Computer program for training neurological disease detection algorithms, method for programming implantable neural stimulation devices, and computer program for the same.

Abstract

The present invention relates to a computer program for training a neurological disease detection algorithm to be used for neurological disease detection in an implantable neurostimulator device having a target electrode configuration, the computer program comprising: a) inputting EEG data into a computer executing the computer program, the EEG data being recorded by at least one EEG from at least one patient using an electrode system having multiple electrode channels; b) identifying neurological activity in the EEG data corresponding to the neurological disease based on a neurological disease identification tag included in the EEG data and / or input into the computer; c) selecting a subset of electrode channels from among the available electrode channels in the EEG data according to c1) the identified neurological activity and / or c2) characteristic data of the target electrode configuration; and d) training the neurological disease detection algorithm by using only the EEG data of the selected subset of electrode channels.

Description

Technical Field

[0001] The present invention relates to a computer program for training a neurological disease detection algorithm for use in detecting epileptic seizures, for example, in an implantable nerve stimulation device having a target electrode arrangement. This computer program is also referred to as a training computer program. The present invention further relates to a method of programming an implantable nerve stimulation device using such a training computer program, and to a computer program in the form of a neurological disease detection algorithm or classifier for detecting neurological diseases from EEG data that has been trained and / or is being trained by such a training computer program. This computer program is also referred to as a neurological disease detection computer program. The present invention further relates to a nerve stimulation device that executes such a neurological disease detection computer program. The present invention further relates to a method of treating neurological diseases using a nerve stimulation device that executes such a neurological disease detection computer program.

Background Art

[0002] Despite progress in pharmaceutical development, the majority of epileptic patients show resistance to treatment with anti-epileptic drugs. Since only a small fraction of these patients are candidates for surgical treatment, innovative treatment methods are needed. An alternative treatment concept for these patients is to apply electrical stimulation at the early stages of epileptic seizures to prevent the spread of epileptic seizures across the brain. This can be achieved using a closed-loop system, where electrical brain activity is recorded by a set of electrodes and continuously or intermittently monitored using an epileptic seizure detector that triggers electrical stimulation to the seizure onset zone (SOZ) by the same or different electrodes.

[0003] Therefore, there is a need for a highly reliable and energy-efficient epileptic seizure detector. The objective of this invention is to provide a solution to this need. [Overview of the Initiative]

[0004] Embodiments of the present invention include a computer program for training a neurological disorder detection algorithm, which is used, for example, for epileptic seizure detection in an implantable neurostimulator having a target electrode configuration, and this computer program is a) A step of inputting EEG data into a computer running a computer program, wherein the EEG data is recorded by at least one invasive and / or non-invasive EEG from at least one patient using a recording electrode system having multiple electrode channels, for example, using a 10-20 or 10-10 or any other high-density EEG electrode system; b) A step of identifying neurological activity in the EEG data that corresponds to a neurological disorder, based on neurological disorder identification tags included in the EEG data and / or entered into the computer, c) Select a subset of electrode channels from the available electrode channels in the EEG data. c1) Identified neurological activity, and / or c2) Characteristic data of the target electrode configuration Steps to select according to, d) The step of training a neurological disease detection algorithm by using only EEG data from a subset of selected electrode channels.

[0005] Neurological disorders identified by neurological activity may include epileptic seizures, stroke, neuropathic pain, dementia, Parkinson's disease, tinnitus, aphasia, or any other designated neurological event. In the following, the example of epileptic seizures will be used primarily to illustrate the present invention. However, this will also cover other types of neurological events.

