A multimodal fusion continuous neuromuscular monitoring system and storage medium

Abstract

The present application belongs to the field of multi-modal signal processing, and particularly relates to a multi-modal fusion continuous muscle relaxation monitoring system and a storage medium. The present application applies standardized electrical stimulation to respiratory muscle innervation nerves, synchronously collects stimulation-induced electromyographic signals, myoacoustic signals, muscle mechanical vibration signals and muscle oxygen signals, and performs time synchronization, feature extraction and fusion analysis on the multi-modal signals to construct a comprehensive evaluation index reflecting the recovery state of respiratory muscle nerve-muscle function, thereby realizing continuous monitoring of muscle relaxation metabolism and safe anesthesia management. Not only does the present application solve the core pain points of the prior art, such as "non-respiratory muscle, non-continuous and easy to interfere", but also improves muscle relaxation monitoring from "single parameter judgment" to "functional state evaluation" through electrical-mechanical coupling quantization and continuous trend analysis, thereby providing key technical support for precise anesthesia and patient safety.

Description

Technical Field

[0001] This invention belongs to the field of multimodal signal processing, specifically relating to a multimodal fusion continuous muscle relaxation monitoring system and storage medium. Background Technology

[0002] Neuromuscular relaxation monitoring technology is mainly used to assess the degree of neuromuscular blockade to guide the rational use of perioperative neuromuscular relaxants and prevent respiratory complications caused by residual neuromuscular relaxation after surgery. Early clinical practice relied mainly on subjective signs such as head lifting, handshake, and eye opening for empirical judgment; this method was unreliable and difficult to accurately identify residual neuromuscular relaxation. Since the 1950s, peripheral nerve stimulators have been used clinically, enabling electrical stimulation assessment of neuromuscular function. In the 1980s, neuromuscular relaxation monitoring entered a period of quantitative development; by the 1990s, a neuromuscular relaxation monitor based on the acceleration principle was successfully developed, achieving continuous, quantitative, and real-time measurement of the degree of neuromuscular relaxation. Currently, commonly used quantitative monitoring techniques in clinical practice include accelerometer (AMG), electromyography (EMG), piezoelectric sensor methods, and diaphragmatic ultrasound, among which the TOF (four-stage stimulation) ratio is the gold standard for assessing the degree of neuromuscular relaxation recovery. Compared with traditional qualitative assessment, quantitative neuromuscular relaxation monitoring can significantly reduce the incidence of postoperative residual neuromuscular relaxation and improve perioperative patient safety.

[0003] Traditional qualitative muscle relaxation monitoring involves various transcutaneous nerve stimulation modalities, including singletwitch stimulation, train-of-four (TOF) stimulation, post-tetanic count, and double-burst stimulation. Traditional quantitative muscle relaxation monitoring, developed from qualitative methods, quantifies perioperative muscle relaxation and is currently the most commonly used clinical approach. This includes mechanomyography (MG), acceleration-motor kinetic recording (AMG), electromyography (EMG), kinemyography (KMG), and phonomyography (PMG). However, traditional techniques still have limitations, such as: I. Traditional quantitative muscle relaxation monitoring uses peripheral muscles (such as the adductor thumb) as the target muscles, which makes it difficult to truly reflect the functional status of respiratory muscle groups such as the diaphragm and intercostal muscles.

[0004] Second, traditional techniques require manual, intermittent operation and cannot provide continuous monitoring data throughout the perioperative period.

[0005] Third, when using a single sensing mode, it is easily affected by external interference such as surgical electrosurgical unit and environmental noise in complex operating room environments. Summary of the Invention

[0006] To address the problems of existing technologies, this invention provides a multimodal fusion continuous muscle relaxation monitoring system, method, and storage medium.

[0007] A multimodal fusion continuous muscle relaxation monitoring system includes the following modules: The neural electrical stimulation module is configured to apply standardized electrical stimulation to the nerves innervating the respiratory muscles; and to generate a corresponding stimulation event labeling signal each time standardized electrical stimulation is applied. The signal acquisition module is configured to acquire multimodal signals from patients before anesthesia induction or before the action of muscle relaxants, and process them through this system to use them as baseline data; and to acquire multimodal signals induced by standardized electrical stimulation. The signal synchronization and preprocessing module is configured to perform time alignment of multimodal signals based on stimulus event labeling signals, and to preprocess the multimodal signals to generate multimodal stable signals. The feature extraction and fusion module is configured to: quantitatively analyze the multimodal stable signal to obtain the muscle relaxation depth; extract features from the multimodal stable signal to obtain multimodal signal features; perform temporal consistency checks on the multimodal signal features to determine their validity; adjust the multimodal signal feature fusion weights based on the muscle relaxation depth; fuse the valid multimodal signal features according to the fusion weights to obtain the coupling strength; and obtain the coupling index based on the coupling strength. The muscle relaxation state assessment and decision-making module is configured to map the obtained coupling index and the temporal information of multimodal signal characteristics into functional state indicators that can reflect changes in muscle relaxation state; analyze the absolute level of the functional state indicators and their trend over time; continuously assess the effects and metabolic processes of muscle relaxant drugs; and determine the state of respiratory muscle function, whether the state is in a state of blockade, recovery, or stable recovery. The human-computer interaction and data storage module is configured to receive respiratory muscle function status information from the muscle relaxation status assessment and decision module, as well as real-time multimodal stabilization signals from the signal synchronization and preprocessing module, and to visualize the status information, multimodal stabilization signals, and functional status indicators. The multimodal signals include: electromyographic signals, muscle sound signals, muscle mechanical vibration signals, and muscle oxygenation signals.

[0008] Preferably, the preprocessing includes: artifact suppression, noise suppression, quality control, and normalization; Artifact suppression: When the electromyography (EMG) signal channel detects high-frequency, sudden, non-physiological electrosurgery interference, the weight of the EMG signal is transferred to the myophone signal. When electrical stimulation artifacts cause instantaneous saturation of the EMG signal, the system uses an adaptive template reduction algorithm to remove artifacts. If a certain modality remains in a low confidence state for several consecutive stimulation cycles, the system uses the current reliable myophone signal to reverse predict and output the predicted EMG feature value based on the EMG-myophone coupling relationship established in the previous few cycles before the interference. Noise suppression: For electromyography signals: an adaptive notch filter is used, with the center frequency automatically set to 50Hz based on geographical location, bandwidth ±0.5 Hz, and notch depth ≥40 dB, effectively suppressing power supply network coupling interference; For muscle sound signals and muscle mechanical vibration signals: a high-pass filter with a cutoff frequency of 0.5 Hz and a second-order order is used to filter out extremely low-frequency components caused by patient breathing, body position changes, and sensor thermal drift. For muscle oxygenation signals: a modulation-demodulation optical detection scheme is adopted, that is, the LED driving current is modulated at a fixed frequency, and the receiver extracts the same frequency component through a lock-in amplifier, effectively eliminating DC and low-frequency interference from ambient light; a digital high-pass filter is used to filter out slow baseline drift caused by tissue edema and probe pressure changes; at the same time, an adaptive common-mode suppression differential structure can be selected to compensate for the drift correlation between adjacent wavelength channels and achieve slow drift suppression. Quality control: Quality control is performed on multimodal signals that have undergone artifact suppression and noise suppression processing. The quality evaluation indicators include: signal-to-noise ratio, stability, and waveform consistency. Normalization: Use any one or more combinations of the following normalization methods: Relative baseline normalization, maximum-minimum normalization, and Z-value standardization.

