Muscle relaxation depth prediction anesthesia system based on electromyographic signals
By performing spectral transformation and inertia calculation on electromyographic signals, a muscle relaxation trend prediction result is generated, which solves the problems of discontinuity and interpretability in the existing technology of muscle relaxation depth monitoring and achieves highly reliable muscle relaxation trend prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG CANCER HOSPITAL
- Filing Date
- 2026-02-05
- Publication Date
- 2026-05-12
AI Technical Summary
In existing technologies, muscle relaxation depth monitoring methods cannot provide continuous monitoring data, and the decision-making process of machine learning models is difficult to interpret, resulting in low clinical reliability.
By acquiring electromyographic signals, performing preprocessing, and then performing spectral transformation to calculate spectral inertia, the results of muscle relaxation trend prediction are generated by combining the muscle relaxation depth and displayed using an interpretable algorithm.
It achieves low latency, low power consumption, and high reliability in predicting muscle relaxation trends, reducing computational complexity and enhancing the credibility of clinical applications.
Smart Images
Figure CN122004906A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present invention relate to the field of medical monitoring, and more specifically, to an anesthesia system for predicting the depth of muscle relaxation based on electromyographic signals. Background Technology
[0002] In general anesthesia, precise control of the patient's muscle relaxation level is crucial. Insufficient muscle relaxation may lead to intraoperative movement, interfering with the surgical process and even causing serious complications such as coughing and laryngospasm; while excessive muscle relaxation may prolong postoperative recovery time and increase the risk of respiratory depression. Therefore, continuous and accurate monitoring of muscle relaxation depth and prediction of future trends are core aspects of ensuring anesthetic safety.
[0003] Currently, clinical assessment of muscle relaxation primarily relies on stimulating peripheral nerves and observing the contraction response of the corresponding muscles, such as through Train-of-Four (TOF) counting. However, these methods are typically intermittent, failing to provide continuous monitoring data, and their resolution and sensitivity significantly decrease in deep muscle relaxation (TOF count of 0). To achieve continuous monitoring, academia and industry have begun researching the use of patients' spontaneous electromyographic (EMG) signals to assess muscle relaxation. Some approaches attempt to use machine learning models, such as Long Short-Term Memory (LSTM) networks, to analyze the temporal characteristics of EMG signals in an attempt to predict trends in muscle relaxation. However, the "black box" nature of these models makes their decision-making process difficult to interpret, and clinicians cannot intuitively understand the basis for the predictions, limiting their reliability in critical medical decisions. Summary of the Invention
[0004] This invention provides an anesthesia system for predicting muscle relaxation depth based on electromyographic signals, which at least addresses the problem of low decision-making reliability in related technologies.
[0005] According to one embodiment of the present invention, an anesthesia system for predicting muscle relaxation depth based on electromyography signals is provided, comprising: Electromyography (EMG) signal acquisition module, used to acquire EMG signals; A signal preprocessing module is used to preprocess the electromyographic signals; The spectrum state analysis module is used to perform windowing and spectrum transformation on the preprocessed electromyographic signal to generate spectrum characteristics for each time window, and to calculate the spectrum inertia reflecting the dynamic changes of the spectrum characteristics based on a preset spectrum inertia model; wherein, the spectrum characteristics include median frequency and total power. The trend prediction and display module is used to combine the current muscle relaxation depth with the spectral inertia to generate and display the muscle relaxation trend prediction result.
[0006] In one exemplary embodiment, it further includes: Before the step of windowing and spectral transformation of the preprocessed electromyographic signal to generate the spectral characteristics of each time window, the baseline electromyographic signal is acquired when the muscle relaxant reaches its peak effect. Based on the baseline electromyographic signal, the baseline median frequency and baseline total power are determined.
[0007] In one exemplary embodiment, it further includes: For the current time window, a normalized spectral state vector is constructed based on the median frequency, the total power, the baseline median frequency, and the baseline total power. The spectral velocity vector is calculated based on the normalized spectral state vector of the current time window and the normalized spectral state vector of the previous time window. The spectral inertia is updated by applying an exponentially weighted moving average to the norm of the spectral velocity vector.
[0008] In one exemplary embodiment, generating a muscle relaxation trend prediction result by combining the current muscle relaxation depth with the spectral inertia includes: The spectral inertia is compared with a preset inertia threshold. If the spectral inertia is greater than or equal to the inertia threshold, the muscle relaxation state is determined to be changing, and a corresponding muscle relaxation trend prediction result is generated; or, based on the component directions of the spectral velocity vector, the type of muscle relaxation recovery is distinguished; wherein, if all components of the spectral velocity vector are positive, the muscle relaxation trend prediction result is determined to be rapid recovery. If the spectral inertia is less than the inertia threshold, the muscle relaxation state is determined to be stable, and a corresponding muscle relaxation trend prediction result is generated.
