CN114869301A - Method and apparatus for detecting epileptiform discharges - Google Patents

Abstract

The present application proposes a method for detecting epileptiform discharges, comprising: acquiring an electroencephalographic signal of a patient during an interval of an epileptic seizure, wherein the electroencephalographic signal includes one or more signal segments on at least one lead channel, each signal segment including one or more signal sub-segments; performing feature matching on the signal sub-segments to identify the signal sub-segments with epileptiform discharges; and determining statistics of epileptiform discharges. The method can effectively reduce the workload and the false detection probability of medical staff and improve the diagnosis efficiency by using an automatic detection method.

Description

Method and device for detecting epileptiform discharges

technical field

The present application relates to signal processing, and in particular, to methods, apparatus, and computer storage media for detecting epileptiform discharges in electroencephalographic signals.

Background technique

An EEG is a form of signal data commonly used to record electrical activity produced by the human cerebral cortex. The high temporal resolution, non-invasiveness and low cost of EEG signals make it widely used in the preoperative evaluation of patients with epilepsy lesions. In the long-term EEG signal recording, the seizure period only accounts for a small part of the EEG signal in time. In fact, the patient spends more time in a state of seizure-free (ie, interictal) during the monitoring process. .

The detection of epileptiform discharges between seizures is very important for the judgment of epilepsy syndrome and the localization of lesions. At present, the analysis and judgment of whether the long-term EEG signal contains epileptiform discharge information is completed by the doctor's observation and diagnosis based on experience. This manual method is time-consuming and labor-intensive.

In addition, the traditional spike detection scheme has low sensitivity, high false detection rate, and cannot be accurate to specific lead channels. The traditional scheme does not specifically exclude the physiological waveform in the signal.

Therefore, there is a need for improved protocols for the detection of epileptiform discharges from interictal EEG signals.

SUMMARY OF THE INVENTION

In order to solve at least one of the above problems, embodiments of the present application propose an automated solution for detecting epileptiform discharges in EEG signals.

According to an aspect of the present application, a method for detecting epileptiform discharges is proposed, comprising: acquiring an electroencephalogram signal of a patient between seizures, wherein the electroencephalogram signal includes one or more on at least one lead channel Each signal segment includes one or more signal sub-segments; feature matching is performed on the signal sub-segments to identify signal sub-segments with epileptiform discharges; and statistics of epileptiform discharges are determined.

According to another aspect of the present application, a computer-readable storage medium is proposed on which a computer program is stored, the computer program comprising executable instructions that, when executed by a processor, perform the method according to the above .

According to yet another aspect of the present application, an electronic device is provided, including a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the executable instructions to implement the above the method described.

According to the scheme of detecting epileptiform discharges in EEG signals according to the embodiments of the present application, compared with the common scheme of diagnosing epileptic seizures of patients by manual observation of EEG signals by doctors, automated EEG signals are introduced The detection method can be used as a computer-aided detection tool to output the detection results and indicators related to the presence of epileptiform discharges in the EEG signals monitored by the patient during the interictal period, quantitative statistical data and visualized presentation methods for the reference of doctors, effective Reduce the workload of medical staff and the probability of false detection and improve the efficiency of diagnosis. At the same time, the automated detection scheme of the present application can be accurate to the specific lead channel of the EEG signal, and optimize and update detection standards and templates based on accumulated historical data, further improving the accuracy and speed of detection.

Description of drawings

The above and other features and advantages of the present application will become more apparent by describing in detail exemplary embodiments thereof with reference to the accompanying drawings.

FIG. 1 is a schematic flowchart of a method for detecting epileptiform discharges according to an embodiment of the present application.

FIG. 2 is a comparison diagram of signal waveform processing for identifying epileptiform discharges in an electroencephalogram signal according to an embodiment of the present application.

FIG. 3 is an exemplary time-frequency template for time-frequency matching according to an embodiment of the present application.

FIG. 4 is an exemplary visual presentation of statistics based on identification results of epileptiform discharges according to embodiments of the present application.

FIG. 5 is a schematic structural block diagram of an electronic device according to an embodiment of the present application.

Detailed ways

Example embodiments will now be described more fully with reference to the accompanying drawings. Exemplary embodiments can, however, be embodied in various forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will provide exemplary embodiments The concept will be fully conveyed to those skilled in the art. In the drawings, the size of most elements may be exaggerated or deformed for clarity. The same reference numerals in the drawings denote the same or similar structures, and thus their detailed descriptions will be omitted.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided in order to give a thorough understanding of the embodiments of the present application. However, those skilled in the art will realize that the technical solutions of the present application may be practiced without including one or more of the specific details, or other methods, elements, etc. may be adopted to practice the technical solutions of the present application. In other instances, well-known structures, methods, or operations are not shown or described in detail to avoid obscuring aspects of the application.

Electroencephalogram (EEG) signals of patients are collected clinically using EEG equipment (eg, EEG amplifiers). The basic rhythm of the human EEG signal is composed of signal parts such as delta wave and theta wave, alpha wave and beta wave corresponding to different frequencies, usually there is no sharp wave or spike wave or the sharp wave or spike wave is not obvious. The spike in the EEG signal is a sudden, transient wave waveform, generally lasting 20-70ms. The main component of spike waves is negative phase, and the amplitude is variable. A typical spike has a steep rising edge and can have a slope on its falling edge. Spike waves are mostly pathological EEG waveforms, which are common in localized epilepsy, grand mal seizures, myoclonic seizures, and diencephalic epilepsy. Therefore, spike waveforms in EEG signals can be used to detect information related to epileptic foci. A spike is also a bursty fluctuating waveform, but the fluctuating duration of a spike is relatively long compared to a spike that is short in duration (eg, during a transient period of 20-70 ms). During the seizure and interictal period of epilepsy, the pathological EEG waveform not only produces corresponding changes of spike waves or sharp waves, but also can produce corresponding changes of slow wave waveforms including, for example, spike and slow waves and sharp and slow waves. Thus, epileptiform discharges associated with epileptic lesions can be detected by spike waveforms, spike waveforms, and waveforms of slow waves such as spike-slow waves and sharp-slow waves in the EEG signal. In the actual reading diagnosis of EEG signals, these four types of epileptiform discharge-related waveforms account for most of the epileptiform discharges. The waveforms associated with epileptiform discharges in the EEG signal are referred to herein as abnormal discharges (waveforms) associated with epileptic symptoms or lesions.