[0009] Preferably, the multimodal stable signal includes: Let the first The second nerve electrical stimulation occurred at time The multimodal stable signal is: This is the preprocessed electromyographic signal. This is the preprocessed muscle sound signal. This is the preprocessed muscle mechanical vibration signal. This is the pre-processed muscle oxygen signal; The feature extraction includes: Electromyographic features were extracted from the preprocessed electromyographic signals. in, The effective amplitude of the processed electromyographic signal. Individualized baseline amplitude of electromyography signals before anesthesia induction. These are normalized values ​​after comparison with the baseline. for The time elapsed from the moment the electromyographic signal was generated; Muscle sound features were extracted from the preprocessed muscle sound signal. in, The effective amplitude of the processed muscle sound signal. for The time elapsed from the moment the muscle sound signal is generated; Muscle mechanical motion features were extracted from the preprocessed muscle mechanical vibration signals. in, The effective amplitude of the processed muscle mechanical vibration signal. for The time elapsed from the moment the muscle mechanical movement occurs; Muscle oxygen features were extracted from the preprocessed muscle oxygen signal. in, The effective amplitude of the processed muscle oxygenation signal. The length of the time window used to calculate the average value of the muscle oxygen line before the stimulus occurs; This refers to the length of the time window used to calculate the average muscle oxygen response after the stimulus occurs. for The time elapsed from the moment when the muscle oxygen signal changes; After feature extraction, a temporal consistency check is performed on the multimodal signal features, including: Define the response delay of a multimodal signal: the response delay of an electromyographic signal is... The response delay of the muscle sound signal is The response delay of the muscle mechanical vibration signal is The response delay of muscle oxygenation signal is ; Assume the temporal relationship of the complete physiological chain: < ≤ < ; The acquired multimodal stable signal is considered valid if the complete physiological chain temporal relationship is met; otherwise, the multimodal stable signal is considered invalid, as shown below: Assign values ​​to the timing consistency check results; Specifically, when a certain modal signal is missing, timing consistency checks are performed only on stable signals that have the modality.

[0010] Preferably, the fusion of multimodal signal features includes: Calculate the coupling strength between different modalities based on temporal relationships: Calculate the coupling strength between electromyography and muscle sounds. : ; Calculate the coupling strength between muscle sounds and muscle mechanical motion. : ; Calculate the coupling strength between muscle mechanical movement and muscle oxygen metabolism. : ; in It is a tiny constant introduced to prevent calculation instability caused by a denominator that is zero or too small, and its value is 0.001.

[0011] Preferably, the coupling index is calculated as follows: (a) When all four modal signals are complete, calculate the coupling index: ; in, The coupling index; (ii) When modal signals are missing, adaptive calculation is performed: The fusion weights for different modalities are defined as follows: ,in The signal quality index is determined based on the contact impedance. The coupling strength between data from different modalities; M It is electromyography Muscle sounds Mechanical vibration With muscle oxygen Any subset of the set consisting of the four modes.

[0012] Preferably, the fusion weights are further adjusted based on the depth of muscle relaxation, including the following steps: Step 1: Calculate the normalized response intensity for each mode: For the k-th electrical stimulation, the following three dimensionless indices are defined: Normalized electromyographic response: ;in The effective amplitude of the electromyographic signal. Individualized baseline amplitude of electromyography signals before anesthesia induction; Normalized machine response: ;in The effective amplitude of the processed muscle mechanical vibration signal. Individualized baseline amplitude of muscle mechanical vibration signals before anesthesia induction; Normalized muscle sound response: ;in The effective amplitude of the processed muscle sound signal. Individualized baseline amplitude of muscle sound signals before anesthesia induction; Step 2: Calculate the overall retardation depth index: in This is the depth of blockage index, ranging from 0 to 1, where 0 represents complete recovery and no blockage, and 1 represents no response and complete blockage; weighting coefficients. and The default value is 1 for all values, and it can be dynamically adjusted when the quality of a certain modal signal is below the threshold; when... ≥0.8 indicates deep muscle relaxation; 0.4≤ <0.8, muscle relaxation depth is moderate blockade. <0.4, muscle relaxation depth is mild blockade; Step 3: Depth-based adaptive fusion weights: Based on the muscle relaxation depth determined above, the fusion weights of different modal data are adjusted as follows: In the deep block phase, increase the fusion weight of myophone signals in the coupling index calculation; in the moderate block phase, increase the fusion weight of electromyography (EMG) weight in the coupling index calculation; in the mild block phase, focus on monitoring the stability of coupling strength. To assess the stability of the coupling index, a stability index is adopted. Used to quantify the volatility of the coupling index over multiple consecutive stimulation cycles; stability index The calculation method is as follows: Where W is the length of the sliding window, which is the most recent 5 to 10 electrical stimulation cycles; This represents the mean of the coupling index within the window; The standard deviation of the coupling index within the window; To prevent small constants from being divided by zero, a value of 0.001 is used; The value range is 0~1, when If the preset threshold is exceeded, it is determined that the electromechanical coupling has been firmly established, and quantitative basis is provided for the conditions for tube removal.

[0013] Preferably, a unified time axis for multimodal signals is constructed through interpolation, resampling, or time window mapping to achieve time alignment of the multimodal signals.

[0014] Preferably, the muscle relaxation status assessment and decision-making module also performs the following operation: comparing baseline data with functional status indicators to eliminate individual differences between different patients.

[0015] A computer-readable storage medium, characterized in that: it stores a computer program thereon for implementing the multimodal fusion continuous muscle relaxation monitoring system.

[0016] The technical solution of the present invention achieves the following beneficial technical effects: I. By directly assessing respiratory muscle function, the risk of postoperative pulmonary complications is reduced. This invention directly stimulates the T5-6 intercostal nerves and monitors the intercostal muscles, achieving anatomical-functional consistency and improving the clinical safety of muscle relaxation monitoring.

[0017] Second, by continuously collecting and processing autonomous physiological signals, seamless and continuous monitoring of neuromuscular function is achieved, providing a complete trend change map, enabling anesthesiologists to grasp the depth of muscle relaxation and recovery process in real time, and to make timely interventions, effectively avoiding the risk blind spots that may exist in intermittent monitoring.

[0018] Third, by using surface electromyography and muscle sound / vibration dual-modal sensing, complementary information on the electrical and mechanical dimensions of neuromuscular activity was obtained. Combined with an adaptive fusion algorithm, the system can maintain a stable and reliable evaluation output when a certain modal signal is subjected to specific interference (such as electrosurgical interference and environmental noise interference), relying on the anti-interference processing capabilities of the other modality and the algorithm itself, thus exhibiting stronger robustness in complex operating room environments.

[0019] Obviously, based on the above description of the present invention, and according to common technical knowledge and conventional methods in the field, various other modifications, substitutions, or alterations can be made without departing from the basic technical concept of the present invention.

