A method for localizing epileptogenic zones based on high-frequency oscillations and connectivity

Abstract

This invention discloses a method for locating epileptogenic zones based on high-frequency oscillations and connectivity, comprising the following steps: 1. Acquiring SEEG data and selecting representative channels; 2. Filtering to acquire 54-200Hz signals and selecting baseline and target data; 3. Acquiring normalized high-frequency energy; 4. Calculating time and energy coefficients to obtain a high-frequency epileptogenicity index; 5. Filtering to acquire 12-45Hz signals; 6. Calculating nonlinear regression analysis; 7. Calculating total intensity; 8. Defining and calculating the connectivity high-frequency epileptogenicity index and evaluating performance. This invention can accurately locate epileptogenic zones in patients with different seizure patterns, and has potential application value in epilepsy treatment.

Description

Technical Field

[0001] This invention relates to the field of multichannel neural signal analysis in patients with epilepsy, specifically to high-frequency oscillations and a correlation index based on brain network connectivity for locating the epileptogenic zone in patients with different seizure patterns. Background Technology

[0002] Epilepsy is a serious chronic neurological disorder that disrupts the normal activity of neurons in the brain. During a seizure, patients may be injured or even killed in emergencies, causing immense psychological stress and difficulties in their work and daily life. For patients with drug-resistant epilepsy, epilepsy surgery can effectively treat seizures. The epileptogenic zone is defined as the brain region where a seizure occurs. Typically, assessment of the epileptogenic zone involves multichannel intracranial electroencephalography (iEEG) recordings, particularly stereotactic electroencephalography (SEEG). Identifying the epileptogenic zone using these recordings is crucial before surgery, as the goal of the procedure is to remove it. In fact, precise localization of the epileptogenic zone has always been a major challenge in epilepsy surgery.

[0003] In recent years, researchers have developed methods such as the epileptogenicity index (EI), epileptogenicity maps, high-frequency epileptogenicity index (HFEI), epileptogenicity grading (ER), and epileptogenic zone fingerprinting to locate epileptogenic zones. These methods, based on the detection of the temporal or spatial characteristics of high-frequency oscillations to locate epileptogenic zones, have received considerable attention and achieved significant results in many studies. However, they are not ideal for locating epileptogenic zones in certain situations. In epilepsy patients, approximately 75% of seizure patterns involve high-frequency oscillations (HFOs), while the remainder are slow-seizure patterns. The efficiency of EI, HFEI, and other HFO-based methods becomes lower for slow-seizure patterns. Furthermore, when similar seizure activity is presented in the recording channel, these methods cannot distinguish their temporal sequence or energy intensity differences, resulting in poor estimation results.

[0004] Epilepsy can generally be considered a network disorder, and hyperexcitatory foci (HFOs) cannot capture the network properties of the brain by processing each channel individually. That is, when HFOs are not observed in the seizure zone, relying solely on the electrophysiological characteristics of HFOs is insufficient. On the other hand, previous research has shown that removing hyperexcitatory sites is not the optimal method for reducing seizure rates. Instead, based on the concepts of network structure and connectivity, removing normal sites located at key network points—the "driving factors"—is often more effective. This is because the connectivity between SEEG signals increases before a seizure, gradually decreases in the early stages of a seizure, and gradually increases again in the later stages.

[0005] For example, Wang Haixiang. Application of stereotactic electroencephalography epileptogenic index analysis in epileptogenic zone localization and epileptogenic network evaluation [J]. Chinese Journal of Neurology. 2017; 6:362-367. This paper proposes a new quantitative method for SEEG based on the analysis of high-frequency activity (60-90Hz) during the seizure phase of stereotactic electroencephalography (SEEG), calculates the high-frequency epileptogenic index (HFEI), and thus locates the epileptogenic zone of epilepsy patients and evaluates the epileptogenic network. However, this paper only utilizes the high-frequency oscillation of 60-90Hz and ignores the slow-seizure pattern of epilepsy, which will lead to poor predictive effect when locating the epileptogenic zone of epilepsy patients with a slower seizure pattern. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides an improved method for locating epileptogenic zones—the Connectivity High Frequency Epileptogenicity Index (cHFEI). This method combines the connectivity of the high frequency epileptogenic zone (HFO) with that of the brain network, utilizing the connectivity of the HFO and the brain network to more accurately locate the epileptogenic zone. The method is accurate and highly stable.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0008] A method for locating epileptogenic zones based on high-frequency oscillations and connectivity, characterized by the following steps:

