Acute cerebral trauma prognosis detection system

Abstract

The invention discloses an acute cerebral trauma prognosis detection system, which comprises a multi-source data synchronous acquisition and extraction part, a data preprocessing part, a machine learning prediction model and acute traumatic cerebral injury result display, 19-lead electroencephalogram (EEG) data distributed in the whole brain and left and right cerebral middle artery (MCA) data are subjected to multi-source data synchronous acquisition and extraction under a mixed auditory induction normal form task, and clinical data, EEG, MCA and classification tag Grassco prognostic score (GOS) of a patient with cerebral trauma are subjected to data preprocessing; inputting the preprocessed data into a machine learning prediction model, and outputting the probability of poor prognosis of the TBI patient after calculation by the prediction model; the TBI patient reject ratio prediction result shows the probability of poor prognosis of the TBI patient. The method can be used for early, simply and accurately detecting the acute cerebral trauma patient.

Description

Technical Field

[0001] The invention relates to the field of trauma prognosis, and in particular to an acute brain trauma prognosis detection system. Background Art

[0002] Traumatic brain injury (TBI) refers to the injury caused by direct mechanical force on the head, with an annual incidence of 939 / 100,000 worldwide. [1] The disability and mortality rates of TBI are high. 10% of mild TBI patients and 66% to 100% of moderate to severe TBI patients have long-term cognitive dysfunction. Early and rapid judgment of the severity of TBI can help clinicians formulate treatment plans in a timely manner and improve prognosis. [2] Transcranial Doppler (TCD) has a high value in evaluating nerve damage. Increased cerebral blood flow velocity can early predict the occurrence of neurological complications such as hypoxic-ischemic encephalopathy and neurocognitive decline. [3] The middle cerebral artery is the largest branch of the internal carotid artery, responsible for 80% of the blood supply to the brain tissue, and can directly reflect the changes in brain tissue hemodynamics.

[0003] Currently, cerebral hemodynamics and imaging omics have been used to evaluate the condition and prognosis of spontaneous cerebral hemorrhage. [4,5] However, clinical evaluation of the severity of TBI is often based on cerebral hemodynamics and clinical experience, which is prone to misdiagnosis and missed diagnosis, and is somewhat subjective. [6,7] Multimodal detection has become an important trend in medical diagnosis. Among the many neural information detection methods, with the development of BCI research, many studies have found that there is a certain connection between EEG characteristics and disease pathology. EEG is widely used due to its simple operation, non-invasiveness, high time resolution and other advantages.

[0004] References

[0005] [1]Dewan MC, Rattani A, et al. Estimating the global incidence of traumatic brain injury. JNeurosurg. 2018Apr 27;130(4):1080-1097.

[0006] [2]Maas AIR, Menon DK, et al. Traumatic brain injury: integrated approaches to improve prevention, clinical care, and research. Lancet Neurol. 2017Dec; 16(12):987-1048.

[0007] [3]Millet A,Evain JN,et al.Clinical applications of transcranialDoppler in non-trauma critically ill children:a scoping review.Childs NervSyst.2021Sep;37(9):2759-2768.

[0008] [4]Ye H,Gao F,et al.Precise diagnosis of intracranial hemorrhage andsubtypes using a three-dimensional joint convolutional and recurrent neuralnetwork.Eur Radiol.2019Nov;29(11):6191-6201.

[0009] [5]Jain V,de Godoy LL,et al.Cerebral hemodynamic and metabolicdysregulation in the postradiation brain.J Neuroimaging.2022Nov;32(6):1027-1043.

[0010] [6]Koyama T,Uchiyama Y,et al.Comparison of Fractional Anisotropy fromTract-Based Spatial Statistics with and without Lesion Masking in Patientswith Intracerebral Hemorrhage:ATechnical Note.J Stroke CerebrovascDis.2019Nov;28(11):104376.

