of burst electrical activity of the muscles (RU 2741233 C1, IRE named after V.A. Kotelnikov RAS, IVNDiNF RAS, FGBNU NTSN, 22.01.2021 - prototype). The method includes calculating wavelet spectrograms of accelerometer signals and envelopes of EMG signals, calculating burst parameters on wavelet spectrograms, calculating the number of bursts with parameters specified in the table, calculating coefficients based on the number of bursts, and comparing the calculated coefficients with known threshold values. The method does not take into account the phase shift of the EMG signals of antagonist muscles, as a result of which the accuracy of the differential diagnosis of ET and PD is only 90%. Other disadvantages of this method are the need to use a large number of sensors (EMG and accelerometers), as well as the need to measure signals in two different patient positions, which significantly increases the complexity and duration of the patient examination.

The patented invention is aimed at differential diagnosis of ET and the first stage of PD by means of new quantitative features obtained in the study of bursts on the cross-wavelet spectrum of the envelopes of EMG signals of the antagonist muscles of patients, which is the technical result.

The method for differential diagnosis of essential tremor and the first stage of Parkinson's disease includes the joint registration of electromyograms (EMG) on the antagonist muscles of the patient's limb, the calculation of the cross-wavelet spectrum (CWS) of the envelopes of the EMG signals of the antagonist muscles, the calculation of local maxima on the CV and the selection of bursts on the CV .

The difference is as follows:

the patient is diagnosed in a sitting position in a relaxed state, while the arms are bent at the elbow joints and lie freely on the armrests of the chair, the palms hang down relaxed;

at the same time, EMG electrodes are placed on the antagonist muscles, preferably on the radial flexors of the wrist and the long radial extensor of the wrist;

then, according to the data obtained, the FAC of the envelopes of the EMG signals of the antagonist muscles is calculated;

calculate the local maxima on the FAC of the envelopes of the EMG signals, which emit bursts on the FAC;

calculate the parameters of the bursts on the FAC of the envelopes of the recorded EMG signals, namely, the central frequency, the maximum power spectral density (PSD), the duration at a height of 1/sqrt(2), the bandwidth at a height of 1/sqrt(2), the instantaneous phase;

calculate the number of bursts per FAC per second, the parameters of which correspond to the ranges predetermined from the data of patients with a clinically confirmed diagnosis;

calculate the value of the coefficient F according to the formula:

F=Q ₁ ×A ₁ +Q ₂ ×A ₂ +Q ₃ ×A ₃ +Q ₄ ×A ₄ ,

where Q _n is the number of bursts per FAC per second, the parameters of which correspond to the range n, predetermined according to the data of patients with a clinically confirmed diagnosis, A _n is the value of the coefficient, predetermined according to the data of patients with a clinically confirmed diagnosis, corresponding to the range n;

the factor F is compared with a threshold value T equal to -0.31026 for the patient's left hand and 0.52174 for the patient's right hand;

when the value of the coefficient F is less than the threshold value T, the first stage of Parkinson's disease is diagnosed, and

when the value of the coefficient F is not less than the threshold value T, the diagnosis is essential tremor.

When calculating the envelopes of EMG signals, the Hilbert transform is used.

The CVS is calculated as the product of complex conjugate wavelet spectrograms.

When calculating wavelet spectrograms, the complex Morlet wavelet is used. To smooth the CVS, an adaptive two-dimensional Gaussian window is used.

When calculating the instantaneous phase of the burst, the four-quadrant arc tangent of the imaginary and real parts of the complex value of the FAC is used.

The time of one EMG recording is at least 90 seconds.

The method is carried out as follows:

1. EMG signals are recorded in a relaxed state of the patient.

2. Recording is carried out for 1 minute 30 seconds. The patient is placed electromyographic electrodes on antagonist muscles, for example, on the muscles of the wrist joint of the hand (extensor muscle Musculus extensor carpi radialis longus and flexor muscle Musculus flexor carpi radialis), as shown in figure 1 (position 11 - location of the active bipolar EMG electrode on the flexor muscle of the patient's left arm, position 12 - location of the reference bipolar EMG electrode on the flexor muscle of the patient's left arm, position 13 - location of the active bipolar EMG electrode on the extensor muscle of the patient's left arm, position 14 - location of the reference bipolar EMG- electrode on the extensor muscle of the patient's left arm). The patient, sitting in a chair, puts his hands on the armrests and hangs his palms relaxedly down. The legs are relaxed and stand on the floor with the entire foot. The patient sits with his eyes closed during the recording. Conduct one EMG recording.