[0009] According to another embodiment of the present invention, an anesthesia method for predicting muscle relaxation depth based on electromyographic signals is provided, comprising: Acquire and preprocess electromyographic signals; The preprocessed electromyographic signal is windowed and subjected to spectral transformation to generate spectral features for each time window, wherein the spectral features include median frequency and total power. Based on a preset spectral inertia model, the spectral inertia that reflects the dynamic changes of the spectral characteristics is calculated. By combining the current muscle relaxation depth with the spectral inertia, a muscle relaxation trend prediction result is generated.
[0010] In an exemplary embodiment, before performing windowing and spectral transformation on the preprocessed electromyographic signal to generate spectral features for each time window, the method further includes: Baseline electromyography (EMG) signals were acquired when the muscle relaxant reached its peak effect. Based on the baseline electromyographic signal, the baseline median frequency and baseline total power are determined.
[0011] In an exemplary embodiment, the step of calculating the spectral inertia reflecting the dynamic changes of the spectral characteristics based on a preset spectral inertia model includes: For the current time window, a normalized spectral state vector is constructed based on the median frequency, the total power, the baseline median frequency, and the baseline total power. The spectral velocity vector is calculated based on the normalized spectral state vector of the current time window and the normalized spectral state vector of the previous time window. The spectral inertia is updated based on the spectral velocity vector.
[0012] In one exemplary embodiment, generating a muscle relaxation trend prediction result by combining the current muscle relaxation depth with the spectral inertia includes: The spectral inertia is compared with a preset inertia threshold. If the spectral inertia is greater than or equal to the inertia threshold, the muscle relaxation state is determined to be changing, and a corresponding muscle relaxation trend prediction result is generated; or, based on the component directions of the spectral velocity vector, the type of muscle relaxation recovery is distinguished; wherein, if all components of the spectral velocity vector are positive, the muscle relaxation trend prediction result is determined to be rapid recovery. If the spectral inertia is less than the inertia threshold, the muscle relaxation state is determined to be stable, and a corresponding muscle relaxation trend prediction result is generated.
[0013] According to yet another embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.
[0014] According to yet another embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.
[0015] This invention predicts the trend of muscle relaxation by quantifying the dynamic rate of change of its spectral characteristics, i.e., "spectral inertia." This approach significantly reduces computational complexity and dependence on hardware resources, making it easy to deploy in embedded monitoring devices. Furthermore, because each step of the algorithm has a clear physical meaning, its prediction results are highly interpretable, enhancing the reliability for clinical applications. Therefore, it solves the problems of high computational resource consumption, poor interpretability, and low decision-making reliability, achieving low-latency, low-power, and high-reliability prediction of muscle relaxation trends. Attached Figure Description
[0016] Figure 1 This is a structural block diagram of an anesthesia system for predicting muscle relaxation depth based on electromyography signals according to an embodiment of the present invention; Figure 2 This is a flowchart of an anesthesia method for predicting muscle relaxation depth based on electromyography signals according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the normalized spectrum state vector according to an embodiment of the present invention; Figure 4 This is a partial flowchart according to Embodiment 3 of the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.
[0018] In the following description, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0019] Furthermore, in this application, directional terms such as "upper," "lower," "left," and "right" may be defined relative to the orientation of the components shown in the accompanying drawings. It should be understood that these directional terms can be relative concepts, used for relative description and clarification, and may change accordingly depending on the orientation of the components in the accompanying drawings.
[0020] In this application, unless otherwise expressly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral part; it can be a direct connection or an indirect connection through an intermediate medium. Furthermore, the term "coupled" can refer to a method of electrical connection for signal transmission.
[0021] As used herein, “about,” “approximately,” or “approximately” includes the stated value and the average value within an acceptable range of deviation from the given value, wherein the acceptable range of deviation is determined by a person skilled in the art taking into account the measurement under discussion and the error associated with the measurement of the given quantity (i.e., the limitations of the measurement system).
[0022] Example 1
[0023] Reference Figure 1This embodiment provides an anesthesia system for predicting muscle relaxation depth based on electromyography (EMG) signals. In a specific implementation, this system employs an EMG signal acquisition module 110, a signal preprocessing module 120, a spectrum state analysis module 130, and a trend prediction and display module 140. This enables real-time and continuous monitoring of the patient's muscle relaxation state and prediction of its changing trends. This system solves the technical problems of high computational resource consumption and poor interpretability caused by reliance on machine learning models in existing technologies, achieving the beneficial effect of high-reliability muscle relaxation trend prediction on low-power embedded devices.
[0024] Among them, the electromyography signal acquisition module 110 is used to non-invasively capture weak voltage signals reflecting muscle activity from the patient's body surface. This module is physically composed of a set of surface electrodes and a preamplifier circuit.
[0025] The surface electrodes can be medical-grade silver chloride (Ag / AgCl) electrodes, which have low polarization potential and good stability, making them suitable for long-term monitoring of bioelectrical signals. In use, a three-electrode configuration is typically employed: two differential measurement electrodes and one reference electrode. For example, the two measurement electrodes can be placed along the muscle fiber direction of the target muscle (such as the adductor pollicis, a standard muscle used to assess the degree of neuromuscular blockade), with a spacing of approximately 2 cm, while the reference electrode can be placed in a nearby area with low electrical activity, such as the styloid process of the ulna. This differential input configuration effectively suppresses common-mode noise, such as 50 / 60Hz power frequency interference from the surrounding environment. Before electrode application, the skin needs to be pretreated, for example, by cleaning with alcohol swabs to remove oil and gently exfoliating with a scrub to reduce stratum corneum impedance, ensuring that the contact impedance between the electrode and the skin is less than 5kΩ.