[0011] [7]Muravskiy A, Polischuk M, et al. Changes in cerebral hemodynamics inboxers with repeated traumatic brain injury. Pol Merkur Lekarski. 2020Oct 23;48(287):312-317. Summary of the invention

[0012] The present invention provides an acute brain trauma prognosis detection system, which can perform early, simple and accurate detection on patients with acute brain trauma, as described below:

[0013] An acute brain injury prognosis detection system, the system comprising: a multi-source data synchronous acquisition and extraction part, a data preprocessing part, a machine learning prediction model and an acute traumatic brain injury result display,

[0014] The 19-channel EEG data distributed throughout the whole brain and the left and right MCA data were synchronously collected and extracted under the mixed auditory evoked paradigm task, and the clinical data, EEG, MCA, and classification label GOS of patients with brain trauma were preprocessed;

[0015] The preprocessed data is input into a machine learning prediction model, and the prediction model calculates and outputs the probability of a poor prognosis for a TBI patient;

[0016] The adverse rate prediction results of TBI patients show the probability of poor prognosis of TBI patients.

[0017] Wherein, the paradigm described is a hybrid auditory evoked paradigm in which the auditory steady-state evoked response (ASSR) is integrated into the auditory P300 system;

[0018] The carrier frequencies of sound stimulation were 500Hz and 2000Hz, and the modulation frequency was 40Hz. The auditory stimulation was amplitude modulated AM with a modulation depth of 100% to obtain standard stimulation and deviant stimulation respectively. In each task trial, standard stimulation and deviant stimulation appeared randomly in a ratio of 4:1, thereby inducing auditory P300. The EEG and left and right MCA data under the mixed auditory evoked paradigm were collected synchronously, and the user's EEG feature parameters were extracted by offline calculation.

[0019] The collection and extraction of acute traumatic brain injury data are as follows: patients are excluded and multiple physiological data are extracted. Inclusion criteria: age ≥ 18 years old; Glasgow coma score (GCS) ≤ 12 points; exclusion criteria: open craniocerebral trauma; severe multiple injuries; patients transferred to rehabilitation hospitals from other hospitals; patients who died upon admission; patients with brain trauma before this disease; patients who were lost to follow-up or had incomplete data; patients with liver and kidney failure; patients with hemorrhagic diseases; coagulation disorders; patients with severe infections; patients with malignant tumors;

[0020] Data collection: ERP and ASSR EEG characteristic parameters, MCA hemodynamic parameters and basic patient information at 1 day, 1 week and 2 weeks after hospitalization; cerebral blood flow parameters and EEG data were collected and extracted simultaneously;

[0021] After the data are collected and extracted, the data are preprocessed: data cleaning and normalization.

[0022] The machine learning prediction model was constructed using a support vector machine machine learning algorithm, and a total of 300 TBI patients were retrospectively collected, with 210 TBI cases as a training set and 90 TBI cases as a test set;

[0023] The preprocessed data was input into the prediction model, and the parameters of the SVM classification model were set as follows: the penalty coefficient of the error term was 200, the kernel function type was Gaussian kernel, the kernel function coefficient was 5, whether to enable probability estimation was False, whether to use heuristic shrinkage method was True, the error accuracy of the model stopping training was 0.001, the memory required for training was specified to be 400, the weight of different penalty parameters for each category was set to None, the maximum number of iterations was -1, the cross-validation parameter was 5, and the model evaluation criteria were ROC and AUC.

[0024] The beneficial effects of the technical solution provided by the present invention are:

[0025] 1. The present invention uses a brain-computer interface method based on a hybrid auditory evoked paradigm of auditory event-related potential (ERP) and auditory steady-state response (ASSR) to establish a machine learning (ML) model to predict the 6-month poor prognosis rate of acute TBI using EEG parameters and middle cerebral artery hemodynamic parameters;

[0026] 2. The present invention has been trained and verified in the training set and the test set, ensuring that the prediction model established by the present invention can make early, simple and accurate predictions for patients with acute brain trauma, thereby clinically assisting in the formulation of clinical decisions;

[0027] 3. The prognostic detection system based on SVM machine learning can make early prognostic predictions for TBI patients, help clinicians provide personalized auxiliary treatment plans for TBI patients earlier, and provide advice for the treatment of TBI patients.