Electrodes are placed on the surface of a person's head when acquiring EEG signals using an EEG amplifier that meets clinical standards. The placement of the electrodes on the body surface (head) of the human body and the connection between the electrodes and the EEG signal amplifier are called the leads of the EEG. The leads of an EEG include multiple lead channels. For example, a 10-20 lead system has 19 lead channels, in which the Fp1 and Fp2 lead channels detect EEG EEG signals on the left and right forehead, the Cz lead channel detects EEG signals on the top of the head, and There are also F7, F8 lead channels, etc.

An example process of the method for detecting epileptiform discharges according to an embodiment of the present application is described below with reference to FIG. 1 .

The method first acquires the acquired EEG signal in step S110. The sampling rate of the EEG amplifier ranges from 200 to 1000 Hz. The number of lead channels of an EEG amplifier is between 16-32. In the embodiment of the present application, some lead channels are reference lead channels without detection signals. Actually, there are 19 lead channels with detection signals of EEG signals. Clinically, the monitoring time of EEG signal is usually about 2 hours or longer, and its purpose is to detect the EEG signal during epileptic seizure. Automated monitoring of EEG signals between seizures does not need to wait until the onset of seizures, so shorter monitoring durations such as 30 minutes, 15 minutes, 10 minutes or less can be used.

In order to facilitate processing, in the solution described in the embodiments of the present application, the EEG signal of the electroencephalogram is divided into signal segments with the same time length according to each lead channel. For example, the time length of the signal segment can be selected to be 30 seconds, 1 minute, 2 minutes or more in length. It is also possible to choose a shorter time length than 30 seconds, such as 20 seconds, 10 seconds, etc. In this way, the EEG signal on each lead channel includes one or more signal segments. Those skilled in the art should understand that the purpose of segmenting the EEG signal into a plurality of signal segments in the signal processing process is to facilitate signal analysis and processing (eg, determination of relevant thresholds). In actual operation, the EEG signal may not be segmented, but the corresponding part of the continuous EEG signal may be directly selected for calculation and analysis according to the duration of the signal segment of interest.

After acquiring the EEG signal of the patient, a preprocessing operation may be performed on the original waveform data of the EEG signal in step S120. The preprocessing operation is mainly used to reduce noise, reduce the amount of processed data to improve the detection speed and improve the detection accuracy.

The preprocessing step S120 may include a sub-step S121 of reducing the sampling frequency, a sub-step S122 of removing power frequency interference, a sub-step S123 of removing baseline drift interference, a sub-step S124 of removing myoelectric interference, and shielding the faulty lead channel. Sub-step S125, re-reference processing sub-step S126, and sub-step S127 of removing artifact data. One or more of the sub-steps S121 to S127 may be selected for preprocessing, and the order of these sub-steps may be adjusted according to requirements during the execution process.

The sub-step S121 of reducing the sampling frequency may, for example, reduce the sampling rate of the EEG amplifier from more than 500 Hz to about 200 Hz. In the process of detecting epileptiform discharges, reducing the sampling frequency of the acquired EEG signals can speed up the calculation. The sampling frequency of epileptiform discharge detection can generally be reduced to 1/2 or lower of the sampling frequency of the EEG amplifier. However, according to the requirements of detection accuracy, the detection frequency should also be high enough. In the embodiment of the present application, the sampling frequency of epileptiform discharge detection is set to be more than 150 Hz, for example, 200 Hz.

The sub-step S122 of removing power frequency interference is used to eliminate the interference from the power system in the EEG signal. The frequency of the power system is generally 50Hz or 60Hz, so a notch filter with an operating frequency of 50Hz and an integer multiple of 50Hz can be designed to remove power frequency noise.

Baseline drift disturbances come from slow movements of the patient, etc., such as electrode displacement caused by the patient moving the body. Based on experience, the frequency of baseline drift generally does not exceed 0.5 Hz, so in sub-step S123, a high-pass filter with an operating frequency of 0.5 Hz can be used to remove baseline drift interference.

EMG disturbances include, for example, disturbances produced by the patient's clenching of the teeth, which must be present in the EEG signal of the electroencephalogram. Electromyography (EMG) is the temporal and spatial superposition of motor unit action potentials (MUAPs) in numerous muscle fibers. Surface electromyography (SEMG) is the combined effect of superficial muscle EMG and nerve trunk electrical activity on the skin surface, which can reflect neuromuscular activity to a certain extent. The EMG signals were collected together. The EMG interference signal is usually a signal with a higher frequency. Combined with the spike or sharp wave frequency related to epileptiform discharge in the EEG, a low-pass filter with a working frequency of 70 Hz can be designed in sub-step S124 to remove the EMG. electrical interference.

Sub-step S125 relates to equipment failure detection and processing. There may be some lead channel failures when EEG equipment is in use. When it is found that some lead channels are faulty and there is no EEG signal waveform, the faulty lead channels may be masked or replaced in sub-step S125.

