CN107029351B

CN107029351B - System and method for extracting global LFP parkinsonism characteristic value

Info

Publication number: CN107029351B
Application number: CN201710245882.1A
Authority: CN
Inventors: 孙齐峰; 赵德春; 王力; 李文瀚; 方程; 赵兴; 田银
Original assignee: Chongqing University of Post and Telecommunications
Current assignee: Chongqing University of Post and Telecommunications
Priority date: 2017-04-14
Filing date: 2017-04-14
Publication date: 2021-01-15
Anticipated expiration: 2037-04-14
Also published as: CN107029351A

Abstract

The application discloses a system and a method for extracting a global LFP Parkinson's disease characteristic value, which comprises a signal receiving and transmitting module for simultaneously generating a stimulation pulse and measuring an LFP signal; a characteristic extraction module for performing energy analysis on the LFP signal measured by the signal transceiver module and extracting characteristic values of each frequency band; the comprehensive processing module is used for calculating the correlation coefficient of each frequency band and the energy baseline spectrum according to the difference value of each frequency band characteristic value and each corresponding point on the energy baseline spectrum; and the output module is used for adding the products of the characteristic value of each frequency band and the correlation coefficient to obtain a global LFP Parkinson's disease characteristic value. The LFP signal is preprocessed into three frequency bands of low frequency, intermediate frequency and high frequency; and converting the preprocessed signals from a time domain to a frequency domain for analysis, establishing an autoregressive model and a neural network model by taking the characteristic value of each frequency band as a parameter and the correlation coefficient of each frequency band as a weight, and obtaining an objective and accurate global characteristic value independent of doctor experience.

Description

System and method for extracting global LFP parkinsonism characteristic value

Technical Field

The invention relates to the field of brain wave electric digital data processing, in particular to a system and a method for extracting a global LFP (linear frequency shift keying) Parkinson disease characteristic value for a brain pacemaker.

Background

Parkinson's Disease (PD) is a common degenerative disease of the nervous system, and brain pacemaker surgery is one of the major means of treating Parkinson's disease. A brain pacemaker is a small and exquisite set of microelectronic devices that include a pulse generator, a set of electrodes, and extension leads connected between the pulse generator and the electrodes.

The brain pacemaker operation is commonly known as Deep Brain Stimulation (DBS), and weak electric pulses are distributed through electrodes implanted in the brain to stimulate related nerve nuclei in the brain for controlling movement and inhibit abnormal brain nerve signals causing Parkinson's disease symptoms, so that the Parkinson's disease symptoms are eliminated, and the original activity and self-care ability of a patient can be restored.

However, the following drawbacks exist in using a brain pacemaker at present:

1. the stimulation parameters are obtained mainly by depending on the observation of the patient and the description of the patient, and the stimulation frequency and the stimulation intensity are obtained through experience, so that the time difference is large when different people operate the cerebral pacemaker; the subjective dependence of stimulation parameter adjustment is strong, and too much physician experience is relied on for accumulation, which is not beneficial to the comprehensive popularization of DBS technology.

2. The adjustment of the parameters lacks time sensitivity, and cannot be timely adjusted according to the change of the physical condition of the patient, the current average adjustment period is 3-12 months, and the targeted treatment cannot be carried out according to the change of the illness state of the patient;

to overcome the drawbacks of the current cerebral pacemaker, Local Field Potential (LFP) is the most ideal stimulation reference signal for closed-loop DBS. Local field potential is a special type of electrophysiological signal, in a living body, dendritic synaptic activity in a certain volume of biological tissue induces a current, when the current flows through an extracellular space with certain impedance, a certain electric field distribution is formed, and a local voltage value recorded at a certain point is called local field potential. The overall situation of the electrical activity of all the neurons near the LFP nuclear mass is measured, and different activity states of the brain can be reflected by the obtained LFP change situation.

However, the use of LFP as the stimulation reference signal for closed-loop DBS is only an assumption, and there is no complete system for actually realizing the automatic change of weak electric pulses delivered by the cerebral pacemaker by using LFP as the stimulation reference signal. The reason is that there is no way to extract an objective parkinson feature value from the LFP signal, which does not depend on the experience of the doctor, can accurately reflect the activity of the brain, and is predictable. By this characteristic value, the pulse generator can be made to output an appropriate pulse stimulation for parkinson's disease. The objective and independent of the doctor's experience as used herein means not only the objective requirement on the method and route of generation of the eigenvalues, but also the objective requirement on the extraction and processing of the original LFP signal. Even if the same LFP signal is obtained, there is a large technical bias in processing the LFP signal. The frequency bands with small surface information are basically discarded when the LFP signal is processed, but there is no objective judgment criterion for which frequency bands have small information content. Moreover, it is not known how much the effective information contained in the LFP signal frequency band discarded to be unused affects the entire characteristic value. Both from the LFP signal extraction and processing, an objective system and method that does not rely on physician experience is needed to enable the final parkinson signature to accurately reflect the patient's treatment.

