Auxiliary decision-making system of transcranial magnetic stimulation mode

Abstract

The invention discloses an auxiliary decision-making system of a transcranial magnetic stimulation mode, and the system comprises an obtaining module which is used for obtaining the personal feature information of a patient; the determination module is used for calling a preset evaluation model to calculate the brain function retention degree of the patient by adopting the personal characteristic information, and determining the treatment mode of the patient according to the brain function retention degree; wherein the preset evaluation model is a classification model constructed based on demographic characteristic parameters, behavioral evaluation characteristic parameters, neural electrophysiological characteristic parameters and brain image characteristic parameters as independent variables. The brain function retention degree is calculated by using the model, and the transcranial magnetic stimulation mode is determined according to the brain function retention degree without relying on subjective calculation of a doctor, so that the accuracy of determining the treatment mode can be improved; meanwhile, multiple evaluation, detection and calculation are not needed for processing, the processing time can be shortened, the processing efficiency and precision can be further improved, the diagnosis and subsequent treatment of the patient are prevented from being affected, and precise treatment is achieved.

Description

Technical Field

[0001] This invention relates to the technical field of brain function preservation classification, and more particularly to a transcranial magnetic stimulation modality-assisted decision-making system. Background Technology

[0002] Brain injuries include concussion, diffuse axonal injury, cerebral contusion and laceration, brainstem injury, and central nervous system injury. For acute, focal, or diffuse central nervous system injuries, one effective treatment is transcranial magnetic stimulation (TMS), which can effectively promote functional recovery in patients.

[0003] Transcranial magnetic stimulation (TMS) is divided into excitatory and inhibitory modes. The degree of brain function preservation is an important indicator reflecting the residual brain function of patients after the onset of the disease, and can be used as an important basis for selecting the TMS treatment mode. For patients with high brain function preservation, the inhibitory TMS treatment mode can be selected, while for patients with low brain function preservation, the excitatory TMS treatment mode is selected.

[0004] Currently, the common practice relies on the experience and judgment of clinicians, who determine the degree of brain function preservation in patients based on their personal experience, and then select the appropriate treatment mode. However, this approach has the following technical problems: different clinicians have different experience and understanding, resulting in inconsistent judgments. This not only lacks objective evaluation standards, leading to deviations in judgment accuracy, but also requires doctors to make judgments one by one each time, involving a large amount of information and taking a long time, thus delaying the patient's treatment cycle. Summary of the Invention

[0005] This invention proposes an auxiliary decision-making system based on transcranial magnetic stimulation (TMS) mode. This system can solve the technical problems of existing technologies, which delay patient diagnosis and subsequent treatment due to large assessment biases and long processing times.

[0006] A first aspect of the present invention provides an auxiliary decision-making system for transcranial magnetic stimulation (TMS) patterns, the system comprising:

[0007] The acquisition module is used to acquire the patient's personal characteristic information;

[0008] The determination module is used to call a preset assessment model to calculate the patient's brain function retention using the personal characteristic information, and determine the patient's transcranial magnetic stimulation mode based on the brain function retention.

[0009] The preset evaluation model is a classification model obtained by constructing a training set based on demographic characteristic parameters, behavioral evaluation characteristic parameters, neurophysiological characteristic parameters, and brain imaging characteristic parameters as independent variables, and using the training set for training operations.

[0010] The training operation includes: normalizing the feature parameters of the training set using the Z-transform method to obtain normalized parameters; filtering the normalized parameters using the P-value test method to obtain filtered parameters; and training the model using the filtered parameters.

[0011] In conjunction with the first aspect, in one implementation, the construction process of the preset evaluation model includes the following steps:

[0012] Obtain a parameter set, which includes: demographic characteristic parameters, behavioral assessment characteristic parameters, neurophysiological characteristic parameters, and brain imaging characteristic parameters;

[0013] The parameter set is divided into a training set and a test set by using the leave-one-out method;

[0014] The preset support vector machine is trained using the training set, and the preset support vector machine is cross-validated using the test set to obtain the preset evaluation model.

[0015] The default kernel function of the support vector machine is a polynomial kernel, and the default optimal hyperparameters of the support vector machine are constructed using a grid search method.

[0016] In conjunction with the first aspect, in one implementation, the neurophysiological characteristic parameters include: TMS assessment index values;

[0017] The operation of obtaining the neurophysiological characteristic parameters includes the following steps:

[0018] Acquire the patient's electromyographic signals;

[0019] Based on the electromyographic signals, the resting movement threshold, active movement threshold, and interhemispheric inhibition value are determined to obtain TMS evaluation index values.

[0020] In conjunction with the first aspect, in one implementation, the neurophysiological characteristic parameters include: electroencephalogram (EEG) index values;

[0021] The operation of obtaining the neurophysiological characteristic parameters includes the following steps:

[0022] Collecting electroencephalogram (EEG) signals;

[0023] The power spectral density value, the EEG phase lag index value, and the potential distribution parameters are calculated using the EEG signals to obtain the EEG index values.

[0024] In conjunction with the first aspect, in one implementation, the power spectral density value is calculated as follows:

[0025]

[0026] In the above formula, PSD[k] represents the power spectral density value, X[k] represents the discrete Fourier transform coefficient of the k-th frequency point, and N is the number of data points of the EEG signal;

[0027] The discrete Fourier transform coefficients X[k] are calculated as follows:

[0028]

[0029] In the above formula, x[n] is the EEG signal at the nth time point, and j is the imaginary unit;

[0030] The EEG phase lag index value is calculated as follows:

[0031]

[0032] In the above formula, PLI represents the EEG phase lag index, t represents the phase difference at time t, and T is the total number of time points.

[0033] Wherein, the phase difference at time t The calculation is shown in the following formula:

[0034]

[0035] In the above formula, Im represents the imaginary part, Re represents the real part, and x(t)y * (t) represents the complex conjugate product of signals x(t) and y(t);

[0036] The potential distribution parameters are calculated as follows:

[0037] Φ = KJ + ε;

[0038] In the above formula, Φ represents the potential distribution measured on the scalp, K represents the forward field matrix, J represents the current density distribution in the source space, and ε is the measurement noise.