The re-referencing process in sub-step S126 involves calibration of the EEG signal based on the reference signal. The EEG signals detected on multiple or all lead channels can be averaged as a reference signal. Subtracting the calculated reference signal from the detected signal on all lead channels removes the common error present on all lead channels. For example, a patient may have epileptiform discharges on the lead channel on the left forehead but not on the lead channel on the right Detection of epileptiform discharges on lead channels.

The removal of artifact data operation in sub-step S127 is used to remove artifacts with significant noise in the EEG signal. Artifacts are deviation components in the detected EEG signals that are significantly different from the real EEG signals, and belong to obvious errors exceeding a preset threshold. Artifacts may be present on one or more lead channels of the EEG signal. For example, EEG signal segments with excessive amplitude are usually considered to be signal interference caused by electrode falling off, or physiological artifacts caused by patients' large-scale movements during EEG monitoring. According to experience, EEG signal segments with signal amplitudes above 300uV can be identified as signal segments with artifacts and not detected. The removal of the artifact data can be done by discarding the signal segment with the artifact.

After preprocessing, the method performs feature matching on the signal segments in step S140 to identify signal segments with epileptiform discharges.

Before the feature matching is performed, a step S130 of preliminarily screening the signal segments may be included to reduce the computational complexity of the feature matching. According to the characteristic that the spikes or sharp waves in the EEG signal have high-frequency signal energy that exceeds the background signal, the normal EEG signal can be used as the background signal for preliminary screening of signal fragments. According to the embodiment of the present application, the preprocessed EEG signal is filtered to obtain a high-frequency EEG signal _SH with a frequency of 10-40 Hz. For adults, epileptiform discharges have high-frequency spikes or sharp waveforms between 14-40Hz, and the frequency of epileptiform discharges in children is slightly lower, so EEGs in the 10-40Hz frequency range are selected. The waveform is filtered and initially screened.

The corresponding high frequency energy E _H is calculated based on the high frequency EEG signal _SH . For example, the high frequency energy E _H can be calculated as the square of the high frequency EEG signal _SH . Generally speaking, signal segments with high frequency energy E _H exceeding the threshold may include other discharges than epileptiform discharges, but the high frequency energy of signal segments without epileptiform discharges must not exceed the threshold. For a certain signal segment of the EEG signal, the signal segment includes n signal sub-segments, assuming that the high-frequency energy of the signal sub-segment i (i is an integer and 0<i≤n) is E _Hi , then the signal sub-segment i The high-frequency energy threshold of the associated signal segment can be calculated by:

Threshold=mean(E _Hi )+std(E _Hi )

Among them, Threshold is the high-frequency energy threshold of the signal segment to which the signal sub-segment i belongs, mean(E _Hi ) is the average value of the high-frequency energy E _Hi of all the signal sub-segments i in the signal segment, and std(E _Hi ) is the signal Standard deviation of the high frequency energy E _Hi of all signal sub-segments i within a segment.

If there is at least one signal sub-segment in the signal sub-segment included in the signal segment, so that the high-frequency energy of the signal sub-segment exceeds the corresponding threshold of the signal segment to which it belongs, then the signal sub-segment whose high-frequency energy in the signal segment exceeds the corresponding threshold Fragments are preserved for further processing. At this time, the signal sub-segments included in the pre-screened signal segments are all signal sub-segments whose high-frequency energy exceeds the corresponding threshold. If the high-frequency energy of all signal sub-segments included in the signal segment does not exceed the threshold value of the signal segment, these signal sub-segments do not have epileptiform discharges and are eliminated, that is, the signal segment is eliminated. It should be understood that, for each signal segment, the high-frequency energy thresholds of all signal sub-segments in the signal segment can be used to calculate the high-frequency energy threshold of the signal segment according to the above formula, and each signal in the signal segment can be calculated as The high-frequency energy of the sub-segment is compared with the high-frequency energy threshold of the signal segment to complete the filtering and preliminary screening of the signal sub-segments within the signal segment. According to the embodiment of the present application, the signal after preliminary screening in step 130 is subjected to subsequent feature matching in units of signal sub-segments.

The step S130 of preliminarily screening the signal segments is used to reduce the number of signal sub-segments for feature matching. In the subsequent processing steps, the primary screening based on the original EEG EEG signals of all frequencies or the preprocessed EEG signals is still used. signal subsegments to detect epileptiform discharges.

Next, the method proceeds to step S140 to perform feature matching on the screened signal sub-segments that may have epileptiform discharges.

Feature matching can be done based on a comparison of the signal features of the signal sub-segments with feature templates. Signal features include waveform features and time-frequency features, so feature matching may specifically include waveform matching between waveform features and corresponding waveform templates and/or time-frequency matching between time-frequency features and corresponding time-frequency templates. As shown in the figure, the feature matching includes sub-step S141 of waveform matching based on waveform features of signal sub-segments and sub-step S142 of time-frequency matching based on time-frequency features of signal sub-segments.

In sub-step S141, the following waveform characteristics of the EEG signal in the signal sub-segment are mainly investigated according to the morphology of the spike/sharp wave and the slow wave (including the spike and slow wave and the sharp and slow wave):

1) The amplitude difference A ₁ and the duration D ₁ of the spike wave or the rising edge of the sharp wave;

2) The amplitude difference A ₂ and the duration D ₂ of the spike wave or the falling edge of the sharp wave;

3) The amplitude difference A ₃ of the rising edge of the slow wave followed by the falling edge of the spike or sharp wave;

4) The standard deviation Std _pre of the amplitude before the rising edge of the spike or sharp wave.

