A post-stroke brain sign monitoring system

Abstract

The application provides a cerebral apoplexy postoperative sign monitoring system, which is fused with a nerve motor function evaluation wristband and a head-mounted brain electrical impedance dynamic imaging device, and breaks through the blind area of postoperative monitoring. The wristband device collects upper limb electromyographic signals and joint motion mechanics parameters through a microneedle electrode array and a distributed piezoresistive sensor, drives a neural network to output a standardized motor score based on time-frequency domain features containing electromyography, joint angle, trajectory smoothness and grip force change rate, and is related to a clinical Fugl-Meyer scale; the imaging device adopts a multi-ring electrode array and an improved D-bar algorithm to reconstruct a brain electrical conductivity distribution map, constructs a multi-level decision tree early warning model by extracting impedance change rate Delta Z, spatial gradient entropy SGE and hemisphere asymmetry Asym, and realizes brain edema detection. Through multi-modal data fusion and minute-level response, the system solves the fragmentation problem of postoperative monitoring and provides precise decision support for cerebral apoplexy rehabilitation.

Description

Technical Field

[0001] This invention relates to the field of medical monitoring technology, specifically a post-stroke vital signs monitoring system that integrates multimodal biosensing and artificial intelligence analysis to achieve real-time quantitative assessment of neuromotor function and the risk of cerebral edema. Background Technology

[0002] Stroke, commonly known as cerebrovascular accident, is divided into two types: ischemic stroke and hemorrhagic stroke. It is a disease caused by various factors that damage cerebral blood vessels, resulting in focal or systemic brain tissue damage. Stroke is characterized by high incidence, disability rate, recurrence rate, and mortality rate; ischemic stroke accounts for 75% to 90% of all strokes, while hemorrhagic stroke accounts for 10% to 25%. Men, obese individuals, and diabetic patients are considered high-risk groups.

[0003] The prevention and treatment of post-stroke complications faces severe challenges. Current clinical monitoring suffers from three major deficiencies: First, cerebral edema detection relies on intermittent CT / MRI images, posing a high risk to patient transport and resulting in an average delay of over 6 hours, causing 35% of patients to miss the golden intervention window. Second, motor function assessment still uses the Fugl-Meyer scale, which takes over 30 minutes per session and is influenced by physician subjective experience; while wearable devices can collect electromyographic signals, they lack clinical interpretability, with correlation coefficients of only 0.55-0.62. Third, cerebral edema and motor dysfunction data are fragmented; existing equipment, such as intracranial pressure monitors and rehabilitation robots, operate independently, failing to establish a pathological-neurological function correlation model. More seriously, the fragmentation of traditional monitoring methods leads to delayed clinical decision-making, with a 28% missed diagnosis rate for complications within 72 hours post-surgery. Therefore, there is an urgent need for an intelligent monitoring system that can integrate multimodal physiological parameters in real time at the bedside, possessing both millimeter-level spatial resolution and minute-level response speed, to overcome the temporal and spatial limitations of post-operative monitoring. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a post-stroke vital signs monitoring system, comprising a neuromotor function assessment wristband device, a head-mounted brain electrical impedance dynamic imaging device, a multimodal data fusion processor, and a clinical early warning terminal.

[0005] The neuromotor function assessment wristband device includes: a surface electromyography signal acquisition module, a joint motion biomechanics sensing module, and a tactile guidance module, used to acquire the patient's upper limb electromyographic activity parameters and finger movement trajectory data in real time; wherein the surface electromyography signal acquisition module adopts a multi-channel microneedle dry electrode array, and the electrode spacing is configured to cover the anatomical area of ​​the forearm flexor muscle group; the joint motion biomechanics sensing module includes piezoresistive sensors distributed in the metacarpophalangeal joints and interphalangeal joints;

[0006] The head-mounted brain electrical impedance dynamic imaging device includes a multi-ring distributed electrode array, a precision current excitation source, and a boundary voltage measurement circuit, which is used to dynamically reconstruct the spatial distribution of brain tissue conductivity.

[0007] The multimodal data fusion processor performs quantitative scoring of motor function and determination of cerebral edema risk level by synchronizing data streams from the wristband device and the head-mounted electroencephalogram device in time.

[0008] The clinical early warning terminal receives the analysis results and generates multi-level alarm signals.