[0028] 4. The present invention realizes the detection of brain consciousness level, which can be extended to the fields of neuroscience, life science, etc., and obtain considerable social and economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a structural schematic diagram of an acute brain injury prognosis detection system;

[0030] Figure 2 Schematic diagram of the timing design for the hybrid auditory evoked paradigm. DETAILED DESCRIPTION

[0031] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention are described in further detail below.

[0032] The embodiment of the present invention provides an acute brain trauma prognosis detection system, the main components of which include: a multi-source data synchronous acquisition part, a data preprocessing part, a machine learning model prediction, and an acute traumatic brain injury (TBI) prediction result display. The embodiment of the present invention is suitable for improving the accuracy of predicting the poor prognosis rate of patients in the intensive care unit of a hospital after 6 months, assisting doctors in clinical decision-making, and improving patient prognosis. To this end, the embodiment of the present invention aims to design a hybrid auditory evoked paradigm, synchronously collect EEG and MCA cerebral blood flow parameters, integrate the patient's clinical physiological information, and then build an acute brain trauma prediction system.

[0033] The technical process is: by inputting the ERP, ASSR EEG characteristic parameters and MCA hemodynamic parameters of TBI patients 1 day, 1 week and 2 weeks after hospitalization, based on multi-physiological information fusion analysis, the poor prognosis rate of TBI patients can be predicted efficiently, quickly, accurately and timely, providing strong support for clinical decision-making.

[0034] 1. Design of the prognosis prediction system for acute brain trauma based on SVM

[0035] Study subjects: Clinical data of patients in the craniocerebral trauma ward. According to the following inclusion and exclusion criteria, eligible patients with acute brain trauma were screened. Inclusion criteria: age ≥ 18 years old; Glasgow Coma Score (GCS) ≤ 12 points. Exclusion criteria: open craniocerebral trauma; severe multiple injuries; patients transferred to rehabilitation from other hospitals; patients who died on admission; patients with brain trauma before this disease; patients who were lost to follow-up or had incomplete data; liver and kidney failure; accompanied by hemorrhagic diseases; coagulation dysfunction; combined with severe infection; accompanied by malignant tumors.

[0036] Data collection: ERP and ASSR EEG characteristic parameters, middle cerebral artery hemodynamic parameters and basic patient information (age, gender, GCS score at admission) at 1 day, 1 week and 2 weeks after hospitalization, cerebral blood flow parameters and EEG data are collected synchronously. The present invention constructs a feature set, all of which include basic patient information: age, gender, Glasgow Coma Score (GCS); middle cerebral artery hemodynamic parameters: end-diastolic blood flow velocity (EDV), peak systolic blood flow velocity (PSV), mean maximum blood flow velocity (MV) and pulsatility index (PI); EEG characteristic parameters: amplitude, latency, fast Fourier transform (FFT), event-related spectral perturbation EEG time-frequency analysis (Event-related spectral perturbation, ERSP), phase-locked value (PLV), signal-to-noise ratio (SNR), relative power (RP). This multi-source data fusion method can comprehensively reflect the patient's physiological state and disease progression, and provide richer and more accurate information for prediction.

[0037] The system of the embodiment of the present invention is as follows Figure 1 As shown in the figure, its system architecture and technical process include: behavioral data collection part (Glasgow Coma Scale GCS), multi-source data synchronous collection and extraction part, neuroelectrophysiological index data preprocessing part, machine learning model prediction part, and acute brain trauma poor prognosis rate prediction result display part.

