RU2738583C1 - Diagnostic technique for focal epilepsy based on electroencephalogram analysis - Google Patents

Abstract

FIELD: medicine.

SUBSTANCE: invention refers to medicine, namely to a method for diagnosing focal epilepsy based on an electroencephalogram (EEG) analysis. Value of the oscillation period threshold is evaluated with preservation of fractal properties and a scaling index in the transition zone of the fluctuation curve in the frontal and temporal leads. If the oscillation period is more than 0.2 s and the scaling factor is less than or equal to 0.3, focal epilepsy is diagnosed.

EFFECT: provided is diagnosing focal epilepsy based on EEG analysis using fluctuation analysis of a signal with a remote trend.

1 cl, 9 dwg, 3 tbl, 2 ex

Description

The invention relates to medicine, namely to clinical medicine (neurology).

One of the promising methods of mathematical analysis in electroencephalography is the identification and description of the fractal properties of the electroencephalographic signal and the application of methods for assessing its nonlinear dynamics [Kaffashi F., 2007; Lopes da Silva F.H. et al. 2003]. The now "traditional" method of spectral analysis of the electroencephalogram (EEG) involves the allocation of a fairly narrow frequency range of analyzed bioelectric processes using fast Fourier transform and characteristics of spectral power and frequency to describe, interpret as physiological and pathological identified phenomena [Alexandrov M.V., Ivanov L.B., Lytaev S.A. et al., 2019].

At the same time, EEG as a complex nonlinear process has the property of self-similarity (fractality), suggesting the inclusion of simpler electroencephalogram graphoelements (patterns) into more complex patterns, which differ primarily in time scales [Buzsaki G., 2006].

The practical implementation of this approach in clinical neurology can be a description of the fractal properties of EEG in epilepsy.

EFFECT: development of a method for diagnosing focal epilepsy based on EEG analysis.

The method closest to the intended purpose to the claimed invention in terms of the totality of features is the use the human operator of the method of fluctuation analysis of electroencephalograms in the problems of stress state control, in which, based on the analysis of the crossover of the EEG fluctuation curve, interpreted in the form of a scaling index and the scales (periods) of EEG oscillations, it is proposed to assess the functional state of the operators.

We have proposed a Method for the diagnosis of focal epilepsy based on the analysis of an electroencephalogram, including the use of the method of fluctuation analysis of a signal with a distant trend, characterized in that, on the basis of DFA analysis (detrended fluctuation analysis), the value of the threshold of the oscillation period is estimated with the preservation of fractal properties in the frontal and temporal leads with an oscillation period of more than 0.2 s and a scaling indicator in the transitional the zone of the fluctuation curve in the frontal and temporal leads is less than or equal to 0.3.

The main difference between this method and the prototype is the use of specific threshold values for the classification of fractal properties of the EEG in patients with focal epilepsy.

For analysis, it is recommended to use artifact-free EEG fragments recorded in monopolar commutation with reference electrodes on the ears in the background state (the state of relaxed wakefulness with closed eyes); analysis epoch duration is 20.48 seconds. The EEG signal is exported in ASCII format (8-bit computer encoding for representing text, decimal digits); the processing of data series corresponding to the EEG recording channels is carried out in accordance with the formula

where F (n) is the value of the fluctuation function, N is the number of EEG counts, k is the number of the current count; y - instantaneous (at the moment of counting) value of the EEG potential (μV).

The data processing with the calculation of the scaling index, the values of the data scales for crossover 1 and crossover 2 was carried out using the MathLAB program, data presentation was performed in an Excel spreadsheet processor.

The theoretical justification of this method is the methodology of fluctuation analysis with the removal of the trend. The method involves studying the fractal properties of processes and signals by comparing the range of fluctuations when the trend is removed on different time scales. It is shown that the degree of slope of the curve approximating the range of oscillations at different time scales of the process under study (scaling index - a) reflects various phenomena of self-organization within the framework of the theory of random signals and noise (Table 1).