[0009] Furthermore, in some embodiments, the neuromotor function assessment wristband device further includes a signal conditioning unit, an embedded control unit, and a wireless transmission unit; the signal conditioning unit uses a high input impedance amplifier as a preamplifier after a multi-channel microneedle dry electrode array, and the subsequent stage is connected to a programmable gain amplifier and a high-precision analog-to-digital converter to acquire the patient's upper limb electrophysiological activity parameters; the mechanical sensing channel is equipped with a constant voltage source circuit, and the mechanical parameters are inverted by measuring the resistance change of the piezoresistive sensor, and the conversion formula is:

[0010] ;

[0011] Where F represents the force on the joint; R represents the real-time resistance value of the piezoresistive sensor. The zero-force reference resistance value is represented; k represents the material property constant; the embedded control unit performs real-time kinematic modeling and feature extraction; the wireless transmission unit uses a low-latency wireless protocol to transmit data.

[0012] Furthermore, in some implementations, the quantitative scoring of motor function includes the following steps: calculating the fingertip spatial coordinates based on a kinematic chain model; the input feature vector of the motor function scoring model includes electromyographic time-domain features, electromyographic frequency-domain features, joint angle values, and grip strength change rate; and outputting a standardized score through a multi-layer neural network to obtain the motor function score, specifically including:

[0013] Electromyography temporal feature extraction: Calculate the average amplitude (MAV) and zero-crossing rate (ZC) within a sliding window for surface electromyography signals;

[0014] ;

[0015] in This represents the discrete sampled values ​​of the electromyographic signal, where N represents the number of sampling points within the window;

[0016] Electromyography frequency domain feature extraction: The power spectral density is calculated by fast Fourier transform, and the median frequency MF is solved;

[0017] Joint kinematic feature extraction: The angles of the metacarpophalangeal and interphalangeal joints are calculated based on piezoresistive sensor signals, and the smoothness of the motion trajectory is also calculated.

[0018] ;

[0019] Where r(t) represents the fingertip spatial position function, Indicates the duration of the action;

[0020] Grip force dynamic feature extraction: Calculation of grip force change rate: ;

[0021] Indicates the maximum grip strength during the task period. Indicates the minimum grip strength, Indicates the duration of the task;

[0022] The following features were combined into an input vector: MAV, ZC, MF values ​​of 8-channel electromyography, angles of 5 metacarpophalangeal joints, angles of 8 interphalangeal joints, trajectory smoothness, and grip strength change rate.

[0023] The feature vectors are input into a three-layer fully connected neural network. The network structure is as follows:

[0024] Input layer: 39 nodes;

[0025] Hidden layer 1: 64 nodes, activation function is ReLU;

[0026] Dropout layer: Drop rate 20%;

[0027] Hidden layer 2: 32 nodes, activation function is ReLU;

[0028] Output layer: 1 node, linear activation;

[0029] The neural network outputs a continuous score from 0 to 100, which is linearly correlated with the total score of the clinical Fugl-Meyer scale. The neural network score is then mapped to the standard Fugl-Meyer action score.

[0030] The neural network scoring interval [90, 100] corresponds to a Fugl-Meyer motion score of 2 points, with the criteria being joint angle error < 5° and trajectory smoothness < 0.5; the neural network scoring interval [80, 90) corresponds to a Fugl-Meyer motion score of 1 point, with the criteria being joint angle error 5°-15° or smoothness 0.5-1.0; the neural network scoring interval [0, 80) corresponds to a Fugl-Meyer motion score of 0 points, with the criteria being joint angle error > 15° or smoothness > 1.0.

[0031] Furthermore, in some embodiments, the head-mounted electroencephalogram (EEG) device specifically includes a current excitation module, a voltage detection module, and a control logic module; the current excitation module uses a current pump topology to output a sinusoidal current of a specific frequency; the voltage detection module includes an amplifier, a bandpass filter, a synchronous demodulator, and a low-pass filter; the control logic module performs multi-electrode scanning to achieve high-frequency data acquisition.

[0032] Furthermore, in some implementations, electrical impedance image reconstruction specifically involves: constructing a three-dimensional head model based on medical imaging data; and using a scattering transform algorithm to solve for the conductivity distribution.

[0033]

[0034] Where k represents the space wavenumber vector; The Dirichlet-to-Neumann mapping representing the actual conductivity distribution; ψ represents a uniform conductivity mapping; ψ represents a characteristic function.

[0035] Furthermore, in some implementations, time synchronization is achieved by the following methods: the two devices share a high-precision clock source; the imaging device generates an electrical pulse trigger signal at the start of each frame of data acquisition; and the wristband device records a precise timestamp through an interrupt mechanism to establish a data acquisition time window.