[0006] The present invention enables the optimal adaptation of a neurological disorder detection algorithm to “actual” electrode configurations, specifically patterns of target electrode configurations. In this way, the neurological disorder detection algorithm can be optimally adapted to the individual neurological disorder onset patterns of a patient and to individual EEG recording sites of implantable electrodes where optimal detection of neurological events is expected. The target electrode configuration may comprise only the recording electrode used in a neurostimulation device to record EEG signals from the patient. It is also possible that the target electrode configuration may comprise only the stimulating electrode used in a neurostimulation device to output stimulating signals to the patient. Another possibility is that the target electrode configuration may comprise a combination of recording and stimulating electrodes. In some cases, one, more, or all of the electrodes in the target electrode configuration may be used for both the purpose of recording EEG signals and outputting stimulating signals, and as a result, they are a combined recording and stimulating electrode.

[0007] Typically, steps of a computer program are executed by a computer. However, the present invention also relates to a method comprising the steps described above and / or below. The target electrode configuration must be defined before the training computer program is executed. For example, characteristic data of the target electrode configuration may be input or programmed into the training computer program. Characteristic data of the target electrode may include geometric data of the location, size, and / or configuration of the electrodes of the target electrode configuration. Generally, the EEG data used for training includes several measurements over time, per recording electrode channel and per patient. The training computer program may require some further manual input depending on the details contained in the EEG dataset input into the computer program. For example, the EEG data may already be tagged or labeled with information that identifies the time and / or location of a neurological disorder, such as epileptic seizure activity. Alternatively, information that identifies the time and / or location of a neurological disorder may be manually entered. In the training step, a neurological disorder detection algorithm performs the detection of neurological activity in the EEG data, and during the course of training, the results of the detection by the neurological disorder detection algorithm are compared with the tagged information. Based on the comparison results, the parameters of the neurological disease detection algorithm are recursively optimized until a neurological disease detection algorithm with a sufficient detection level is achieved.

[0008] According to embodiments of the present invention, such electrode channels are selected from among the available electrode channels that are closest to the location of identified neurological activity corresponding to the neurological disorder. This allows for an optimal spatial correlation between the patient's expected neurological activity and the location of the electrodes in the target electrode configuration.

[0009] According to embodiments of the present invention, such electrode channels are selected from among the available electrode channels having the closest geometric match to the electrodes of the target electrode configuration. In this way, neurological disease detection using the electrodes of the target electrode configuration can be optimized to a level comparable to neurological disease detection using relatively complex 10-20 or 10-10 EEG electrode systems, without requiring such a large number of electrodes within the target electrode configuration.

[0010] According to embodiments of the present invention, the electrodes of the target electrode configuration are arranged in a pseudo-Laplacian pattern. This enables highly localized EEG recording. Through the pseudo-Laplacian configuration, external signal noise and / or artifacts from muscle activity or movement are significantly reduced, improving the detection of neurological disorders.

[0011] According to embodiments of the present invention, such an electrode channel is selected from among the available electrode channels arranged in a pseudo-Laplacian pattern. This allows for a close match of the selected electrode channel with the pattern of the target electrode structure in the pseudo-Laplacian pattern. In the pseudo-Laplacian pattern, the selected electrode channel comprises a central electrode and at least two surrounding electrodes surrounding the central electrode.

[0012] According to embodiments of the present invention, step d) is d1) A step of calculating a linear combination of EEG data of a subset of selected electrode channels, such as calculating a linear combination representing a bipolar or quadrupole electrode channel, d2) The step of training a neurological disease detection algorithm by using a calculated linear combination of EEG data.

[0013] In this approach, additional virtual electrode channels may be generated by such computational steps. This allows for fine-tuning of the spatial signal resolution of the electrode pattern without requiring additional hardware electrode channels.

[0014] According to embodiments of the present invention, exactly five electrode channels are selected from the available electrode channels. This means that the target electrode configuration also has five electrodes, enabling optimal matching of the selected electrode channels.

