Parkinson's disease cognitive impairment early judgment system based on transcranial magnetic-electroencephalogram multi-dimensional characteristics

Abstract

The invention discloses an early judgment system for cognitive impairment of Parkinson's disease based on transcranial magnetic-electroencephalogram multi-dimensional features. The early judgment system comprises a transcranial magnetic-electroencephalogram combined acquisition module, a synchronous signal preprocessing module, a multi-dimensional feature extraction module, a feature intelligent fusion module and a storage and display module, the transcranial magnetic-electroencephalogram combined acquisition module is used for disturbing the neural activity of a patient through monopulse transcranial magnetic stimulation coding and acquiring an evoked state electroencephalogram signal; the synchronizing signal preprocessing module is used for denoising and filtering the evoked state electroencephalogram signals; a multi-dimensional feature extraction module performs feature extraction on the processed evoked state electroencephalogram signals by utilizing multi-dimensional electroencephalogram feature analysis to obtain multi-parameter electroencephalogram neural activity features; the feature intelligent fusion module is used for carrying out weight distribution and feature fusion on the multi-parameter electroencephalogram neural activity features to obtain cognitive disorder multi-modal neural indexes; and the storage and display module is used for storing the cognitive disorder multi-modal nerve indexes and visualizing the cognitive disorder multi-modal nerve indexes.

Description

Technical Field

[0001] The present invention belongs to the field of brain-computer interface technology and medical signal technology, and specifically relates to an early judgment system for cognitive impairment of Parkinson's disease based on the multi-dimensional characteristics of transcranial magnetic and electroencephalographic (TEE) technology. Background Art

[0002] Cognitive impairment in Parkinson's disease is characterized by a progressive decline in cognitive function, which worsens with age, disease duration and severity. It is clinically heterogeneous and involves deficits in one or more cognitive areas, including executive dysfunction, accompanied by deficits in language, visual-spatial ability and memory. Cognitive impairment in Parkinson's disease is common in both the early and late stages of the disease course in Parkinson's disease patients, who are at high risk of developing Parkinson's dementia. Its onset is insidious and greatly affects the quality of life, increasing social and medical burdens. In recent years, the incidence of cognitive impairment in Parkinson's disease has been on the rise, and has become an increasingly serious public health problem.

[0003] There is currently no effective treatment for Parkinson's disease cognitive impairment, and early diagnosis is crucial to delay the progression of the disease. Existing resting-state EEG features such as time-frequency oscillatory activity and changes in brain network functional connectivity are associated with cognitive decline in Parkinson's disease patients. However, resting-state EEG indicators are easily disturbed by patients' speech or motor symptoms, and the cortical defect indicators judged by imaging are not obvious in the early stages of cognitive decline. In the past, early judgment of Parkinson's disease cognitive impairment was based on single-dimensional brain neural characteristics, and there was a lack of stable multi-parameter neural indicators. As an emerging method of brain activity assessment, single-pulse transcranial magnetic stimulation combined with electroencephalography (TMS-EEG) can directly record cortical activation caused by transcranial magnetic stimulation in specific areas of the cortex non-invasively, reflecting multi-dimensional cortical responses such as cortical excitability, oscillation dynamics and effective connectivity. It does not require the subject to participate in the task, has the advantages of safe operation, stable indicators and strong repeatability, and combined with multi-dimensional EEG feature analysis methods, it can be used as an effective means for early assessment of Parkinson's disease cognitive impairment.

[0004] Therefore, this study specifically aims at the early judgment of cognitive impairment in Parkinson's disease patients and designs an early judgment system for cognitive impairment in Parkinson's disease based on transcranial magnetic and electroencephalography. Summary of the invention

[0005] The present invention aims to solve the deficiencies of the prior art and provides the following solutions:

[0006] An early judgment system for cognitive impairment of Parkinson's disease based on multi-dimensional features of transcranial magnetic and electroencephalogram, comprising: a transcranial magnetic and electroencephalogram joint acquisition module, a synchronous signal preprocessing module, a multi-dimensional feature extraction module, a feature intelligent fusion module and a storage and display module;

[0007] The transcranial magnetic and electroencephalogram (EEG) joint acquisition module is used to disturb the patient's neural activity through single-pulse transcranial magnetic stimulation encoding and acquire evoked EEG signals;

[0008] The synchronization signal preprocessing module is used to perform denoising filtering on the evoked EEG signal to obtain a processed evoked EEG signal;

[0009] The multi-dimensional feature extraction module uses multi-dimensional EEG feature analysis to extract features from the processed evoked EEG signal to obtain multi-parameter EEG neural activity features, wherein the multi-parameter EEG neural activity features include: time domain features, time-frequency domain features, and brain network multi-dimensional brain neural features;

[0010] The feature intelligent fusion module is used to construct a fusion feature machine learning model, and use the fusion feature machine learning model to perform weight assignment and feature fusion on the multi-parameter EEG neural activity features to obtain a multimodal neural index of cognitive impairment;

[0011] The storage and display module is used to store the cognitive impairment multimodal neural index and visualize the cognitive impairment multimodal neural index.