[0036] Furthermore, in some implementations, the cerebral edema risk warning model specifically involves: extracting global impedance change rate, spatial gradient entropy, and left-right hemisphere asymmetry features from the electrical impedance distribution map; and using a multi-level decision tree architecture risk warning model for early warning.

[0037] Furthermore, in some implementations, the global impedance change rate ΔZ, spatial gradient entropy SGE, and left-right hemisphere asymmetry Asymmetry are calculated as follows:

[0038] ;

[0039] Where N represents the total number of pixels in the electrical impedance distribution map; This represents the conductivity value of the i-th pixel at time t; This indicates the baseline electrical conductivity value after surgery;

[0040] ;

[0041] Where M represents the number of gradient direction partitions; This represents the gradient magnitude within the j-th partition and its proportion of the global gradient energy; the gradient magnitude is calculated using the Sobel operator:

[0042] ;

[0043] ;

[0044] in This represents the average conductivity of the left hemisphere; This indicates the average electrical conductivity of the right hemisphere; the division of the hemispheres is based on the sagittal suture of the brain.

[0045] Furthermore, in some implementations, the risk warning model employs a multi-level decision tree architecture:

[0046] First-level decision: If ΔZ > α, proceed to the second-level decision; otherwise, output a safe state.

[0047] Second-level judgment: If Asym ≤ γ and SGE ≤ β, a low-risk alarm is triggered; if Asym > γ and SGE ≤ β, a medium-risk alarm is triggered; if SGE > β, a high-risk alarm is triggered.

[0048] Level 3 judgment: When ΔZ > δ and the motor function score decreases by more than η within 24 hours, a comprehensive risk alarm is triggered;

[0049] Where α, β, γ, δ, and η are clinically validated thresholds, and δ > α and β·γ > 1.

[0050] Furthermore, in some implementations, a baseline adaptive update module is included:

[0051] Initial baseline Established during the 24-hour stable period post-surgery;

[0052] Baseline calibration is performed automatically every morning at midnight:

[0053] ;

[0054] in The average conductivity at rest during the night; Forgetting factor;

[0055] When device displacement is detected, real-time baseline recalibration is triggered:

[0056] ;

[0057] If the shift index If the threshold θ is exceeded, the warning will be suspended and the electrode contact inspection process will be initiated.

[0058] The beneficial effects of this invention are as follows:

[0059] This invention innovatively integrates a neuromotor function assessment wristband with head-mounted impedance dynamic imaging, achieving three major breakthroughs:

[0060] First, by using a 39-dimensional feature vector that includes electromyographic time-frequency domain parameters, joint kinematics, and grip dynamics to drive neural network scoring, the Fugl-Meyer assessment is made more objective and continuous, the correlation between the score and the gold standard is improved, the time for a single assessment is reduced from 30 minutes to a few seconds, and it is accurately mapped to clinical movement scores, thus solving the problem of rehabilitation quantification.

[0061] Secondly, the improved D-bar algorithm was used to reconstruct the brain impedance distribution map, with a spatial resolution of up to 8mm. Combined with the impedance change rate ΔZ, spatial gradient entropy SGE, and hemispherical asymmetry Asym multi-parameter decision tree, the sensitivity of cerebral edema detection was improved compared with CT, and the warning time was several hours earlier on average than traditional imaging.

[0062] Most importantly, the first dynamic correlation model between "decline in motor score and progression of edema" was established. A comprehensive alarm is triggered when ΔZ > 0.15 and the decline in motor score > 20% within 24 hours, improving the detection rate of neurological function deterioration. Clinical validation shows that this system reduces disability rates and unnecessary ICU transfers. In addition, flexible microneedle electrodes and adaptive baseline calibration technology further reduce false alarms, providing a precise tool for post-stroke monitoring. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Appendix Figure 1 This is a schematic diagram of the system architecture of the present invention;

[0065] Appendix Figure 2 This is a schematic diagram of the neuromotor function assessment wristband device of the present invention;

[0066] Appendix Figure 3 This is a schematic diagram of the structure of the head-mounted brain electrical impedance dynamic imaging device of the present invention;

[0067] Appendix Figure 4 This is an external view of the head-mounted brain electrical impedance dynamic imaging device of the present invention;

[0068] Appendix Figure 5 This is a schematic diagram of the neuromotor function assessment wristband device of the present invention. Detailed Implementation

[0069] Example 1:

[0070] Please see the appendix Figure 1 To be continued Figure 5A post-stroke vital signs monitoring system includes a neuromotor function assessment wristband, a head-mounted brain electrical impedance dynamic imaging device, a multimodal data fusion processor, and a clinical early warning terminal.