[0009] Step S1: Collect SEEG data and select representative channels;

[0010] Step S2: Filter to acquire the 54-200Hz signal and select baseline and target data;

[0011] Step S3: Obtain the normalized high-frequency energy NHFE;

[0012] Step S4: Calculate the time coefficient TC and energy coefficient EC based on NHFE, and obtain the high-frequency epileptogenicity index HFEI;

[0013] Step S5: Filter to obtain the 12-45Hz signal;

[0014] Step S6: Calculate nonlinear regression analysis;

[0015] Step S7: Calculate the total strength TOT;

[0016] Step S8: Define and calculate the connectivity high-frequency epileptogenicity index (cHFEI), and locate the epileptogenic zone based on the cHFEI.

[0017] A further improvement of the technical solution of the present invention is that: in step S1, stereotactic electroencephalogram (SEEG) data of twenty patients with different epileptic seizure patterns are collected according to the stereotactic method, and 10-30 representative channels are selected to cover the main areas of epileptic seizures according to the patients' epileptic seizure conditions.

[0018] A further improvement of the technical solution of the present invention is that: in step S2, a second-order IIR notch digital filter and a fifth-order IIR Butterworth digital filter are used to filter the original signal to the high-frequency band of 54-200Hz; baseline data and target data are manually selected. The principle for selecting baseline data is that there are no obvious abnormal signals before the epileptic seizure, and the target data must cover the entire seizure process.

[0019] A further improvement of the technical solution of the present invention is that: in step S3, the bandpass signal is converted into a high-frequency energy spectrum by means of amplitude square and window smoothing, and the average value of the high-frequency energy of the baseline data is calculated as the baseline value of the high-frequency energy; the high-frequency energy of all channels is divided by the high-frequency energy baseline value of the channel to obtain the normalized high-frequency energy NHFE.

[0020] A further improvement to the technical solution of the present invention is that the specific operation of step S4 is as follows:

[0021] Calculate the onset time threshold for each channel i, defined as the maximum baseline NHFE plus ten times the standard deviation of the baseline NHFE:

[0022] threshold onset time =max(NHFE) BL )+10σ(NHFE BL ),

[0023] When the normalized energy of the target data in each channel exceeds the threshold, this moment is determined to be the start time of abnormal activity, and the channels are sorted according to the start time of each channel. The time coefficient TC is defined as the reciprocal of the order of each channel.

[0024] The average energy within 250ms before and after the earliest start of each channel is calculated using NHFE as the energy coefficient EC, and finally the HFEI of each channel i is obtained:

[0025]

[0026] A further improvement of the technical solution of the present invention is that: in step S5, the original data is bandpass filtered, with the passband being the β-γ frequency band, 12-45Hz, and then downsampled to 256Hz.

[0027] A further improvement to the technical solution of the present invention is that: in step S6, based on h2 The nonlinear correlation index is used to calculate nonlinear regression analysis, performing piecewise linear regression between each pair of signals to test all offsets of one signal relative to the other signal within the maximum hysteresis; for two signals x and y, the piecewise linear approximation f of the transfer function between x and y is defined as:

[0028] y n-τ =f(x) n )+e n ,

[0029]

[0030] h 2 The goodness of fit of a regression is measured using the approximate value f.

[0031]

[0032] A further improvement to the technical solution of the present invention is that: in step S7, all pairs of h are calculated. 2 The values ​​are calculated using a sliding window of length 3s and step size 1s, with a maximum delay of 0.1s, to generate a binary connection matrix; then, for each channel within each time window, the total intensity TOT, independent of direction, is summarized.

[0033] A further improvement to the technical solution of this invention lies in: In step S8, for each patient's SEEG channel, the average total intensity and HFEI of the 8 seconds before seizure are summarized and normalized respectively. Combined with TOT and HFEI, the combined index cHFEI of the connectivity high-frequency epileptogenicity index is obtained:

[0034]

[0035] A further improvement of the technical solution of the present invention is that it provides accurate and stable localization of the epileptogenic zone.

[0036] The technological advancements achieved by this invention due to the adoption of the above technical solutions are as follows:

[0037] This invention can more accurately locate the epileptogenic zone, solving the problem that the localization method based solely on HFO has low accuracy in slower seizure modes.