[0012] Preferably, the transcranial magnetic-electroencephalogram joint acquisition module comprises: a resting motor threshold measurement unit, a transcranial magnetic stimulation unit and an electroencephalogram acquisition unit;

[0013] The resting motor threshold measurement unit is used to apply transcranial magnetic stimulation to the primary motor cortex of the affected side of the patient in a 45° front-to-back direction to induce muscle contraction on the opposite side to generate motor evoked potentials, and the minimum intensity required to induce an electromyographic peak-to-peak amplitude greater than or equal to 50 μV in at least 5 of 10 consecutive stimulations is used as the resting motor threshold;

[0014] The transcranial magnetic stimulation unit is used to apply 80 single pulses to the right posterior parietal cortex of the patient in a 45° front-to-back direction, with each single pulse stimulation interval being 2 to 4 seconds, and the pulse intensity being set to 90% of the resting motor threshold;

[0015] The EEG acquisition unit acquires the evoked EEG signal at a sampling rate of 1000 Hz.

[0016] Preferably, the workflow of the synchronization signal preprocessing module includes:

[0017] Importing the evoked EEG signal and transcranial magnetic stimulation electrode information, locating electrodes based on the electrode information, and deleting useless electrodes;

[0018] Extracting the events of interest in the evoked EEG signal and locking the corresponding evoked EEG signal, and then performing baseline correction to remove transcranial magnetic stimulation pulse artifacts and induced electromyographic artifacts;

[0019] Performing a first missing data interpolation around a transcranial magnetic stimulation pulse to minimize artifacts caused by a low-pass and anti-aliasing filter, then replacing the interpolated data of the first missing data interpolation with the first constant amplitude EEG signal data, and then using independent component analysis and automatic component selection to remove voltage decay artifacts caused by scalp muscle activation and transcranial magnetic stimulation-induced electromyography in the evoked state EEG signal;

[0020] The evoked EEG signal within 2 ms before and 10 ms after the transcranial magnetic stimulation pulse is deleted, a second missing data interpolation is performed around the transcranial magnetic stimulation pulse, and a 1-100 Hz band-pass filter and a 48-52 Hz band-stop filter are performed, and then the interpolated data of the second missing data interpolation are replaced with the second constant EEG signal amplitude data, and then independent component analysis and automatic component selection are used again to remove the non-transcranial magnetic stimulation locked artifacts remaining in the evoked EEG signal;

[0021] A third missing data interpolation is performed around the transcranial magnetic stimulation pulse, the electrode channel CPz of the missing EEG signal data is interpolated, and the electrode information is re-referenced to a common average reference to obtain the processed evoked EEG signal.

[0022] Preferably, the multidimensional feature extraction module includes: a time domain feature extraction unit, a time-frequency domain feature extraction unit and a brain network feature extraction unit;

[0023] The time domain feature extraction unit is used to extract the global average field power and transcranial magnetic stimulation evoked potential of each electrode at each time point in the processed evoked state EEG signal;

[0024] The time-frequency domain feature extraction unit is used to extract the average power of each frequency band of each electrode channel of the whole brain of the processed evoked EEG signal within a certain time period;

[0025] The brain network feature extraction unit is used to extract the weighted phase lag index matrix and brain network attribute graph theory parameters of the processed evoked state electroencephalogram signal.

[0026] Preferably, the workflow of the time domain feature extraction unit includes:

[0027] The global mean field power is calculated by evaluating the standard deviation of the EEG signal amplitude of each electrode at each time point in the processed evoked EEG signal:

[0028]

[0029] Among them, n represents the total number of channels of EEG signals, i represents the number of channels of EEG signals, and A i (t) represents the amplitude of the EEG signal of the ith channel, A mean (t) represents the average amplitude of all channels at time t;

[0030] The peak values ​​of the transcranial magnetic stimulation evoked potential of each EEG signal channel in the five time windows of interest in the evoked EEG signal after the processing are extracted as the transcranial magnetic stimulation evoked potential, which are P30, N45, P60, N100 and P180 respectively:

[0031]

[0032] Where x represents each subject, It represents the average potential peak value of the event in all channels of all subjects during the period from t1 to t2. P_tep represents the positive potentials P30, P60 and P180, and N_tep represents the negative potentials N45 and N100.

[0033] Preferably, the workflow of the time-frequency domain feature extraction unit includes:

[0034] Morlet wavelet convolution is used to perform time-frequency decomposition of the five frequency bands of δ, θ, α, β and γ in the processed evoked EEG signal, extract the signal amplitude, and obtain the energy of specific time-frequency points. Then, the time window is divided into intervals of 100 ms, and the average power of each frequency band of each electrode channel in the whole brain in the processed evoked EEG signal during the period of 0 to 400 ms after transcranial magnetic stimulation is calculated.