Therefore, there is an urgent need to provide a system and method for extracting parkinson's characteristic values that can provide accurate, appropriate and predictable pulse stimulation frequency and intensity for a patient.

Disclosure of Invention

The invention aims to provide a system for extracting a global LFP Parkinson's disease characteristic value, so as to solve the problem that the existing pacemaker cannot provide accurate, proper and predictable pulse stimulation frequency and intensity for a patient because the pacemaker is too dependent on the experience of a person when in use because no objective Parkinson's disease characteristic value which is not dependent on the experience of a doctor exists.

In order to solve the above problems, the following scheme is provided:

the first scheme is as follows: a system for global LFP parkinson's disease feature value extraction, comprising:

the signal transceiver module is used for simultaneously generating stimulation pulses and measuring LFP signals;

the characteristic extraction module is connected with the signal transceiver module and used for carrying out energy analysis on the LFP signals measured by the signal transceiver module, carrying out frequency spectrum segmentation on the LFP signals according to the analysis result and extracting characteristic values of all frequency bands;

the comprehensive processing module is connected with the characteristic extraction module and is used for subtracting the characteristic value of each frequency band extracted by the characteristic extraction module from each corresponding point value on the energy baseline spectrum stored in the comprehensive processing module in advance; calculating a correlation coefficient between each frequency band and the energy baseline spectrum according to the difference value between the characteristic value of each frequency band and each corresponding point on the energy baseline spectrum;

and the output module is connected with the comprehensive processing module and the brain pacemaker, the correlation coefficient of each frequency band obtained by the comprehensive processing module is used as a weight value, and the products of the characteristic value and the correlation coefficient of each frequency band are added to obtain a global LFP Parkinson characteristic value.

The system principle is as follows:

when the system runs, the signal transceiver module generates stimulation pulses and simultaneously measures the LFP signals, and an acquisition end for measuring the LFP signals is not required to be additionally arranged, so that the number of devices is reduced, and the stimulation signals can be quickly sent and the feedback signals can be quickly received. After a period of time during which the transceiver module generates a stimulation pulse to the brain, the transceiver module may acquire a varying LFP signal. The signal transceiver module transmits the LFP signal to the feature extraction module. The feature extraction module processes the LFP signal and then performs energy analysis, divides the LFP signal into a plurality of frequency bands with different energy contents according to the result of the energy analysis, and extracts feature values of the frequency bands. And when receiving a plurality of discrete characteristic values of each frequency band extracted by the characteristic extraction module, the comprehensive processing module makes difference values between the characteristic values and corresponding point values on the corresponding energy baseline spectrum respectively, and calculates the correlation coefficient between the frequency spectrum of the frequency band and the energy baseline spectrum according to the obtained difference values. And taking the correlation coefficient as a weight value, respectively multiplying the weight value by the characteristic value of each frequency band, and adding the products of the frequency bands to obtain the global LFP Parkinson characteristic value which can be input into the cerebral pacemaker.

Has the advantages that:

the signal transceiver module has the functions of generating stimulation pulses and measuring LFP signals, so that the additional module arrangement can be reduced, the operation steps can be reduced, and the corresponding LFP signal response can be obtained immediately while the stimulation signals are sent to the brain.

The invention comprehensively considers the characteristic value of each frequency band on the LFP signal, can comprehensively reflect all information on the LFP signal, not only increases the diversity of the information and the information capacity, but also can make the characteristic of the whole LFP signal more obvious; the finally obtained Parkinson characteristic value can objectively represent the information content to be expressed by the whole LFP signal, so that the extracted Parkinson characteristic value can be closer to the actual treatment condition of a patient. Compared with the method that other people only select a certain frequency band in the LFP signals in the past, all characteristic values on the full frequency band are adopted to reflect the relation between the LFP signals and the Parkinson's disease better, and a more accurate, proper and objective global LFP Parkinson characteristic value can be obtained, so that the stimulation pulse output by the cerebral pacemaker after referring to the global LFP Parkinson characteristic value can be matched with the symptoms of the current patient.

The invention effectively solves the problem that the existing cerebral pacemaker is too dependent on the experience of doctors and can not provide accurate, proper and predictable pulse stimulation frequency and strength for patients.

Scheme II: on the basis of the first scheme, the signal receiving and transmitting module is a microelectrode at least comprising two groups of contacts, and a filter circuit is connected between the two groups of contacts.