Table 1. Values of the scaling index and characteristics of the phenomenon.

Scaling value (a) Process characteristics 0 <a <0.5 Abrupt signal changes a = 0.5 Random process (white noise / chaotic oscillation) 0.5 <a <1.0 The presence of fractal (self-similar) properties a = 1 Strict self-similar (fractal) properties a> 1 Transient process (random walk) a = 2 Strictly organized signal

When studying the fractal properties of bioelectric signals reflecting the dynamics of physiological processes (including EEG) in practically healthy individuals, it was found [2, 3] that the plot of the dependence of DFA values on the scale has 2 breaks (crossovers) characterizing a sharp change in the fractal properties of the signal ( from self-similar properties to a sharp loss of fractal properties of the signal), which was interrelated with the standard spectral analysis of electroencephalograms: the interval between 23 counts and 128 counts corresponds to the transition zone between the first (23 counts) and the second (128 counts) crossover (Fig. 1).

In this regard, it was proposed to describe not only the values of the scaling index, but also the periodometric (frequency) characteristics of the crossovers.

We used this technique to describe the EEG signal in patients with focal epilepsy in order to highlight additional criteria for the diagnosis of this neurological disorder.

The practical substantiation of the method was carried out on 50 patients suffering from structural focal frontal and temporal lobe epilepsy in accordance with the ILAE classification, 2017 [5, 8] and 50 practically healthy individuals (practically healthy volunteers, with no epileptiform activity on the EEG at 30-minute registration, no history of paroxysmal disorders). The average age of practically healthy persons was 33.1 years (average error 1.22 years), patients with epilepsy 34.2 years (average error 1.15 years); both groups had the same number of men and women (25 men and women in each group). The average duration of the illness was 3.5 years; The study included patients with at least 1 seizure in the last year of follow-up.

EEG registration was carried out using the Neuron-Spectrum.NET hardware and software complex for 30 minutes; background recording, recording with opening and closing of eyes, photostimulation, 3-minute hyperventilation, after hyperventilation were carried out.

The data on the EEG analysis in the background recording are presented. Fragments without epileptiform activity were analyzed in patients with focal epilepsy and in the control group. The duration of the fragment for analysis was 5 minutes. The analysis was carried out on artifact-free areas without epileptiform activity (analysis epoch 20.48 seconds). Signal processing was carried out using the MathLab software package.

The values of the scaling index (a) in the transition zone between crossover 1 and crossover 2 and the period of oscillations corresponding to the crossover (transition with a pronounced change in the fractal properties of the EEG signal) were determined.

Descriptive statistics of indicators are presented as median, upper and lower quartile (Me; LQ; UQ), comparative analysis was carried out using the Mann-Whitney test. Operating characteristic curves (ROC analysis) were used to study the sensitivity, specificity and accuracy / diagnostic significance of the indicators. ROC curves were built using the Statistica 10.0 software package

Table 2 shows significant differences in the values of the scaling index in the studied leads in the transition zone.

Table 2. Values of the scaling indicator in the frontal and temporal leads in patients with focal epilepsy and in apparently healthy individuals (alpha)

Indicators Epilepsy Control group Z p Me; LQ; Uq Me; LQ; Uq and in the transition zone in lead F3 0.198 ; 0.127; 0.296 0.411; 0.297; 0.588 2,712 0.007 and in the transition zone in lead F4 0.246; 0.174; 0.314 0.382; 0.293; 0.490 2,285 0.022 and in the transition zone in lead T3 0.222 ; 0.158; 0.282 0.362; 0.286; 0.496 2,322 0.020

As follows from the table, in patients with focal epilepsy, a significantly lower value of the scaling index in the transition zone is determined, while its values correspond to sharp changes in the signal with a greater tendency to the formation of a random process (chaotic oscillations).