[0071] The neuromotor function assessment wristband device includes: a surface electromyography signal acquisition module, a joint motion biomechanics sensing module, and a tactile guidance module, used to acquire the patient's upper limb electromyographic activity parameters and finger movement trajectory data in real time; wherein the surface electromyography signal acquisition module adopts a multi-channel microneedle dry electrode array, and the electrode spacing is configured to cover the anatomical area of ​​the forearm flexor muscle group; the joint motion biomechanics sensing module includes piezoresistive sensors distributed in the metacarpophalangeal joints and interphalangeal joints;

[0072] The head-mounted brain electrical impedance dynamic imaging device includes a multi-ring distributed electrode array, a current excitation source, and a boundary voltage measurement circuit, which is used to dynamically reconstruct the spatial distribution of brain tissue conductivity.

[0073] The multimodal data fusion processor performs quantitative scoring of motor function and determination of cerebral edema risk level by synchronizing data streams from the wristband device and the head-mounted electroencephalogram device in time.

[0074] The clinical early warning terminal receives the analysis results and generates multi-level alarm signals.

[0075] Furthermore, in some embodiments, the neuromotor function assessment wristband device further includes a signal conditioning unit, an embedded control unit, and a wireless transmission unit; the signal conditioning unit uses a high input impedance amplifier as a preamplifier after a multi-channel microneedle dry electrode array, and the subsequent stage is connected to a programmable gain amplifier and a high-precision analog-to-digital converter to acquire the patient's upper limb electrophysiological activity parameters; the mechanical sensing channel is equipped with a constant voltage source circuit, and the mechanical parameters are inverted by measuring the resistance change of the piezoresistive sensor, and the conversion formula is:

[0076] ;

[0077] Where F represents the force on the joint; R represents the real-time resistance value of the piezoresistive sensor. The zero-force reference resistance value is represented; k represents the material property constant; the embedded control unit performs real-time kinematic modeling and feature extraction; the wireless transmission unit uses a low-latency wireless protocol to transmit data.

[0078] Furthermore, in some implementations, the quantitative scoring of motor function includes the following steps: calculating the fingertip spatial coordinates based on a kinematic chain model; the input feature vector of the motor function scoring model includes electromyographic time-domain features, electromyographic frequency-domain features, joint angle values, and grip strength change rate; and outputting a standardized score through a multi-layer neural network to obtain the motor function score, specifically including:

[0079] Electromyography temporal feature extraction: Calculate the average amplitude (MAV) and zero-crossing rate (ZC) within a sliding window for surface electromyography signals;

[0080] ;

[0081] in This represents the discrete sampled values ​​of the electromyographic signal, where N represents the number of sampling points within the window;

[0082] Electromyography frequency domain feature extraction: The power spectral density is calculated by fast Fourier transform, and the median frequency MF is solved;

[0083] Joint kinematic feature extraction: The angles of the metacarpophalangeal and interphalangeal joints are calculated based on piezoresistive sensor signals, and the smoothness of the motion trajectory is also calculated.

[0084] ;

[0085] Where r(t) represents the fingertip spatial position function, Indicates the duration of the action;

[0086] Grip force dynamic feature extraction: Calculation of grip force change rate: ;

[0087] Indicates the maximum grip strength during the task period. Indicates the minimum grip strength, Indicates the duration of the task;

[0088] The following features were combined into an input vector: MAV, ZC, MF values ​​of 8-channel electromyography, angles of 5 metacarpophalangeal joints, angles of 8 interphalangeal joints, trajectory smoothness, and grip strength change rate.

[0089] The feature vectors are input into a three-layer fully connected neural network. The network structure is as follows:

[0090] Input layer: 39 nodes;

[0091] Hidden layer 1: 64 nodes, activation function is ReLU;

[0092] Dropout layer: Drop rate 20%;

[0093] Hidden layer 2: 32 nodes, activation function is ReLU;

[0094] Output layer: 1 node, linear activation;

[0095] The neural network outputs a continuous score from 0 to 100, which is linearly correlated with the total score of the clinical Fugl-Meyer scale. The neural network score is then mapped to the standard Fugl-Meyer action score.

[0096] The neural network scoring interval [90, 100] corresponds to a Fugl-Meyer motion score of 2 points, with the criteria being joint angle error < 5° and trajectory smoothness < 0.5; the neural network scoring interval [80, 90) corresponds to a Fugl-Meyer motion score of 1 point, with the criteria being joint angle error 5°-15° or smoothness 0.5-1.0; the neural network scoring interval [0, 80) corresponds to a Fugl-Meyer motion score of 0 points, with the criteria being joint angle error > 15° or smoothness > 1.0.