[0035] Preferably, the workflow of the brain network feature extraction unit includes:

[0036] The phase information of each electrode in the processed evoked EEG signal is extracted based on the time-frequency matrix after time-frequency decomposition, and the phase matrix is ​​subtracted to obtain the phase difference matrix, and the cross spectrum between electrodes is obtained. The weighted phase lag index matrix is ​​calculated based on the cross spectrum:

[0037]

[0038] Among them, Im[·] represents the imaginary part, G i，j (t, f) represents the cross spectrum of two time series of the channel of the i-th EEG signal and the channel of the j-th EEG signal, t represents time, and f represents frequency;

[0039] The characteristics of the brain network are quantitatively described using graph theory methods, and the graph theory parameters of the brain network attributes are calculated. The graph theory parameters of the brain network attributes include: clustering coefficient, average shortest path length, global efficiency and local efficiency:

[0040]

[0041] Where CC represents the clustering coefficient, ASPL represents the average shortest path length, GE represents the global efficiency, LE represents the local efficiency, N represents the number of all nodes in the brain network, and h a is the number of nodes adjacent to node a, f a Yes a The number of actual connecting edges between nodes, l ab is the shortest path length between node a and node b, G a is a subgraph consisting of the adjacent nodes of node a, l bh is the shortest path length from node b to node h.

[0042] Preferably, the feature intelligent fusion module includes: a data processing unit, a feature selection unit, a weight analysis unit, a model building unit and a weighted fusion unit;

[0043] The data processing unit is used to pre-process the time domain features, the time-frequency domain features and the brain network multi-dimensional cranial nerve features to obtain processed features;

[0044] The feature selection unit uses an elastic network regression feature selection method to determine the optimal feature selection parameters;

[0045] The weight analysis unit is used to assign weights to the time domain features, the time-frequency domain features and the multi-dimensional cranial nerve features of the brain network in the optimal feature selection parameters to obtain weight proportions of different features;

[0046] The model building unit builds a multi-core support vector machine fusion model by combining different kernel functions to obtain the fusion feature machine learning model;

[0047] The weighted fusion unit uses the fusion feature machine learning model to perform feature matching between different kernel functions and features of different dimensions, and fuses the features of different kernel functions based on the weight proportions to obtain the multimodal neural index of cognitive impairment.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. Aiming at the problem of instability of existing resting-state EEG indicators, the present invention independently designs a single-pulse transcranial magnetic stimulation encoding scheme, selects the right posterior parietal cortex of the cognitive and motor integration brain area as the magnetic stimulation target, and synchronously collects the evoked EEG signals of the subjects under disturbance, thereby improving the objectivity of the disease cognitive function detection and being able to more accurately reflect the symptoms and characteristics of mild cognitive impairment of Parkinson's disease;

[0050] 2. The present invention utilizes a time-frequency-space multi-dimensional brain neural feature extraction algorithm, which can provide multi-parameter EEG features of cortical excitability, neural oscillation dynamics and brain network connectivity compared to the existing single-dimensional EEG features, thereby increasing the characteristic dimensions and indicator credibility of Parkinson's disease mild cognitive impairment.

[0051] 3. The present invention adopts artificial intelligence algorithm to construct a fusion feature machine learning model. Different from the existing recognition method based on quantitative EEG, it can optimize the selection of disease features, determine the weight distribution of multimodal disease features, remove redundant information, and improve the recognition specificity of mild cognitive impairment in Parkinson's disease patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0053] Figure 1 A schematic diagram of the system structure of an embodiment of the present invention;

[0054] Figure 2 Schematic diagram of a transcranial magnetic and electroencephalographic joint acquisition module according to an embodiment of the present invention;

[0055] Figure 3 It is a schematic diagram of the experimental results of an embodiment of the present invention, wherein A represents the average butterfly diagram of the subject's transcranial magnetic stimulation evoked potential waveform of all electrode channels, B represents the time-frequency diagram of the mean value of the whole-brain oscillation power induced after transcranial magnetic stimulation, and C represents the statistical difference display diagram of two groups of brain network functional connections constructed by weighted phase lag index in the Theta and Alpha frequency bands. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0058] Embodiment 1

[0059] In this embodiment, if Figure 1 As shown, a system for early judgment of cognitive impairment of Parkinson's disease based on the multi-dimensional features of transcranial magnetic and electroencephalogram (TME), including: a transcranial magnetic and electroencephalogram (TME) joint acquisition module, a synchronous signal preprocessing module, a multi-dimensional feature extraction module, a feature intelligent fusion module and a storage and display module.

[0060] The transcranial magnetic and electroencephalogram (EEG) combined acquisition module is used to disturb the patient's neural activity through single-pulse transcranial magnetic stimulation encoding and to collect evoked EEG signals.

[0061] The transcranial magnetic and electroencephalogram (EEG) combined acquisition module includes: a resting motor threshold measurement unit, a transcranial magnetic stimulation unit and an EEG acquisition unit; the resting motor threshold measurement unit is used to apply transcranial magnetic stimulation to the primary motor cortex on the affected side of the patient in a 45° front-to-back direction, causing the contralateral muscle to contract and produce motor evoked potentials, and the minimum intensity required to cause the inter-peak amplitude of the electromyography to be greater than or equal to 50μV in at least 5 out of 10 continuous stimulations is used as the resting motor threshold; the transcranial magnetic stimulation unit is used to apply 80 single pulses to the right posterior parietal cortex of the patient in a 45° front-to-back direction, with the stimulation interval of each single pulse being 2 to 4 seconds, and the pulse intensity being set to 90% of the resting motor threshold; the EEG acquisition unit acquires evoked EEG signals at a sampling rate of 1000Hz.