[0015] According to embodiments of the present invention, the computer program includes at least two training cycles of a neurological disease detection algorithm. e) In the first training cycle, general training of the neurological disorder detection algorithm is performed using EEG data from one or more patients. f) In the second training cycle, patient-specific training of the neurological disorder detection algorithm is performed using only the EEG data of the patient to whom the neurological disorder detection algorithm should be applied, and / or using the EEG data of another patient having a similar neurological disorder onset pattern to the patient to whom the neurological disorder detection algorithm should be applied.

[0016] In this approach, training of neurological disease detection algorithms can be further optimized and accelerated. According to embodiments of the present invention, a computer program is designed to evaluate data tags assigned to EEG data input to a computer running the computer program, and the data tags are used to select a subset of electrode channels from among the available electrode channels. The data tags may contain information about the location of electrode channels in the patient's head during clinical measurement of the EEG signal.

[0017] Another embodiment of the present invention is a method for programming an implantable nerve stimulation device, the method being g) The step of running the above-mentioned type of training computer program on a computer, h) The step of programming a neurological disease detection algorithm, which is trained by a computer program, into an implantable neurostimulator.

[0018] This approach can achieve the same advantages as those previously mentioned. In addition, for training purposes, a different computer may be used than the one used for the implanted neural stimulation device. This allows for the use of a more powerful computer to run the training computer program, thereby saving considerable time.

[0019] According to embodiments of the present invention, the implantable nerve stimulator is a closed-loop nerve stimulator designed to record an EEG signal, calculate a stimulation signal based on the recorded EEG signal, and output a stimulation signal. In this manner, nerve stimulation therapy can be optimized and specifically tailored to the needs of individual patients. The stimulation signal is transmitted via electrodes of a target electrode configuration. The same electrodes of the target electrode configuration may be used to record the EEG signal. It is also possible to use different electrodes for recording the EEG signal and for generating the stimulation signal, for example, electrodes specifically optimized for EEG signal recording and electrodes specifically optimized for signal output.

[0020] According to embodiments of the present invention, the electrodes of the target electrode configuration are arranged in a pseudo-Laplacian configuration. This allows for optimal stimulation success when the electrodes are used for nerve stimulation. Through the pseudo-Laplacian pattern, deep stimulation can be delivered to the patient even if the electrode pattern is planar. For example, the electrode may comprise a central electrode surrounded by at least two stimulating electrodes. It is also possible to have a larger number of stimulating electrodes surrounding the central electrode, for example, four stimulating electrodes.

[0021] According to embodiments of the present invention, in step g), the computer program is executed on an external computer that is not part of the implantable nerve stimulation device. Another embodiment of the present invention is in the form of a neurological disease detection algorithm or classifier for detecting neurological diseases from EEG data trained by and / or being trained by a training computer program of the above-described type. The computer program can be configured for implementation on a microcontroller, or on a digital signal processor (DSP), or on a field programmable gate array (FPGA), or on an application specific integrated circuit (ASIC). According to an embodiment of the present invention, the computer program is optimized for minimum power consumption. According to another embodiment of the present invention, the computer program is optimized for maximum performance. According to another embodiment of the present invention, the computer program is optimized for minimum power consumption under defined performance constraints. According to another embodiment of the present invention, the computer program is optimized for maximum performance under defined power constraints.

[0022] According to an embodiment of the present invention, the neurological disease detection algorithm or classifier for detecting neurological diseases is an artificial intelligence algorithm, for example, random forest, support vector machine, multi-layer perceptron, convolutional neural network, for example, long short-term memory network, or attention-based network. This enables easy and optimal adaptation of the neurological disease detection algorithm to typical neurological disease onset patterns occurring in actual patients.

[0023] The computer can be located within a cloud (server type computer), or any commercially available computer such as a PC, laptop, notebook, tablet, or smartphone, or a microprocessor, microcontroller, DSP or FPGA, or a combination of such elements. The computer program can be stored on a non-transitory computer-readable medium.

[0024] As far as closed-loop control is concerned, closed-loop control is different from open-loop control in that closed-loop control has feedback, or feedback of measured values or internal values, and the generated output value of the closed-loop control is affected in the sense of the closed-loop control circuit. In a closed-loop control system, only variables are controlled with or without such feedback.