[0020] The following detailed embodiments further illustrate the above-described content of the present invention. However, this should not be construed as limiting the scope of the present invention to the following examples. All technologies implemented based on the above-described content of the present invention fall within the scope of the present invention. Attached Figure Description

[0021] Figure 1 Framework diagram of a multimodal fusion continuous muscle relaxation monitoring system. Detailed Implementation

[0022] It should be noted that the algorithms for data acquisition, transmission, storage and processing steps not specifically described in the embodiments, as well as the hardware structures and circuit connections not specifically described, can all be implemented using content already disclosed in the prior art.

[0023] Example 1: A multimodal fusion continuous muscle relaxation monitoring system A multimodal fusion continuous muscle relaxation monitoring system includes the following modules: The neural electrical stimulation module is configured to apply standardized electrical stimulation to the nerves innervating the respiratory muscles; and to generate a corresponding stimulation event labeling signal each time standardized electrical stimulation is applied. The signal acquisition module is configured to acquire multimodal signals from patients before anesthesia induction or before the action of muscle relaxants, and process them through this system to use them as baseline data; and to acquire multimodal signals induced by standardized electrical stimulation. The signal synchronization and preprocessing module is configured to perform time alignment of multimodal signals based on stimulus event labeling signals, and to preprocess the multimodal signals to generate multimodal stable signals. The feature extraction and fusion module is configured to: quantitatively analyze the multimodal stable signal to obtain the muscle relaxation depth; extract features from the multimodal stable signal to obtain multimodal signal features; perform temporal consistency checks on the multimodal signal features to determine their validity; adjust the multimodal signal feature fusion weights based on the muscle relaxation depth; fuse the valid multimodal signal features according to the fusion weights to obtain the coupling strength; and obtain the coupling index based on the coupling strength. The muscle relaxation status assessment and decision-making module is configured to map the obtained coupling index and its time series information into functional status indicators that can reflect changes in muscle relaxation status; analyze the absolute level of the functional status indicators and their trend over time; continuously assess the effects and metabolic processes of muscle relaxant drugs; and determine whether respiratory muscle function is in a state of blockade, recovery, or stable recovery. The human-computer interaction and data storage module is configured to receive respiratory muscle function status information from the muscle relaxation status assessment and decision-making module, as well as real-time multimodal stabilization signals from the signal synchronization and preprocessing module, and to visualize the status information, multimodal stabilization signals, and functional status indicators.

[0024] In practical implementation, the framework of a multimodal fusion continuous muscle relaxation monitoring system is as follows: Figure 1 As shown. Specifically, it includes: I. Neural Electrical Stimulation Module The neural electrical stimulation module is used to apply electrical stimulation to the nerves innervating the respiratory muscles, including but not limited to: the phrenic nerve, intercostal nerves, and nerves related to accessory respiratory muscles. The preferred stimulation method is transcutaneous electrical stimulation, but adhesive or semi-invasive stimulation electrodes may also be used as needed clinically.

[0025] This module consists of a stimulation signal generator, a constant current output circuit, and an integrated stimulation electrode. Stimulation signal generator: Employs a voltage-controlled constant current source technology, supporting free switching and intelligent combination of multiple clinical standard stimulation modes. Single Twitch (ST) mode: pulse frequency 0.1 Hz (once every 10 seconds), pulse width 200 μs, current intensity adjustable from 30 to 60 mA, used for rapid assessment of the "all or nothing" state of neuromuscular junction conduction.

[0026] The Train-of-Four (TOF) mode features a stimulation train frequency of 0.1 Hz (once every 10 seconds), with each train containing four rectangular wave pulses of 200 μs width and a pulse interval of 500 ms. The current intensity is automatically set to an ultra-strong stimulation level (typically 50 mA). This mode is used to calculate the TOF ratio (T4 / T1) to quantify the depth of muscle relaxation blockade.

[0027] Tetanic Stimulation (TS) mode: Used for neuromuscular function monitoring or assessment of post-tetanic facilitation during deep blockade. Stimulation frequency: 50 Hz or 100 Hz, pulse width: 200 μs, duration: 5 seconds, current intensity: 50 mA. This mode can be triggered manually or automatically at preset time intervals (e.g., every 5 minutes).

[0028] Post-Tetanic Count (PTC) mode: A precise quantitative assessment of deep muscle relaxation (TOF count of 0). A 50 Hz tonic stimulus is applied for 5 seconds, followed by a 3-second interval, and then 15-20 single stimuli are released at a frequency of 1 Hz. The number of induced muscle twitches (PTC count) is recorded. PTC values ​​are highly correlated with the time to first TOF response, accurately predicting the muscle relaxation recovery time window to accommodate different monitoring density requirements.

[0029] Constant current output circuit: Based on the HOWLAND architecture, the constant current source has an output current intensity that is continuously adjustable from 0 to 60 mA. The actual working current is set to 30-50 mA according to individual patient differences to ensure that a strong stimulation is generated in the intercostal nerve pathway area, while limiting the single pulse charge to <12 μC, which complies with the IEC 60601-1 medical electrical safety standard.

[0030] Integrated stimulation electrode: Utilizing a patch gel electrode, the outer ring diameter is 20 mm as the anode, and the inner core diameter is 3 mm as the cathode, with a 5 mm gap between the two electrodes. It contacts the skin through a hydrogel conductive layer. A temperature sensor (NTC thermistor) and impedance monitoring circuit are integrated on the back of the electrode, providing real-time feedback on skin contact impedance (target impedance <25 kΩ). When the impedance is too high, the output voltage is automatically adjusted or a re-attachment prompt is given to ensure stimulation stability.

[0031] II. Signal Acquisition Module The signal acquisition module is used to synchronously acquire the respiratory muscle physiological response signals induced by nerve stimulation after the stimulation is triggered, including at least two of the following, preferably three or more: Electromyography (EMG) signal acquisition unit: Used to acquire EMG signals from the surface of respiratory muscles, which reflect the electrical activity characteristics after nerve impulses are transmitted to the muscle. The acquisition unit includes differential electrodes, front-end amplification and filtering circuits, and analog-to-digital conversion circuits.

[0032] Muscle sound signal acquisition unit: Used to acquire acoustic signals generated during muscle contraction, which reflect the mechanical efficiency of muscle fiber contraction. The muscle sound signals are acquired via an acoustic sensor. This unit consists of a microphone, a front-end amplification and filtering circuit, and an analog-to-digital conversion circuit.

[0033] Muscle mechanical vibration signal acquisition unit: used to acquire local mechanical displacement, vibration or acceleration changes caused by respiratory muscle contraction, in order to reflect the amplitude and dynamic characteristics of muscle contraction, and is acquired by an accelerometer.

[0034] The muscle oxygenation signal acquisition unit is used to obtain oxygenation status information of local respiratory muscle tissue to reflect the corresponding metabolic level, oxygen supply status, and functional recovery capacity of the muscle during nerve stimulation and muscle relaxant metabolism. The muscle oxygenation signal acquisition unit employs local tissue optical detection, preferably near-infrared spectroscopy, to obtain parameters reflecting the local muscle oxygenation status by analyzing the absorption differences of different wavelengths of light in muscle tissue. This unit consists of an optical emission module, an optical reception module, an analog signal conditioning module, and an analog-to-digital conversion module.