[0062] In this embodiment, if Figure 2 As shown, the transcranial magnetic-electroencephalogram joint acquisition module includes a transcranial magnetic stimulation device and a synchronous electroencephalogram acquisition device. The transcranial magnetic stimulation device includes a frameless computerized stereotactic neuronavigation system, an electromyography acquisition device, a magnetic stimulation generating device, and a rapid magnetic biphasic stimulator. The electroencephalogram acquisition device includes an electroencephalogram cap and an electroencephalogram signal acquisition device. The frameless computerized stereotactic neuronavigation system is responsible for tracking and recording the position of the coil, the electromyography acquisition device is responsible for online monitoring of electromyographic activity to determine the resting motion threshold, and the magnetic stimulation generating device is responsible for generating magnetic stimulation of selected parameters. Transcranial magnetic stimulation is performed by a rapid magnetic biphasic stimulator equipped with an 8-shaped coil (70mm) for emitting transcranial magnetic stimulation pulses. Equipped with a synchronous electroencephalogram acquisition device, a 64-channel electroencephalograph with wet electrodes that meet clinical standards is used to continuously record the induced electroencephalogram signals, using Ag / AgCl electrodes, and the electrode placement follows the 10-20 international standard, which is used for real-time synchronous acquisition of electroencephalogram signals induced by transcranial magnetic stimulation, and a display screen is provided to synchronously display the collected electroencephalogram signals.

[0063] In this embodiment, the workflow of the transcranial magnetic and electroencephalogram (EEG) combined acquisition module includes: (1) Navigation positioning: Based on the frameless computerized stereotactic neuronavigation system, the position of the coil is tracked and recorded, and the signal acquisition brain area and electrode position are determined in combination with a 64-channel EEG system compatible with transcranial magnetic stimulation; (2) Acquisition instructions: Parkinson's patients are asked to sit on a chair with their forearms on the armrests, remain relaxed, and stare at the "black cross" on the computer screen 70 cm in front; (3) Resting motor threshold measurement: Transcranial magnetic stimulation is applied to the primary motor cortex on the affected side of the patient in a 45° front-to-back direction to induce contralateral muscle contraction to generate motor evoked potentials, which are obtained through Ag / AgCl surface electrodes in a ventral tendon connection manner, where the active electrode is placed on the muscle belly and the reference electrode is placed on the metacarpophalangeal joint of the index finger. Electromyographic activity was continuously monitored online, and the resting motor threshold was defined as the minimum intensity required to elicit an inter-peak amplitude of 50 μV or more in at least 5 out of 10 consecutive trials. (4) Transcranial magnetic stimulation was used to synchronize EEG acquisition: 80 single pulses were applied to the patient's right posterior parietal cortex at a 45° anterior-posterior direction. The interval between each single pulse was 2 to 4 seconds, and the pulse intensity was set to 90% of the resting motor threshold. (5) Whole-brain EEG signals were acquired at a sampling rate of 1000 Hz, and the impedance value was kept below 5 kΩ.

[0064] The synchronous signal preprocessing module is used to perform denoising and filtering on the evoked EEG signal to obtain the processed evoked EEG signal.

[0065] The workflow of the synchronous signal preprocessing module includes: importing evoked EEG signals and transcranial magnetic stimulation electrode information, locating electrodes based on the electrode information, and deleting useless electrodes; extracting events of interest in evoked EEG signals and locking the corresponding evoked EEG signals, and then performing baseline correction to remove transcranial magnetic stimulation pulse artifacts and induced electromyographic artifacts; performing the first missing data interpolation around the transcranial magnetic stimulation pulse to minimize the artifacts caused by the low-pass and anti-aliasing filters, and then using the first constant amplitude EEG signal data to replace the interpolated data of the first missing data interpolation, and then using independent component analysis and automatic component selection to remove voltage attenuation artifacts caused by scalp muscle activation and transcranial magnetic stimulation-induced electromyography in the evoked EEG signals; the period from 2 ms before to after the transcranial magnetic stimulation pulse is interpolated. The evoked EEG signals within 10ms are deleted to minimize the artifacts caused by the low-pass and anti-aliasing filters, and the second missing data interpolation is performed around the transcranial magnetic stimulation pulse, and a band-pass filter of 1 to 100 Hz and a band-stop filter of 48 to 52 Hz are performed. Then, the interpolated data of the second missing data interpolation is replaced by the second constant amplitude data, and then the independent component analysis and automatic component selection are used again to remove the non-transcranial magnetic stimulation locked artifacts remaining in the evoked EEG signals; the third missing data interpolation is performed around the transcranial magnetic stimulation pulse, and the electrode channel CPz of the missing EEG signal data is interpolated, that is, the average EEG signal of the electrodes around it is used as its EEG signal, and the electrode information is re-referenced to the common average reference to obtain the processed evoked EEG signal.