The horizontal axis of the EEG signal is time, and the vertical axis is the amplitude of the signal waveform. First, find the moment when the waveform signal in the signal sub-segment is at the maximum amplitude (maximum value, here is the absolute value) in the signal sub-segment, and then examine the waveform characteristics of the EEG signal in the time period before and after this moment. . Among them, the amplitude difference A1 _of the rising edge of the spike or sharp wave is the continuous rising process before the waveform of the EEG signal of the EEG reaches the maximum value of the signal sub-segment. The amplitude difference (vertical axis height difference) between the minimum value (ascending process minimum value) and the maximum value (ascending process maximum value) during the time period defined by the end time. The continuous rise process includes a continuous rise, and it can also include a continuous rise after the waveform amplitude remains unchanged for a short period of time and then continues to rise again. The duration D ₁ of the rising edge of the spike or sharp wave is the length of the time period (horizontal axis length) between the times corresponding to the rising process minimum value and the rising process maximum value of the rising spike/spike waveform. The amplitude difference A ₂ of the falling edge of the spike or sharp wave is defined by the falling spike/spike waveform from the start time to the end time of the falling process during the continuous falling process after the waveform of the EEG signal reaches the maximum value of the signal sub-segment The amplitude difference between the maximum value of the descent process and the minimum value of the descent process during the time period (vertical axis height difference). A continuous fall process is similar to the situation covered by a continuous rise process, including both a continuous fall and a continuous fall, where the amplitude of the waveform remains the same for a short period of time and then continues to fall again. The duration D ₂ of the falling edge of the spike or sharp wave is the length of the time period (horizontal axis length) between the times corresponding to the maximum value of the falling process and the minimum value of the falling process of the falling spike/spike waveform. The amplitude difference A3 _of the rising edge of the slow wave followed by the falling edge of the spike or sharp wave is that in the falling edge of the spike or sharp wave, the waveform reaches the minimum value of the falling process and then there is a new continuous rising process. The rising spike/spike waveform in the process is between the minimum value (new ascent process minimum value) and the maximum value (new ascent process maximum value) during the time period defined from the start time to the end time of the ascent process Amplitude difference (vertical height difference). The maximum value of the new rising process is generally smaller than the maximum value of the signal sub-segment, that is, less than the maximum value of the rising process of the previous spike or sharp wave rising process, and the new rising process may exist after the spike or sharp wave is involved. Therefore, the new rising process is called the "rising edge of the slow wave" followed by the falling edge of the spike or sharp wave. The standard deviation of the amplitude before the rising edge of the spike or sharp wave, Std _pre , refers to the standard deviation of the amplitude of the signal waveform for a period of time before the process of the rising edge of the spike or sharp wave of the signal sub-segment. For example, the standard deviation of the amplitude of the signal in a time period of about 100 ms before the time corresponding to the minimum value of the rising edge of the spike wave or sharp wave rising edge of the signal sub-segment can be selected to calculate the amplitude standard deviation.

Corresponding waveform feature thresholds may be set for each waveform shape to form a waveform feature threshold set, that is, the waveform template 14A. The aforementioned waveform characteristics of the signal sub-segments are compared to the waveform template 14A to determine whether epileptiform discharges are present in the candidate signal sub-segments. For example, when the measured value of the waveform feature exceeds the corresponding threshold value in the waveform template 14A, it is considered that the threshold value requirement is met at that waveform feature. The degree of importance of the aforementioned waveform features may be the same. According to the embodiments of the present application, when at least one waveform feature meets the threshold requirement, it can be determined that there is an epileptiform discharge in the signal sub-segment. The greater the number of waveform features that meet the threshold requirements, the higher the likelihood that epileptiform discharges are present. The probability that more than 4 waveform features meet the threshold requirement is greater than the case of less than 3 waveform features. According to requirements, signal sub-segments of which 5 or all 6 waveform features meet the threshold requirement in the waveform template 14A may be selected as candidate signal sub-segments. Those signal sub-segments whose number of waveform features satisfying the threshold requirement are below the set number (eg, 4, 5) are determined as waveform mismatches and filtered out.

In sub-step S142, the signal sub-segments that may have epileptiform discharges are screened based on another matching rule. The sub-step S142 further includes the sub-step S1421 of calculating the energy of the signal sub-segments in the low frequency, intermediate frequency and high frequency bands respectively to form the time-frequency feature of the signal sub-segment, and determining the time-frequency of the signal sub-segment based on the time-frequency feature and the time-frequency template 14B. sub-step S1422 of frequency similarity, and sub-step S1423 of identifying whether there is epileptiform discharge in the signal sub-segment based on the comparison of the time-frequency similarity with the corresponding threshold.

According to the commonly used division criteria for EEG signal analysis, the EEG signal is calculated separately according to the definition of low frequency δ, θ (1.6-8 Hz), intermediate frequency α (8-15 Hz) and high-frequency β (15-40 Hz) frequency bands. Low frequency energy E _L , medium frequency energy E _M and high frequency energy E _H . The energy combination of each frequency band in each time period of 120 ms before and after the time corresponding to the maximum amplitude value of the signal sub-segment determined in sub-step S141 can be selected to constitute the time-frequency feature of the signal sub-segment. The time length of the 120 ms period was chosen based on theoretical or empirical coverage of the spike/spike waveform of the EEG signal. In general, a time period of 120 ms each before and after the moment corresponding to the amplitude maximum (240 ms in total) is sufficient to cover the spike/spike waveform features.

In the time sub-segment of 240ms, according to the detection sampling frequency above, it can be known that it includes 48 sampling moments (sampling points). Correspondingly, a time-frequency energy matrix of 48*3 dimensions (energy data of 48 sampling instants, and energy values corresponding to 3 frequency bands exists at each sampling instant) can be obtained by combining the energy of the sub-bands. Correspondingly, the preset time-frequency template 14B is also a time-frequency energy matrix template based on the sub-band energy at the sampling moment, and has a dimension of 48*3. The time-frequency template 14B is a sub-band energy template that characterizes the reference time-frequency characteristics of epileptiform discharges, and can be presented in the form of a visual image, as shown in FIG. 3 . In Figure 3, the horizontal axis is time, and the vertical axis is the sub-band energy value. The darker the color, the higher the sub-band energy, and the lighter the color, the lower the sub-band energy. In the time-frequency energy matrix template, the value of each element is the energy weight of the corresponding sub-band at this moment.