[0039] In conjunction with the first aspect, in one implementation, the brain imaging feature parameters include: diffusion tensor imaging index values;

[0040] Obtaining the brain imaging feature parameters includes the following steps:

[0041] Acquire echo plane imaging;

[0042] The echo-planar imaging was used to determine the brain region activation statistical parameter map, ReHo map, imaging function matrix and anisotropy fraction value, respectively, and the diffusion tensor imaging index value was obtained.

[0043] In conjunction with the first aspect, in one implementation, the brain imaging feature parameters include: functional near-infrared index values;

[0044] The operation of obtaining the brain imaging feature parameters includes the following steps:

[0045] Acquire functional near-infrared spectral signals;

[0046] The near-infrared blood oxygen concentration, near-infrared contralateral index, and near-infrared functional connectivity values ​​are calculated using the functional near-infrared spectral signals to obtain functional near-infrared index values.

[0047] In conjunction with the first aspect, in one implementation, the near-infrared blood oxygen concentration value is calculated as follows:

[0048]

[0049] In the above formula, Δc represents the near-infrared blood oxygen concentration, ΔOD represents the change in optical density, ∈ represents the absorption coefficient, DPF represents the diffuse path length factor, and L represents the optical path length.

[0050] The optical density is calculated as follows:

[0051]

[0052] In the above formula, OD represents optical density, I represents the light intensity after passing through brain tissue, and I0 represents the initial light intensity;

[0053] The near-infrared contralateral index value is calculated as follows:

[0054] LI = (ci) / (c+i);

[0055] In the above formula, LI represents the near-infrared contralateral index value, c represents the oxyhemoglobin concentration in the healthy hemisphere, and i represents the oxyhemoglobin concentration in the stroke hemisphere.

[0056] The near-infrared functional connectivity value is calculated as follows:

[0057]

[0058] In the above formula, R z R represents the near-infrared functional connectivity value, and R represents the Pearson correlation coefficient of the electrode pair.

[0059] In conjunction with the first aspect, in one implementation, the brain imaging feature parameters include: functional magnetic resonance imaging (fMRI) index values;

[0060] The operation of obtaining the brain imaging feature parameters includes the following steps:

[0061] Acquire nuclear magnetic resonance (NMR) scan images;

[0062] The nuclear magnetic resonance scan images are converted to NIFTI format and then preprocessed to obtain functionally processed images.

[0063] Based on the functionally processed images, the MRI blood oxygen concentration value, the MRI contralateral index value, and the MRI functional connectivity value are calculated to obtain the functional MRI index values.

[0064] In conjunction with the first aspect, in one implementation, the image preprocessing includes: removal of time points, temporal layer correction, head motion correction, registration and spatial normalization, delinear drift removal, low-frequency filtering, and smoothing.

[0065] Compared to existing technologies, the transcranial magnetic stimulation (TMS) modality auxiliary decision-making system provided in this invention offers the following advantages: The invention acquires the patient's personal characteristic information via an acquisition module; then, a determination module calls a preset assessment model to calculate the patient's brain function retention using this personal characteristic information, and determines the patient's TMS modality based on this retention. Since the preset assessment model is a classification model constructed based on demographic parameters, behavioral assessment parameters, neurophysiological parameters, and brain imaging parameters as independent variables, it can calculate the brain function retention and determine the TMS modality accordingly, eliminating reliance on subjective calculations and judgments by doctors, thereby improving the accuracy of modality determination. Simultaneously, the processing eliminates the need for multiple assessment and calculation calculations, shortening processing time and further improving efficiency and accuracy, avoiding impact on patient diagnosis and subsequent treatment, and achieving precision medicine. Attached Figure Description

[0066] Figure 1 This is a schematic diagram of the structure of an auxiliary decision-making system based on transcranial magnetic stimulation mode according to an embodiment of the present invention;

[0067] Figure 2 This is a flowchart of the model training operation provided in an embodiment of the present invention;

[0068] Figure 3 This is a schematic diagram illustrating the process of establishing a transcranial magnetic stimulation mode-assisted decision-making system according to an embodiment of the present invention. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] Brain injuries include concussion, diffuse axonal injury, cerebral contusion and laceration, brainstem injury, and central nervous system injury. For acute, focal, or diffuse central nervous system injuries, one effective treatment is transcranial magnetic stimulation (TMS), which can effectively promote functional recovery in patients.

[0071] Transcranial magnetic stimulation (TMS) is divided into excitatory and inhibitory modes. The degree of brain function preservation is an important indicator reflecting the residual brain function of patients after the onset of the disease, and can be used as an important basis for selecting the TMS treatment mode. For patients with high brain function preservation, the inhibitory TMS treatment mode can be selected, while for patients with low brain function preservation, the excitatory TMS treatment mode is selected.

[0072] Currently, the common practice relies on the experience and judgment of clinicians, who determine the degree of brain function preservation in patients based on their personal experience, and then select the appropriate treatment mode. However, this approach has the following technical problems: different clinicians have different experience and understanding, resulting in inconsistent judgments. This not only lacks objective evaluation standards, leading to deviations in judgment accuracy, but also requires doctors to make judgments one by one each time, involving a large amount of information and taking a long time, thus delaying the patient's treatment cycle.

[0073] To address the aforementioned issues, the following detailed embodiments will be used to illustrate and explain a transcranial magnetic stimulation modal auxiliary decision-making system provided in this application.

[0074] Reference Figure 1 The diagram shows a schematic representation of a transcranial magnetic stimulation modal auxiliary decision-making system provided in an embodiment of the present invention.

[0075] In one embodiment, the transcranial magnetic stimulation (TMS) modal decision support system can be installed on a smart device, including a computer or mobile terminal. This system can assist doctors in selecting appropriate treatment plans.