[0066] The multi-dimensional feature extraction module uses multi-dimensional EEG feature analysis to extract features from the processed evoked EEG signals to obtain multi-parameter EEG neural activity features. The multi-parameter EEG neural activity features include: time domain features, time-frequency domain features and multi-dimensional brain neural features of the brain network.

[0067] The multi-dimensional feature extraction module includes: a time domain feature extraction unit, a time-frequency domain feature extraction unit and a brain network feature extraction unit.

[0068] The time domain feature extraction unit is used to extract the global average field power and transcranial magnetic stimulation evoked potential of each electrode at each time point in the processed evoked EEG signal.

[0069] The workflow of the time domain feature extraction unit includes: calculating the global average field power by evaluating the standard deviation of the EEG signal amplitude of each electrode at each time point in the processed evoked EEG signal:

[0070]

[0071] Among them, n represents the total number of channels of EEG signals, i represents the number of channels of EEG signals, and A i (t) represents the amplitude of the EEG signal of the ith channel, A mean(t) represents the average amplitude of all channels at time t; the peak value of the transcranial magnetic stimulation evoked potential of each EEG signal channel in the five time windows of interest in the evoked state EEG signal after processing is extracted as the transcranial magnetic stimulation evoked potential, which are P30 (15-35ms), N45 (35-55ms), P60 (55-75ms), N100 (75-150ms) and P180 (150-250ms):

[0072]

[0073] Where x represents each subject, It represents the average potential peak value of the event in all channels of all subjects during the period from t1 to t2. P_tep represents the positive potentials P30, P60 and P180, and N_tep represents the negative potentials N45 and N100.

[0074] The time-frequency domain feature extraction unit is used to extract the average power of each frequency band of each electrode channel in the whole brain of the processed evoked EEG signal within a certain time period.

[0075] The workflow of the time-frequency domain feature extraction unit includes: using Morlet wavelet convolution to perform time-frequency decomposition on the five frequency bands of δ (1-4 Hz), θ (4-7 Hz), α (8-13 Hz), β (14-30 Hz) and γ (31-45 Hz) in the processed evoked EEG signal, extracting the signal amplitude, and obtaining the energy of a specific time-frequency point. The parameters of Morlet wavelet convolution include: bandwidth parameter (Fb) is 7, center frequency parameter is 0.5, total scale is 1024, and the calculation formula is as follows:

[0076]

[0077] Among them, Sca represents the scale value, Fc represents the center frequency parameter, Tsc represents the total scale parameter, m represents the scale parameter, w represents the intermediate variable, and ψ(t) represents the Morley wavelet function; then the time window is divided into intervals of 100ms, and the average power of each frequency band of each electrode channel in the whole brain in the processed evoked EEG signal during the period of 0 to 400ms after transcranial magnetic stimulation is calculated.

[0078] The brain network feature extraction unit is used to extract the weighted phase lag index matrix of the processed evoked EEG signal and the brain network attribute graph theory parameters.

[0079] The workflow of the brain network feature extraction unit includes: calculating the weighted phase lag index to quantify the degree of signal coupling between channels, establishing a brain network with scalp electrodes as network nodes and weighted phase lag index as connection parameter; extracting the phase information of each electrode in the evoked EEG signal after processing based on the time-frequency matrix after time-frequency decomposition, and subtracting the phase matrix to obtain the phase difference matrix, obtaining the cross spectrum between electrodes, and calculating the weighted phase lag index matrix based on the cross spectrum:

[0080]

[0081] Among them, Im[·] represents the imaginary part, G i，j (t, f) represents the cross spectrum of two time series of the channel of the i-th EEG signal and the channel of the j-th EEG signal, t represents time, and f represents frequency; the graph theory method is used to quantitatively describe the characteristics of the brain network, and the graph theory parameters of the brain network attributes are calculated. The graph theory parameters of the brain network attributes include: clustering coefficient, average shortest path length, global efficiency and local efficiency. In this embodiment, the clustering coefficient (CC) is constructed to measure the possibility that the adjacent nodes of a node in the network are adjacent to each other; the efficiency of information transmission within the network is reflected by the average shortest path length (ASPL) and global efficiency (GE) which are reciprocals of each other; and the local efficiency (LE) is constructed to quantify the ability to transmit brain interaction information in the local part of the brain network:

[0082]

[0083] Where CC represents the clustering coefficient, ASPL represents the average shortest path length, GE represents the global efficiency, LE represents the local efficiency, N represents the number of all nodes in the brain network, and h a is the number of nodes adjacent to node a, f a Yes a The number of actual connecting edges between nodes, l ab is the shortest path length between node a and node b, G a is a subgraph consisting of the adjacent nodes of node a, l bh is the shortest path length from node b to node h.

[0084] The feature intelligent fusion module is used to construct a fusion feature machine learning model, and use the fusion feature machine learning model to perform weight assignment and feature fusion on multi-parameter EEG neural activity features to obtain multimodal neural indicators of cognitive impairment.