[0035] III. Signal Synchronization and Preprocessing Module The signal synchronization and preprocessing module performs unified time alignment, artifact suppression, and quality control on the raw signals from the signal acquisition module, providing stable and reliable basic data for subsequent feature extraction and fusion analysis. This module addresses the data incomparability issues caused by differences in sampling rates, transmission delays, and noise characteristics across different modalities by establishing a temporal correspondence between neural electrical stimulation events and multimodal physiological responses.

[0036] Each time a neural electrical stimulation is triggered, the system generates a corresponding stimulation event marker signal, which is sent to the signal synchronization module via a hardware interrupt or a high-priority task.

[0037] The signal synchronization and preprocessing module uses the stimulus event marker as a time reference to perform time alignment processing on each modality signal, so that electromyographic signals, muscle sound signals, muscle mechanical vibration signals, and muscle oxygenation signals can all correspond to the same stimulus-induced muscle response event. Under different sampling rate conditions, the system completes the construction of a unified time axis for multimodal signals through interpolation, resampling, or time window mapping.

[0038] Noise suppression and signal quality control: For the complex environment of the operating room, appropriate noise suppression processing is performed on different modal signals, including but not limited to: power frequency interference suppression and high-frequency noise filtering of electromyographic signals; low-frequency baseline drift removal of muscle sound and mechanical vibration signals; and ambient light and slow-varying drift suppression of muscle oxygenation signals.

[0039] Artifact suppression: When the electromyography (EMG) signal channel detects high-frequency, sudden, non-physiological electrosurgery interference, the weight of the EMG signal is transferred to the myophone signal. When electrical stimulation artifacts cause instantaneous saturation of the EMG signal, the system uses an adaptive template reduction algorithm to remove artifacts. If a certain modality remains in a low confidence state for several consecutive stimulation cycles, the system uses the current reliable myophone signal to reverse predict and output the predicted EMG feature value based on the EMG-myophone coupling relationship established in the previous few cycles before the interference. Noise suppression: For electromyography signals: an adaptive notch filter is used, with the center frequency automatically set to 50Hz based on geographical location, bandwidth ±0.5 Hz, and notch depth ≥40 dB, effectively suppressing power supply network coupling interference; For muscle sound signals and muscle mechanical vibration signals: a high-pass filter with a cutoff frequency of 0.5 Hz and a second-order order is used to filter out extremely low-frequency components caused by patient breathing, body position changes, and sensor thermal drift. For muscle oxygenation signals: a modulation-demodulation optical detection scheme is adopted, that is, the LED driving current is modulated at a fixed frequency, and the receiver extracts the same frequency component through a lock-in amplifier, effectively eliminating DC and low-frequency interference from ambient light; a digital high-pass filter is used to filter out slow baseline drift caused by tissue edema and probe pressure changes; at the same time, an adaptive common-mode suppression differential structure can be selected to compensate for the drift correlation between adjacent wavelength channels and achieve slow drift suppression. The system performs quality assessments on each preprocessed modal signal, with assessment metrics including but not limited to: signal-to-noise ratio, stability, and waveform consistency. When the quality of a certain modal signal falls below a preset threshold, the system may take at least one of the following measures: mark the modal signal as low confidence; reduce the weight of the modality in the fusion analysis; or prompt the user to check the sensor status.

[0040] Quality control: Quality control is performed on multimodal signals that have undergone artifact suppression and noise suppression processing. The quality evaluation indicators include: signal-to-noise ratio, stability, and waveform consistency. As a specific implementation method, the comprehensive judgment of quality indicators adopts the following rules: First, if the signal-to-noise ratio is lower than 10 dB, it is directly judged as low confidence. Second, if the stability index is lower than 0.7 or the waveform consistency is lower than 0.75, it is judged as medium confidence, and the fusion weight is appropriately reduced. Finally, only when all three meet the threshold is it judged as high confidence, and the normal fusion weight is maintained.

[0041] The signal synchronization and preprocessing module is also used to verify the physiological consistency between multimodal signals, including: electromyographic activity precedes mechanical vibration after stimulation; changes in mechanical vibration precede changes in muscle oxygenation; and the above temporal relationship is used to determine whether the signal originates from a real muscle physiological response, thereby further suppressing the influence of abnormal data on subsequent analysis.

[0042] In a preferred embodiment, the system also integrates interference identification and signal reconstruction mechanisms to address common electrosurgical interference and electrical stimulation artifacts in the operating room environment. Specifically, the system monitors the quality status of each modality signal in real time. When the electromyography (EMG) signal channel detects high-frequency, sudden, and non-physiological electrosurgical interference, the system automatically marks the EMG signal for that period as low confidence and temporarily transfers the fusion weights in the fusion analysis to the myophone signal to ensure the continuity of the monitoring output. When electrical stimulation artifacts cause instantaneous saturation of the EMG signal, the system uses an adaptive template reduction algorithm to remove the artifacts. If a modality remains in a low confidence state for several consecutive stimulation cycles, the system uses the EMG-myophone coupling relationship established in the previous few cycles to predict the EMG feature value in reverse using the currently reliable myophone signal, in order to maintain a stable output of the fusion parameters and avoid jumps in evaluation results due to short-term signal quality problems. The system also generates corresponding signal quality evaluation indicators and displays the current monitoring confidence level in the human-computer interaction interface for clinical personnel to refer to.

[0043] In this embodiment, the preprocessing procedure further includes dimensionless processing of the signals from each modality. Specifically, since the original electromyography (EMG) signals, muscle sound signals, muscle mechanical vibration signals, and muscle oxygenation signals have different physical units and magnitude ranges, they cannot be directly subjected to algebraic operations or fusion. Therefore, the system employs one or a combination of dimensionless strategies such as relative baseline normalization, maximum-minimum normalization, and Z-value normalization.

[0044] The synchronized and preprocessed multimodal signals are output to the feature extraction and fusion analysis module in a unified format. The output includes at least: preprocessed signal data, corresponding stimulus event labels, and signal quality evaluation results, to support subsequent continuous and dynamic assessment of muscle relaxation status. The multimodal signals are defined as follows: Let the first The second nerve electrical stimulation occurred at time Let the set of multimodal stable signals be as follows: in, This is the preprocessed electromyographic signal. This is the preprocessed muscle sound signal. This is the preprocessed muscle mechanical vibration signal. This is the pre-processed muscle oxygen signal; Define the set of available modes M :set up M It is electromyography Muscle sounds Mechanical vibration With muscle oxygen Any subset of the set consisting of the four modes.

[0045] IV. Feature Extraction and Fusion Module The feature extraction and fusion module is used to quantitatively analyze the multimodal physiological signals output by the signal synchronization and preprocessing module. Based on the above module, the depth of muscle relaxation (mild block, moderate block, or deep block) is determined. Then, by extracting and fusing key features representing nerve conduction, muscle contraction, and metabolic state in different modal signals, the fusion weight of key physiological signals is dynamically adjusted according to the depth of muscle relaxation to construct a comprehensive evaluation result that can continuously reflect the recovery status of respiratory muscle neuromuscular function.