3. Wavelet spectrograms of EMG signal envelopes are calculated. Various methods can be used to calculate the envelopes of the signals, but the preferred method is to use the Hilbert transform. Various complex wavelets can be used to compute wavelet spectrograms, but the complex Morlet wavelet is preferred.

4. Calculate CVS as a product of complex-conjugate wavelet spectrograms of envelopes of EMG signals.

5. Calculate the absolute value of the FAC. The absolute value of the FAC is smoothed to remove computational artifacts. Various smoothing methods can be used, however, the use of an adaptive 2D Gaussian window is preferred.

6. The bursts on the FIC are identified and their parameters are calculated: burst center frequency, maximum burst power spectral density, burst duration at a height of 1/sqrt(2), measured in the number of periods at the burst center frequency, burst bandwidth at a height of 1/sqrt (2), instantaneous burst phase. To detect bursts, local maxima are used at the absolute value of the FAC. To calculate the instantaneous phase of the burst, the four-quadrant arc tangent of the imaginary and real parts of the complex value of the FAC is used.

7. Calculate the number of bursts per FAC per second, the parameters of which correspond to the conditions specified in table 1. The first column of the table corresponds to the name of the burst parameter (center frequency, maximum power spectral density, duration in periods, bandwidth, instantaneous phase). The second column of the table is the units of these burst parameters. Table columns 3 to 10 contain burst parameters for different frequency ranges and for different patient arms. Table columns 3-6 correspond to the patient's left hand. Table columns 7-10 correspond to the patient's right hand. Parameter values are pre-calculated using two-dimensional AUC-diagrams [Sushkova O.S., Morozov A.A., Gabova A.V., Karabanov A.V., Illarioshkin S.N. A statistical method for exploratory data analysis based on 2D and 3D area under curve diagrams: Parkinson's disease investigation // Sensors / Ernest N. Kamavuako (Eds.). - MDPI, 2021. - V. 21, Issue 14. - P. 4700].

8. Calculate the coefficient F by the number of bursts in different frequency ranges. To calculate the coefficient F, 4 coefficients A _n are used, given in table 2, n is the row number in the table (n=1…4):

F=Q ₁ ×A ₁ +Q ₂ ×A ₂ +Q ₃ ×A ₃ +Q ₄ ×A ₄ ,

where Q _n is the number of bursts per FAC per second. The values of the coefficients are calculated in advance according to the data of patients with a clinically confirmed diagnosis and are shown in Table 2. Index n corresponds to the row number in the table. The left column of table 2 corresponds to the left hand, and the right column of table 2 corresponds to the right hand.

9. The calculated F factor is compared to a threshold value of 7, equal to -0.31026 for the patient's left hand and 0.52174 for the patient's right hand. If the value of the coefficient F is less than the threshold value T, the first stage of Parkinson's disease is diagnosed, and if the value of the coefficient F is not less than the threshold value T, the diagnosis is essential tremor.

The technical result is that:

the accuracy of differential diagnosis (distinguishing) of ET and the first stage of PD increases; accuracy assessment on EMG of patients with a clinically confirmed diagnosis reaches 100%; simplicity and high speed of the survey; to measure the EMG of patients, it is necessary to install only two EMG sensors, that is, four bipolar electrodes and one ground electrode, measurements are carried out in one patient position for 1 minute 30 seconds.

The essence of the invention is illustrated in the figures:

Fig. 1 - EMG electrodes on the patient's left arm.

Fig. 2 - An example of the PIC of the envelopes of the EMG signals of the antagonist muscles of the patient's left arm at the first stage of PD.

Fig. 3 - An example of a burst on the FAC of the envelopes of the EMG signals of the antagonist muscles of the patient's left arm at the first stage of PD.

Fig. 4 - Example of envelopes of EMG signals of antagonist muscles. The splash on the FAC is indicated by a circle. The envelope of the EMG signal of the extensor muscle is shown on the left. The envelope of the EMG signal of the flexor muscle is shown on the right.