When analyzing the oscillation period corresponding to the crossover, a significantly shorter oscillation period in healthy individuals and, accordingly, a longer oscillation period in patients with focal epilepsy, in which the phenomenon of loss of fractal properties occurs, was revealed.

Table 3. The value of the indicators of the periods of fluctuations (s) in

crossover point in groups of subjects

Indicators Epilepsy Control group Z p Me; LQ; Uq Me; LQ; Uq Crossover period in F3, s 0.187; 0.148; 0.206 0.135; 0.098; 0.187 2,532 0.011 Crossover period in T3, s 0.187; 0.163; 0.285 0.132; 0.097; 0.175 3.11 0.002 Crossover period in T4, s 0.201; 0.163; 0.236 0.135; 0.102; 0.196 2.940 0.004

Thus, in the group of patients with focal epilepsy, a sharp loss of fractal properties occurs when switching to the scale of fluctuations with a significantly longer period compared to the control group. In other words, in the group of practically healthy individuals, the fractal properties of the electroencephalogram (self-similarity) are retained for oscillations of a higher frequency than in the group of epilepsy patients. As a significant indicator, we chose the indicator of the oscillation period in the frontal and temporal leads. The average oscillation frequency, below which there is a loss of fractal properties of the EEG in patients with focal epilepsy, corresponded to 6-7 Hz.

We performed an ROC analysis to assess the sensitivity, specificity and accuracy of the distribution of subjects into groups of patients with epilepsy and a control group based on the presented DFA scores.

ROC analysis was carried out using the artificial neural networks module of the Statistica 10.0 Ru software package; to analyze the sample, linear neural networks (1 neuron in the hidden layer) with a logistic activation function were created.

FIG. Figure 2 shows the ROC curve for the scaling exponent in the transition zone in lead F3. The threshold for classification is 0.3; the area under the curve was 0.820; which corresponds to a sufficiently high quality of the model (Fig. 2).

Similar ROC curves have been proposed for the scaling factor in lead T3: area under the curve 0.84; threshold 0.3 (Fig. 3).

FIG. 4 and 5 show the ROC curves for the crossover period in leads F3 (area under the curve 0.791; threshold 0.215 s) and T3 (area under the curve 0.844; threshold 0.214 s), respectively.

According to the ROC analysis, the classification threshold for the scaling indicator was 0.3, and for the oscillation period - more than 0.2 s.

Let's give examples of using this method.

Example 1.

Patient A., 32 years old, was consulted on an outpatient basis at the outpatient department of the Ryazan State Medical University of the Ministry of Health.

Describes seizures from 28 years of age, mostly nocturnal: oropharyngeal automatism (in the form of an unusual "throat sound", wheezing) with a transition to bilateral tonic-clonic seizures; some of the seizures are of the nature of hypermotor automatisms: high-amplitude automated movements of the arms and legs with impaired consciousness.

He took Finlepsin up to 800 mg per day for 3 years, then, due to a decrease in the frequency of attacks, he independently stopped taking the drug. 1 week before the visit, a focal motor seizure developed at night with a transition to a bilateral tonic-clonic seizure.

Examined: MRI (magnetic resonance imaging) (1.5 T), assumed to be DNET (dysembyroplastic neuroepithelial tumor - highly epileptogenic substrate) in the right temporal and frontal lobes.

During the appointment, an outpatient EEG study (30 minutes) was performed with functional tests. During the study, epileptiform activity was not registered.

It should be noted that frontal lobe epilepsy is characterized by the predominance of seizures at night and a low index of epileptiform activity in the waking state.

Conclusion: Structural focal epilepsy with focal motor seizures from transitions to bilateral tonic-clonic seizures.

A fragment of the EEG of patient A in the background recording is shown in Fig. 6 (bipolar switching is used for analysis in epileptology; DFA analysis was carried out on data sets in monopolar switching)

DFA analysis performed for epoch 20.48 seconds. Figure 7 shows the graphs of the fluctuation curve of patient A: at the top - the general graph of the fluctuation curve with crossovers and the presence of a transition zone; below - the analysis of the scaling index (multiplier for the argument x in the linear equation) for the transition zone.