[0025] Further exemplary embodiments of the present invention are described using the following figures.

Brief Description of the Drawings

[0026] [Figure 1] A diagram showing a patient having an implantable nerve stimulation device. [Figure 2] A diagram showing details of an implantable nerve stimulation device. [Figure 3] A diagram showing an example of the result of automatic electrode selection using the first method. [Figure 4] A diagram showing an example of the result of automatic electrode selection using the second method.

Modes for Carrying Out the Invention

[0027] Figure 1 shows a nerve stimulation system implanted in a patient. The nerve stimulation system includes an implantable nerve stimulation device 1 connected to a target electrode assembly 2 by a lead 12. The target electrode assembly 2 can be placed outside the patient's skull, under the patient's scalp.

[0028] Figure 1 further illustrates an external computer 13 and a patient controller 11 as another external device. Both the computer 13 and the patient controller 11 can communicate wirelessly with the implantable neurostimulator 1. For example, the computer 13 may be used as a clinical system to be used by a physician to program the implantable neurostimulator 1. The patient controller 11 may be used by the patient to check the status of the implantable neurostimulator 1 or to activate a specific stimulation mode. The patient controller 11 may also function as a recorder to log epileptic seizure events or other events entered by the patient or transmitted from the implantable neurostimulator 1. The patient controller 11 may also function as an alarm system to provide the patient with tactile and / or luminous and / or auditory feedback in the event of a neurological event. The computer 13 may also be used to run the training computer program of the present invention for training a neurological disorder detection algorithm. The computer 13 may also be used to program the trained neurological disorder detection algorithm into the implantable neurostimulator 1.

[0029] Figure 2 shows further details of the implantable neurostimulator 1 and the target stimulation and / or recording electrode assembly 2 in a block diagram representation similar to an electrical circuit diagram. The neurostimulator 1 comprises a control processor 6, a signal generation circuit 3, a charge balancing circuit 4, a protection circuit 5, sensors 7 and 8, a battery pack 9, and a user input element 10. The neurostimulator 1 is connected to the electrode assembly 2 via a cable 12. As can be seen, the electrode assembly 2 comprises a central electrode 20 and four stimulation electrodes 21, 22, 23, and 24 positioned around the central electrode 20. The central electrode 20 may be a common ground electrode, meaning that the central electrode 20 is connected to the common ground of each neurostimulator 1.

[0030] The control processor 6 may be a microcontroller device (MCU) or any other device that can perform control steps through the processing of a computer program, for example, in the form of hardware, firmware, or a software program.

[0031] The signal generation circuit 3 can generate stimulation pulses and deliver them to the stimulation electrodes 21, 22, 23, and 24 in response to commands from the control processor 6. The signal generation circuit 3 may include amplifier components.

[0032] The control processor 6 may detect nerve signals and / or brain activity through sensors 7 and 8. The detected nerve signals and / or brain activity may be processed and used for event-driven delivery of stimulation pulses to any of the stimulation electrodes 21, 22, 23, and 24.

[0033] The battery pack 9 supplies electrical energy to the aforementioned elements of the power supply unit 1. The battery pack 9 may include a rechargeable battery. The control processor 6 is designed to execute the trained neurological disease detection algorithm. The training of the neurological disease detection algorithm is performed on the computer 13, as described below.

[0034] Neurological disease detection algorithms for such systems must be trained on recordings configured similarly to those of subgalea aponeurotic electrodes. Because of the high similarity between EEG data from subcutaneous and proximal scalp electrodes within patients with neocortical epilepsy, the inventors use surface EEG recordings obtained from electrode configurations representing implantable subgalea aponeurotic electrodes for training and evaluating neurological disease detection algorithms. This approach may be used to pre-train personalized detectors using scalp-based neurological disease recordings prior to device implantation.