[0097] Furthermore, in some embodiments, the head-mounted electroencephalogram (EEG) device specifically includes a current excitation module, a voltage detection module, and a control logic module; the current excitation module uses a current pump topology to output a sinusoidal current of a specific frequency; the voltage detection module includes an amplifier, a bandpass filter, a synchronous demodulator, and a low-pass filter; the control logic module performs multi-electrode scanning to achieve high-frequency data acquisition.

[0098] Furthermore, in some implementations, electrical impedance image reconstruction specifically involves: constructing a three-dimensional head model based on medical imaging data; and using a scattering transform algorithm to solve for the conductivity distribution.

[0099]

[0100] Where k represents the space wavenumber vector; The Dirichlet-to-Neumann mapping representing the actual conductivity distribution; ψ represents a uniform conductivity mapping; ψ represents a characteristic function.

[0101] Furthermore, in some implementations, time synchronization is achieved by the following methods: the two devices share a high-precision clock source; the imaging device generates an electrical pulse trigger signal at the start of each frame of data acquisition; and the wristband device records a precise timestamp through an interrupt mechanism to establish a data acquisition time window.

[0102] Furthermore, in some implementations, the cerebral edema risk warning model specifically involves: extracting global impedance change rate, spatial gradient entropy, and left-right hemisphere asymmetry features from the electrical impedance distribution map; and using a multi-level decision tree architecture risk warning model for early warning.

[0103] Furthermore, in some implementations, the global impedance change rate ΔZ, spatial gradient entropy SGE, and left-right hemisphere asymmetry Asymmetry are calculated as follows:

[0104] ;

[0105] Where N represents the total number of pixels in the electrical impedance distribution map; This represents the conductivity value of the i-th pixel at time t; This indicates the baseline electrical conductivity value after surgery;

[0106] ;

[0107] Where M represents the number of gradient direction partitions; This represents the gradient magnitude within the j-th partition and its proportion of the global gradient energy; the gradient magnitude is calculated using the Sobel operator:

[0108] ;

[0109] ;

[0110] in This represents the average conductivity of the left hemisphere; This indicates the average electrical conductivity of the right hemisphere; the division of the hemispheres is based on the sagittal suture of the brain.

[0111] Furthermore, in some implementations, the risk warning model employs a multi-level decision tree architecture:

[0112] First-level decision: If ΔZ > α, proceed to the second-level decision; otherwise, output a safe state.

[0113] Second-level judgment: If Asym ≤ γ and SGE ≤ β, a low-risk alarm is triggered; if Asym > γ and SGE ≤ β, a medium-risk alarm is triggered; if SGE > β, a high-risk alarm is triggered.

[0114] Level 3 judgment: When ΔZ > δ and the motor function score decreases by more than η within 24 hours, a comprehensive risk alarm is triggered;

[0115] Where α, β, γ, δ, and η are clinically validated thresholds, and δ > α and β·γ > 1.

[0116] Furthermore, in some implementations, a baseline adaptive update module is included:

[0117] Initial baseline Established during the 24-hour stable period post-surgery;

[0118] Baseline calibration is performed automatically every morning at midnight:

[0119] ;

[0120] in The average conductivity at rest during the night; Forgetting factor;

[0121] When device displacement is detected, real-time baseline recalibration is triggered:

[0122] ;

[0123] If the shift index If the threshold θ is exceeded, the warning will be suspended and the electrode contact inspection process will be initiated.

[0124] Example 2:

[0125] Creation of a neuromotor function assessment wristband

[0126] Take a 0.2mm thick polyimide flexible circuit board and etch 8 channels of electromyography signal traces;

[0127] Install the microneedle electrode array (model BioFlex-M8): locate the center of the belly of the flexor carpi radialis muscle, with an electrode spacing of 12mm;

[0128] Soldering signal chain components: Connect INA333 pins 1-2 to the positive and negative terminals of the electrodes, connect the output terminal to the IN+ of PGA280 after connecting a 10Hz high-pass filter in series, and connect the PGA output to the AIN1 channel of ADS1299.

[0129] Adhesive force sensors: FSR-402 thin-film pressure sensors (range 0-50N) are installed at the metacarpophalangeal joints, and Flex-3A micro-bending sensors are installed at the interphalangeal joints;

[0130] The main control board integrates an STM32F407VGT6 minimum system board connected to a DRV2605L haptic driver chip, with a vibration motor (model LRA2010) placed inside the wristband.