[0038] In terms of evaluation criteria, the embodiments of the present invention adopt the following two main dimensions: (1) Assessment of the severity of brain trauma: Based on the Glasgow Coma Scale (GCS), a GCS score of 9 to 12 at the time of hospitalization is considered moderate brain trauma, 6 to 8 points is considered severe, and 5 points or less is considered extremely severe. The GCS score at the time of hospitalization is used as one of the features of the model. (2) Prognosis outcome assessment: The Glasgow Outcome Score (GOS) 6 months after injury is used for measurement, where patients with a GOS score of 11-15 points are considered to have a good prognosis, while those with a GOS score of 10 points or less are considered to have a poor prognosis. The Glasgow Outcome Score Scale (GOS) of the patient 6 months later is used as a classification label.

[0039] Subsequently, the embodiment of the present invention uses the support vector machine (SVM) machine learning algorithm to fuse analysis and construct a 6-month prognosis prediction model for patients. To ensure the accuracy and reliability of the model, the patients were randomly divided into a training set and a test set. The embodiment of the present invention was trained on the test set and verified on the test set. It is ensured that the prediction model established by the embodiment of the present invention can predict the prognosis of patients with brain trauma early, simply and accurately, and assist clinical decision-making.

[0040] The timing design of the mixed auditory evoked paradigm is as follows Figure 2 As shown, the embodiment of the present invention designs a mixed auditory evoked paradigm of ERP and ASSR. By integrating ASSR into the auditory P300 system, an EEG synchronous acquisition device is built, and the user's EEG characteristic parameters are extracted by offline calculation. Auditory stimulation uses amplitude modulation AM with a modulation depth of 100%, the sound stimulation carrier frequency is 500Hz and 2000Hz, and the modulation frequency is 40Hz. In each task trial, the standard stimulus and the deviant stimulus appear randomly in a ratio of 80%:20%, thereby inducing auditory P300.

[0041] Sinusoidal amplitude modulation is to control the amplitude of the carrier signal with the modulation frequency to be transmitted, so that the spectrum of the baseband signal is moved to a higher frequency, the carrier frequency changes periodically, and the envelope shape is similar to the modulation signal. The modulation depth is the ratio of the amplitude of the modulation signal to the amplitude of the DC signal, expressed as a percentage; the modulation signal m(t) is superimposed on the DC signal A and multiplied by the carrier to form an amplitude modulation signal, and its time domain expression is:

[0042] S AM (t) = (A + m (t)) * cos (2 * pi * f c *t)

[0043]

[0044] Among them, S AM (t) is the modulated signal, A is the DC signal, m(t) is the modulated signal, f c is the carrier signal frequency, md is the modulation depth of the signal, in %, peak(m(t)) is the peak value of the modulation signal, for sound stimulation, f c=1000Hz; the modulation signal m(t) is a sinusoidal signal with a modulation frequency of 40Hz. In the embodiment of the present invention, the sound is generated by the PsychPortAudio function in the Psychtoolbox toolbox in Matlab, and is generated in a waveform audio file format (*.wav) with a sampling rate of 44.1kHz, and is played binaurally through headphones. In order to accurately control the presentation of auditory stimulation, the time error of sound playback is controlled within 5ms to control the time error within the minimum range. In order to prevent the sound stimulation from producing obvious blasts, the sound is faded in and out for 10ms.

[0045] 2. Functions of each module of the brain trauma prediction and prognosis system

[0046] (1)EEG-cerebral blood flow data acquisition module

[0047] EEG acquisition uses a 19-lead EEG acquisition system and its dedicated acquisition software. The data acquisition sampling rate is set to 1000Hz and 50Hz power frequency notch. Different task modes correspond to different event codes, which is convenient for later synchronous data interception and feature analysis. The parameters of the bilateral middle cerebral arteries (MCA) are collected through bilateral temporal windows using the transmitter of the color transcranial Doppler (TCD) pulse wave Doppler probe. TCD is a cerebrovascular disease examination method that uses the ultrasonic Doppler effect to examine the hemodynamics of each major artery on the intracranial cerebral artery circle (Willis circle) and various blood flow physiological parameters. TCD has the advantages of non-invasive, portable, radiation-free, and repeatable monitoring, and can directly obtain the hemodynamic parameters of the middle cerebral artery: EDV, PSV, MV, PI. The embodiment of the present invention innovatively adds EEG features to the method of multi-source data fusion of cerebral hemodynamic parameters, which can comprehensively reflect the patient's physiological state and disease progression, and provide richer and more accurate information for prediction.