In sub-step S1422, the time-frequency similarity of the signal sub-segments may be determined based on the calculated time-frequency energy matrix. The time-frequency similarity represents the degree of matching between the signal sub-segments and the reference signal sub-segments with epileptic discharges in time-frequency features. When calculating the unit time-frequency similarity of each unit of the time-frequency energy matrix, binarization processing may be performed on the sub-band energy measurement value represented by each unit in the time-frequency energy matrix. Calculated at each sampling time by comparing the sub-band energy measurement (the value of the corresponding cell of the time-frequency energy matrix) with the sub-band energy threshold (the sub-band energy threshold of the corresponding cell of the time-frequency energy matrix template) at each sampling instant. Unit time-frequency similarity at units. A corresponding energy threshold may be set for each of the multiple frequency bands, and a corresponding sub-band energy threshold may be further set for each unit. The energy threshold can be calculated by smoothing or averaging the sub-band energy of the EEG waveform in the presence of epileptiform discharges, or the energy threshold parameters of each unit in the time-frequency energy matrix template can be set based on experience. If the measured value of the sub-band energy of each cell of the matrix exceeds the corresponding sub-band energy threshold, the cell time-frequency similarity indicates a high probability of the presence of epileptiform discharges, otherwise within the energy threshold, the probability is low. Different unit time-frequency similarity can be set according to the comparison result between the measured value and the energy threshold. For example, the unit that exceeds the energy threshold has a larger unit time-frequency similarity value, and if it does not exceed the energy threshold, it has a smaller unit time-frequency similarity value. And further, it can be set that the greater the degree of exceeding the threshold, the greater the unit time-frequency similarity value, and the like. The binarization process described above uses two unit time-frequency similarity values such as 0 and 1 to characterize the likelihood that a measure of subband energy involves epileptiform discharges, where 1 indicates that the unit (i.e., the difference between the unit and the unit). The sampling time corresponding to the frequency band energy measurement value) involves epileptiform discharge or the epileptiform discharge process includes the sampling time corresponding to the measurement value of the unit, and 0 represents the unit (that is, the sampling time corresponding to the subband energy measurement value of the unit) The epileptiform discharge is not involved or the epileptiform discharge process does not include the sampling time corresponding to the measurement value of the unit. The unit time-frequency similarity of the unit whose measurement value exceeds the energy threshold is 1, and the unit time-frequency similarity of the unit that does not exceed the energy threshold is 0, then the time-frequency similarity of the corresponding unit of each unit in the time-frequency energy matrix is [ A value in 0,1]. The time-frequency energy matrix template 14B sets different weights for each unit based on the importance of detection for the sub-band energy at different times (the weight value is, for example, in [0, 1], or further in [-0.5, 1]. real value). Then, the time-frequency energy matrix is multiplied by the time-frequency energy matrix template 14B to obtain a weighted time-frequency energy matrix. Wherein, the unit time-frequency similarity value of each unit is respectively multiplied by the weight value of the corresponding unit in the matrix template 14B to obtain the weighted unit time-frequency similarity of each unit.

The time-frequency similarity of the signal sub-segments is calculated by summing the weighted unit time-frequency similarity of each unit in the time-frequency energy matrix. Then, in sub-step S1423, the time-frequency similarity of the signal sub-segment is compared with the corresponding time-frequency similarity threshold to identify whether there is epileptiform discharge in the signal sub-segment. For example, if the time-frequency similarity exceeds the time-frequency similarity threshold, it indicates that there is epileptiform discharge in the signal sub-segment, otherwise, it is considered that there is no epileptiform discharge. Whether the time-frequency similarity threshold is satisfied or not is equivalent to judging the degree of agreement between the signal sub-segment and the template of the reference signal sub-segment representing epileptic discharges in the interictal period. Signal sub-segments determined to meet the requirements by time-frequency matching are retained.

As shown in Fig. 2, for the original EEG signal waveform of the upper part, in the preliminary screening of step S130, it is found that the high-frequency energy of the filtered signal sub-segments 201 to 205 exceeds the high-frequency energy threshold, and the relevant preliminary screening results are in The middle part is rendered. Therefore, these signal sub-segments are reserved for feature matching in step S140. The lower part particularly shows the case where the time-frequency feature matching of multi-band energy is performed in sub-step S142. Compared with the time-frequency template shown in FIG. 3, only the time-frequency similarity of the signal sub-segment 203 meets the corresponding threshold requirements. , the signal sub-segment 203 is considered to have epileptiform discharges and remains, while the signal sub-segments 201, 202, 204 and 205 are screened out. Those skilled in the art can understand that the epileptiform discharge detection process shown in FIG. 2 is only an example and not a limitation, and sub-steps S141 and S142 in step S140 related to feature matching can be performed individually or in combination.

After feature matching is performed on the signal sub-segments, the identification result may be further corrected through step S150. The correction step can screen out those signal sub-segments that are mistakenly identified as having epileptiform discharges and originally belong to the normal EEG, and can also retrieve the missed signal sub-segments for re-detection.

Step S150 includes a sub-step S151 of screening out physiological waveforms, a sub-step S152 of screening out non-stereotypical repetitive waveforms, and/or a sub-step S153 of updating a feature template, and so on.