[0131] Sealing treatment: Medical silicone is used to pot the circuit area, with a Shore hardness of A35 and a curing time of 24 hours.

[0132] Fabrication of a head-mounted electrical impedance tomography (EEG) device

[0133] 3D printed headband frame: using an EOS P396 laser sintering machine, nylon 12 powder, layer thickness 0.1mm;

[0134] Upper ring diameter 22cm (frontal-parietal bone layer): 8 electrodes evenly distributed; Middle ring diameter 20cm (parietal bone layer): 16 electrodes; Lower ring diameter 18cm (occipital bone layer): 8 electrodes;

[0135] Electrode installation: Insert the titanium alloy electrode post (3mm in diameter) into the reserved hole and weld the wire to the signal board.

[0136] Circuit assembly:

[0137] Current excitation board: AD822 op-amp forms Howland current pump, feedback resistor 10kΩ, accuracy 0.1%;

[0138] Voltage detection board: AD8421 gain 50x → LTC1562 bandpass filter (45-55kHz) → AD630 demodulator.

[0139] Conductive gel filling: Inject hydroxyethyl cellulose gel (formulation: 0.9% NaCl + 2% HEC) into the syringe, 0.2 ml per electrode cavity.

[0140] System parameter configuration

[0141] Electromyography (EMG) signal acquisition: The bandpass filter cutoff frequency is set to a specific low-frequency to a specific high-frequency value, and the notch filter center frequency is set to the power frequency interference value. Mechanical sensor calibration: A reference resistance value is recorded under zero-load conditions; a known force is applied, and the corresponding resistance change is recorded; material property constants are calculated. Electrical impedance imaging: The excitation current amplitude follows medical electrical safety standards, and the scanning protocol uses an adjacent electrode drive measurement mode.

[0142] Kinematic model: Finger bone length parameters are set based on human anatomical data. Neural network model: The number of nodes in the input layer corresponds to the dimension of the feature vector, the number of nodes in the hidden layer is set according to empirical formulas, and the output layer is configured with a linear activation function. Image reconstruction: The convolutional neural network structure contains multiple convolutional layers and downsampling layers, and the mean squared error loss function is used during training.

[0143] Clinical operating procedures

[0144] The patient assumes a standard sitting position with the forearm horizontal. Electromyography baseline acquisition is initiated: surface electrical signals are recorded for a specific time period in a static state. Mechanometry zero-point calibration is performed: sensor output values ​​are read in the joint's natural extension position. Electrical impedance reference frame acquisition: initial electrical impedance data are acquired in a stable postoperative state.

[0145] Tactile guidance protocol: Phase 1: Continuous vibration prompts the initiation of shoulder abduction movement; Phase 2: Pulsed vibration prompts the initiation of wrist dorsiflexion movement; Phase 3: Double pulse prompts the initiation of finger pinching movement.

[0146] Synchronous data acquisition: The electrical impedance imaging device starts the scanning cycle at fixed time intervals, and the wristband device opens a specific time window to acquire electromyographic and mechanical data after receiving the synchronous trigger signal.

[0147] Motor function score calculation:

[0148] The time-domain features of the electromyography (EMG) signal are extracted, including the average amplitude and zero-crossing rate. The frequency-domain features of the EMG signal are extracted, including the median frequency. The trajectory of the change in joint angle is calculated, the gradient of the change in grip force is calculated, and the multidimensional feature vector is input into the neural network model to output the score value.

[0149] Cerebral edema risk assessment:

[0150] Reconstruct the current brain impedance distribution map, calculate the global impedance relative change rate, calculate the information entropy of the impedance spatial gradient distribution, calculate the impedance asymmetry between the left and right hemispheres, and apply multi-level threshold judgment rules to generate risk levels.

[0151] Comprehensive decision-making:

[0152] Establish a correlation model between the trend of motor score changes and the risk of cerebral edema, and trigger an alarm when the preset risk combination conditions are met.

[0153] Clinical report generation

[0154] Data format: The data uses a structured medical information format to encapsulate fields such as timestamps, exercise scores, and risk levels.

[0155] Transmission protocol: The information is pushed to the hospital information system via a standard medical information exchange protocol.

[0156] Visual interface: Displays a patient status matrix diagram on the nurse station terminal.

[0157] Thus far, the description of the above embodiments has been provided for illustrative and descriptive purposes. This is not intended to be exhaustive or limiting of the present disclosure. Individual elements or features of particular embodiments are generally not limited to those particular embodiments, but may be interchanged and used in selected embodiments where applicable, even if not specifically shown or described. In many respects, the same elements or features may also be varied. Such variations are not considered a departure from this disclosure, and all such modifications are intended to be included within the scope of this disclosure.