[0076] As an example, the decision support system for the transcranial magnetic stimulation mode may include:

[0077] Module 101 is used to acquire the patient's personal characteristic information;

[0078] The determination module 102 is used to call a preset evaluation model to calculate the patient's brain function retention using the personal characteristic information, and determine the patient's transcranial magnetic stimulation mode based on the brain function retention.

[0079] The pre-defined assessment model is a classification model constructed based on demographic characteristic parameters, behavioral assessment characteristic parameters, neurophysiological characteristic parameters, and imaging characteristic parameters as independent variables.

[0080] In one embodiment, the patient's personal characteristic information can be acquired. Specifically, the doctor can input the patient's personal characteristic information on a mobile terminal. The acquisition module of the transcranial magnetic stimulation (TMS) mode auxiliary decision-making system receives the personal characteristic information, and the determination module of the TMS mode auxiliary decision-making system can call a preset assessment model to calculate the patient's brain function retention rate using the personal characteristic information, and determine the patient's treatment mode based on the brain function retention rate.

[0081] In practice, personal characteristics information may include the patient's demographic characteristics, behavioral assessment characteristics, neurophysiological characteristics, and / or brain imaging characteristics.

[0082] Demographic information may include age, sex, stroke location, and disease course.

[0083] Behavioral assessment features may include NIHSS (National Institute of Health Stroke Scale), BI (Barthel Index), MRS (Modified Rankin Scale), FM (Fast Moving Average) score, ARAT (Action Research Arm Test) score, etc.

[0084] Neurophysiological characteristics can include: electroencephalogram (EEG) signals and electromyogram (EMG) signals.

[0085] Brain imaging features can include: diffusion tensor imaging index values, near-infrared spectral signals, functional magnetic resonance imaging (fMRI) information, etc.

[0086] In one embodiment, the preset assessment model is a classification model constructed based on demographic characteristic parameters, behavioral assessment characteristic parameters, neurophysiological characteristic parameters, and brain imaging characteristic parameters as independent variables.

[0087] The pre-set assessment model can calculate the patient's brain function retention rate based on individual characteristics. It should be noted that the brain function retention rate is an important indicator reflecting the patient's residual brain function after the onset of the disease. It can be used as an important basis for selecting the transcranial magnetic stimulation mode. That is, for patients with a high brain function retention rate, the inhibitory transcranial magnetic stimulation mode can be determined, while for patients with a low brain function retention rate, the excitatory transcranial magnetic stimulation mode is determined.

[0088] After calculation, the preset assessment model can determine whether the degree of brain function preservation is greater than a preset threshold. If so, the patient is identified as having a high degree of brain function preservation, and the treatment mode can be determined as inhibitory transcranial magnetic stimulation (TMS). Conversely, if the degree of preservation is less than a preset threshold, the treatment mode can be determined as excitatory TMS. TMS, also known as transcranial magnetic stimulation (TMS) therapy, is a new technology for treating post-stroke functional impairment and can effectively promote functional recovery in stroke patients. The determined inhibitory and excitatory TMS modes can then be displayed on a smart terminal for doctors' reference. By directly referring to the mode recommended by the TMS auxiliary decision-making system, doctors can not only standardize the assessment mode and improve its accuracy but also shorten processing time and improve diagnostic efficiency.

[0089] To improve the processing accuracy of the model, the construction process of the preset evaluation model, as an example, may include the following sub-steps:

[0090] The first step is to obtain a set of parameters, which includes: demographic characteristic parameters, behavioral assessment characteristic parameters, neurophysiological characteristic parameters, and brain imaging characteristic parameters.

[0091] The second step is to use the leave-one-out method to divide the parameter set into a training set and a test set.

[0092] The third step is to train the preset support vector machine using the training set and to cross-validate the preset support vector machine using the test set to obtain the preset evaluation model.

[0093] The default kernel function of the support vector machine is a polynomial kernel, and the default optimal hyperparameters of the support vector machine are constructed using a grid search method.

[0094] In one embodiment, demographic parameters may include age, sex, stroke location, and disease duration. Behavioral assessment parameters may include the NIHSS (National Institute of Health Stroke Scale), Barthel Index (BI), Modified Rankin Scale (MRS), FM (Fast Moving Average) score, and Action Research Arm Test (ARAT) score. Neurophysiological parameters may include electroencephalogram (EEG) and electromyography (EMG) signals. Brain imaging parameters may include diffusion tensor imaging parameters, near-infrared spectral signals, and functional magnetic resonance imaging (fMRI) information.

[0095] This invention can call a model to calculate the degree of brain function preservation in stroke patients based on their personal characteristics, and then determine the corresponding treatment mode based on the degree of brain function preservation. This eliminates the need for subjective calculations by doctors, thereby improving the accuracy of determining the treatment mode. At the same time, the process does not require multiple assessment and testing calculations, which can shorten the processing time and further improve the efficiency and accuracy of the process, avoiding affecting the patient's diagnosis and subsequent treatment, and achieving precision treatment.

[0096] Reference Figure 2-3 The diagrams show the operation flowchart of model training provided in one embodiment of the present invention and the schematic diagram of the establishment process of a transcranial magnetic stimulation mode auxiliary decision-making system provided in one embodiment of the present invention.

[0097] After obtaining the parameter set, a classification model is constructed using the parameter set as the independent variable and the subject group as the dependent variable, thus obtaining the preset evaluation model.

[0098] During training, the leave-one-out method can be used to divide the parameter set into a training set and a test set. Then, the preset support vector machine is trained and cross-validated sequentially using the training set and the test set to obtain the preset evaluation model.

[0099] It should be noted that the default Support Vector Machine (SVM) kernel function is a multinomial kernel. The optimal hyperparameters for constructing the SVM—the penalty factor C and the kernel function parameter gamma—are explored using the grid search method. Their values ​​are set to a range of 2-10-210, and the search process is set to 5-fold cross-validation.