[0035] Electrode selection can be performed to represent target electrode configuration 2 during long-term video EEG monitoring of neurological disorders. A skilled epilepsy specialist can define the seizure-initiating zone (SOZ) by visually exploring the seizure-initiating electrode and examining the inter-electrode distance. Our proposed method automates electrode selection by examining the geometric dimensions of the system shown in Figure 1, ensuring optimal electrode selection that simulates the placement of subgalea aponeurotica recording electrodes. Next, a neurological disorder detection algorithm is needed that can achieve good results with a reduced number of electrodes with limited spatial coverage and has low power consumption. This would allow the neurological disorder detector to be integrated into a fully implantable intervention device. Several publications exist on the development of neurological disorder detection algorithms for either online or offline applications. However, the number of studies examining the limitations of closed-loop applications is limited.

[0036] Following the development and implementation of two automated electrode selection methods in the inventors' system, the inventors designed and implemented four high-energy-efficiency neurological disorder detection algorithms using a random forest (RF) classifier, support vector machine (SVM), multilayer perceptron (MLP), and convolutional neural network (CNN) that can be reliably performed with a reduced electrode set for the detection of focal epileptic seizures. Finally, the inventors compared their detection performance to evaluate their suitability for implantable devices.

[0037] The EEG dataset used for training contained surface EEG recordings from 50 patients with a total of 358 epileptic seizures. Patients with a SOZ covered by electrodes positioned according to a 10-10 system in addition to the conventional 10-20 electrode layout were selected. EEG data were recorded at a 250 Hz sampling rate on a 256-channel DC amplifier with 24-bit resolution. Electrodes were rereferenced to the centrally located electrode. A low-pass filter with a 100 Hz cutoff frequency was applied for anti-aliasing. EEG data from 10 patients were used to train a hybrid model of the CNN. Therefore, to maintain dataset consistency across all classifiers, the remaining 40 patients with a total of 286 epileptic seizures were used to evaluate the classifiers.

[0038] To improve data quality and remove data contaminated with artifacts, the data was filtered using a Chebyshev II bandpass (order: 10, bandstop = 40 dB) with low and high cutoff frequencies of 0.1 Hz and 48 Hz. Then, very high amplitudes representing artifacts were removed from the analysis.

[0039] Electrode selection was performed to obtain an electrode set with the maximum number of electrodes covering the seizure-causing region, with inter-electrode distances below a threshold that maps to the design of the implantable system shown in Figure 1. For this purpose, for each epileptic seizure, a list of electrodes—considered to cover the SOZ (or plural)—was determined by a skilled epilepsy specialist. Subsequently, based on this list, the minimum electrode set containing all seizure-causing electrodes was determined. If the size of this electrode set was smaller than the number of implantable electrodes (n=5), electrodes from the remaining seizure-causing electrodes most frequently involved in seizure onset were added to this electrode set. In this process, a list of the five electrodes most likely to capture all epileptic seizures was established for each patient. Method 1 In this method, the average of the coordinates of the five selected electrodes was first calculated to find the scalp electrode closest to this position. This electrode was considered the central electrode. Any of the five selected electrodes that were at a distance greater than the threshold distance from the central electrode were removed from the list. As depicted in Figure 3, this can be visualized by drawing a sphere of threshold radius with its center at the selected central electrode and checking whether the other selected electrodes are inside this sphere. Here, the electrodes remaining in the list included the central electrode and nearby epileptic seizure-inducing electrodes. The remaining electrodes with the smallest distance from the central electrode replaced the removed electrode from the list of five electrodes. Method 2 In this method, at each step, one electrode was selected as the central electrode, and, as in Method 1, the number of electrodes from the initial SOZ electrodes whose distance from the central electrode was less than the threshold distance was counted. This process was repeated for all EEG electrode locations across the scalp to generate a list of selected initial SOZ electrodes surrounded by each electrode. The electrode containing the largest number of initial SOZ electrodes was selected as the optimal central electrode. Where necessary, electrodes with the minimum distance from the central electrode were added to the list to obtain exactly five electrodes for epileptic seizure detection.