[0048] (2) Extraction of middle cerebral artery hemodynamic parameters

[0049] Transcranial Doppler ultrasound (TCD) is based on the principle of the Doppler effect. It emits ultrasonic waves that penetrate the skull and are reflected by red blood cells moving through the cerebral blood vessels, thereby obtaining cerebral hemodynamic parameters. The parameters are obtained using the Delica TCD monitor with the data acquisition software ICM+ software, a dual-channel transcranial Doppler instrument, and a probe frequency of 2MHz. The patient lies in a supine position, the examiner stands at the patient's head position and fixes the patient's head, holds the probe in hand and explores along the temporal window, with a depth of 55 to 65mm, finds the blood flow in the direction of the probe, fine-tunes the probe angle, tries to make the probe sound wave direction parallel to the blood flow direction, and takes the maximum measurement speed. To obtain the hemodynamic parameters of the middle cerebral artery, the steps include:

[0050] 1) Use TCD blood flow imaging technology to detect the distribution and direction of intracranial blood vessels and locate the MCA position based on the depth, speed, and blood flow direction;

[0051] 2) Use ICM+ software to synchronously obtain MCA hemodynamic parameters;

[0052] 3) Define and quantify MCA blood flow parameters, including EDV, PSV, MV, and PI.

[0053] (3) Extraction of EEG neurophysiological features

[0054] Fast Fourier Transform FFT: ASSR is a neural response based on the frequency domain. In order to explore the ASSR induction situation, it is necessary to convert the EEG signal from the time domain to the frequency domain. The embodiment of the present invention uses fast Fourier transform for conversion. The formula is as follows:

[0055]

[0056] Wherein, e is the base of natural logarithm, N represents the sampling rate, and x[n] represents the collected discrete EEG signal, which is in complex form and is expressed as x^[k]=a+bi.

[0057] Event-related spectral perturbation EEG time-frequency analysis (ERSP): Sound stimulation with a modulation frequency of 40 Hz will cause synchronous oscillations in the corresponding auditory cortex neurons of the brain. ERSP is used to explore the effects of auditory stimulation in different tasks on auditory cortex activation. The ERSP of multiple trials (n trials) is defined as:

[0058]

[0059] Where n represents the total number of all trials, F k (f,t) 2 Represents the energy spectrum estimation of the kth trial at frequency f and time t. ERSP calculation with baseline is adopted. The data segment of sound stimulation (standard stimulation and deviant stimulation) from 200ms before stimulation to 1000ms after stimulation is selected, and the spectrum mean value 200ms before stimulation is used as the baseline to calculate the ERSP value in the frequency range of 20-50Hz.

[0060] Phase-locked value: The phase-locked value is represented by a non-random phase or phase difference distribution, which can be used to analyze nonlinear and non-stationary systems. The brain can be regarded as a nonlinear dynamic system, and PLV can be used to analyze the synchronization of neural activity. PLV is defined as the average vector length calculated based on the phase angle difference:

[0061]

[0062] Where n is the number of trials, e is the natural logarithm, i is the imaginary unit, and p is the radian phase difference of each t trial. The value range of PLV is between 0 and 1, representing the phase difference between the two signals. When the PLV value is 0, it means that the phase difference is evenly distributed on the unit circle of the complex plane according to time, and there is no synchronization. The closer the PLV is to 1, the stronger the synchronization between the two signals. When the PLV value is 1, it means that the two signals are completely synchronized.