The scaling index in the transition zone (0.228) reflects the loss of the fractal properties of the EEG signal, which is typical for interictal EEG in patients with focal epilepsy. The oscillation period at which the loss of fractal properties occurs was 0.355 s, which is also typical for interictal EEG of patients with focal epilepsy.

Thus, this method of EEG analysis confirmed the diagnosis of focal epilepsy in this patient.

Example 2.

Patient B., 30 years old, turned to the outpatient department of the Ryazan State Medical University of the Ministry of Health of the Russian Federation about myofascial syndrome in the neck.

A history of complaints of a single undifferentiated episode of a short-term disorder of consciousness at the height of emotional stress, in its structure corresponding to a psychogenic non-epileptic seizure.

Neurological status without pathology.

FIG. 8 shows a fragment of the EEG recording of patient B (bipolar switching is used for analysis in epileptology; DFA analysis was carried out using data arrays in monopolar switching).

MRI of the brain (high resolution): no changes were found. Video-EEG monitoring was recommended: with 6-hour video EEG monitoring in the waking state and stages 1-2 of nonREM sleep, no epileptiform activity was recorded.

With the subsequent follow-up (2.5 years), no paroxysmal events were recorded.

Conclusion: A single psychogenic non-epileptic seizure in history .

Patient B's fluctuation curve plots are shown in FIG. 9: top - general graph of the fluctuation curve with crossovers and the presence of a transition zone; below - the analysis of the scaling index (multiplier for the argument x in the linear equation) for the transition zone.

In DFA analysis, the scaling index (0.793) reflects the presence of distinct fractal (self-similar) properties of the electroencephalogram at different time scales, the oscillation period was 0.178 s, which is typical for the EEG of patients without epileptiform activity.

The proposed method of EEG analysis confirmed the absence of focal epilepsy in this patient.

Thus, the satisfactory diagnostic significance of the indicators suggests their use for description, mathematical analysis, automated analysis of electroencephalograms as an auxiliary diagnostic technology in epileptology.

Bibliography

1.SPb .: SpetsLit, 2019.200 p.

2. Alpatov A.V. et al. Method of EEG fluctuation analysis in the problems of stress state control of a human operator // Biotechnosphere. 2013. T. 26, No. 26. S. 54-60. PROTOTYPE

3. Alpatov A.V., Mitrofanova M.Yu. Method of fluctuation analysis of heart rate in real time // Biomedical Radioelectronics. 2011. No. 7. S. 66-71.

4. Buzsaki G. Rhytms of the brain. New York: Oxford University Press, 2006.462 p.

5. Fisher R.S. et al. ILAE official report: a practical clinical definition of epilepsy. Epilepsia. Vol 55 (4): 475-482.

6. Kaffashi, F. Variability analysis & Its application to physiological time series data. Cleveland: Case Western Reserve University, 2007.124 p. URL: https://etd.ohiolink.edu/rws_etd/document/get/case1181072302/inline

7. Lopes da Silva F.H. et al. Dynamical diseases of brain systems: different routes to epileptic seizures // IEEE transactions on bio-medical engineering. 2003. Vol. 50. No. 5.p. 540-548.

8. Scheffer I.E. et al. ILAE classification of the epilepsies: Position paper of the ILAE Comission for Classification and Terminology. Epilepsia. - Vol 58 (4): 512-521.

Claims (1)

A method for diagnosing focal epilepsy based on the analysis of an electroencephalogram, including the use of the method of fluctuation analysis of a signal with a distant trend, characterized in that the value of the threshold of the oscillation period is estimated with preservation of fractal properties and a scaling index in the transition zone of the fluctuation curve in the frontal and temporal leads and at values of the oscillation period more than 0.2 s and a scaling index less than or equal to 0.3 diagnose focal epilepsy.

Images