The filter circuit can eliminate the interference influence of the stimulation pulse on the LFP signal during acquisition, so that the acquired LFP signal has no noise interference as much as possible, and a feature extraction module is convenient to process the LFP signal.

The third scheme is as follows: on the basis of the first scheme, the feature extraction module comprises a microcontroller and is respectively connected with the microcontroller:

the frequency spectrum conversion module is used for converting the frequency spectrum of the LFP signal into an energy spectrum through a prestored formula;

the frequency band segmentation module is connected with the frequency spectrum transformation module, carries out cluster analysis on the received energy frequency spectrum according to the energy value on the ordinate of the energy frequency spectrum, and divides the energy frequency spectrum into at least three sub-frequency bands according to the analysis result;

the frequency band division characteristic value extraction module is connected with the frequency band division module and the comprehensive processing module, and takes an energy value extracted by average distribution of each frequency band on a frequency domain as a characteristic value of the frequency band; the number of the characteristic values of each frequency band is equal.

The LFP signal time domain frequency spectrum obtained directly through the signal receiving and sending module is gradually converted into an energy frequency spectrum through the frequency spectrum conversion module, so that the characteristic value extraction is convenient to carry out later. When the energy frequency spectrum is segmented, clustering analysis is carried out according to the difference of the energy value of each frequency band, a section of continuous frequency spectrum with similar energy values is divided into a segment, and the whole energy frequency spectrum is decomposed into at least three sub-segments. Since the energy values of each sub-band are almost the same, it can be assumed that the brain activity in this sub-band is similar, and the features exhibited from this sub-band are similar, so that the entire LFP energy spectrum band can be divided into sub-bands that are more conducive to analysis and feature extraction.

And the scheme is as follows: on the basis of the third scheme, the comprehensive processing module comprises:

the difference module is connected with the sub-band characteristic value extraction module; the characteristic values on all sub-bands received from the sub-band characteristic value extraction module are subtracted from the pre-stored energy baseline spectrum to obtain difference values, and absolute values of the difference values are output;

and the correlation coefficient module is respectively connected with the difference module and the output module, calculates the correlation coefficient of the difference received from the difference module through a pre-stored cross-correlation function and outputs the correlation coefficient to the output module.

And calculating a correlation coefficient in each sub-frequency band through a cross-correlation function, and taking the correlation coefficient as a weight value in each sub-frequency band to participate in the extraction calculation of the global LFP Parkinson characteristic value in the output module.

Another object of the present invention is to provide a method for global LFP parkinson's disease feature value extraction, comprising the steps of:

step one, signal setting: LFP signals under different states are measured through a signal receiving and sending module, and the LFP signals are respectively defined as a first LFP signal under the normal state of a Parkinson's disease patient, a second LFP signal under the state of Parkinson's disease and without any DBS stimulation and drug therapy, a third LFP signal under the state of Parkinson's disease and only subjected to DBS stimulation therapy, a fourth LFP signal under the state of Parkinson's disease and only subjected to drug therapy, and a fifth LFP signal under the state of Parkinson's disease and simultaneously subjected to DBS stimulation and drug therapy;

step two, LFP signal preprocessing: sequentially filtering, denoising and down-sampling the acquired LFP signals to respectively obtain low-frequency LFP signals of 1-7Hz, intermediate-frequency LFP signals of 8-35Hz and high-frequency LFP signals above 150 Hz;

step three, spectrum analysis: the preprocessed signal is converted from the time domain to the frequency domain by a spectral transformation module and a frequency band segmentation module for analysis: analyzing by combining the frequency spectrum, the power spectrum and the spectrum to respectively obtain frequency domain characteristics of the low-frequency LFP signal, the medium-frequency LFP signal and the high-frequency LFP signal;

step four, extracting and standardizing the characteristics of each frequency band: quantizing the frequency domain characteristics on each frequency band obtained in the third step by a sub-band characteristic value extraction module to obtain discrete values within 0-100;

step five, global modeling and feature standardization: the comprehensive processing module takes the characteristic values of the frequency bands obtained in the front as parameters, the correlation between the frequency bands represented by the comprehensive processing module and the Parkinson's disease symptoms of the frequency bands is taken as weight, modeling is carried out, and the output module outputs the global characteristic values.

The method objectively collects the whole LFP signal and objectively and comprehensively extracts the Parkinson global characteristic value reflecting the LFP signal through the steps of signal setting, LFP signal preprocessing, spectrum analysis, characteristic extraction and standardization of each frequency band, global modeling and characteristic standardization. The invention does not abandon the information of a certain frequency band with small surface information amount, but obtains a global characteristic value which can be used for directly judging whether the generated pacemaker electrical stimulation parameters are suitable for accuracy and appropriateness by analyzing and modeling all LFP signals on a low frequency band, a middle frequency band and a high frequency band in the information frequency band supplied by people through the technology in the field. The stability of closed-loop DBS is improved. When local characteristics are disordered, the normal electrical stimulation parameters obtained by the DBS are not influenced, and the normal DBS operation is not influenced; when the signal characteristics of a certain frequency band are disordered, the system can provide a reasonable and effective global characteristic value according to the characteristics of other frequency bands.