[0046] During feature extraction, the system first processes the physiological response induced by each nerve electrical stimulation independently based on the stimulus event label and the corresponding analysis time window. For electromyographic signals, the system performs time-domain and frequency-domain analysis on the short-term response signal after stimulation to obtain electrophysiological characteristics reflecting the degree of motor unit recruitment, synchronization level, and neuromuscular conduction efficiency. These characteristics can reflect the effectiveness of nerve impulse transmission to muscle fibers in a relaxed state and its trend over time.

[0047] For myosalpingography (MGM) signals and muscle mechanical vibration signals, the system extracts mechanical characteristics that reflect the actual contractile strength, mechanical output capacity, and contraction duration of muscle fibers. These mechanical characteristics are used to characterize whether the muscle can produce a sufficient and stable mechanical response given that a nerve impulse has arrived, thereby distinguishing the difference between simple electrophysiological recovery and functional contractile recovery.

[0048] For muscle oxygenation signals, the system focuses on the changes in local muscle oxygenation status before and after stimulation, as well as during continuous stimulation. By analyzing the magnitude, rate of change, and recovery trend of oxygenation levels, it assesses the metabolic supply and demand balance of muscles under the current neuromuscular functional state. This characteristic can reflect whether muscles have the metabolic basis for sustained work after neuromuscular conduction is restored.

[0049] The signals after feature extraction for each modality are defined as follows: Electromyographic features were extracted from the preprocessed electromyographic signals. in, The effective amplitude of the processed electromyographic signal. Individualized baseline amplitude of electromyography signals before anesthesia induction. These are normalized values ​​after comparison with the baseline. for The time elapsed from the moment the electromyographic signal was generated; Muscle sound features were extracted from the preprocessed muscle sound signal. in, The effective amplitude of the processed muscle sound signal. for The time elapsed from the moment the muscle sound signal is generated; Muscle mechanical motion features were extracted from the preprocessed muscle mechanical vibration signals. in, The effective amplitude of the processed muscle mechanical vibration signal. for The time elapsed from the moment the muscle mechanical movement occurs; Muscle oxygen features were extracted from the preprocessed muscle oxygen signal. in, The effective amplitude of the processed muscle oxygenation signal. The length of the time window used to calculate the average value of the muscle oxygen line before the stimulus occurs; This refers to the length of the time window used to calculate the average muscle oxygen response after the stimulus occurs. for The time elapsed from the moment when the muscle oxygen signal changes; After extracting features for each individual modality, the feature extraction and fusion module performs time correspondence and correlation analysis on the features of different modalities. It uses a unified temporal constraint model to determine whether the temporal series are consistent, thereby judging the validity of the data. When a certain modality is missing, the temporal constraint judgment is only performed on the existing modalities. The specific model is as follows: Define the response delay of each modality signal: the response delay of electromyographic signals is... The response delay of the muscle sound signal is The response delay of the muscle mechanical vibration signal is The response delay of muscle oxygenation signal is ; Assume the temporal relationship of the complete physiological chain: < ≤ < The acquired multimodal signal is considered valid if the complete physiological chain temporal relationship is met; otherwise, the multimodal signal is considered invalid, as shown below: Assign a value to the timing consistency check result.

[0050] The system jointly models electrical activity characteristics, mechanical response characteristics, and metabolic characteristics based on the physiological temporal relationship of stimulus-induced response, in order to form a fusion feature expression that reflects the overall functional state of the neuromuscular system.

[0051] Furthermore, the depth of muscle relaxation is determined based on the signal synchronization and preprocessing module, enabling adaptive multimodal fusion weighting for each muscle relaxation stage. Specifically, in the deep block stage, muscle sound signals are the primary fusion modality to avoid false negatives in electromyography (EMG); in the moderate block stage, the weight of EMG fusion is increased to provide early warning of muscle relaxation recovery trends; in the mild block stage, the stability of the electromechanical coupling index is monitored to determine the muscle relaxation recovery trend and whether extubation is feasible. Muscle oxygenation signals are monitored consistently throughout the process to continuously reflect changes and trends in muscle oxygenation. The fusion process can be implemented based on a preset rule model, weight model, or adaptive learning model, and allows for dynamic adjustment of the contribution of different modal signals to the fusion analysis based on their real-time quality. The determination of muscle relaxation depth includes the following steps: Step 1: Calculate the normalized response intensity of each mode. For the k-th electrical stimulation, the following three dimensionless indices are defined: Normalized electromyographic response: ;in The effective amplitude of the electromyographic signal. Individualized baseline amplitude of electromyography signals before anesthesia induction; Normalized machine response: ;in The effective amplitude of the processed muscle mechanical vibration signal. Individualized baseline amplitude of muscle mechanical vibration signals before anesthesia induction; Normalized muscle sound response: ;in The effective amplitude of the processed muscle sound signal. Individualized baseline amplitude of muscle sound signals before anesthesia induction; Step 2: Calculate the overall retardation depth index: in This is the depth of blockage index, ranging from 0 to 1, where 0 represents complete recovery and no blockage, and 1 represents no response and complete blockage; weighting coefficients. and The default value is 1 for all values, and it can be dynamically adjusted when the quality of a certain modal signal is below the threshold; when... ≥0.8 indicates deep muscle relaxation; 0.4≤ <0.8, muscle relaxation depth is moderate blockade. <0.4, muscle relaxation depth is mild blockade; Step 3: Depth-based adaptive fusion weights Based on the determined depth of muscle relaxation, the system automatically adjusts the fusion weights of each modality in subsequent feature fusion: In the deep block stage, the electromyographic signal may have extremely low amplitude or even be a false negative due to the deep block; in this case, the muscle sound signal is the primary fusion basis, as it can still detect weak mechanical activity. In the moderate block stage, the electromyographic fusion weight is increased, utilizing the early recovery trend of the electromyographic signal to warn of impending muscle relaxation. In the mild block stage, the stability of the coupling strength is monitored to determine whether the electromechanical coupling has been firmly established, providing a quantitative basis for extubation conditions.

[0052] To assess the stability of the coupling index, a stability index is adopted. This method is used to quantify the fluctuation of the coupling index over multiple consecutive stimulation cycles. The sliding window coefficient of variation method is employed. Where W is the length of the sliding window, which is the most recent 5 to 10 electrical stimulation cycles; This represents the mean of the coupling index within the window; Standard deviation; To prevent small constants from being divided by zero, a value of 0.001 is used; The value ranges from 0 to 1, with a larger value indicating greater stability.

[0053] Multimodal coupling computation is used in the fusion process, and the specific process is as follows: First, the coupling strength between different modalities is calculated based on the temporal relationship. (1) Calculate the coupling strength between electromyography and muscle sounds. : ; (2) Calculate the coupling strength between muscle sounds and muscle mechanical movements. : ; (3) Calculate the coupling strength between muscle mechanical movement and muscle oxygen metabolism. : ; in It is a tiny constant introduced to prevent calculation instability caused by a denominator that is zero or too small, and its value is 0.001.

[0054] Secondly, different processing methods are applied based on the integrity of the modal data: (a) When all four modal signals are complete, calculate the coupling index: ; (ii) When modal signals are missing, adaptive calculation is performed: Define the fusion weights for different modalities: ,in The signal quality index is determined based on the contact impedance; then, the coupling index is constructed. .in This represents the coupling strength between data from different modalities.