Fig. 5 - An example of the original EMG signals. The splash on the FAC is indicated by a circle. On the left is the EMG signal of the extensor muscle. The EMG signal of the flexor muscle is shown on the right.

Fig. 6 - Histograms of the values of the F coefficient calculated for the left hands of patients with the first stage of PD and patients with ET.

Fig. 7 - Histograms of the values of the coefficient F, calculated for the right hands of patients with the first stage of PD and patients with ET.

The basis of the invention is the registration and analysis of bursts on FAC signals. FAC is a function of two variables - time and frequency. When calculating the CVS, the complex-conjugate wavelet spectrograms of two signals are multiplied. Thus, the PBC shows not only the degree of connectivity of two signals, but also the value of the spectral power density of the signals at that moment in time and at the frequency where the connectivity of the signals is estimated. A burst on the FAC is an increase in the absolute value of the FAC localized in time and frequency. The following parameters of bursts on the FAC are analyzed: burst center frequency, maximum burst power spectral density, burst duration at a height of 1/sqrt(2), measured in the number of periods at the burst center frequency, burst bandwidth at a height of 1/sqrt(2), instantaneous burst phase, average number of bursts per second.

In FIG. Figure 2 shows an example of the CVS of the envelopes of the EMG signals of the antagonist muscles of the left arm of a patient with the first stage of PD. Trembling hyperkinesis is observed on the arm. Time is plotted along the abscissa axis, frequency is plotted along the ordinate axis, and the power spectral density is plotted along the applicate axis. Position 21 - the position of the bursts on the FAC in the frequency-time domain.

In FIG. Figure 3 shows an example of a burst on FAC. Position 31 - the position of the burst on the FAC in the time-frequency domain. Burst center frequency 5.3 Hz, maximum burst power spectral density 20292 μV ² /Hz, burst duration 1.39 cycles, burst bandwidth 1 Hz, burst instantaneous phase -3.02 radians.

In FIG. 4 shows the envelopes of the EMG signals, from which the FAC shown in FIG. 3. On the left is the envelope of the EMG signal of the extensor muscle (position 41). On the right is the envelope of the EMG signal of the flexor muscle (position 43). Positions 42 and 44 - the position of the burst on the envelopes of the EMG signals in time. On the envelopes of the EMG signals (positions 41 and 43, respectively), 2 periods of oscillations can be seen.

In FIG. Figure 5 shows the original EMG signals from which the envelopes shown in FIG. 4. On the left is the EMG signal of the extensor muscle (position 51). On the right is the EMG signal of the flexor muscle (position 53). Positions 52 and 54 - the position of the burst on the EMG signals in time. On the EMG signals (positions 51 and 53, respectively), 2 volleys of trembling hyperkinesis can be seen.

Tremor in ET and PD differs in a number of parameters of bursts on CV, however, the amount of overlap between groups of patients for each parameter is significant and does not allow the use of any individual parameter of bursts on CV for differential diagnosis. This circumstance necessitated the invention of a method aimed at reliably distinguishing between ET and PD by a combination of several parameters of bursts on the FAC of the EMG envelopes of antagonist muscles.

Accuracy score. To record EMG, a 41-channel multifunctional complex for neurophysiological studies "Neuron-Spectrum-5" (Neurosoft) was used. The EMG sampling rate was 500 Hz. For EMG, a high-pass filter with a cutoff frequency of 0.5 Hz and a notch filter with a frequency of 50 Hz were used. In addition, a Butterworth filter with a bandwidth of 60 to 240 Hz was used for EMG. The Hilbert transform was applied to the EMG signals after filtering to extract the signal envelope. The duration of the recording was 1 minute 30 seconds. The records were analyzed as is, without selecting individual areas in the signal.

To test the efficiency of the patented method, we used data from 12 patients with PD with right hand tremor, 10 patients with PD with left hand tremor, and 13 patients with ET. All patients were examined at the Scientific Center of Neurology, and they were given a clinical diagnosis.

Using the patent-pending method, out of 12 PD patients with right hand tremor, 12 patients were diagnosed with PD, out of 10 PD patients with left hand tremor, 10 patients were diagnosed with PD, out of 13 patients with ET, 13 patients were diagnosed with THIS. In FIG. Figure 6 shows histograms of the values of the F coefficient calculated for the left hands of patients with the first stage of PD (position 61) and patients with ET (position 62). In FIG. Figure 7 shows histograms of the values of the F coefficient calculated for the right hands of patients with the first stage of PD (position 71) and patients with ET (position 72). In FIG. Figures 6 and 7 show that the histograms of patients with PD and ET do not overlap. Thus, the accuracy of differential diagnosis was 100%, which confirms the technical results claimed in the patent.