The invention not only considers the characteristic value of a certain frequency band, but also comprehensively and comprehensively considers the LFP signal change of the whole frequency band, thereby obtaining the accurate and proper global characteristic value which is positively correlated with the Parkinson's disease treatment response, being beneficial to reducing the excessive dependence on the personal experience of doctors during the DBS operation, and objectively adopting the selection and the adjustment of the electrical stimulation parameters of the cerebral pacemaker through the global characteristic value. Meanwhile, the invention analyzes the characteristics of the local field potential under the multi-environment to finally obtain a global characteristic value, thereby improving the application range of the global characteristic value.

Further, in the third step, the preprocessed signals are converted by adopting an FFT algorithm to obtain frequency spectrums and power spectrums of the signals, and energy on each frequency band is quantitatively evaluated; or, in the third step, the signals of each frequency band are respectively subjected to frequency domain conversion by using Hilbert-Huang transform, and the characteristics of each frequency band are observed according to the power spectrum and the spectrum; or, in the third step, first, half of the data of the acquired data volume is taken, the characteristics of the whole power spectrum under each environment are integrally analyzed, classification and sorting are performed according to the analysis result, the frequency bands under each environment are solidified according to the sorting result, and then the data are processed by using a Hilbert-Huang transform method.

The FFT algorithm is the most basic algorithm, and can obtain global frequency domain characteristics more efficiently.

Hilbert-Huang transform, which mainly acquires time-frequency characteristics of signals and monitors instantaneous change characteristics of the signals.

Further, in the fourth step, the energy peak values of all the collected signals are taken as standard values to be subjected to normalization processing; mapping each energy value in a fluctuation range of 0-100, and taking the average value of all energy of the energy values of the sub-frequency bands as a characteristic value; or, in the fourth step, the statistical processing is performed on each frequency band according to the respective condition, the energy of the normal LFP signal on each frequency band is combined to calculate the baseline on each frequency band, the energy value on each frequency band is subtracted from the baseline, and the sum of the difference values of each frequency band is output as the characteristic value; the characteristic values are quantized between 0 and 100 with the signal peak as a criterion.

Further, in the fifth step, a linear regression model is adopted, and the correlation between the frequency band represented by the linear regression model and the Parkinson's disease is calculated and used as the coefficient of the linear model, and the coefficient is quantized to the fluctuation range of 0-1; taking the parameters of each frequency band as input to be multiplied by the coefficient, and taking the finally obtained result as a global characteristic value to be output; or in the fifth step, an artificial neural network model is adopted, the acquired offline data are used as training samples, the number of neural network nodes is designed according to the number of frequency division, and the correlation coefficient is used as the weight of the nodes to learn the network; and after learning is finished, the characteristic values of all frequency bands are used as input and input to corresponding nodes, and the model output is used as the global characteristic value to be output.

Further, in the fifth step, the characteristic parameters of the low frequency band are superposed with one DBS stimulation disturbance parameter, the characteristic parameters of the medium frequency band are superposed with one drug influence disturbance parameter, and the characteristic parameters of the high frequency band are superposed with the DBS stimulation and the drug influence disturbance parameter; the DBS stimulation disturbance parameter, the drug influence disturbance parameter and the DBS stimulation and drug influence disturbance parameter are all in the range of 0-1.

The signal peak herein refers to the peak of the highest peak.

The DBS stimulation disturbance parameter, the drug influence disturbance parameter, the DBS stimulation and drug influence disturbance parameter are respectively used as one of the weight coefficients of the global modeling to participate in calculation, only the DBS stimulation disturbance parameter is considered in the low frequency band, only the drug influence disturbance parameter is considered in the middle frequency band, and only the DBS stimulation and the drug influence disturbance parameter are considered in the high frequency band. The characteristic parameters of each frequency band are obtained through the difference value of the first LFP signal and the second LFP signal in the first step; the DBS stimulation disturbance parameters are obtained through the third LFP signals in the step one; the drug-affected perturbation parameter is obtained through the fourth LFP signal in the first step; DBS stimulation and drug-affecting perturbation parameters are obtained from the fifth LFP signal in step one. Through the spectral analysis and observation in the third step, disturbance parameters suitable for the frequency band are respectively superposed on each frequency band, so that the finally obtained characteristic values of each frequency band can be more fit with the actually highlighted information content of each frequency band.