[0040] Several features in the time and frequency domains were selected for epileptic seizure detection. Time-domain features included mean, maximum, mean absolute deviation, variance, skewness, kurtosis, line length, autocorrelation, and entropy. Frequency-domain features included mean, maximum, and variance of the power spectrum, power in the theta band (4–8 Hz), beta band (13–30 Hz), and gamma band (30–45 Hz), as well as the epileptogenicity index. These features were used for classification by random forest, SVM, and MLP classifiers. For SVM and MLP, the features were scaled for classification because the range of the calculated features affects their weights and subsequent decision boundaries. 1) Random Forest The number of binary decision trees was set to 100. The inventors selected entropy for impurity measurement as the branching exponent. Four features were randomly selected at each node. To keep the tree size limited, the maximum depth of the tree was limited to 10. Bootstrap samples were used while constructing the decision trees. Sample weights for each class were adjusted inversely proportional to the class frequency in the input data. The “leave-one-out” method was used for cross-validation. 2) Support Vector Machines (SVMs) The inventors selected a Gaussian radial basis function (RBF) as the kernel function to handle the nonlinearity between features and class labels. For this purpose, two hyperparameters needed to be set: the kernel coefficient of the Gaussian function was set to 0.01, and the penalty parameter of the error term, which acts as a regularization parameter for the SVM, was set to 0.1. The sample weights for each class were adjusted so that they were inversely proportional to the class frequency in the input data. A “leave-one-out” method was used for cross-validation. 3) Multilayer perceptron (MLP) An MLP network consists of at least three layers of nodes: an input layer, one or more hidden layers, and an output layer. The inventors implemented an MLP classifier consisting of one hidden layer with 20 neurons. The inventors selected "Adam" as the solver for weight optimization. The logistic sigmoid function was selected as the activation function, and the adaptive learning rate was selected to schedule weight updates. The L2 penalty (regularization term) parameter was set to 10 -4 The settings were configured as follows. The “Leave-one-out” method was used for cross-validation. 4) Convolutional Neural Networks (CNNs) A CNN consists of an input layer, multiple hidden layers, and an output layer. The hidden layers consist of convolutional layers, pooling layers, and fully connected layers. The CNN architecture proposed by the inventors is shown in Table I. In the first layer, to efficiently learn spatial-temporal patterns, the inventors used a kernel size that extended to all channels and the detection time window (2 seconds = 500 data points). In all hidden layers, the inventors used batch normalization after convolution and used Rectified Linear Units (ReLU) as the activation function. Dropout regularization was applied during training to reduce overfitting. In the two final layers, the inventors used two fully connected layers.

[0041] The following table shows the preferred architecture of the CNN proposed by the inventors.

[0042] [Table 1]

[0043] During training, due to the limited available data for each patient, the inventors employed a transfer learning method consisting of freezing the bottom layers in the model and training only the top layers. Thus, the inventors pre-trained the network on data from 10 patients, and then, for each of the remaining 40 patients, fine-tuned the final convolutional layer and two fully connected layers using patient-specific data. Because the classes were imbalanced, the class index was weighted during training to maintain balance in the weighting of the loss function. Each model was trained with a batch size of 512 for 500 epochs. For weight optimization, the inventors 10 -3 The Adam solver was used with a training rate of . The inventors used binary cross-entropy as the loss function. For evaluation, the inventors used 3-fold cross-validation.