[0100] Meanwhile, during training, the leave-one-out method can be used to divide the data into training and test sets for cross-validation. The area under the curve (AUC) of the receiver operating characteristic curve (ROC) and the classification accuracy are used to judge the effectiveness of the model.

[0101] In subsequent use, the model's judgment results can be used to classify patients into those with high and low brain function preservation, thereby determining the corresponding transcranial magnetic stimulation (TMS) mode. For patients with high brain function preservation, an inhibitory TMS mode can be determined, while for those with low brain function preservation, an excitatory TMS mode can be determined, and then recommended to the doctor.

[0102] In one embodiment, the neurophysiological characteristic parameters include: TMS assessment index values;

[0103] As an example, the operation of obtaining the neurophysiological characteristic parameters may include the following sub-steps:

[0104] S21. Obtain the patient's electromyographic signals.

[0105] S22. Based on the electromyographic signals, determine the resting movement threshold, active movement threshold, and interhemispheric inhibition value to obtain the TMS evaluation index value.

[0106] In practice, electromyographic signals can be applied using a figure-eight coil connected to a transcranial magnetic stimulator (e.g., NS5000 transcranial magnetic stimulator), with a maximum magnetic field strength of 2.5T.

[0107] Disposable Ag / AgCl electrodes were attached to the first digital interosseous (FDI) muscle to record electromyographic (EMG) signals, with a reference electrode placed at the radial process. The EMG signals were recorded using an EE-410eego acquisition amplifier (Eemagine Medical Imaging Solutions GmbH, Germany), with a sampling rate of 2048 Hz, a filtering range of 3-3000 Hz, and a 50 Hz notch filter to remove power line interference. TMS stimulation was applied using a figure-eight coil, positioned at the projection of the primary motor cortex onto the body surface, tangent to the scalp, with the handle pointing posteriorly at a 45-degree angle to the midline.

[0108] Next, the resting movement threshold, active movement threshold, and interhemispheric inhibition value can be calculated based on the electromyographic signals to obtain the TMS evaluation index values.

[0109] The resting motor threshold (RMT) can be determined using a neuronavigation system (Visor2, ANT Neuro, Netherlands). The initial stimulation point is positioned at the cortical representation of the hand. The stimulation intensity is then adjusted in 2% increments, starting at 80% of the maximum stimulator output (MSO), until a motor evoked potential (MEP) of approximately 50 μV is induced at the first dorsal interosseous muscle. The coil is then moved 10-15 mm parallel to the marked point, maintaining its angle and scalp contact. Three to four single-pulse TMS stimulations are applied with an interval of at least 5 seconds, and the amplitude of the induced MEP is observed. This process is repeated until the scalp site that elicits the maximum MEP at the minimum stimulation intensity is found; this is the motor hotspot, and it is marked on the neuronavigation system. The resting motor threshold is determined by applying 10 single-pulse TMS stimulations to the motor hotspot. The resting motion threshold was also determined using relative frequency, defined as the minimum intensity of at least five out of ten stimuli sufficient to cause the MEP to reach an interpeak amplitude of >50 μV when the subject is at rest.

[0110] The active motor threshold (AMT) can be determined using the same relative frequency and is defined as the minimum intensity of at least five out of ten stimuli sufficient to cause the MEP to reach an interpeak amplitude of >200 μV while the subject is in a state of muscle contraction (approximately 20% of maximum muscle strength).

[0111] Inter-hemispheric inhibition (IHI) can be used to determine bilateral motor hotspots and resting motor thresholds. One TMS coil is placed on the healthy side's motor cortex hotspot (conditioning stimulus, CS), and the other on the affected side (test stimulus, TS). The pulse interval is set to 10 ms. The stimulation intensities of CS and TS are set to 120% of the healthy side's resting motor threshold and the 1 mV MEP threshold (i.e., the minimum stimulation intensity required to induce 1 mV MEP) on the affected side, respectively. TS is stimulated alone 15 times as a baseline, followed by 15 CS-TS paired stimulations, with pulse intervals >5 s. The percentage difference between the average MEP amplitude induced by CS-TS paired stimulation and the average MEP amplitude induced by TS alone is defined as the inter-hemispheric inhibition level.

[0112] In an optional embodiment, after calculating the resting motor threshold, active motor threshold, and interhemispheric inhibition value, the changes in excitability of the patient's motor cortex can be revealed by comparing the motor thresholds (including static and dynamic) between the affected and healthy sides. If the resting and active motor thresholds on the affected side are significantly higher than those on the healthy side, it indicates a significant decrease in the excitability of the motor cortex, suggesting a lower degree of brain function preservation; conversely, if the resting and active motor thresholds on the affected side are close to the levels of the healthy side, it indicates that the motor cortex still maintains a relatively intact functional state. A significant decrease in the level of interhemispheric inhibition indicates a weakened neural regulatory effect of the healthy side on the affected side, which is also an important indicator of decreased brain function preservation.

[0113] In one embodiment, the neurophysiological characteristic parameters include: electroencephalogram (EEG) index values;

[0114] As an example, the operation of obtaining the neurophysiological characteristic parameters may include the following sub-steps:

[0115] S31. Collect EEG signals.

[0116] S32. Calculate the power spectral density value, brain phase lag index value and potential distribution parameters using the brain signals to obtain brain wave index values.

[0117] In practice, the patient's electroencephalogram (EEG) signals can be collected using the Dutch ANT64 EEG recording system. The electrode placement conforms to the international 10-20 standard distribution. Reference electrodes are placed bilaterally on the mastoid process, and the resistance of all electrodes is less than 5kΩ. The EEG signal sampling rate is 2048Hz, with a filtering range of 0.1Hz-100Hz. During EEG signal acquisition, resting-state EEG data is collected, and various indicators are calculated.

[0118] Next, the power spectral density value, the EEG phase lag index value, and the potential distribution parameters can be calculated using the EEG signals to obtain the neurophysiological characteristic parameters.