In sub-step S151 , those signal sub-segments that may be erroneously identified as having the physiological waveform form of epileptiform discharges are excluded. The morphological characteristics and time-frequency characteristics of some physiological waveforms are very similar to the abnormal discharges of epileptiform discharges, but physiological waveforms usually have specific appearance positions, so when more abnormal signal sub-segments are detected in the corresponding specific positions, These signal sub-segments can be excluded through the physiological waveform exclusion sub-step S152 to reduce the influence on the detection result.

Physiological waveforms can specifically include:

1) Physiological waveforms of eye blinks and eye movements: Blinks may have similar physiological waveforms in the symmetrical left and right forehead lead channels Fp1 and Fp2, while eye movements may be in the four lead channels Fp1, Fp2, F7, and F8. Similar physiological waveforms exist. Since the waveforms of the contralateral lead channels have high similarity during eye blinking and eye movement, they can be excluded by the waveform correlation between lead channels.

2) Physiological waveform of apex wave: apex wave is related to sleep and is not abnormal discharge, but the shape of apex wave is also similar to abnormal discharge. The apical wave physiological waveform usually appears in lead channel Cz and nearby lead channels, which can be excluded by the correlation of lead channel waveforms near the center of the head.

3) Physiological waveform of alpha activity in the occipital region: The alpha rhythm in the occipital region is a normal physiological waveform, but its shape and abnormal discharge also have a certain similarity. time to exclude.

Sub-step S152 is to filter out the signal sub-segments where the non-stereotyped repetitive waveforms do not have stereotyped repetitive waveform characteristics. Abnormal discharges such as epileptiform discharges often have stereotyped repetitive waveform characteristics. Stereotyped repetitions are repeated/multiple repetitions of similar spikes or sharp waves in the EEG signal waveform of the same lead channel. The characteristics of spikes or sharp waves in EEG waveforms of different patients are generally different, so screening of non-stereotyped repetitive waveforms cannot be done between EEG signals of different patients, and usually only EEG signals on the same lead channel of the same patient are compared. Non-stereotyped repetitive waveforms can be excluded by calculating correlations between detected signal sub-segments believed to have abnormal discharges, while retaining stereotypical repetitive waveforms that are believed to indeed have stereotyped repetitive characteristics. For example, the correlation data between different signal sub-segments on the same lead channel, especially the signal sub-segments with abnormal waveforms, can be calculated, and the correlation data can be compared with the corresponding thresholds to determine whether the signal sub-segments have stereotyped repetitions. waveform characteristics.

For the correlation between signal sub-segments, the pearson correlation value can be used for calculation, and the formula is as follows:

Among them, ρ _{X, Y} is the pearson correlation value. X and Y represent different abnormal discharge signal sub-segments on the same lead channel, respectively. A corresponding threshold is set for the pearson correlation value. When the average value of the correlation between a signal sub-segment with abnormal discharge and other signal sub-segments with abnormal discharge on the same lead channel is less than the corresponding correlation threshold, the signal sub-segment is determined. Fragments are considered free of stereotyped repetitive waveforms. On the contrary, if the average value of the correlation of the signal sub-segment is greater than the corresponding correlation threshold, it is considered that the signal sub-segment has a stereotyped repetitive waveform and belongs to an abnormal waveform, which is reserved for the detection of epileptic discharge. Other algorithms for calculating correlations between signal sub-segments and/or averaging algorithms can be used to calculate the overall correlation between a signal sub-segment and other signal sub-segments on multiple other channels of the same lead to screen for non-stereotyped repetitive waveforms .

Sub-step S153 detects the abnormal discharge (especially epileptiform discharge) waveform information in the key lead channel to update the feature matching template such as the waveform template and the time-frequency template, so that the template in the feature matching process can be more accurately used as a patient with patients. Individually characterized reference to monitor epileptiform discharges, reintroducing signal sub-segments that were previously incorrectly screened out to the set of candidate signal sub-segments. Sub-step S153 further includes sub-step S1531 of determining key lead channels and sub-step S1532 of determining abnormal discharge characteristic template.

In sub-step S1531, the lead channel with more signal sub-segments for which abnormal waveforms are detected is determined as the key lead channel. For example, a lead channel with a total time length of more than 20 seconds (greater than 20 s/h) in each hour of signal sub-segments with abnormal waveforms is regarded as a focus lead channel. If the length of each abnormal discharge signal sub-segment is 480ms (approximately 500ms), it is determined that a key lead channel needs to have at least 40 signal sub-segments with abnormal discharge waveform per hour on the lead channel.

Since abnormal discharge waveforms on the same lead channel usually have stereotyped repetitive signal features, which further include waveform features and/or time-frequency energy features, the abnormal discharge signal sub-segments detected on the same key lead channel can be divided into sub-segments. Perform waveform averaging to obtain an abnormal discharge waveform template corresponding to the patient, and extract at least one of the abnormal discharge time-frequency templates of abnormal discharge signal sub-segments. On each lead channel, the waveform template and time-frequency template of the abnormal discharge of the patient on the lead channel can be generated, so as to find the common waveform characteristics and time-frequency characteristics of the epileptic discharge on the lead channel. The characteristic template of the abnormal discharge waveform may be a waveform template or a time-frequency template, so the generated abnormal discharge characteristic template may update the waveform template in step S141 and/or the time-frequency template in step S142. Further, the similarity between the previously screened signal sub-segments on the key lead channel and the abnormal discharge waveform template and/or time-frequency template can be calculated, or the sub-step S141 of waveform matching and/or the sub-step S141 of time-frequency matching can be performed again. In step S142, those signal sub-segments with a high degree of similarity and/or meeting the requirements of the waveform characteristic threshold and/or the time-frequency characteristic threshold are re-included into the subsequent abnormal discharge detection and statistics. Therefore, the sub-step S153 of updating the feature template can be regarded as a remedial measure for missed detection, which is different from the step of screening out the wrong detection in the sub-steps S151 and S152.