[0038] This invention solves the problem that localization methods based solely on the time-frequency characteristics of HFO cannot detect the temporal sequence (or seizure delay) or energy intensity of each channel when similar activities are observed within the channels, resulting in low accuracy in locating the epileptogenic zone. Attached Figure Description

[0039] Figure 1 This is a flowchart of the invention;

[0040] Figure 2 This is a comparison chart of the area under the ROC curve (AUC) for different methods. Detailed Implementation

[0041] The present invention will be further described in detail below with reference to embodiments:

[0042] This invention is a method for locating epileptogenic zones based on high-frequency oscillations and connectivity, comprising the following steps, such as... Figure 1 As shown:

[0043] Step S1: Using stereotactic methods, collect stereotactic electroencephalogram (SEEG) data from twenty patients with different seizure patterns, and select 10-30 channels to cover the main seizure areas based on the patients' seizure patterns.

[0044] Step S2: Filter the original signal to the high-frequency band of 54-200Hz using a second-order IIR notch digital filter and a fifth-order IIR Butterworth digital filter. Manually select baseline and target data. The baseline data should be SEEG data without obvious abnormal signals before the seizure, and the target data should cover the entire seizure process.

[0045] Step S3: Convert the bandpass signal into a high-frequency energy spectrum using amplitude squared and window smoothing, and calculate the average high-frequency energy of the baseline data as the baseline value of the high-frequency energy. Further, divide the high-frequency energy of all channels by the baseline value of the high-frequency energy of that channel to obtain the normalized high-frequency energy (NHFE).

[0046] Step S4: Calculate the time coefficient (TC) and energy coefficient (EC) based on NHFE, and calculate the high-frequency epileptogenicity index (HFEI) for each channel.

[0047] Calculate the onset time threshold for each channel i, defined as the maximum baseline NHFE plus ten times the standard deviation of the baseline NHFE:

[0048] threshold onset time =max(NHFE) BL )+10σ(NHFE BL );

[0049] When the normalized energy of the target data in each channel exceeds the threshold, we determine this moment as the start time of abnormal activity, and sort the channels according to their start time. The time coefficient TC is defined as the reciprocal of the channel order.

[0050] The average energy within 250 ms before and after the earliest start of each channel is calculated using NHFE as the energy coefficient EC. Finally, the HFEI of each channel i is obtained:

[0051]

[0052] Step S5: Perform bandpass filtering on the original data, with the passband being the β-γ band (12-45Hz), and then downsample to 256Hz.

[0053] Step S6: Based on h 2 The nonlinear correlation index is used to calculate nonlinear regression analysis, performing piecewise linear regression between each pair of signals to test all offsets of one signal relative to the other within the maximum lag. For two signals x and y, the piecewise linear approximation f of the transfer function between x and y is defined as:

[0054] y n-τ =f(x) n )+e n ,

[0055]

[0056] h 2 The measure of goodness of fit in regression is equivalent to the r-squared value used in linear regression. 2 This can be explained using the approximation f:

[0057]

[0058] Step S7: Calculate all pairs of h 2 The values ​​are calculated using a sliding window of length 3s, step size 1s, and a maximum delay of 0.1s, generating a binary connectivity matrix. Then, within each time window, for each channel (graph node), the orientation-independent total intensity (TOT) is summarized.

[0059] Step S8: For each patient's SEEG channel, we summarize the average total intensity and HFEI in the 8 seconds before seizure, and normalize them separately. Combining TOT and HFEI, we obtain the combined index called the connectivity high-frequency epileptogenicity index (cHFEI):

[0060]

[0061] The performance of this invention was evaluated by calculating the area under the ROC curve (AUC) of different epileptogenic zone localization methods.

[0062] Comparison of the area under the ROC curve (AUC) results of different methods is as follows: Figure 2 As shown. In this invention, SEEG data from 20 epilepsy patients with different seizure patterns were analyzed. Figure 2(A) Comparison of overall AUC values ​​for different methods across all patients. Compared to detection methods based solely on HFO (EI, HFEI), the mean AUC value of cHFEI is significantly higher than that of HFEI and EI, indicating the best predictive performance. Furthermore, HFEI and EI exhibit large variances, while cHFEI shows relatively small variances, demonstrating the high stability of this invention and its applicability to different types of epileptic seizures. Figure 2 (B) Comparison of AUC values ​​for each patient using different methods. Except for patients s1 and s19, whose AUC values ​​were slightly lower than HFEI, the AUC values ​​for the remaining patients were higher than the other two methods. In particular, the AUC values ​​for patients s9, s16, and s20 were all 1, highly consistent with the epileptogenic zones marked by clinicians.