[0119] The power spectral density value is calculated as follows:

[0120]

[0121] In the above formula, PSD[k] represents the power spectral density value, X[k] represents the discrete Fourier transform coefficient of the k-th frequency point, and N is the number of data points of the EEG signal;

[0122] The discrete Fourier transform coefficients X[k] are calculated as follows:

[0123]

[0124] In the above formula, x[n] is the EEG signal at the nth time point, and j is the imaginary unit, representing the square root of -1;

[0125] Power spectral density (PSD) values ​​can be obtained by converting EEG signals into frequency domain signals using the Fast Fourier Transform (FFT), a process known as spectral analysis. The Fast Fourier Transform is an efficient algorithm for estimating the spectral power and frequency composition of EEG signals. It works by decomposing the time-domain EEG waveform into a series of sine and cosine waves with different frequencies and amplitudes. By squaring these spectral amplitudes, the signal power at each frequency can be quantified, thus generating the power spectral density (PSD).

[0126] Specifically, the Fast Fourier Transform (FFT) decomposes a continuous EEG signal time series into N discrete samples. Each EEG data point is treated as a complex number. Then, N unique sine waves are constructed, with frequencies ranging from 0 to the Nyquist frequency (i.e., half the sampling rate). The amplitude and phase of each complex sine are calculated so that their sum best matches the actual EEG signal when combined.

[0127] The EEG phase lag index can be used as a measure of functional connectivity (FC), specifically as an indicator to quantify this connectivity. PLI is calculated by assessing the asymmetry in the phase lag distribution between two signals. Specifically, for each pair of electrode channels, the instantaneous phase of each electrode is obtained using a Hilbert transform. The phase difference at each time point is calculated, and the proportion of time points with a phase difference greater than zero is counted. This is then normalized using the total number of time points, resulting in a value between 0 and 1, where 0 indicates that the two signals are completely out of sync, and 1 indicates that the two signals are completely synchronized.

[0128] Specifically, the EEG phase lag index value is calculated as follows:

[0129]

[0130] In the above formula, PLI represents the EEG phase lag index, t represents the phase difference at time t, and T is the total number of time points.

[0131] Wherein, the phase difference at time t The calculation is shown in the following formula:

[0132]

[0133] In the above formula, Im represents the imaginary part, Re represents the real part, and x(t)y * (t) represents the complex product of the complex conjugate of signals x(t) and y(t).

[0134] Source location (SL) analysis can be performed on the potential distribution parameters. Specifically, the fieldtrip toolkit in MATLAB can be used to perform source location analysis for low-resolution electromagnetic tomography (LORETA). LORETA is a source location method based on a distributed source model. It assumes that the cerebral cortex is composed of a large number of densely distributed electric dipoles. The orientation of these dipoles is perpendicular to the cortical surface and is constrained by the minimum norm, thereby estimating the current density distribution of these dipoles and realizing the conversion of EEG signals from sensor space to source space.

[0135] The potential distribution parameters are calculated as follows:

[0136] Φ = KJ + ε;

[0137] In the above formula, Φ represents the potential distribution measured on the scalp, K represents the forward field matrix, which describes the mapping relationship from the source space to the sensor space, J represents the current density distribution in the source space, and ε is the measurement noise.

[0138] The optimization of current density in the source space can be achieved using the following method:

[0139] min||Φ-KJ|| 2 ,subject to||J|| 2 <λ

[0140] Where λ is the regularization parameter, the optimization objective is to minimize the sum of squared reconstruction errors, so that the estimated scalp potential distribution KJ is as close as possible to the actual measured potential distribution Φ.

[0141] Optionally, after calculating the power spectral density, EEG phase lag index, and potential distribution parameters, if the spectral power on the affected side is significantly lower than that on the healthy side, it indicates weakened brain activity and suggests lower brain function retention. When using PLI to quantify FC, if the synchronicity between the affected brain region and other regions is significantly reduced, it indicates weakened neural network synergy, which is also an important indicator of decreased brain function retention. In addition, using LORETA for SL can assess the current density distribution in the affected brain region; a significant decrease in current density indicates weakened neuronal activity, further corroborating the decline in brain function retention.

[0142] In one embodiment, the brain imaging feature parameters include: diffusion tensor imaging index values;

[0143] As an example, obtaining the brain image feature parameters may include the following sub-steps:

[0144] S41. Obtain echo plane imaging.

[0145] S42. Using the echo plane imaging, determine the brain region activation statistical parameter map, ReHo map, imaging function matrix and anisotropy fraction value respectively, and obtain the diffusion tensor imaging index value.

[0146] In one operating mode, echo plane imaging (EPI) can be used to acquire an echo plane imaging image sequence, including: TR / TE = 12500 / 112ms; FOV = 256mm′256mm; matrix = 128′128; 2mm′2mm planar resolution; number of slices = 75; slice thickness 2mm, without gaps.

[0147] Specifically, each scan yields 33 images: 32 diffusion-weighted images (b = 1000 s / mm²) and 1 undiffusion-weighted image (b = 0 s / mm²).

[0148] After acquiring the image sequences, the brain region activation statistics map can be obtained by performing a first-level analysis on the echo-planar imaging sequence using a general linear model. Specifically, a general linear model (GLM) can be used for first-level analysis to obtain the brain region activation statistics map for each subject.

[0149] For ReHo maps (Regional homogeneity, ReHo), the ReHo value of the brain can be calculated by performing low-frequency filtering on the echo plane imaging sequence and then calling the consistency coefficient, and the ReHo map can be generated using the ReHo value.

[0150] In practice, the data can be low-frequency filtered (0.01–0.1 Hz), and Kendall's consistency coefficient can be used to calculate the ReHo value of the entire brain at the voxel level to assess the similarity of a given voxel's time series to its 26 nearest neighbor time series, generating a ReHo map for each subject. The ReHo of each voxel is divided by the global average ReHo value of the entire brain for each subject to reduce intra-subject variability.