[0055] Through the multimodal fusion calculations described above, a coupling index was constructed to characterize the degree of complete recovery of respiratory muscle neuromuscular function. This coupling index simultaneously reflects whether nerve impulses were successfully transmitted to the muscle, whether the muscle produced effective mechanical contraction, and whether the muscle possessed the corresponding metabolic support capacity, thereby avoiding misjudgments that may be caused by single-modal assessments.

[0056] During continuous monitoring, the feature extraction and fusion module performs time-series analysis on the coupling index to generate dynamic trend information that changes with the anesthesia process. By analyzing the absolute level, rate of change, and stability of this trend information, the system can identify different stages of muscle relaxation blockade deepening, maintenance, and recovery, and provide quantitative basis for subsequent muscle relaxation status assessment and extubation decisions.

[0057] V. Muscle Relaxation Status Assessment and Decision-Making Module The muscle relaxation status assessment and decision-making module is used to continuously assess the neuromuscular function status of the respiratory muscles based on the comprehensive evaluation results output by the feature extraction and fusion module, and generate status judgment information reflecting the degree of action of muscle relaxant drugs and the recovery process, thereby providing objective technical support for perioperative muscle relaxation management and extubation timing.

[0058] During operation, this module receives time-series information of coupling index and multimodal signal features output by the feature extraction and fusion module, and maps them into functional state indicators that reflect changes in muscle relaxation. Specifically, the mapping involves: converting the coupling index... Divided by its individualized baseline value Then, multiply by 100 and limit the value to the range of 0-100 to obtain the functional status index; the higher the index value, the more fully the respiratory muscle function has recovered. The functional status index is based on the comprehensive performance of nerve conduction effectiveness, muscle mechanical contraction ability, and metabolic support ability, and is used to describe whether the respiratory muscles have the functional conditions required to complete effective spontaneous breathing at the current moment.

[0059] The muscle relaxation status assessment module continuously tracks the entire process of muscle relaxant drugs from onset, maintenance, to metabolic decline by analyzing the absolute levels of functional status indicators and their trends over time. During the deepening of muscle relaxation blockade, the system identifies the synchronous decline in nerve conduction and mechanical response; during the maintenance phase, the system monitors the relative stability of functional status indicators; and during the recovery phase, the system focuses on the continuous upward trend and stability of functional status indicators to determine whether respiratory muscle function has entered a recoverable range.

[0060] In a preferred embodiment, the module compares the current functional status indicators with an individualized baseline status established preoperatively or during the initial stage of anesthesia induction, thereby eliminating the influence of anatomical differences, differences in basic muscle function, and differences in anesthetic medications among different patients on the assessment results, thus achieving individualized muscle relaxation status assessment. This baseline status may be derived from spontaneous respiratory signals at rest or from the initial response signals evoked by standardized stimuli.

[0061] Based on continuous assessment, the muscle relaxation status assessment and decision-making module further analyzes the stability of functional status indicators to distinguish between transient fluctuations and true functional recovery. When the comprehensive evaluation results remain within the preset functional range and the fluctuation amplitude is below the threshold for multiple consecutive stimulation cycles, the system determines that the respiratory muscle neuromuscular function has reached a stable recovery state; when the indicators reach the target range but show obvious instability or a downward trend, the system determines that there is still a risk of residual muscle relaxation or insufficient function.

[0062] Based on the above assessment results, the muscle relaxation status assessment and decision-making module generates corresponding respiratory muscle function status information and provides this information to the human-computer interaction module for display or prompting. This respiratory muscle function status information is used to assist clinicians in determining whether the technical conditions for extubation or discontinuation of mechanical ventilation are met, but it does not directly replace the comprehensive judgment of clinicians, thus avoiding limiting the system's function to purely medical decision-making.

[0063] In another embodiment, the module can also analyze the rate of change of functional status indicators by combining historical monitoring data to predict the time interval required for further metabolism of muscle relaxants. When the prediction results show that the functional status indicators are about to enter the stable recovery range, the system can generate a prompt message in advance to assist clinicians in preparing for extubation or adjusting anesthesia management strategies.

[0064] Through the above methods, the muscle relaxation status assessment and decision-making module realizes a continuous technical assessment of the respiratory muscle neuromuscular function from "whether there is a response" to "whether there is functional recovery ability". This transforms muscle relaxation monitoring from traditional discrete threshold judgment to a dynamic assessment process based on the trend of functional status changes, thereby effectively reducing the risk of postoperative respiratory complications caused by residual muscle relaxation.

[0065] VI. Human-Computer Interaction and Data Storage Module The human-computer interaction and data storage module is used to realize the visualization and secure storage of system operation status, monitoring results and historical data, providing clinical personnel with intuitive, continuous and traceable technical information output, and providing a data foundation for postoperative analysis and system optimization.

[0066] In terms of human-computer interaction, this module receives respiratory muscle function status information from the muscle relaxation status assessment and decision-making module and real-time multimodal stable signals from the multimodal signal processing module, and visualizes the monitoring process through a display interface. The displayed content includes real-time updated multimodal physiological signal waveforms, functional status indicators obtained from fusion analysis, and their trends over time, allowing users to simultaneously observe the original signals and processing results on the same interface, thereby understanding the source and change process of the system's evaluation results.

[0067] In the implementation of human-computer interaction, the system adopts an adaptive display method for different monitoring stages. During the establishment and maintenance of muscle relaxation blockade, the interface focuses on presenting the overall level and trend of functional status indicators; during the muscle relaxation recovery stage, the interface emphasizes the stability changes of functional status indicators and the comparison results with the individualized baseline, thereby helping users quickly identify whether functional recovery has entered a stable range. This display method does not directly provide clinical operation instructions, but rather assists users in making comprehensive judgments through status information and trend changes.

[0068] In terms of operation and interaction, the human-computer interaction module supports the setting and adjustment of system operating parameters, including stimulus-related parameters, monitoring cycle, and display mode. The system ensures that parameter adjustments will not affect system security or monitoring continuity by validating the user input, and automatically updates the corresponding signal processing and evaluation process after parameter changes.

[0069] Regarding data storage, the module is used for unified management and storage of key data generated during the monitoring process. The stored data includes at least stimulus event information, preprocessed multimodal signal data, fusion analysis results, and corresponding timestamp information. The data storage process runs in parallel with the real-time monitoring process, without affecting the system's real-time performance.

[0070] In a preferred embodiment, the data storage module establishes independent data recording units according to patients or monitoring tasks, and timestamps and status tags on each record to support postoperative review, trend analysis, and personalized model optimization. The data records can be stored on local storage media or transmitted to external systems or information platforms, provided that security and privacy requirements are met.

[0071] Through the aforementioned human-computer interaction and data storage modules, the system achieves a complete technical closed loop from real-time monitoring and status assessment to historical data management, enabling the multimodal continuous muscle relaxation monitoring process to have good interpretability, traceability, and engineering feasibility, thereby improving the reliability and applicability of the system in actual clinical applications.