Clinical example 1. Patient K., 61 years old, clinical diagnosis - PD, first stage, tremor is observed on the left hand. Registration of EMG signals was carried out. The coefficient F=-0.7133 for the left hand was calculated. The F value is less than the -0.31026 threshold set for the left hand F factor. It follows from the above that this patient has PD.

Clinical example 2. Patient P., 62 years old, clinical diagnosis - PD, first stage, tremor is observed on the right hand. Registration of EMG signals was carried out. The coefficient F=-0.0548 for the right hand was calculated. The F value is less than the 0.52174 threshold set for the F factor for the right hand. It follows from the above that this patient has PD.

Clinical example 3. Patient B., 66 years old, clinical diagnosis - ET, tremor is observed on the left and right hands. Registration of EMG signals was carried out. The coefficient F=-0.2263 for the left hand was calculated. The F value is greater than the -0.31026 threshold set for the left hand F factor. The coefficient F=0.6063 for the right hand was calculated. The F value is greater than the 0.52174 threshold set for the F factor for the right hand. It follows from the above that this patient has ET.

Clinical example 4. Patient D., aged 36, clinical diagnosis - ET, tremor is observed on the left and right hands. Registration of EMG signals was carried out. The coefficient F=-0.2609 for the left hand was calculated. The F value is greater than the -0.31026 threshold set for the left hand F factor. The coefficient F=0.5412 for the right hand was calculated. The F value is greater than the 0.52174 threshold set for the F factor for the right hand. It follows from the above that this patient has ET.

Clinical example 5. Patient U., 54 years old, clinical diagnosis - ET, turning into PD. During the examination, a tremor was visually observed on the right hand, a tremor on the left hand was not observed. Registration of EMG signals was carried out. The coefficient F=0.2681 for the right hand was calculated. The F value is less than the 0.52174 threshold set for the F factor for the right hand. From the above, it follows that this patient does show signs of PD.

Thus, the patented method makes it possible to achieve the technical result - the possibility of differential diagnosis of two diseases, ET and the first stage of PD, with an accuracy of up to 100%.

Claims (7)

1. A method for the differential diagnosis of essential tremor and the first stage of Parkinson's disease, including the simultaneous registration of electromyograms (EMG) on the antagonist muscles of the patient's limb, characterized in that the patient is diagnosed in a sitting position in a relaxed state with eyes closed, while the arms are bent at the elbows joints and lie on the armrests of the chair, the palms hang down, and the EMG electrodes are placed on the radial flexors of the wrist and the long radial extensor of the wrist of the patient, the recording is carried out for 1 minute 30 seconds; extract the envelopes of EMG signals and calculate the cross-wavelet spectrum (CWS) for them, calculate the local maxima on the CWS, determine the parameters of the bursts corresponding to the local maxima on the CWS: the central frequency, the maximum spectral power density, the duration at a height of 1/sqrt(2), the width frequency bands at a height of 1/sqrt(2), instantaneous phase and calculate the number of bursts per FAC per second, the parameters of which correspond to the conditions specified in table 1 contained in the description; calculate the value of the coefficient F for the left and right hands according to the formula: F=Q ₁ ×A ₁ +Q ₂ ×A ₂ +Q ₃ ×A ₃ +Q ₄ ×A ₄ , where Q _n (n=1-4) - the number of bursts per FAC per second; A _n - the value of the coefficient determined from table 2 contained in the description, the coefficient F is compared with the threshold value T equal to -0.31026 for the patient's left hand and 0.52174 for the patient's right hand; if the value of the coefficient F is less than the threshold value T for the corresponding arm of the patient, the first stage of Parkinson's disease is diagnosed, and if the value of the coefficient F is not less than the threshold value T, the diagnosis is essential tremor. 2. The method according to claim 1, characterized in that when calculating the instantaneous phase of the burst, a four-quadrant arc tangent of the imaginary and real parts of the complex value of the FAC is used.

Images