Claims

1. A computer program for training a neurological disease detection algorithm to be used for neurological disease detection in an implantable nerve stimulation device having a target electrode configuration, wherein the computer program is a) A step of inputting EEG data into a computer that executes the computer program, wherein the EEG data is recorded by at least one EEG from at least one patient using an electrode system having multiple electrode channels, b) The step of identifying neurological activity in the EEG data that corresponds to a neurological disorder, based on neurological disorder identification tags included in the EEG data and / or entered into the computer; c) A subset of electrode channels from among the available electrode channels in the EEG data c1) The identified neurological activity, and / or c2) A step of selecting according to characteristic data of the target electrode structure, d) A step of training a neurological disease detection algorithm by using only the EEG data of the selected subset of electrode channels. Includes, A computer program selects the electrode channel from among the available electrode channels that are closest to the location of the identified neurological activity corresponding to the neurological disorder.

2. A computer program for training a neurological disease detection algorithm to be used for neurological disease detection in an implantable nerve stimulation device having a target electrode configuration, wherein the computer program comprises: a) A step of inputting EEG data into a computer that executes the computer program, wherein the EEG data is recorded by at least one EEG from at least one patient using an electrode system having multiple electrode channels, b) The step of identifying neurological activity in the EEG data that corresponds to a neurological disorder, based on neurological disorder identification tags included in the EEG data and / or entered into the computer; c) A subset of electrode channels from among the available electrode channels in the EEG data c1) The identified neurological activity, and / or c2) A step of selecting according to characteristic data of the target electrode structure, d) A step of training a neurological disease detection algorithm by using only the EEG data of the selected subset of electrode channels. Includes, A computer program selects the electrode channel from among the available electrode channels having the closest geometric match to the electrode of the target electrode configuration.

3. A computer program for training a neurological disease detection algorithm to be used for neurological disease detection in an implantable nerve stimulation device having a target electrode configuration, wherein the computer program comprises: a) A step of inputting EEG data into a computer that executes the computer program, wherein the EEG data is recorded by at least one EEG from at least one patient using an electrode system having multiple electrode channels, b) The step of identifying neurological activity in the EEG data that corresponds to a neurological disorder, based on neurological disorder identification tags included in the EEG data and / or entered into the computer; c) A subset of electrode channels from among the available electrode channels in the EEG data c1) The identified neurological activity, and / or c2) A step of selecting according to characteristic data of the target electrode structure, d) A step of training a neurological disease detection algorithm by using only the EEG data of the selected subset of electrode channels. Includes, A computer program in which the electrode channels are selected from the available electrode channels arranged in a pseudo-Laplacian pattern.

4. A computer program for training a neurological disease detection algorithm to be used for neurological disease detection in an implantable nerve stimulation device having a target electrode configuration, wherein the computer program comprises: a) A step of inputting EEG data into a computer that executes the computer program, wherein the EEG data is recorded by at least one EEG from at least one patient using an electrode system having multiple electrode channels, b) The step of identifying neurological activity in the EEG data that corresponds to a neurological disorder, based on neurological disorder identification tags included in the EEG data and / or entered into the computer; c) A subset of electrode channels from among the available electrode channels in the EEG data c1) The identified neurological activity, and / or c2) A step of selecting according to characteristic data of the target electrode structure, d) A step of training a neurological disease detection algorithm by using only the EEG data of the selected subset of electrode channels. Includes, Step d) is d1) A step of calculating a linear combination of the EEG data of the selected subset of electrode channels, such as calculating a linear combination representing a bipolar or quadrupole electrode channel, d2) A computer program comprising the step of training the neurological disease detection algorithm by using a calculated linear combination of the EEG data.

5. A computer program for training a neurological disease detection algorithm to be used for neurological disease detection in an implantable nerve stimulation device having a target electrode configuration, wherein the computer program comprises: a) A step of inputting EEG data into a computer that executes the computer program, wherein the EEG data is recorded by at least one EEG from at least one patient using an electrode system having multiple electrode channels, b) The step of identifying neurological activity in the EEG data that corresponds to a neurological disorder, based on neurological disorder identification tags included in the EEG data and / or entered into the computer; c) A subset of electrode channels from among the available electrode channels in the EEG data c1) The identified neurological activity, and / or c2) A step of selecting according to characteristic data of the target electrode structure, d) A step of training a neurological disease detection algorithm by using only the EEG data of the selected subset of electrode channels. Includes, A computer program in which exactly five electrode channels are selected from the available electrode channels.