[0151] The function matrix can be functional connectivity (FC). The fluctuation coefficients of each pair of regions in the echo plane imaging image sequence can be identified using the Harvard-Oxford probabilistic map, and the function matrix can be generated using the fluctuation coefficients based on the automatic low-frequency fluctuation method. The fluctuation coefficients are calculated after removing non-neural oxygenation level-dependent fluctuations.

[0152] In practice, for each paired region identified using the Harvard-Oxford probabilistic mapping, the correlation coefficient is calculated after removing non-neural oxygenation level-dependent fluctuations. The automatic low-frequency fluctuation method is used to calculate the correlation between ROIs, generating a correlation matrix, which in turn yields the function matrix.

[0153] For fractional anisotropy (FA), image feature values ​​can be extracted from the echo-plane imaging sequence using Oxford FMRIB, and the fractional anisotropy value can be calculated using the image feature values.

[0154] In practice, Oxford fMRIB can be used for feature extraction. By bending the hind limb of a template of the internal capsule volume of interest onto the patient image, the mean anisotropy fraction (FA) of each hind limb of the internal capsule is calculated. The structural integrity of the hind limb of the internal capsule is quantified by the anisotropy fraction asymmetry index. The anisotropy fraction asymmetry index is calculated as: (Healthy side FA - Affected side FA) / (Healthy side FA + Affected side FA).

[0155] In one approach, activation levels are analyzed using a general linear model (GLM). If the activation intensity on the affected side is significantly lower than that on the healthy side, it indicates weakened functional activity in the affected brain region, suggesting a lower degree of brain function preservation. Conversely, if the activation level is close to that on the healthy side, it suggests a higher degree of brain function preservation.

[0156] Secondly, a significant decrease in ReHo on the affected side indicates weakened local functional synergy, suggesting lower retention of brain function. Simultaneously, functional connectivity is assessed using correlation coefficients between ROIs; a significant decrease in functional connectivity between the affected brain region and other regions indicates weakened synergy between the affected side and other brain regions, suggesting lower retention of brain function. Furthermore, white matter fiber integrity is assessed using anisotropy scores; a higher anisotropy score asymmetry index indicates more severe white matter damage on the affected side, suggesting lower retention of brain function.

[0157] In one embodiment, the brain imaging feature parameters include: functional near-infrared index values;

[0158] As an example, the operation of obtaining the brain image feature parameters may include the following sub-steps:

[0159] S51. Acquire functional near-infrared spectral signals.

[0160] S52. Calculate the near-infrared blood oxygen concentration value, near-infrared contralateral index value, and near-infrared functional connectivity value using the functional near-infrared spectral signal to obtain the functional near-infrared index value.

[0161] In practice, a 46-channel fNIRS instrument can be used to acquire functional near-infrared spectral signals. The channels between each transmitter and receiver are positioned according to a 10-20 system. After data acquisition, the data is saved and analyzed offline. The multi-channel fNIRS instrument can acquire raw light intensity and calculate near-infrared blood oxygen saturation, near-infrared contralateral index, and near-infrared functional connectivity, thereby obtaining functional near-infrared index values.

[0162] In this process, near-infrared blood oxygen concentration values ​​can be obtained by first converting the raw light intensity into optical density. During the measurement of functional near-infrared spectral signals, the transmitter emits light of a specific wavelength into the brain tissue, and the detector records the light intensity after passing through the brain tissue. These recorded light intensity values ​​constitute the raw light intensity data sequence, which is then converted using the Beer-Lambert law to quantify the attenuation of light intensity as it passes through the brain tissue.

[0163]

[0164] Where OD is optical density, I is the light intensity after passing through brain tissue, and I0 is the initial light intensity.

[0165] The optical density is then converted into near-infrared blood oxygen concentration. By measuring the intensity change of light of a specific wavelength after passing through brain tissue, and using the known absorption coefficient and diffuse path length factor, the modified Beer-Lambert Law (MBLL) can be applied to calculate the changes in oxyhemoglobin and deoxyhemoglobin concentrations caused by brain activity, thus obtaining the near-infrared blood oxygen concentration value.

[0166]

[0167] Where Δc represents the change in hemoglobin concentration, or near-infrared blood oxygen concentration, ΔOD represents the change in optical density, ∈ represents the absorption coefficient, DPF represents the diffuse path length factor, and L represents the optical path length.

[0168] For the near-infrared laterality index (LI), the oxyhemoglobin concentration in the bilateral regions of interest can be selected to calculate the near-infrared laterality index.

[0169] The near-infrared contralateral index value is calculated as follows:

[0170] LI = (ci) / (c+i);

[0171] In the above formula, LI represents the near-infrared contralateral index value, c represents the oxyhemoglobin concentration in the healthy hemisphere, and i represents the oxyhemoglobin concentration in the stroke hemisphere.

[0172] To assess connectivity within the network, Pearson correlation coefficients R for all electrode pairs were calculated for near-infrared functional connectivity values. Fisher's Z-transform was then applied to these coefficients to obtain the functional connectivity value for each electrode pair. The Fisher's Z-transform approximates the distribution of the correlation coefficients as a normal distribution.

[0173] Specifically, the near-infrared functional connectivity value is calculated as follows:

[0174]

[0175] In the above formula, R z R represents the near-infrared functional connectivity value, and R represents the Pearson correlation coefficient of the electrode pair.

[0176] In one embodiment, the brain imaging feature parameters include: functional magnetic resonance imaging (fMRI) index values;

[0177] As an example, the operation of obtaining the brain image feature parameters may include the following sub-steps:

[0178] S61. Acquire nuclear magnetic resonance scan images.

[0179] S62. After converting the nuclear magnetic resonance scan image into NIFTI format, image preprocessing is performed to obtain a functionally processed image. The image preprocessing includes: removal of time points, time layer correction, head motion correction, registration and spatial normalization, delinear drift removal, low-frequency filtering and smoothing.