[0063] Signal-to-Noise Ratio: Signal-to-Noise Ratio (SNR) is one of the commonly used indicators to measure brain-computer interface systems. The signal-to-noise ratio can be determined by the ratio of the target frequency amplitude to the noise, where the noise is defined by the average value of n adjacent amplitudes on both sides of the target frequency on the frequency domain spectrum:

[0064]

[0065] Wherein, y(f) represents the amplitude of the target frequency, Δf represents the frequency resolution, and n represents the number of points near the signal. In the embodiment of the present invention, the target frequency is 40 Hz, the frequency resolution is 1.25, and n is 5, that is, the signal-to-noise ratio in the embodiment of the present invention is determined by the ratio of the amplitude of 40 Hz to the average value of the 5 adjacent amplitudes on both sides of 40 Hz.

[0066] Relative power: Relative power (RP) can be used to evaluate the degree of ASSR induction. Studies have found that there is a significant positive correlation between the value of relative power and the CRS-R score. The RP calculation formula for a certain frequency band is:

[0067]

[0068] Among them, P relative represents the relative power value of a frequency band, P(f) represents the signal power spectrum of the entire frequency band, [f1,f2] represents the upper and lower limits of a frequency band, [f L ,f H ] represents the upper and lower limits of the full frequency band. Relative power is a frequency domain feature. To calculate the RP of 40 Hz, the frequency band value range is 40 Hz. To reduce the interference of low-frequency ERP and power frequency interference of equipment on RP, the value range of the full frequency band is 35-45 Hz. That is, the RP in the embodiment of the present invention is calculated by dividing the power of 40 Hz by the total power of the frequency range near 40 Hz (35-45 Hz).

[0069] 3. Data preprocessing:

[0070] The following steps are involved:

[0071] 1) The data of the first hospitalization due to brain trauma of the same patient are retained;

[0072] 2) Delete the patient's name, examination date, patient ID and other information, and only save the cerebral blood flow data;

[0073] 3) Unify the age data format into integers;

[0074] 4) For the digital processing of gender, "1" represents male and "0" represents female;

[0075] 5) For the patient's GCS score, "-1" indicates moderate brain trauma, "0" indicates severe, and "1" indicates extremely severe;

[0076] 6) Patients with poor prognosis according to the GOS scale were labeled as "1", and patients with good prognosis were labeled as "0";

[0077] 7) The label of patients with poor prognosis rate of brain trauma is 1, and the label of patients without poor prognosis of brain trauma is 0;

[0078] 8) The missing data are filled by KNN (K Nearest Neighbors) algorithm;

[0079] 9) For the processing of outliers, the upper quartile and the lower quartile are first used for judgment (the normal value range is between [Q2-1.5×Q3] and [Q1+1.5×Q2], where Q1 is the upper quartile, Q2 is the lower quartile, and Q3 is the median difference Q3=Q1-Q2). The median corresponding to the indicator is used to replace the outliers;

[0080] 10) Normalize the data, that is, (original data - minimum value) / (maximum value - minimum value).

[0081] 4. Model building:

[0082] This model uses the support vector machine (SVM) machine learning algorithm to build a prediction model. A total of 300 TBI patients were retrospectively collected, 210 TBI patients were used as training sets, and 90 TBI patients were used as test sets. The brain MCA hemodynamic parameters, EEG neuroelectrophysiological characteristics, and basic patient information (age, gender, GCS score) were feature fused, and the results divided the patients into good prognosis and poor prognosis according to the GOS score at discharge. The effectiveness of the present invention has been fully demonstrated through the training process of the training set and the verification process of the test set.

[0083] The parameters of the SVM classification model are set as follows: the penalty coefficient of the error term is 200, the kernel function type is Gaussian kernel, the kernel function coefficient is 5, whether to enable probability estimation is False, whether to use heuristic shrinkage method is True, the error accuracy of the model stopping training is 0.001, the memory required for training is specified to be 400, the weight of each category is set to None, the maximum number of iterations is -1, the cross-validation parameter is 5, and the model evaluation criteria are ROC and AUC (AUC is used for model evaluation).