[0085] The feature intelligent fusion module includes: data processing unit, feature selection unit, weight analysis unit, model construction unit and weighted fusion unit. The data processing unit is used to preprocess the time domain features, time-frequency domain features and multi-dimensional brain neural features of the brain network to obtain the processed features; the feature selection unit uses the elastic network regression feature selection method to determine the optimal feature selection parameters; the weight analysis unit is used to assign weights to the time domain features, time-frequency domain features and multi-dimensional brain neural features of the brain network in the optimal feature selection parameters to obtain the weight proportions of different features; the model construction unit constructs a multi-core support vector machine fusion model by combining different kernel functions to obtain a fusion feature machine learning model; the weighted fusion unit uses the fusion feature machine learning model to match different kernel functions with features of different dimensions, and fuses the features of different kernel functions based on the weight proportions to obtain multimodal neural indicators of cognitive impairment.

[0086] In this embodiment, the data processing unit constructs feature matrices, removes outliers, normalizes data, and performs sign tests on time domain features, time-frequency domain features, and multi-dimensional brain nerve features of brain networks; specifically, the feature matrices are constructed based on the global average field strength, transcranial magnetic stimulation evoked potentials, weighted phase lag index, and graph theory parameters, and the label vectors for each type of feature matrix training are set; the abnormal values ​​in the feature matrix, such as null values ​​and infinite values, are removed to avoid computer memory overflow and model training failure; the zero mean method is used to normalize the data to avoid the dimension problem between different feature data affecting the feature weight calculation; the sign test is used for preliminary feature screening to avoid individual errors from being introduced into the model. The feature selection unit adopts the elastic network regression feature selection method, and sets two layers of parameter optimization iteration to determine the optimal feature selection parameters; specifically, one layer of regularization is used for feature selection and model simplification, and the second layer of regularization prevents model overfitting. In the iterative process, the weight coefficient corresponding to each feature is trained, and the features with zero coefficients, i.e., redundant and invalid features, are removed to improve the iteration speed. The weight parameter introduced by the weight analysis unit in the process of constructing the multi-core support vector machine model is to measure the weight proportion of time-frequency-space multidimensional features in identifying Parkinson's disease cognitive impairment, that is, the contribution (weight) of different dimensional features to the classification results of the fusion model; in order to obtain the weight parameter more accurately, a cross-validation method is adopted, and a joint parameter optimization strategy is used to determine the optimal feature selection parameters through internal iteration, and the optimal parameter classification model and its weight parameters are extracted; in order to measure the importance of sub-features under each dimensional feature, a random repeated eight-fold cross-validation method is adopted to count the optimal feature selection results corresponding to the optimal fusion classifier obtained in each group of tests, calculate the frequency of each dimensional feature and sort it from high to low to determine the feature weight of each feature vector. The model construction unit constructs a multi-core support vector machine fusion model by combining different kernel functions to obtain a fusion feature machine learning model. The weighted fusion unit uses the fusion feature machine learning model to match different kernel functions with features of different dimensions, and fuses the features of different kernel functions based on the weight proportion to obtain a multimodal neural index of cognitive impairment.

[0087] The storage and display module is used for storing the multimodal neural indicators of cognitive impairment and visualizing the multimodal neural indicators of cognitive impairment.

[0088] Embodiment 2

[0089] In this embodiment, in order to verify the effectiveness and implementation effect of this application, the following scientific experiments were conducted, specifically including:

[0090] (1) Subject enrollment: 45 Parkinson's disease patients (21 males, 24 females, 62.11±6.95 years old) were invited to participate in the experiment, and all subjects signed the informed consent before enrollment. The inclusion criteria are as follows: 1) Hoehn-Yahr stage ≤ 3; 2) stable vital signs, no obvious cardiopulmonary or osteoarthritis diseases; 3) stable medication, no adjustment in the past 3 months; 4) no other special treatment required during hospitalization; 5) no deep brain stimulation or implantation therapy; 6) able to understand the informed consent form, willing to sign, and committed to complete the evaluation and treatment. Exclusion criteria are: 1) patients with fractures or psychiatric symptoms; 2) patients with clinical dementia score > 0.5 or visual or hearing impairment; 3) severe resting tremor; 4) the presence of metal implants in the body, such as pacemakers and brain pacemakers; 5) history of epilepsy; 6) other serious systemic diseases.

[0091] All enrolled subjects were classified according to the diagnostic criteria for Parkinson's disease mild cognitive impairment level 1 developed by the Movement Disorders Society Working Group. Parkinson's disease patients with Montreal Cognitive Assessment (MoCA) scores ≤ 25 were divided into the mild cognitive impairment group (22 cases), and Parkinson's disease patients with MoCA scores > 25 were divided into the normal cognitive group (23 cases).

[0092] (2) Experimental process: The judgment system of the present invention is used to complete transcranial magnetic and electroencephalographic joint acquisition, synchronous signal preprocessing, multi-feature extraction, and intelligent fusion of disease features for all Parkinson's disease patients.