[0072] Example 2: A multimodal fusion method for continuous muscle relaxation monitoring A multimodal fusion method for continuous muscle relaxation monitoring includes the following steps: Step 1: Collect multimodal signals from patients before anesthesia induction or before the action of muscle relaxants, and process them using this system as baseline data; Step 2: Apply standardized electrical stimulation to the nerves innervating the respiratory muscles; generate corresponding stimulus event marker signals for each application of standardized electrical stimulation; Step 3: Acquire multimodal signals induced by standardized electrical stimulation; Step 4: Based on the stimulus event labeling signal, perform time alignment of the multimodal signal and preprocess the multimodal signal to generate a stable multimodal signal; Step 5: Quantitatively analyze the multimodal stable signal to obtain the muscle relaxation depth; extract features from the multimodal stable signal to obtain multimodal signal features; perform temporal consistency checks on the multimodal signal features to determine their effectiveness; adjust the multimodal signal feature fusion weights according to the muscle relaxation depth; fuse the effective multimodal signal features according to the fusion weights to obtain the coupling strength; obtain the coupling index based on the coupling strength. Step 6: Map the obtained coupling index and its time series information into functional state indicators that can reflect changes in muscle relaxation status; analyze the absolute level of the functional state indicators and their trend over time to continuously assess the effects and metabolic processes of muscle relaxant drugs, and determine whether respiratory muscle function is in a state of blockade, recovery, or stable recovery. Step 7: Receive respiratory muscle function status information from the muscle relaxation status assessment and decision-making module, as well as real-time multimodal stabilization signals from the signal synchronization and preprocessing module, and generate corresponding prompt information.

[0073] In this embodiment, the specific steps include: Before system operation, the system is started and a self-test is completed. The system checks the working status of the nerve stimulation module, multimodal signal acquisition module, and data processing module to confirm that each functional unit is in normal working condition. After completing the self-test, the system enters the monitoring state.

[0074] At the beginning of the monitoring phase, the system establishes the corresponding monitoring subject information and performs a baseline acquisition process. Baseline acquisition is used to obtain the respiratory muscle physiological characteristics of the patient before anesthesia induction or before the action of muscle relaxants. The baseline data serves as an individualized reference for subsequent muscle relaxation status assessment, eliminating the influence of differences in basic function between different patients on the assessment results.

[0075] After baseline acquisition is completed, the system enters the perioperative continuous monitoring process. The neurostimulation module applies standardized electrical stimulation to the nerves innervating the respiratory muscles according to preset parameters, and generates a corresponding stimulation event marker signal at the same time as each stimulation trigger.

[0076] Following nerve electrical stimulation, the multimodal physiological signal acquisition module simultaneously acquires respiratory muscle response signals induced by the stimulation, including electromyographic signals, muscle sound signals, muscle mechanical vibration signals, and muscle oxygenation signals. Each modality of signal maintains a time correlation with the stimulation event during acquisition to ensure that different signals originate from the same physiological response process.

[0077] The acquired multimodal signals first enter the signal synchronization and preprocessing module. This module performs time alignment of each modality signal based on stimulus event labels, and performs artifact suppression, noise filtering, and quality assessment on the signals, thereby obtaining stable physiological signal data suitable for subsequent analysis.

[0078] The preprocessed signals are input into the feature extraction and fusion module. The system extracts key features reflecting nerve conduction, muscle mechanical contraction, and metabolic state from different modal signals, and constructs a comprehensive evaluation result through multimodal fusion analysis to characterize the overall state of respiratory muscle neuromuscular function.

[0079] After obtaining the fusion analysis results, the muscle relaxation status assessment and decision-making module analyzes the comprehensive evaluation results and their trends over time, continuously assesses the effects and metabolic processes of muscle relaxant drugs, and determines whether respiratory muscle function is in a state of blockade, recovery, or stable recovery.

[0080] The assessment results are displayed in real time through the human-computer interaction module and continuously updated in the form of trends. When the system detects that the respiratory muscle function status meets the preset stable recovery conditions, it generates corresponding prompts to assist clinicians in determining whether the technical conditions for extubation or discontinuation of mechanical ventilation are met. Simultaneously, the system synchronously stores stimulation parameters, physiological signals, fusion results, and status assessment information throughout the monitoring process to support postoperative review analysis and system performance optimization.

[0081] Through the above process, the system realizes a complete closed loop from nerve electrical stimulation, physiological signal acquisition, data processing to muscle relaxation status assessment and information output, ensuring that the muscle relaxation monitoring process is continuous, traceable, and engineering feasible.

Claims

1. A multimodal fusion continuous neuromuscular monitoring system, comprising: Includes the following modules: The neural electrical stimulation module is configured to apply standardized electrical stimulation to the nerves innervating the respiratory muscles; and to generate a corresponding stimulation event labeling signal each time standardized electrical stimulation is applied. The signal acquisition module is configured to acquire multimodal signals from patients before anesthesia induction or before the action of muscle relaxants, and process them through this system to use them as baseline data; and to acquire multimodal signals induced by standardized electrical stimulation. The signal synchronization and preprocessing module is configured to perform time alignment of multimodal signals based on stimulus event labeling signals, and to preprocess the multimodal signals to generate multimodal stable signals. The feature extraction and fusion module is configured to: perform quantitative analysis on the multimodal stable signal to obtain the muscle relaxation depth; extract features from the multimodal stable signal to obtain multimodal signal features; and perform temporal consistency checks on the multimodal signal features to determine the validity of the multimodal signal features. The multimodal signal feature fusion weights are adjusted according to the muscle relaxation depth; the effective multimodal signal features are fused according to the fusion weights to obtain the coupling strength; and the coupling index is obtained based on the coupling strength. The muscle relaxation state assessment and decision-making module is configured to map the obtained coupling index and the temporal information of multimodal signal characteristics into functional state indicators that can reflect changes in muscle relaxation state; analyze the absolute level of the functional state indicators and their trend over time; continuously assess the effects and metabolic processes of muscle relaxant drugs; and determine the state of respiratory muscle function, whether the state is in a state of blockade, recovery, or stable recovery. The human-computer interaction and data storage module is configured to receive respiratory muscle function status information from the muscle relaxation status assessment and decision module, as well as real-time multimodal stabilization signals from the signal synchronization and preprocessing module, and to visualize the status information, multimodal stabilization signals, and functional status indicators. The multimodal signals include: electromyographic signals, muscle sound signals, muscle mechanical vibration signals, and muscle oxygenation signals.