[0084] The embodiment of the present invention designs an auditory ERP and ASSR mixed auditory evoked paradigm, which innovatively integrates non-invasive EEG features, clinical MCA cerebral hemodynamic parameters and basic physiological information of patients to predict the prognosis of patients with acute brain trauma. This invention has great potential and far-reaching impact in acute brain injury, emergency treatment, etc. Further research can obtain a perfect human brain trauma prediction system, which is expected to obtain considerable social and economic benefits.

[0085] Unless otherwise specified, the models of the components in the embodiments of the present invention are not limited, and any device that can perform the above functions may be used.

[0086] Those skilled in the art will appreciate that the accompanying drawing is only a schematic diagram of a preferred embodiment, and the serial numbers of the embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0087] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An acute brain injury prognosis detection system, characterized in that: The system includes: a multi-source data synchronous acquisition and extraction part, a data preprocessing part, a machine learning prediction model and an acute traumatic brain injury result display. The 19-channel EEG data distributed throughout the whole brain and the left and right middle cerebral artery data were synchronously collected and extracted under the mixed auditory evoked paradigm task, and the clinical data, EEG, MCA, and classification label GOS of patients with brain trauma were preprocessed; The preprocessed data is input into a machine learning prediction model, and the prediction model calculates and outputs the probability of a poor prognosis for a TBI patient; The adverse rate prediction results of TBI patients show the probability of poor prognosis of TBI patients.

2. The acute brain injury prognosis detection system according to claim 1, characterized in that: The hybrid auditory evoked paradigm is a hybrid auditory evoked paradigm in which ASSR is integrated into the auditory P300 system; The carrier frequencies of sound stimulation were 500Hz and 2000Hz, and the modulation frequency was 40Hz. The auditory stimulation was amplitude modulated AM with a modulation depth of 100% to obtain standard stimulation and deviant stimulation respectively. In each task trial, in each trial, the standard stimulation and deviant stimulation appeared randomly in a ratio of 4:1, the standard stimulation appeared four times, and the deviant stimulation appeared once, thereby inducing auditory P300. The EEG and left and right MCA data under the mixed auditory evoked paradigm were synchronously collected, and the user's EEG feature parameters were extracted by offline calculation.

3. The acute brain injury prognosis detection system according to claim 1, characterized in that: The collection and extraction of the acute traumatic brain injury data are as follows: patients were excluded and multiple physiological data were extracted, and the inclusion criteria were: age ≥ 18 years old; the patient's Glasgow Coma Scale score ≤ 12 points; Exclusion criteria: open craniocerebral trauma; severe multiple injuries; patients transferred from other hospitals to rehabilitation; patients who died upon admission; patients with brain trauma before this disease; patients who were lost to follow-up or had incomplete data; patients with liver and kidney failure; patients with hemorrhagic diseases; patients with coagulation disorders; patients with severe infection; patients with malignant tumors; Data collection: ERP and ASSR EEG characteristic parameters, MCA hemodynamic parameters and basic patient information at 1 day, 1 week and 2 weeks after hospitalization; cerebral blood flow parameters and EEG data were collected and extracted simultaneously; After the data are collected and extracted, the data are preprocessed: data cleaning and normalization.

4. The acute brain injury prognosis detection system according to claim 1, characterized in that: The machine learning prediction model was constructed using a support vector machine machine learning algorithm, and a total of 300 TBI patients were retrospectively collected, with 210 TBI cases as a training set and 90 TBI cases as a test set; The preprocessed data was input into the prediction model, and the parameters of the SVM classification model were set as follows: the penalty coefficient of the error term was 200, the kernel function type was Gaussian kernel, the kernel function coefficient was 5, whether to enable probability estimation was False, whether to use heuristic shrinkage method was True, the error accuracy of the model stopping training was 0.001, the memory required for training was specified to be 400, the weight of different penalty parameters for each category was set to None, the maximum number of iterations was -1, the cross-validation parameter was 5, and the model evaluation criteria were ROC and AUC.

Images