[0093] (3) Experimental results: Figure 3 As shown, Figure 3 (A) The average butterfly diagram of the TMS-evoked potential waveforms from all electrode channels in the PD-NC group (left) and the PD-MCI group (right). The EEG neural activity feature components such as P30 and N45 used to construct the fusion feature machine learning model have been marked in the figure; Figure 3 (B) Time-frequency diagram of the mean power of whole-brain oscillations induced by transcranial magnetic stimulation in the PD-NC group (left) and the PD-MCI group (right). The dotted boxes highlight the characteristic components of EEG neural activity such as alpha of 0-100ms and theta of 200-300ms used to construct the fusion feature machine learning model; Figure 3 (C) The statistical difference of brain network functional connectivity between the two groups constructed by weighted phase lag index in the Theta and Alpha frequency bands. The bars mark the electrode connections with significant differences, which will be used as the characteristic components of EEG neural activity to build the fusion feature machine learning model.

[0094] like Figure 3As shown, in terms of cortical activity, the P30, P60, P180, N45, and N100 components of time-domain TEP of PD-NC and PD-MCI patients were significantly different, and the cortical activity of PD-MCI was stronger; the time-frequency results showed that PD-NC and PD-MCI patients had significant differences in θ (4-7Hz) and α (8-13Hz) oscillations; brain network results showed that the network connections of the frontal lobe, parietal lobe, and central area of ​​PD-NC and PD-MCI patients were significantly different. The above results can provide effective potential multi-dimensional neural indicators for the early identification of cognitive impairment in clinical Parkinson's disease.

[0095] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. An early diagnosis system for cognitive impairment in Parkinson's disease based on multi-dimensional features of transcranial magnetic and electroencephalographic (TEE), characterized in that: include: Transcranial magnetic and electroencephalogram joint acquisition module, synchronous signal preprocessing module, multi-dimensional feature extraction module, feature intelligent fusion module and storage and display module; The transcranial magnetic and electroencephalogram (EEG) joint acquisition module is used to disturb the patient's neural activity through single-pulse transcranial magnetic stimulation encoding and acquire evoked EEG signals; The synchronization signal preprocessing module is used to perform denoising filtering on the evoked EEG signal to obtain a processed evoked EEG signal; The multi-dimensional feature extraction module uses multi-dimensional EEG feature analysis to extract features from the processed evoked EEG signal to obtain multi-parameter EEG neural activity features, wherein the multi-parameter EEG neural activity features include: time domain features, time-frequency domain features, and brain network multi-dimensional brain neural features; The feature intelligent fusion module is used to construct a fusion feature machine learning model, and use the fusion feature machine learning model to perform weight assignment and feature fusion on the multi-parameter EEG neural activity features to obtain a multimodal neural index of cognitive impairment; The storage and display module is used to store the cognitive impairment multimodal neural index and visualize the cognitive impairment multimodal neural index.

2. According to claim 1, a system for early diagnosis of cognitive impairment in Parkinson's disease based on multi-dimensional features of transcranial magnetic and electroencephalographic (TEE) characteristics: The transcranial magnetic and electroencephalogram (EEG) joint acquisition module includes: a resting motor threshold measurement unit, a transcranial magnetic stimulation unit and an EEG acquisition unit; The resting motor threshold measurement unit is used to apply transcranial magnetic stimulation to the primary motor cortex of the affected side of the patient in a 45° front-to-back direction to induce muscle contraction on the opposite side to generate motor evoked potentials, and the minimum intensity required to induce an electromyographic peak-to-peak amplitude greater than or equal to 50 μV in at least 5 of 10 consecutive stimulations is used as the resting motor threshold; The transcranial magnetic stimulation unit is used to apply 80 single pulses to the right posterior parietal cortex of the patient in a 45° front-to-back direction, with each single pulse stimulation interval being 2 to 4 seconds, and the pulse intensity being set to 90% of the resting motor threshold; The EEG acquisition unit acquires the evoked EEG signal at a sampling rate of 1000 Hz.

3. According to claim 1, a system for early diagnosis of cognitive impairment in Parkinson's disease based on multi-dimensional features of transcranial magnetic and electroencephalographic (TEE) characteristics: The workflow of the synchronization signal preprocessing module includes: Importing the evoked EEG signal and transcranial magnetic stimulation electrode information, locating electrodes based on the electrode information, and deleting useless electrodes; Extracting the events of interest in the evoked EEG signal and locking the corresponding evoked EEG signal, and then performing baseline correction to remove transcranial magnetic stimulation pulse artifacts and induced electromyographic artifacts; Performing a first missing data interpolation around a transcranial magnetic stimulation pulse to minimize artifacts caused by a low-pass and anti-aliasing filter, then replacing the interpolated data of the first missing data interpolation with the first constant amplitude EEG signal data, and then using independent component analysis and automatic component selection to remove voltage decay artifacts caused by scalp muscle activation and transcranial magnetic stimulation-induced electromyography in the evoked state EEG signal; The evoked EEG signal within 2 ms before and 10 ms after the transcranial magnetic stimulation pulse is deleted, a second missing data interpolation is performed around the transcranial magnetic stimulation pulse, and a 1-100 Hz band-pass filter and a 48-52 Hz band-stop filter are performed, and then the interpolated data of the second missing data interpolation are replaced with the second constant EEG signal amplitude data, and then independent component analysis and automatic component selection are used again to remove the non-transcranial magnetic stimulation locked artifacts remaining in the evoked EEG signal; A third missing data interpolation is performed around the transcranial magnetic stimulation pulse, the electrode channel CPz of the missing EEG signal data is interpolated, and the electrode information is re-referenced to a common average reference to obtain the processed evoked EEG signal.