Next, the method determines statistics of epileptiform discharges and/or generates a detection report in step S160. Statistics may include statistical summing of duration lengths in signal sub-segments in which epileptiform discharges are present on each lead channel. The statistical data can be combined with the position coordinates of each lead channel to generate a visual report graph. In the top view of the patient's head shown in Figure 4, the duration of epileptiform discharges at various locations (which correlate to the electrode positions of the lead channels) are represented in different colors and/or depths (grayscale), The darker the position of the abnormal discharge, the longer the duration of abnormal discharge, indicating that the more epileptiform discharge occurs in this position, and the greater the possibility of epilepsy focus. For example, epileptiform discharges occurred for a longer duration (approximately 20 seconds/hour) in the patient's right crown position 301 than at the median crown position 302 . This visual report graph can provide the user (doctor) with two-dimensional distribution/marking of abnormal discharge conditions and/or probability, and the doctor can also further mark or perform diagnostic description on the two-dimensional distribution report graph of abnormal discharge.

According to the scheme of detecting epileptiform discharges in EEG signals as described above, compared with the common scheme of diagnosing epileptic seizures in patients by manual observation of EEG signals by doctors, an automated EEG signal detection method is introduced , can be used as a computer-aided detection tool to output the detection results and indicators related to whether there is epileptiform discharge in the EEG signal monitored by the patient during the interictal period, quantitative statistical data and visualized and other presentation methods for the reference of doctors, effectively reducing medical costs. It reduces the workload of personnel and the probability of false detection and improves the efficiency of diagnosis, especially for the targeted elimination of physiological waveforms that are prone to false detection. At the same time, the automated detection scheme of the present application can be accurate to the specific lead channel of the EEG signal, and optimize and update detection standards and templates based on accumulated historical data, further improving the accuracy and speed of detection.

In an exemplary embodiment of the present application, there is also provided a computer-readable storage medium on which a computer program is stored, the program including executable instructions, which, when executed by, for example, a processor, can implement any one of the above The steps of the method for detecting epileptiform discharges described in the Examples. In some possible implementations, various aspects of the present application can also be implemented in the form of a program product, which includes program code, which is used to cause the program product to run on a terminal device when the program product is executed. The terminal device performs the steps according to various exemplary embodiments of the present application described in the method for detecting epileptiform discharges in this specification.

The program product for implementing the above method according to the embodiments of the present application may adopt a portable compact disc read only memory (CD-ROM) and include program codes, and may be executed on a terminal device such as a personal computer. However, the program product of the present application is not limited thereto, and in this document, a readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or a combination of any of the above. More specific examples (non-exhaustive list) of readable storage media include: electrical connections with one or more wires, portable disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

The computer-readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, carrying readable program code therein. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A readable storage medium can also be any readable medium other than a readable storage medium that can transmit, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. Program code embodied on a readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out the operations of the present application may be written in any combination of one or more programming languages, including object-oriented programming languages—such as Java, C++, etc., as well as conventional procedural Programming Language - such as the "C" language or similar programming language. The program code may execute entirely on the user computing device, partly on the user device, as a stand-alone software package, partly on the user computing device and partly on a remote computing device, or entirely on the remote computing device or server execute on. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (eg, using an Internet service provider business via an Internet connection).

In an exemplary embodiment of the present application, there is also provided an electronic device, which may include a processor, and a memory for storing executable instructions of the processor. Wherein, the processor is configured to perform the steps of the method for detecting epileptiform discharges in any one of the above embodiments by executing the executable instructions.

As will be appreciated by one skilled in the art, various aspects of the present application may be implemented as a system, method or program product. Therefore, various aspects of the present application can be embodied in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, which may be collectively referred to herein as implementations "circuit", "module" or "system".

The electronic device 500 according to this embodiment of the present application is described below with reference to FIG. 5 . The electronic device 500 shown in FIG. 5 is only an example, and should not impose any limitations on the functions and scope of use of the embodiments of the present application.

As shown in FIG. 5, electronic device 500 takes the form of a general-purpose computing device. Components of the electronic device 500 may include, but are not limited to, at least one processing unit 510, at least one storage unit 520, a bus 530 connecting different system components (including the storage unit 520 and the processing unit 510), a display unit 540, and the like.

Wherein, the storage unit stores a program code, and the program code can be executed by the processing unit 510, so that the processing unit 510 executes the method according to the present application described in the method for detecting epileptiform discharge in this specification. steps of an exemplary embodiment. For example, the processing unit 510 may perform the steps shown in FIG. 1 .

The storage unit 520 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 5201 and/or a cache storage unit 5202 , and may further include a read-only storage unit (ROM) 5203 .

The storage unit 520 may also include a program/utility 5204 having a set (at least one) of program modules 5205 including, but not limited to, an operating system, one or more application programs, other program modules, and programs Data, each or some combination of these examples may include an implementation of a network environment.

The bus 530 may be representative of one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local area using any of a variety of bus structures. bus.

The electronic device 500 may also communicate with one or more external devices 600 (eg, keyboards, pointing devices, Bluetooth devices, etc.), with one or more devices that enable a user to interact with the electronic device 500, and/or with Any device (eg, router, modem, etc.) that enables the electronic device 500 to communicate with one or more other computing devices. Such communication may occur through input/output (I/O) interface 550 . Also, the electronic device 500 may communicate with one or more networks (eg, a local area network (LAN), a wide area network (WAN), and/or a public network such as the Internet) through a network adapter 560 . Network adapter 560 may communicate with other modules of electronic device 500 through bus 530 . It should be understood that, although not shown, other hardware and/or software modules may be used in conjunction with electronic device 500, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives and data backup storage systems.