[0158] Example embodiments are provided so that this disclosure will become thorough and will fully convey the scope to those skilled in the art. Numerous details, such as examples of specific parts, apparatus, and methods, are set forth to provide a thorough understanding of embodiments of this disclosure. It will be apparent to those skilled in the art that the specific details are not required, and the example embodiments may be implemented in many different forms, neither of which should be construed as limiting the scope of this disclosure. In some example embodiments, well-known processes, well-known apparatus structures, and well-known techniques are not described in detail.

[0159] Technical terms are used herein for the purpose of describing specific exemplary embodiments only and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a” and “the” as used herein may also refer to the plural forms. The terms “comprising” and “having” are inclusive and therefore specify the presence of the stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or additional having of one or more other features, integrals, steps, operations, elements, components, and / or combinations thereof. Unless expressly indicated in order of execution, the method steps, processes, and operations described herein are not to be construed as necessarily requiring performance in the specific order discussed and shown. It should also be understood that additional or optional steps may be employed.

Claims

1. A post-stroke vital signs monitoring system, comprising a neuromotor function assessment wristband, a head-mounted brain electrical impedance dynamic imaging device, a multimodal data fusion processor, and a clinical early warning terminal; characterized in that: The neuromotor function assessment wristband device includes: a surface electromyography signal acquisition module, a joint motion biomechanics sensing module, and a tactile guidance module, used to acquire the patient's upper limb electromyographic activity parameters and finger movement trajectory data in real time; wherein the surface electromyography signal acquisition module adopts a multi-channel microneedle dry electrode array, and the electrode spacing is configured to cover the anatomical area of ​​the forearm flexor muscle group; the joint motion biomechanics sensing module includes piezoresistive sensors distributed in the metacarpophalangeal joints and interphalangeal joints; The head-mounted brain electrical impedance dynamic imaging device includes a multi-ring distributed electrode array, a precision current excitation source, and a boundary voltage measurement circuit, which is used to dynamically reconstruct the spatial distribution of brain tissue conductivity. The multimodal data fusion processor performs quantitative scoring of motor function and determination of cerebral edema risk level by synchronizing data streams from the wristband device and the head-mounted electroencephalogram device in time. The clinical early warning terminal receives the analysis results and generates multi-level alarm signals; The cerebral edema risk warning model is as follows: global impedance change rate, spatial gradient entropy, and left-right hemisphere asymmetry features are extracted from the electrical impedance distribution map; a multi-level decision tree architecture risk warning model is used for early warning. The calculation methods for global impedance change rate ΔZ, spatial gradient entropy SGE, and hemispherical asymmetry Asymmetry are as follows: ； Where N represents the total number of pixels in the electrical impedance distribution map; This represents the conductivity value of the i-th pixel at time t; This indicates the baseline electrical conductivity value after surgery; ； Where M represents the number of gradient direction partitions; This represents the gradient magnitude within the j-th partition and its proportion of the global gradient energy; the gradient magnitude is calculated using the Sobel operator: ； ； in This represents the average conductivity of the left hemisphere; This indicates the average electrical conductivity of the right hemisphere; the division of the hemispheres is based on the sagittal suture of the brain. The risk warning model adopts a multi-level decision tree architecture: First-level decision: If ΔZ > α, proceed to the second-level decision; otherwise, output a safe state. Second-level judgment: If Asym ≤ γ and SGE ≤ β, a low-risk alarm is triggered; if Asym > γ and SGE ≤ β, a medium-risk alarm is triggered; if SGE > β, a high-risk alarm is triggered. Level 3 judgment: When ΔZ > δ and the motor function score decreases by more than η within 24 hours, a comprehensive risk alarm is triggered; Where α, β, γ, δ, and η are clinically validated thresholds, and δ > α and β·γ > 1.

2. The postoperative vital signs monitoring system for stroke according to claim 1, characterized in that: The neuromotor function assessment wristband device further includes a signal conditioning unit, an embedded control unit, and a wireless transmission unit. The signal conditioning unit employs a high-input-impedance amplifier as a preamplifier after a multi-channel microneedle dry electrode array, followed by a programmable gain amplifier and a high-precision analog-to-digital converter to acquire electrophysiological parameters of the patient's upper limb muscles. The mechanical sensing channel is equipped with a constant voltage source circuit, which inversely retrieves mechanical parameters by measuring the resistance change of a piezoresistive sensor; the conversion formula is as follows: ； Where F represents the force on the joint; R represents the real-time resistance value of the piezoresistive sensor; R0 represents the zero-force reference resistance value; k represents the material property constant; the embedded control unit performs real-time kinematic modeling and feature extraction; The wireless transmission unit uses a low-latency wireless protocol to transmit data.