2. The multimodal fusion based continuous neuromuscular monitoring system as claimed in claim 1, wherein, The preprocessing includes: artifact suppression, noise suppression, quality control, and normalization; Artifact suppression: When the electromyography (EMG) signal channel detects high-frequency, sudden, non-physiological electrosurgery interference, the weight of the EMG signal is transferred to the myophone signal. When electrical stimulation artifacts cause instantaneous saturation of the EMG signal, the system uses an adaptive template reduction algorithm to remove artifacts. If a certain modality remains in a low confidence state for several consecutive stimulation cycles, the system uses the current reliable myophone signal to reverse predict and output the predicted EMG feature value based on the EMG-myophone coupling relationship established in the previous few cycles before the interference. Noise suppression: For electromyography signals: an adaptive notch filter is used, with the center frequency automatically set to 50 Hz based on geographical location, bandwidth ±0.5 Hz, and notch depth ≥40 dB, effectively suppressing power supply network coupling interference; For muscle sound signals and muscle mechanical vibration signals: a high-pass filter with a cutoff frequency of 0.5 Hz and a second-order order is used to filter out extremely low-frequency components caused by patient breathing, body position changes, and sensor thermal drift. For muscle oxygenation signals: a modulation-demodulation optical detection scheme is adopted, that is, the LED driving current is modulated at a fixed frequency, and the receiver extracts the same frequency component through a lock-in amplifier, effectively eliminating DC and low-frequency interference from ambient light; a digital high-pass filter is used to filter out slow baseline drift caused by tissue edema and probe pressure changes; at the same time, an adaptive common-mode suppression differential structure can be selected to compensate for the drift correlation between adjacent wavelength channels and achieve slow drift suppression. Quality control: Quality control is performed on multimodal signals that have undergone artifact suppression and noise suppression processing. The quality evaluation indicators include: signal-to-noise ratio, stability, and waveform consistency. Normalization: Use any one or more combinations of the following normalization methods: Relative baseline normalization, maximum-minimum normalization, and Z-value standardization.

3. The multimodal fusion continuous muscle relaxation monitoring system according to claim 1, characterized in that, The multimodal stable signal includes: Let the first The second nerve electrical stimulation occurred at time The multimodal stable signal is: This is the preprocessed electromyographic signal. This is the preprocessed muscle sound signal. This is the preprocessed muscle mechanical vibration signal. This is the pre-processed muscle oxygen signal; The feature extraction includes: Electromyographic features were extracted from the preprocessed electromyographic signals. in, The effective amplitude of the processed electromyographic signal. Individualized baseline amplitude of electromyography signals before anesthesia induction. These are normalized values ​​after comparison with the baseline. for The time elapsed from the moment the electromyographic signal is generated; Muscle sound features were extracted from the preprocessed muscle sound signal. in, The effective amplitude of the processed muscle sound signal. for The time elapsed from the moment the muscle sound signal is generated; Muscle mechanical motion features were extracted from the preprocessed muscle mechanical vibration signals. in, The effective amplitude of the processed muscle mechanical vibration signal. for The time elapsed from the moment the muscle mechanical movement occurs; Muscle oxygen features were extracted from the preprocessed muscle oxygen signal. in, The effective amplitude of the processed muscle oxygenation signal. The length of the time window used to calculate the average value of the muscle oxygen line before the stimulus occurs; This refers to the length of the time window used to calculate the average muscle oxygen response after the stimulus occurs. for The time elapsed from the moment the muscle oxygen signal changes; After feature extraction, a temporal consistency check is performed on the multimodal signal features, including: Define the response delay of a multimodal signal: the response delay of an electromyographic signal is... The response delay of the muscle sound signal is The response delay of the muscle mechanical vibration signal is The response delay of muscle oxygenation signal is ; Assume the temporal relationship of the complete physiological chain: < ≤ < ; The acquired multimodal stable signal is considered valid if the complete physiological chain temporal relationship is met; otherwise, the multimodal stable signal is considered invalid, as shown below: Assign values ​​to the timing consistency check results; Specifically, when a certain modal signal is missing, timing consistency checks are performed only on stable signals that have the modality.

4. The multimodal fusion continuous muscle relaxation monitoring system according to claim 3, characterized in that, The fusion of multimodal signal features includes: Calculate the coupling strength between different modalities based on temporal relationships: Calculate the coupling strength between electromyography and muscle sounds. : ; Calculate the coupling strength between muscle sounds and muscle mechanical motion. : ; Calculate the coupling strength between muscle mechanical movement and muscle oxygen metabolism. : ; in It is a tiny constant introduced to prevent calculation instability caused by a denominator that is zero or too small, and its value is 0.

001.

5. The multimodal fusion continuous muscle relaxation monitoring system according to claim 4, characterized in that, The coupling index is calculated as follows: (a) When all four modal signals are complete, calculate the coupling index: ； in, The coupling index; (ii) When modal signals are missing, adaptive calculation is performed: The fusion weights for different modalities are defined as follows: ,in The signal quality index is determined based on the contact impedance. The coupling strength between data from different modalities; M It is electromyography Muscle sounds Mechanical vibration With muscle oxygen Any subset of the set consisting of the four modes.

6. The multimodal fusion continuous muscle relaxation monitoring system according to claim 5, characterized in that, The fusion weights are also adjusted based on the depth of muscle relaxation, including the following steps: Step 1: Calculate the normalized response intensity for each mode: For the k-th electrical stimulation, the following three dimensionless indices are defined: Normalized electromyographic response: ;in The effective amplitude of the electromyographic signal. Individualized baseline amplitude of electromyography signals before anesthesia induction; Normalized machine response: ;in The effective amplitude of the processed muscle mechanical vibration signal. Individualized baseline amplitude of muscle mechanical vibration signals before anesthesia induction; Normalized muscle sound response: ;in The effective amplitude of the processed muscle sound signal. Individualized baseline amplitude of muscle sound signals before anesthesia induction; Step 2: Calculate the overall retardation depth index: in This is the depth of blockage index, ranging from 0 to 1, where 0 represents complete recovery and no blockage, and 1 represents no response and complete blockage; weighting coefficients. and The default value is 1 for all values, and it can be dynamically adjusted when the quality of a certain modal signal is below the threshold; when... ≥0.8 indicates deep muscle relaxation; 0.4≤ <0.8, muscle relaxation depth is moderate blockade. <0.4, muscle relaxation depth is mild blockade; Step 3: Depth-based adaptive fusion weights: Based on the muscle relaxation depth determined above, the fusion weights of different modal data are adjusted as follows: In the deep block phase, increase the fusion weight of myophone signals in the coupling index calculation; in the moderate block phase, increase the fusion weight of electromyography (EMG) weight in the coupling index calculation; in the mild block phase, focus on monitoring the stability of coupling strength. To assess the stability of the coupling index, a stability index is adopted. Used to quantify the volatility of the coupling index over multiple consecutive stimulation cycles; stability index The calculation method is as follows: Where W is the length of the sliding window, which is the most recent 5 to 10 electrical stimulation cycles; This represents the mean of the coupling index within the window; The standard deviation of the coupling index within the window; To prevent small constants from being divided by zero, a value of 0.001 is used; The value range is 0~1, when If the preset threshold is exceeded, it is determined that the electromechanical coupling has been firmly established, and quantitative basis is provided for the conditions for tube removal.

7. The multimodal fusion continuous muscle relaxation monitoring system according to claim 1, characterized in that, A unified time axis for multimodal signals is constructed through interpolation, resampling, or time window mapping, thus achieving time alignment of the multimodal signals.

8. The multimodal fusion continuous muscle relaxation monitoring system according to claim 1, characterized in that, The muscle relaxation status assessment and decision-making module also performs the following operations: comparing baseline data with functional status indicators to eliminate individual differences between different patients.

9. A computer-readable storage medium, characterized in that: It stores a computer program for implementing the multimodal fusion continuous muscle relaxation monitoring system as described in claims 1-8.

Images