Drawings

FIG. 1 is a logic diagram of an embodiment of the present invention.

Fig. 2 is a flowchart of global feature value extraction according to an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail below by way of specific embodiments:

the drawings in the specification are numbered as follows: the device comprises a signal transceiving module 1, a feature extraction module 2, a comprehensive processing module 3, an output module 4, a microcontroller 5, a spectrum transformation module 6, a frequency band segmentation module 7, a sub-band feature value extraction module 8, a difference module 9, a correlation coefficient module 10, an abnormal discharge state evaluation module 11, a control rule base design module 12 and a stimulation generation module 13.

Examples

As shown in fig. 1, the system for extracting feature values of global LFP parkinson's disease includes:

the signal transceiving module 1 is used for acquiring field potential signals of the STN or GPi part through an implanted electrode; simultaneously generating stimulation pulses and measuring LFP signals; the signal receiving and transmitting module 1 is a microelectrode at least comprising two groups of contacts, and a filter circuit is connected between the two groups of contacts.

The characteristic extraction module 2 is used for extracting the characteristics of the field potential signals; the LFP signal processing module is connected with the signal transceiver module 1 and used for carrying out energy analysis on the LFP signal measured by the signal transceiver module 1, carrying out spectrum segmentation on the LFP signal according to an analysis result and extracting a characteristic value of each frequency band;

the comprehensive processing module 3 is connected with the characteristic extraction module 2, and is used for subtracting the characteristic value of each frequency band extracted by the characteristic extraction module 2 from each corresponding point value on the energy baseline spectrum stored in the comprehensive processing module 3 in advance; calculating a correlation coefficient between each frequency band and the energy baseline spectrum according to the difference value between the characteristic value of each frequency band and each corresponding point on the energy baseline spectrum;

and the output module 4 is connected with the comprehensive processing module 3 and the brain pacemaker, the correlation coefficient of each frequency band obtained by the comprehensive processing module 3 is used as a weight value, and the products of the characteristic value and the correlation coefficient of each frequency band are added to obtain a global LFP Parkinson characteristic value.

Wherein, feature extraction module 2 includes microcontroller 5 to and the three modules of being connected with microcontroller 5 respectively:

the frequency spectrum transformation module 6 is used for transforming the time domain frequency spectrum of the LFP signal into an energy frequency spectrum through a prestored formula;

the frequency band segmentation module 7 is connected with the frequency spectrum transformation module 6, carries out cluster analysis on the received energy frequency spectrum according to the energy value on the ordinate, and divides the energy frequency spectrum into at least three sub-frequency bands according to the analysis result;

a sub-band eigenvalue extraction module 8, which is connected with the frequency band segmentation module 7 and the comprehensive processing module 3, and extracts the corresponding energy value on each frequency band according to the average distribution on the frequency domain as the eigenvalue of the frequency band; the number of the characteristic values of each frequency band is equal.

Wherein, the comprehensive processing module 3 includes:

a difference module 9, connected to the sub-band eigenvalue extraction module 8, for subtracting the pre-stored energy baseline spectrum from the eigenvalues on the sub-bands received from the sub-band eigenvalue extraction module 8 to obtain an absolute value of the difference and outputting the absolute value;

and the correlation coefficient module 10 is respectively connected with the difference module 9 and the output module 4, and calculates the correlation coefficient of the difference value received from the difference module 9 through a pre-stored cross-correlation function and outputs the correlation coefficient to the output module 4.

A closed-loop DBS system is constructed through a system for extracting the feature value of the global LFP Parkinson's disease, and the system mainly comprises an abnormal discharge state evaluation module 11 connected with a feature extraction module 2 and a comprehensive processing module 3; a control rule base design module 12 respectively connected with the abnormal discharge state evaluation module 11 and the output module 4; and the stimulation generation module 13 is respectively connected with the control rule base design module 12 and the signal transceiving module 1.

Abnormal discharge state evaluation module 11: evaluating the current morbidity state through the characteristic value obtained after the characteristic extraction; without the abnormal discharge state evaluation module 11, the global feature value output by the output module 4 may be directly input into the control rule base design module 12 for electrical stimulation selection. The addition of the abnormal discharge state evaluation module 11 can enable the input of the control rule base design module 12 to be not only the global characteristic value, but also enable the control rule base design module 12 to output electrical stimulation which is more accurate and conforms to the actual situation of a patient to the stimulation generation module 13.

Control rule base design module 12: setting a corresponding stimulation parameter table according to the brain function state and the morbidity intensity;

the generation stimulation module 13: stimulation is generated by implanted electrodes. The stimulation generating module 13 may be a single microelectrode, or may be a microelectrode combined with the signal transceiver module 1.