4. According to claim 1, a system for early diagnosis of cognitive impairment in Parkinson's disease based on multi-dimensional features of transcranial magnetic and electroencephalographic (TEE) characteristics: The multidimensional feature extraction module includes: a time domain feature extraction unit, a time-frequency domain feature extraction unit and a brain network feature extraction unit; The time domain feature extraction unit is used to extract the global average field power and transcranial magnetic stimulation evoked potential of each electrode at each time point in the processed evoked state EEG signal; The time-frequency domain feature extraction unit is used to extract the average power of each frequency band of each electrode channel of the whole brain of the processed evoked EEG signal within a certain time period; The brain network feature extraction unit is used to extract the weighted phase lag index matrix and brain network attribute graph theory parameters of the processed evoked state electroencephalogram signal.

5. According to claim 4, a system for early diagnosis of cognitive impairment in Parkinson's disease based on multi-dimensional features of transcranial magnetic and electroencephalographic (TEE) characteristics: The workflow of the time domain feature extraction unit includes: The global mean field power is calculated by evaluating the standard deviation of the EEG signal amplitude of each electrode at each time point in the processed evoked EEG signal: Among them, n represents the total number of channels of EEG signals, i represents the number of channels of EEG signals, and A i (t) represents the amplitude of the EEG signal of the ith channel, A mean (t) represents the average amplitude of all channels at time t; The peak values ​​of the transcranial magnetic stimulation evoked potential of each EEG signal channel in the five time windows of interest in the evoked EEG signal after the processing are extracted as the transcranial magnetic stimulation evoked potential, which are P30, N45, P60, N100 and P180 respectively: Where x represents each subject, It represents the average potential peak value of the event in all channels of all subjects during the period from t1 to t2. P_tep represents the positive potentials P30, P60 and P180, and N_tep represents the negative potentials N45 and N100.

6. According to claim 4, a system for early diagnosis of cognitive impairment in Parkinson's disease based on multi-dimensional features of transcranial magnetic and electroencephalographic (TEE) characteristics: The workflow of the time-frequency domain feature extraction unit includes: Morlet wavelet convolution is used to perform time-frequency decomposition of the five frequency bands of δ, θ, α, β and γ in the processed evoked EEG signal, extract the signal amplitude, and obtain the energy of specific time-frequency points. Then, the time window is divided into intervals of 100 ms, and the average power of each frequency band of each electrode channel in the whole brain in the processed evoked EEG signal during the period of 0 to 400 ms after transcranial magnetic stimulation is calculated.

7. According to claim 4, a system for early diagnosis of cognitive impairment in Parkinson's disease based on multi-dimensional features of transcranial magnetic and electroencephalographic (TEE) characteristics: The workflow of the brain network feature extraction unit includes: The phase information of each electrode in the processed evoked EEG signal is extracted based on the time-frequency matrix after time-frequency decomposition, and the phase matrix is ​​subtracted to obtain the phase difference matrix, and the cross spectrum between electrodes is obtained. The weighted phase lag index matrix is ​​calculated based on the cross spectrum: Among them, Im[·] represents the imaginary part, G i，j (t, f) represents the cross spectrum of two time series of the channel of the i-th EEG signal and the channel of the j-th EEG signal, t represents time, and f represents frequency; The characteristics of the brain network are quantitatively described using graph theory methods, and the graph theory parameters of the brain network attributes are calculated. The graph theory parameters of the brain network attributes include: clustering coefficient, average shortest path length, global efficiency and local efficiency: Where CC represents the clustering coefficient, ASPL represents the average shortest path length, GE represents the global efficiency, LE represents the local efficiency, N represents the number of all nodes in the brain network, and h a is the number of nodes adjacent to node a, f a Yes a The number of actual connecting edges between nodes, l ab is the shortest path length between node a and node b, G a is a subgraph consisting of the adjacent nodes of node a, l bh is the shortest path length from node b to node h.

8. According to claim 1, a system for early diagnosis of cognitive impairment in Parkinson's disease based on multi-dimensional features of transcranial magnetic and electroencephalographic (TEE) characteristics: The feature intelligent fusion module includes: a data processing unit, a feature selection unit, a weight analysis unit, a model building unit and a weighted fusion unit; The data processing unit is used to pre-process the time domain features, the time-frequency domain features and the brain network multi-dimensional cranial nerve features to obtain processed features; The feature selection unit uses an elastic network regression feature selection method to determine the optimal feature selection parameters; The weight analysis unit is used to assign weights to the time domain features, the time-frequency domain features and the multi-dimensional cranial nerve features of the brain network in the optimal feature selection parameters to obtain weight proportions of different features; The model building unit builds a multi-core support vector machine fusion model by combining different kernel functions to obtain the fusion feature machine learning model; The weighted fusion unit uses the fusion feature machine learning model to perform feature matching between different kernel functions and features of different dimensions, and fuses the features of different kernel functions based on the weight proportions to obtain the multimodal neural index of cognitive impairment.

Images