3. The postoperative vital signs monitoring system for stroke according to claim 2, characterized in that: The quantitative assessment of motor function includes the following steps: calculating the fingertip spatial coordinates based on a kinematic chain model; the input feature vector of the motor function assessment model includes electromyographic time-domain features, electromyographic frequency-domain features, joint angle values, and grip strength change rate; and outputting a standardized score through a multi-layer neural network to obtain the motor function score, specifically including: Electromyography temporal feature extraction: Calculate the average amplitude (MAV) and zero-crossing rate (ZC) within a sliding window for surface electromyography signals; ； in This represents the discrete sampled values ​​of the electromyographic signal, where N represents the number of sampling points within the window; Electromyography frequency domain feature extraction: The power spectral density is calculated by fast Fourier transform, and the median frequency MF is solved; Joint kinematic feature extraction: The angles of the metacarpophalangeal and interphalangeal joints are calculated based on piezoresistive sensor signals, and the smoothness of the motion trajectory is also calculated. ； Where r(t) represents the fingertip spatial position function, Indicates the duration of the action; Grip force dynamic feature extraction: Calculation of grip force change rate: ; Indicates the maximum grip strength during the task period. Indicates the minimum grip strength, Indicates the duration of the task; The following features were combined into an input vector: MAV, ZC, MF values ​​of 8-channel electromyography, angles of 5 metacarpophalangeal joints, angles of 8 interphalangeal joints, trajectory smoothness, and grip strength change rate. The feature vectors are input into a three-layer fully connected neural network. The network structure is as follows: Input layer: 39 nodes; Hidden layer 1: 64 nodes, activation function is ReLU; Dropout layer: Drop rate 20%; Hidden layer 2: 32 nodes, activation function is ReLU; Output layer: 1 node, linear activation; The neural network outputs a continuous score from 0 to 100, which is linearly correlated with the total score of the clinical Fugl-Meyer scale. The neural network score is then mapped to the standard Fugl-Meyer action score. The neural network scoring interval [90, 100] corresponds to a Fugl-Meyer motion score of 2 points, with the criteria being joint angle error < 5° and trajectory smoothness < 0.5; the neural network scoring interval [80, 90) corresponds to a Fugl-Meyer motion score of 1 point, with the criteria being joint angle error 5°-15° or smoothness 0.5-1.0; the neural network scoring interval [0, 80) corresponds to a Fugl-Meyer motion score of 0 points, with the criteria being joint angle error > 15° or smoothness > 1.

0.

4. The postoperative vital signs monitoring system for stroke according to claim 1, characterized in that: The head-mounted electroencephalography (EEG) device further includes a current excitation module, a voltage detection module, and a control logic module. The current excitation module uses a current pump topology to output a sinusoidal current. The voltage detection module includes an amplifier, a bandpass filter, a synchronous demodulator, and a low-pass filter. The control logic module performs multi-electrode scanning to achieve high-frequency data acquisition.

5. The postoperative vital signs monitoring system for stroke according to claim 4, characterized in that: Electrical impedance image reconstruction specifically involves: constructing a three-dimensional head model based on medical imaging data; and using a scattering transform algorithm to solve for the conductivity distribution. ； Where k represents the space wavenumber vector; The Dirichlet-to-Neumann mapping representing the actual conductivity distribution; ψ represents a uniform conductivity mapping; ψ represents a characteristic function.

6. The postoperative vital signs monitoring system for stroke according to claim 1, characterized in that: Time synchronization is achieved in the following ways: the two devices share a high-precision clock source; the imaging device generates an electrical pulse trigger signal at the start of each frame of data acquisition; and the wristband device records a precise timestamp through an interrupt mechanism to establish a data acquisition time window.

7. The postoperative vital signs monitoring system for stroke according to claim 6, characterized in that: Includes a baseline adaptive update module: Initial baseline Established during the 24-hour stable period post-surgery; Baseline calibration is performed automatically every morning at midnight: ； in The average conductivity at rest during the night; Forgetting factor; When device displacement is detected, real-time baseline recalibration is triggered: ； If the shift index If the threshold θ is exceeded, the warning will be suspended and the electrode contact inspection process will be initiated.

Images