From the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described herein may be implemented by software, or may be implemented by software combined with necessary hardware. Therefore, the technical solutions according to the embodiments of the present application may be embodied in the form of software products, and the software products may be stored in a non-volatile storage medium (which may be CD-ROM, U disk, mobile hard disk, etc.) or on the network , including several instructions to cause a computing device (which may be a personal computer, a server, or a network device, etc.) to perform the method for detecting epileptiform discharges according to embodiments of the present application.

Other embodiments of the present application will readily occur to those skilled in the art upon consideration of the specification and practice of what is disclosed herein. This application is intended to cover any variations, uses or adaptations of this application that follow the general principles of this application and include common knowledge or conventional techniques in the technical field not disclosed in this application . The specification and examples are to be regarded as exemplary only, with the true scope and spirit of the application being indicated by the appended claims.

Claims (17)

1. A method for detecting epileptiform discharges, comprising:

acquiring an electroencephalographic signal of a patient during an interval of an epileptic seizure, wherein the electroencephalographic signal includes one or more signal segments on at least one lead channel, each of the signal segments including one or more signal sub-segments;

performing feature matching on the signal sub-segments to identify the signal sub-segments with epileptiform discharges; and

statistics of epileptiform discharges are determined.

2. The method of claim 1, wherein prior to feature matching the signal sub-segments, the method further comprises:

and screening out the signal sub-segments without the epileptic-like discharge based on the comparison of the energy of the signal sub-segments in the high-frequency band and the energy threshold value of the signal segments to which the signal sub-segments belong in the high-frequency band.

3. The method according to claim 1 or 2, wherein feature matching the signal sub-segments comprises at least one of:

performing waveform matching based on the waveform characteristics of the signal sub-segments;

and performing time-frequency matching based on the time-frequency characteristics of the signal sub-segments.

4. The method of claim 3, wherein the waveform matching comprises identifying whether epileptiform discharges are present in the signal sub-segments based on a comparison of at least one of the following waveform characteristics to a waveform template:

the amplitude difference of the rising edge of the spike wave or sharp wave in the signal sub-segment;

the duration of a spike or spike rising edge in the signal sub-segment;

the amplitude difference of the falling edge of the spike wave or sharp wave in the signal sub-segment;

a duration of a spike or spike falling edge in the signal sub-segment;

a spike or spike in the signal sub-segment followed by a slow wave rising edge;

the standard deviation of the amplitude before the rising edge of the spike or spike in the signal sub-segment.

5. The method of claim 3, wherein the time-frequency matching comprises:

respectively calculating the energy of the signal sub-segments in low frequency, intermediate frequency and high frequency bands to form time-frequency characteristics of the signal sub-segments;

determining the time-frequency similarity of the signal sub-segments based on the time-frequency characteristics and a time-frequency template;

identifying whether epileptiform discharges are present in the signal sub-segments based on a comparison of the time-frequency similarities to respective thresholds.

6. The method according to claim 5, characterized in that the energies of the signal sub-segments in the low, medium and high frequency bands in the time segment with peak amplitude are calculated separately to constitute the time-frequency characteristics of the signal sub-segments.

7. The method of claim 5, wherein the time-frequency template is a time-frequency energy matrix template with sub-band energies based on time, and wherein determining the time-frequency similarity of the signal sub-segments based on the time-frequency features and the time-frequency template comprises:

generating a time-frequency energy matrix as the time-frequency features based on the calculated energy of the signal sub-segments at the low frequency, the intermediate frequency and the high frequency bands at the sampling time in the time period;

and multiplying the time-frequency energy matrix and the time-frequency energy matrix template to calculate the time-frequency similarity.

8. The method of claim 7, wherein multiplying the time-frequency energy matrix by the time-frequency energy matrix template comprises:

carrying out binarization processing based on the value of the unit of the time-frequency energy matrix and the unit threshold value of the corresponding time-frequency energy matrix;

and multiplying the values of the units of the time-frequency energy matrix subjected to binarization by the values of the corresponding units of the time-frequency energy matrix template to calculate the unit time-frequency similarity of each unit of the time-frequency energy matrix.

9. The method of claim 8, wherein the values of the cells of the time-frequency energy matrix template are weight values.

10. The method according to any of claims 7 to 9, characterized in that the time-frequency similarity is calculated by summing the cell time-frequency similarities of all cells of the time-frequency energy matrix.

11. The method according to claim 1 or 2, characterized in that, prior to the feature matching, the acquired electroencephalographic signal is preprocessed, the preprocessing comprising at least one of:

the sampling frequency is reduced;

removing power frequency interference;

removing baseline wander interference;

removing myoelectric interference;

shielding the faulty lead channel;

re-reference processing;

the artifact data is removed.

12. The method of claim 3, further comprising correcting the recognition of epileptiform discharges after feature matching the signal sub-segments, the correcting comprising at least one of:

screening out signal sub-segments belonging to physiological waveforms;

signal sub-segments of the waveform without the engraved repeats are screened out.

13. The method of claim 12, wherein determining whether the signal sub-segments have a stereotyped repeating waveform is based on a comparison of correlations between different signal sub-segments on the same lead channel to respective thresholds.

14. The method of claim 12, wherein the correcting further comprises:

at least one of the waveform template and the time-frequency template is updated based on waveform characteristics and time-frequency characteristics of signal sub-segments on a lead channel having a plurality of signal sub-segments with epileptiform discharges.

15. The method of claim 1, wherein determining statistics of epileptiform discharges comprises:

the duration of epileptiform discharges for each lead channel is counted to generate a detection report.

16. A computer-readable storage medium, having stored thereon a computer program comprising executable instructions that, when executed by a processor, perform the method according to any one of claims 1 to 15.

17. An electronic device, comprising:

a processor; and

a memory for storing executable instructions of the processor;

wherein the processor is configured to execute the executable instructions to implement the method of any of claims 1 to 15.

Images