As shown in fig. 2, after the DBS system is constructed, the global feature values corresponding to the stimulation parameters one to one are extracted, and the specific extraction steps are as follows:

step one, signal setting S1: firstly, implanting an electrode corresponding to a cerebral pacemaker into the Subthalamic nucleus (STN) of a basal ganglia part, wherein the implanted electrode is a rod-shaped electrode rod, four electrode points are arranged at the bottom of the electrode rod and are respectively marked as a first electrode, a second electrode, a third electrode and a fourth electrode from top to bottom, the first electrode and the third electrode are recording electrodes and are responsible for recording a field potential signal of the Subthalamic nucleus, and the second electrode and the fourth electrode are stimulating electrodes and are responsible for generating a stimulating pulse; the electrodes are implanted into the subthalamic nucleus through an operation, generally, the electrodes are implanted bilaterally, two electrode rods are needed, four electrode stimulation points are shared, and four field potential recording electrodes are respectively positioned on two sides. The transmission of the field potential signal and the transmission of the stimulation pulse are completed through the lead. Then, the control end of the DBS is implanted to the part under the sternal clavicle, and the analysis and the processing of the acquired signals are completed at the control end of the DBS. A control end of the DBS needs to be added with a signal acquisition circuit on the basis of the existing equipment, wherein the signal acquisition circuit comprises an A/D conversion circuit, a signal amplification circuit, a hardware filter circuit and the like. The A/D conversion mainly completes the conversion of an analog signal into a digital signal according to a certain sampling rate, because an electroencephalogram signal is weak, the signal needs to be amplified before the conversion, and an A/D conversion circuit is required to have higher resolution, in the embodiment, a 24-bit analog-to-digital converter is adopted, and the maximum amplification factor of the signal is 2000 times. The hardware filter circuit is used for directly carrying out hardware filtering on the LFP signal transmitted from the hardware filter circuit, and mainly filtering interference artifacts under the DBS stimulation environment. The DBS control end is connected with an electrode implanted in the cranium through a lead wire embedded in the subcutaneous part.

Implanting a microelectrode of a brain pacemaker into the intracranial space with different symptoms to acquire LFP signals, and respectively obtaining a first LFP signal of a Parkinson patient in a normal state, a second LFP signal which is in a Parkinson state and does not perform any DBS stimulation and drug treatment, a third LFP signal which is only subjected to DBS stimulation treatment in the Parkinson state, a fourth LFP signal which is only subjected to drug treatment in the Parkinson state, and a fifth LFP signal which is in the Parkinson state and simultaneously subjected to DBS stimulation and drug treatment;

LFP signals are acquired by microelectrodes implanted into the cranium, and special signal acquisition electrodes are not implanted, so that the burden of a patient is reduced. The DBS electrode is used as a signal acquisition electrode, and the influence of a stimulation signal on signal acquisition needs to be solved. The design is carried out from a hardware circuit, a hardware filter circuit is added between the first electrode and the third electrode as well as the second electrode and the fourth electrode, and the artificial artifact interference generated by the stimulation signal is directly removed from the hardware, so that the obtained LFP signal is not interfered by the stimulation signal, and the task of recording the signal while stimulating DBS is completed.

Step two, LFP signal preprocessing S2: sequentially filtering, denoising and down-sampling the acquired LFP signals to respectively obtain 1-7Hz low-frequency LFP signals, 8-35Hz medium-frequency LFP signals and 150Hz high-frequency LFP signals in each state;

the filtering of the signal mainly refers to filtering the interference of a special frequency band by adopting band-pass filtering, and includes designing a filter for each signal segment according to the task of the next stage, and the main purpose is to remove the spike potential of high frequency. Denoising, which mainly means removing artifacts, and further filtering software mainly aiming at the artifacts generated by the DBS stimulation mentioned above; the down-sampling is to reduce the sampling frequency of the signal, reduce the data volume, reduce the difficulty of algorithm processing, improve the efficiency of algorithm operation, and facilitate the transplantation of the algorithm to the control end of the DBS system according to the application requirement and by combining the maximum frequency requirement. And finally, finishing the signal time sequence in the signal acquisition process.

And designing a Kalman filter, filtering the acquired signals, wherein the filter adopts third-order filtering and gives consideration to the operation efficiency and accuracy. Firstly, band-pass filtering is carried out to filter the interference of specific frequency, and the band-pass filter mainly filters the frequency band which is the same as the stimulating frequency of the DBS and is close to 130 Hz; and a low-pass filter is used for filtering high-frequency spike potential interference, the cut-off frequency of the low-pass filter is 200Hz, and frequency components higher than 200Hz are filtered. According to the requirements of the filtering, the signal needs to be down-sampled before being filtered, the sampling frequency of the signal is fixed at 500Hz, and the signal of the signal in the frequency band of 0-200Hz can better keep the characteristics of the analog signal. Finally, the time sequence of the signal is preserved.