[0180] S63. Calculate the magnetic resonance blood oxygen concentration value, magnetic resonance contralateral index value, and magnetic resonance functional connectivity value based on the functionally processed image to obtain the functional magnetic resonance index value.

[0181] Specifically, MRI scans were acquired using a 3.0T MRI scanner with an 8-channel head coil.

[0182] The scanning parameters are as follows: Structural imaging uses a 3D pre-magnetized rapid acquisition gradient-echo MPRAGE sequence, TR / TE = 2530ms / 2.93ms, FOV = 256mm×256mm, matrix = 512×512, slice thickness 1mm, no interval, and time point 1. The acquisition of structural images helps to rule out whether the subject has organic lesions and provides operability for spatial registration of subsequent data analysis.

[0183] Resting-state fMRI scans were performed using echo-planar imaging (EPI) sequences with TR / TE = 2000 ms / 21 ms, FOV = 240 mm × 240 mm, matrix size 64 × 64, voxel size 3.8 mm × 3.8 mm × 4.0 mm, flip angle 90°, 33 slices continuously scanned with 0 intervals, 200 time points, and a scan time of 412 s.

[0184] Preprocessing of resting-state functional imaging data was performed using RESTplus. After converting the DICOM data format to NIFTI format, operations such as time point removal, temporal correction, head motion correction, registration and spatial normalization, linear drift correction, low-frequency filtering, and smoothing were applied to obtain the functionally processed image. Then, referring to the formulas in S51-S52 for calculating near-infrared oxygen saturation, near-infrared contralateral index, and near-infrared functional connectivity, the magnetic resonance oxygen saturation, magnetic resonance contralateral index, and magnetic resonance functional connectivity values ​​were calculated. Simultaneously, the values ​​from the MRI scan images were acquired to obtain the MRI values.

[0185] In an optional embodiment, if the concentration of oxyhemoglobin on the affected side is significantly lower than that on the healthy side, it indicates weakened metabolic activity in the affected brain region, suggesting lower retention of brain function; conversely, if the concentration is close to the level of the healthy side, it suggests higher retention of brain function. Secondly, bilateral hemisphere differences are assessed using the LI (Liquidity Index) method; a LI value closer to 1 indicates poorer function on the affected side, suggesting lower retention of brain function. Furthermore, if the functional connectivity (FC) between the affected brain region and other regions is significantly reduced, it indicates weakened synergy between the affected side and other brain regions, suggesting lower retention of brain function.

[0186] After obtaining the above parameter set, in order to improve the training accuracy of the model, the operation of training using the training set, as an example, may include the following sub-steps:

[0187] S71. The Z-transformation method is used to normalize the feature parameters of the training set to obtain normalized parameters.

[0188] S72. The normalized parameters are screened using the P-value test method to obtain the screened parameters.

[0189] S73. The model is trained using the aforementioned screening parameters.

[0190] In practice, the feature parameters of the training set can be normalized using the Z-transform method to obtain normalized parameters, which are then mapped to the 0-1 range to accelerate model convergence. The Z-transform method can be a Z-alteration, a mathematical transformation applied to discrete sequences, often used to solve linear time-invariant difference equations. Its role in discrete systems is analogous to that of the Laplace transform in continuous systems.

[0191] To avoid model overfitting, a p-value test was used to compare the data (demographic characteristics, motor function, neurophysiological and imaging features, etc.) within the normalized parameters, selecting features with p-values ​​less than 0.05. Then, the Least Absolute Shrinkage and Selection Operator (LASSO) was used to further filter the features, resulting in the selection parameters.

[0192] Using the selected features as independent variables and the subject group as the dependent variable, a classification model was constructed using a Support Vector Machine (SVM) algorithm. Due to the presence of non-linear data in the sample data, a multinomial kernel was considered when selecting the SVM kernel function. A grid search was used to explore the optimal hyperparameters for constructing the SVM—the penalty factor C and the kernel function parameter gamma—with values ​​ranging from 2 to 10 to 210. The search process employed 5-fold cross-validation.

[0193] After training, the model can be validated using a test set. Specifically, leave-one-out cross-validation can be performed using the test set. The effectiveness of the model can be judged by the area under the curve (AUC) of the receiver operating characteristic curve (ROC) and the classification accuracy.

[0194] Based on the model's assessment results, patients are categorized into those with high brain function preservation and those with low brain function preservation. The corresponding transcranial magnetic stimulation (TMS) patterns are then determined and recommended to doctors for their reference.

[0195] In this embodiment, the present invention provides an auxiliary decision-making system for transcranial magnetic stimulation (TMS) modes. Its advantages are as follows: the present invention can acquire the patient's personal characteristic information through an acquisition module; then, through a determination module, it calls a preset assessment model to calculate the patient's brain function retention using the personal characteristic information, and determines the patient's TMS mode based on the brain function retention. Since the preset assessment model is a classification model constructed based on demographic characteristic parameters, behavioral assessment characteristic parameters, neurophysiological characteristic parameters, and brain imaging characteristic parameters as independent variables, the model can be used to calculate the brain function retention and determine the TMS mode based on the brain function retention, without relying on subjective calculations by the doctor, thereby improving the accuracy of determining the treatment mode; at the same time, the processing does not require multiple assessment and testing calculations, which can shorten the processing time, further improve the efficiency and accuracy of processing, avoid affecting the patient's diagnosis and subsequent treatment, and achieve precision treatment.