6. A computer program for training a neurological disease detection algorithm to be used for neurological disease detection in an implantable nerve stimulation device having a target electrode configuration, wherein the computer program comprises: a) A step of inputting EEG data into a computer that executes the computer program, wherein the EEG data is recorded by at least one EEG from at least one patient using an electrode system having multiple electrode channels, b) The step of identifying neurological activity in the EEG data that corresponds to a neurological disorder, based on neurological disorder identification tags included in the EEG data and / or entered into the computer; c) A subset of electrode channels from among the available electrode channels in the EEG data c1) The identified neurological activity, and / or c2) A step of selecting according to characteristic data of the target electrode structure, d) A step of training a neurological disease detection algorithm by using only the EEG data of the selected subset of electrode channels. Includes, The computer program includes at least two training cycles of the neurological disease detection algorithm. e) In the first training cycle, general training of the neurological disease detection algorithm is performed using EEG data from one or more patients. f) A computer program wherein, in a second training cycle, patient-specific training of the neurological disorder detection algorithm is performed using only the EEG data of the patient to whom the neurological disorder detection algorithm should be applied, and / or using the EEG data of another patient having a similar neurological disorder onset pattern to the patient to whom the neurological disorder detection algorithm should be applied.

7. A computer program for training a neurological disease detection algorithm to be used for neurological disease detection in an implantable nerve stimulation device having a target electrode configuration, wherein the computer program comprises: a) A step of inputting EEG data into a computer that executes the computer program, wherein the EEG data is recorded by at least one EEG from at least one patient using an electrode system having multiple electrode channels, b) The step of identifying neurological activity in the EEG data that corresponds to a neurological disorder, based on neurological disorder identification tags included in the EEG data and / or entered into the computer; c) A subset of electrode channels from among the available electrode channels in the EEG data c1) The identified neurological activity, and / or c2) A step of selecting according to characteristic data of the target electrode structure, d) A step of training a neurological disease detection algorithm by using only the EEG data of the selected subset of electrode channels. Includes, The computer program is designed to evaluate data tags to be assigned to the EEG data input to a computer running the computer program, and the data tags are used to select a subset of the electrode channels from the available electrode channels.

8. The computer program according to any one of claims 1 to 7, wherein the neurological disease detection algorithm is an artificial intelligence algorithm.

9. A method for programming an implantable neural stimulation device, g) The step of running the computer program described in any one of claims 1 to 8 on a computer, h) A method comprising the step of programming a neurological disease detection algorithm, which is trained by the computer program, into the implantable neurostimulator.

10. The method according to claim 9, wherein the implantable nerve stimulator is a closed-loop nerve stimulator designed to record an EEG signal, calculate a stimulation signal based on the recorded EEG signal, and output the stimulation signal.

11. The method according to claim 9 or 10, wherein in step g), the computer program is executed on an external computer that is not part of the implantable nerve stimulation device.

12. A computer program in the form of a neurological disorder detection algorithm or classifier for detecting neurological disorders from EEG data trained and / or being trained by a computer program according to any one of claims 1 to 8.

13. The computer program according to claim 12, wherein the computer program is configured for implementation on a microcontroller.

14. The computer program according to claim 12 or 13, wherein the computer program is optimized for minimum power consumption.

15. The computer program according to any one of claims 12 to 14, wherein the neurological disease detection algorithm or classifier for detecting the neurological disease is an artificial intelligence algorithm.

16. An implantable nerve stimulator that executes a computer program according to any one of claims 12 to 15.

Images