Step three, spectrum analysis S3: the preprocessed signal is converted from the time domain to the frequency domain for analysis: analyzing by combining the frequency spectrum, the power spectrum and the spectrum to respectively obtain frequency domain characteristics of the low-frequency LFP signal, the medium-frequency LFP signal and the high-frequency LFP signal; the energy spectra of the second to fifth LFP signals in the respective frequency bands are directly compared with the energy spectrum of the first LFP signal measured under normal conditions in the parkinson's disease patient. And respectively 10 frequency values which are uniformly distributed in each frequency band, wherein the power value corresponding to the 10 frequency values is the characteristic value of each frequency band.

In the third step, the preprocessed signals can be directly converted, the most basic FFT algorithm is adopted, the algorithm spectrum is analyzed on the whole, and the signals of each frequency band are directly observed from the power spectrogram to be evaluated. The FFT algorithm is the most basic algorithm, and can obtain global frequency domain characteristics more efficiently.

Or, in the third step, for the preprocessed signal, firstly, frequency division is performed by using a filter, the signal is divided into a plurality of preset frequency bands, then, power spectrum and spectrum analysis are respectively performed on the signal of each frequency band, characteristics on each frequency band are observed, when frequency domain conversion is performed, hilbert yellow conversion is adopted, and the time-frequency conversion characteristics of the signal are concerned. Hilbert-Huang transform, which mainly acquires time-frequency characteristics of signals and monitors instantaneous change characteristics of the signals.

Or in the third step, after the frequency spectrum is acquired, a classical method is adopted to respectively acquire a frequency spectrum density graph and a frequency spectrum graph of the signal, all frequency spectrum densities adopt a uniform quantization standard, a certain amount of data is firstly acquired, the characteristics of the integral power spectrum under each environment are integrally analyzed, the results are classified and sorted by adopting a least square method, the frequency bands under each environment are solidified according to the sorted results, a theoretical baseline is calculated, a filter is adopted to carry out frequency division, the signal is divided into a plurality of preset frequency bands, then the power spectrum and the frequency spectrum of the signal of each frequency band are respectively analyzed, and the characteristics of each frequency band are observed.

Step four, extracting and standardizing the characteristics of each frequency band S4: quantizing the frequency domain characteristics on each frequency band obtained in the third step to obtain discrete values within 0-100;

in the fourth step, the highest peak value of the energy of all the collected signals can be taken as a standard value to carry out normalization processing; mapping each energy value in a fluctuation range of 0-100, and taking the average value of all the energy values of the sub-frequency bands as a characteristic value.

Or, in the fourth step, the statistical processing is performed on each frequency band according to the respective condition, the energy of the normal LFP signal on each frequency band is combined to calculate the baseline on each frequency band, the energy value on each frequency band is subtracted from the baseline, and the sum of the difference values of each frequency band is output as the characteristic value; the characteristic value is quantized between 0 and 100 by taking the highest peak value of the signal as a standard.

Step five, global modeling and feature standardization S5: and modeling by taking the characteristic values of the frequency bands obtained in the front as parameters and taking the correlation between the frequency bands represented by the characteristic values and the Parkinson symptoms thereof as weights, and outputting global characteristic values.

In the fifth step, a linear regression model can be adopted, the correlation between the frequency band represented by the linear regression model and the Parkinson's disease is calculated and used as a coefficient of the linear model, the coefficient is quantized to a fluctuation range of 0 to 1, the parameter of each frequency band is used as an input to be multiplied by the coefficient, and the finally obtained result is used as a global characteristic parameter to be output.

Or in the fifth step, the acquired offline data is used as a database by using the artificial neural network model, the number of the neural network nodes is designed according to the preset number of the frequency division sections as three, the correlation coefficient is used as the weight of the node, and the acquired offline database is used as test data to learn the network. And after learning is finished, the characteristic values of all frequency bands are used as input and input to corresponding nodes, and the model output is used as a global characteristic value.

Or, in step five, the environmental factors are considered in the modeling, the characteristic parameters of the low frequency band consider their relationship with the DBS stimulation, the Beta band considers their influence by drug interference, and the high frequency band considers the overlap with the DBS stimulation parameters. The disturbance parameter is set for the interference under the three environments, and the parameter is introduced into the two schemes mentioned above, so that the adaptability of the whole algorithm is improved.

Superposing one DBS stimulation disturbance parameter on the characteristic parameter of the low frequency band, superposing one drug influence disturbance parameter on the characteristic parameter of the medium frequency band, and superposing one DBS stimulation and drug influence disturbance parameter on the characteristic parameter of the high frequency band; the DBS stimulation disturbance parameter, the drug influence disturbance parameter and the DBS stimulation and drug influence disturbance parameter are all in the range of 0-1.