[0063] The above results indicate that the present invention can accurately locate the epileptogenic zone in patients with different epileptic seizure patterns, and has potential application value in the treatment of epilepsy.

[0064] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for locating epileptogenic zones based on high-frequency oscillations and connectivity, characterized in that... Includes the following steps: Step S1: Collect SEEG data and select representative channels; Step S2: Filter to acquire the 54-200Hz signal and select baseline and target data; Step S3: Obtain the normalized high-frequency energy NHFE; Step S4: Calculate the time coefficient (TC) and energy coefficient (EC) based on NHFE, and obtain the high-frequency epileptogenicity index (HFEI); the specific operation is as follows: Calculate the onset time threshold for each channel i, defined as the maximum baseline NHFE plus ten times the standard deviation of the baseline NHFE: ， When the normalized energy of the target data in each channel exceeds the threshold, this moment is determined to be the start time of abnormal activity, and the channels are sorted according to the start time of each channel. The time coefficient TC is defined as the reciprocal of the order of each channel. The average energy within 250ms before and after the earliest start of each channel is calculated using NHFE as the energy coefficient EC, and finally the HFEI of each channel i is obtained: Step S5: Filter to obtain the 12-45Hz signal; Step S6: Calculate nonlinear regression analysis; Step S7: Calculate the total strength TOT; Step S8: Define and calculate the connectivity high-frequency epileptogenicity index (cHFEI), and locate the epileptogenic zone based on the cHFEI; For each patient's SEEG channel, the average total intensity and HFEI in the 8 seconds before seizure were summarized and normalized. Combined with TOT and HFEI, the combined index cHFEI of the connectivity high-frequency epileptogenicity index was obtained: 。 2. The method for locating the epileptogenic zone based on high-frequency oscillation and connectivity according to claim 1, characterized in that: In step S1, stereotactic electroencephalogram (SEEG) data of twenty patients with different seizure patterns are collected using the stereotactic method. Based on the patients' seizure patterns, 10-30 representative channels are selected to cover the main areas of seizures.

3. The method for locating the epileptogenic zone based on high-frequency oscillation and connectivity according to claim 1, characterized in that: In step S2, the original signal is filtered to the high-frequency band of 54-200Hz using a second-order IIR notch digital filter and a fifth-order IIR Butterworth digital filter; baseline data and target data are manually selected. The principle for selecting baseline data is that the SEEG data does not have obvious abnormal signals before the epileptic seizure, and the target data must cover the entire seizure process.

4. The method for locating the epileptogenic zone based on high-frequency oscillation and connectivity according to claim 1, characterized in that: In step S3, the bandpass signal is converted into a high-frequency energy spectrum by using amplitude square and window smoothing, and the average value of the high-frequency energy of the baseline data is calculated as the baseline value of the high-frequency energy; the high-frequency energy of all channels is divided by the high-frequency energy baseline value of the channel to obtain the normalized high-frequency energy NHFE.

5. The method for locating the epileptogenic zone based on high-frequency oscillation and connectivity according to claim 1, characterized in that: In step S5, the original data is bandpass filtered with a passband of β-γ frequency band, 12-45Hz, and then downsampled to 256Hz.

6. The method for locating the epileptogenic zone based on high-frequency oscillation and connectivity according to claim 1, characterized in that: In step S6, based on h 2 The nonlinear correlation index is used to calculate nonlinear regression analysis, performing piecewise linear regression between each pair of signals to test all offsets of one signal relative to the other signal within the maximum hysteresis; for two signals x and y, the piecewise linear approximation f of the transfer function between x and y is defined as: h 2 The goodness of fit of a regression is measured using the approximate value f. 。 7. The method for locating the epileptogenic zone based on high-frequency oscillation and connectivity according to claim 1, characterized in that: In step S7, all pairs of h are calculated. 2 The values ​​are calculated using a sliding window of length 3s and step size 1s, with a maximum delay of 0.1s, to generate a binary connection matrix; then, for each channel within each time window, the total intensity TOT, independent of direction, is summarized.

Images