[0196] In the description of the embodiments of the present invention, it should be noted that the terms "above," "below," etc., indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. When an element such as a layer, region, or substrate is referred to as being "above" or "on top of" another element, it may be directly on the other element, or there may be an intermediate element. Conversely, when an element is referred to as being "directly on" or "above" another element, there is no intermediate element. It should also be understood that when an element is referred to as being "below" or "under" another element, it may be directly below or under the other element, or there may be an intermediate element. Conversely, when an element is referred to as being "directly below" or "under" another element, there is no intermediate element. Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0197] Those skilled in the art will understand that embodiments of this application may also include computer program products. Therefore, this application may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0198] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), devices, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0199] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0200] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0201] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A decision support system using transcranial magnetic stimulation (TMS) modality, characterized in that, The system includes: The acquisition module is used to acquire the patient's personal characteristic information; The determination module is used to call a preset assessment model to calculate the patient's brain function retention using the personal characteristic information, and determine the patient's transcranial magnetic stimulation mode based on the brain function retention. The preset evaluation model is a classification model obtained by constructing a training set based on demographic characteristic parameters, behavioral evaluation characteristic parameters, neurophysiological characteristic parameters, and brain imaging characteristic parameters as independent variables, and using the training set for training operations. The training operation includes: normalizing the feature parameters of the training set using the Z-transform method to obtain normalized parameters; filtering the normalized parameters using the P-value test method to obtain filtered parameters; and training the model using the filtered parameters.

2. The decision support system for transcranial magnetic stimulation mode according to claim 1, characterized in that, The construction process of the preset evaluation model includes the following steps: Obtain a parameter set, which includes: demographic characteristic parameters, behavioral assessment characteristic parameters, neurophysiological characteristic parameters, and brain imaging characteristic parameters; The parameter set is divided into a training set and a test set by using the leave-one-out method; The preset support vector machine is trained using the training set, and the preset support vector machine is cross-validated using the test set to obtain the preset evaluation model. The default kernel function of the support vector machine is a polynomial kernel, and the default optimal hyperparameters of the support vector machine are constructed using a grid search method.

3. The decision support system for transcranial magnetic stimulation mode according to claim 2, characterized in that, The neurophysiological characteristic parameters include: TMS assessment index values; The operation of obtaining the neurophysiological characteristic parameters includes the following steps: Acquire the patient's electromyographic signals; Based on the electromyographic signals, the resting movement threshold, active movement threshold, and interhemispheric inhibition value are determined to obtain TMS evaluation index values.

4. The decision support system for transcranial magnetic stimulation mode according to claim 2, characterized in that, The neurophysiological characteristic parameters include: electroencephalogram (EEG) index values; The operation of obtaining the neurophysiological characteristic parameters includes the following steps: Collecting electroencephalogram (EEG) signals; The power spectral density value, the EEG phase lag index value, and the potential distribution parameters are calculated using the EEG signals to obtain the EEG index values.

5. The decision support system for transcranial magnetic stimulation mode according to claim 4, characterized in that, The power spectral density value is calculated as follows: In the above formula, PSD[k] represents the power spectral density value, X[k] represents the discrete Fourier transform coefficient of the k-th frequency point, and N is the number of data points of the EEG signal; The discrete Fourier transform coefficients X[k] are calculated as follows: In the above formula, x[n] is the EEG signal at the nth time point, and j is the imaginary unit; The EEG phase lag index value is calculated as follows: In the above formula, PLI represents the EEG phase lag index, t represents the phase difference at time t, and T is the total number of time points. Wherein, the phase difference at time t The calculation is shown in the following formula: In the above formula, Im represents the imaginary part, Re represents the real part, and x(t)y*(t) represents the complex product of the complex conjugates of signals x(t) and y(t); The potential distribution parameters are calculated as follows: Φ = KJ + ε; In the above formula, Φ represents the potential distribution measured on the scalp, K represents the forward field matrix, J represents the current density distribution in the source space, and ε is the measurement noise.

6. The decision support system for transcranial magnetic stimulation mode according to claim 2, characterized in that, The brain imaging feature parameters include: diffusion tensor imaging index values; Obtaining the brain imaging feature parameters includes the following steps: Acquire echo plane imaging; The echo-planar imaging was used to determine the brain region activation statistical parameter map, ReHo map, imaging function matrix and anisotropy fraction value, respectively, and the diffusion tensor imaging index value was obtained.

7. The decision support system for transcranial magnetic stimulation mode according to claim 2, characterized in that, The brain imaging feature parameters include: functional near-infrared index values; The operation of obtaining the brain imaging feature parameters includes the following steps: Acquire functional near-infrared spectral signals; The near-infrared blood oxygen concentration, near-infrared contralateral index, and near-infrared functional connectivity values ​​are calculated using the functional near-infrared spectral signals to obtain functional near-infrared index values.

8. The decision support system for transcranial magnetic stimulation mode according to claim 7, characterized in that, The near-infrared blood oxygen concentration value is calculated as follows: In the above formula, Δc represents the near-infrared blood oxygen concentration, ΔOD represents the change in optical density, ∈ represents the absorption coefficient, DPF represents the diffuse path length factor, and L represents the optical path length. The optical density is calculated as follows: In the above formula, OD represents optical density, I represents the light intensity after passing through brain tissue, and I0 represents the initial light intensity; The near-infrared contralateral index value is calculated as follows: LI = (ci) / (c+i); In the above formula, LI represents the near-infrared contralateral index value, c represents the oxyhemoglobin concentration in the healthy hemisphere, and i represents the oxyhemoglobin concentration in the stroke hemisphere. The near-infrared functional connectivity value is calculated as follows: In the above formula, R z R represents the near-infrared functional connectivity value, and R represents the Pearson correlation coefficient of the electrode pair.

9. The decision support system for transcranial magnetic stimulation mode according to claim 2, characterized in that, The brain imaging feature parameters include: functional magnetic resonance imaging (fMRI) index values; The operation of obtaining the brain imaging feature parameters includes the following steps: Acquire nuclear magnetic resonance (NMR) scan images; The nuclear magnetic resonance scan images are converted to NIFTI format and then preprocessed to obtain functionally processed images. Based on the functionally processed images, the MRI blood oxygen concentration value, the MRI contralateral index value, and the MRI functional connectivity value are calculated to obtain the functional MRI index values.

10. The decision support system for transcranial magnetic stimulation modes according to claim 9, characterized in that, The image preprocessing includes: removal of time points, temporal layer correction, head motion correction, registration and spatial normalization, delinear drift removal, low-frequency filtering, and smoothing.

Images