The signal peak herein refers to the peak of the highest peak.

Test example:

the same patient is selected, and the extraction system in the embodiment and the existing extraction system are respectively used for extracting the parkinson feature value. At a score of ten, fifty professionals with medical knowledge were selected to score whether the extracted parkinsonian features were consistent with the patient's actual degree of parkinsonism. Wherein, the complete conformity is ten, and the complete nonconformity is zero. Under the same conditions, comparative examples and examples shown in table 1 were constructed. The extraction system in the embodiment is simply referred to as the present extraction system, and the extraction method in the embodiment is simply referred to as the present extraction method. The existing extraction system is only the brain wave acquisition system of the LFP signal which is most commonly used at present, and the existing extraction method is only used for extracting the characteristic value of the LFP signal. In comparative example 2, the feature values of each frequency band in the full frequency band were extracted separately by the existing extraction method, and then directly averaged.

TABLE 1

Name (R)	Extraction system	Extraction method	Extracting frequency bands
				Comparative example 1	Existing extraction system	Existing extraction methods	Intermediate frequency band
Comparative example 2	Existing extraction system	Existing extraction methods	Full frequency band
				Comparative example 3	The extraction system	The extraction method	High frequency band
Comparative example 4	The extraction system	The extraction method	Intermediate frequency band
				Comparative example 5	The extraction system	The extraction method	Low frequency band
Examples	The extraction system	The extraction method	Full frequency band

Under the same conditions of time, space, environment and the like, the parkinsonism characteristic values and the parkinsonism degrees represented by the parkinsonism characteristic values obtained in the table 1 and the actual parkinsonism degrees are scored, and the scores of fifty professionals are averaged to obtain the results shown in the table 2.

TABLE 2

Name (R)	Comparative example 1	Comparative example 2	Comparative example 3	Comparative example 4	Comparative example 5	Examples
							Scoring	6.5	7.3	6.7	8.5	7.2	9.8

In table 2, comparing comparative example 1 with comparative example 2 and comparative examples 3 to 5 with examples, respectively, it can be seen that the signal extraction of the full band more reflects the actual condition of the patient than the signal extraction of the single band. Comparing comparative example 1 and comparative example 2 with comparative example 3 and example, respectively, it can be seen that the parkinson's characteristic values obtained by the extraction system and the extraction method in this example are superior to those obtained by the existing extraction system and the existing extraction method. The reason is considered, and the extraction system and the extraction method in the embodiment not only fully consider the information content of different frequency bands, but also add the disturbance parameters of the characteristic representation of each frequency band in a targeted manner through spectrum analysis, so that the finally obtained characteristic value of each frequency band can represent the main information to be represented by the frequency band. The obtained Parkinson characteristic value is more objective and closer to the actual disease condition of a patient, and the method is favorable for treating the disease in a targeted manner in the later period.

The foregoing is merely an example of the present invention, and common general knowledge in the field of known specific structures and characteristics is not described herein in any greater extent than that known in the art at the filing date or prior to the priority date of the application, so that those skilled in the art can now appreciate that all of the above-described techniques in this field and have the ability to apply routine experimentation before this date can be combined with one or more of the present teachings to complete and implement the present invention, and that certain typical known structures or known methods do not pose any impediments to the implementation of the present invention by those skilled in the art. It should be noted that, for those skilled in the art, without departing from the structure of the present invention, several changes and modifications can be made, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicability of the patent. The scope of the claims of the present application shall be determined by the contents of the claims, and the description of the embodiments and the like in the specification shall be used to explain the contents of the claims.

Claims

1. A system for global LFP parkinsonism characteristic value extraction, characterized by: the method comprises the following steps:

the output module is connected with the comprehensive processing module and the brain pacemaker, the correlation coefficient of each frequency band obtained by the comprehensive processing module is used as a weight value, and the products of the characteristic value of each frequency band and the correlation coefficient are added to obtain a global LFP Parkinson characteristic value;

the feature extraction module comprises a microcontroller and is respectively connected with the microcontroller:

the frequency band division characteristic value extraction module is connected with the frequency band division module and the comprehensive processing module, and takes an energy value extracted by average distribution of each frequency band on a frequency domain as a characteristic value of the frequency band; the number of the characteristic values of each frequency band is equal;

the comprehensive processing module comprises:

2. The system for global LFP parkinson's disease feature value extraction according to claim 1, characterized in that: the signal receiving and transmitting module is a microelectrode at least comprising two groups of contacts, and a filter circuit is connected between the two groups of contacts.