[0026] The preamplifier circuit is responsible for initially amplifying the raw electromyographic signals at the microvolt level captured by the electrodes. Since the electromyographic signals are very weak, typically between 10μV and 5mV, the preamplifier needs to have a very high input impedance (e.g., greater than 10mΩ) to avoid signal attenuation, and a very high common-mode rejection ratio (CMRR, e.g., greater than 100dB) to further eliminate common-mode noise. In particular, the amplification factor of this circuit is typically set at around 1000 times to boost the signal amplitude to the millivolt level for subsequent processing.
[0027] The signal preprocessing module 120 is responsible for further processing the initially amplified analog electromyographic signal to make it a clean signal sequence suitable for digital analysis; the main functions of this module include filtering, re-amplification and analog-to-digital conversion.
[0028] The filtering function is implemented using a bandpass filter. Since the main energy of the electromyography (EMG) signal is concentrated in the frequency range of 20Hz to 500Hz, this module uses a bandpass filter with a lower cutoff frequency set at 20Hz to filter out low-frequency interference such as motion artifacts, and an upper cutoff frequency set at 500Hz to filter out high-frequency noise. This filter can be a hardware-implemented analog filter, such as a 4th-order Butterworth filter composed of an operational amplifier and an RC network, designed to have the maximum flatness of the frequency response within the passband to ensure signal integrity. Alternatively, the filtering function can be implemented using a software-implemented digital filter, such as a finite impulse response (FIR) filter, here set to an order 128 FIR bandpass filter. The amplification function is achieved by a programmable gain amplifier (PGA). Since the amplitude of electromyographic signals may vary greatly among different patients and monitoring sites, the PGA can dynamically adjust the amplification factor according to the actual intensity of the signal, ensuring that the signal will not lose accuracy due to too small an amplitude, nor will it produce clipping distortion due to too large an amplitude, thereby maximizing the use of the dynamic range of the analog-to-digital converter.
[0029] The analog-to-digital conversion (ADC) function is implemented by an ADC with a sampling rate of up to 2048Hz to provide ample oversampling margin. The ADC's resolution determines its quantization accuracy; typically, a 12-bit ADC is chosen, which can divide the analog voltage range into... A quantitative level, for a With an input range of 2.5V, its quantization accuracy can reach approximately 1.22mV. After processing by this module, the continuous analog voltage signal is converted into a series of discrete digital sampling points, such as a data stream consisting of 16-bit signed integers, and sent to the spectrum state analysis module 130.
[0030] The spectrum state analysis module 130 is responsible for executing algorithms based on spectrum inertia analysis. This module receives digital electromyography (EMG) signal sequences from the signal preprocessing module 120 and outputs key parameters characterizing the muscle relaxation state and its changing trends. The calculation process can be implemented by a microcontroller (MCU) with digital signal processing (DSP) capabilities or a dedicated DSP chip. The module internally has a circular buffer to store the latest EMG signal sampling points. As new sampling points continuously flow in, the module extracts data segments from the buffer according to a preset time window length (e.g., 512 points) and overlap length (e.g., 256 points). For each data segment, the module first applies a window function (e.g., a Hamming window) to smooth the boundaries of the data segment, reducing spectral leakage caused by subsequent Fourier transforms. Then, the spectrum state analysis module 130 calls its internal hardware or software-implemented Fast Fourier Transform (FFT) engine to calculate the complex form of the spectrum of the windowed data segment. Subsequently, the spectrum state analysis module 130 calculates the power spectral density (PSD) based on the complex spectrum, and further calculates two core spectral features based on the power spectral density: median frequency (MDF) and total power (TP). Finally, the spectrum state analysis module 130 performs the calculation and update of spectral inertia, including the normalization of the state vector, the difference calculation of the velocity vector, and the exponentially weighted moving average of the velocity vector norm. Ultimately, the results of all these calculations, including the current MDF, TP, muscle relaxation depth, spectral inertia, etc., are packaged into a structured data frame and periodically (e.g., every 125 milliseconds) sent to the trend prediction and display module 140.
[0031] The trend prediction and display module 140 serves as the system's interface with the user, receiving calculation results from the spectrum state analysis module 130 and converting them into visual information easily understood by clinicians. This module can consist of a general-purpose processor and a liquid crystal display (LCD). Software running on the processor parses the received data frames. First, it updates a prominent number or a dynamic dashboard on the screen based on the received current muscle relaxation depth value (e.g., a value between 0 and 100). Then, it executes a judgment logic, comparing the received spectrum inertia value with a preset inertia threshold. If the inertia value is below the threshold, the software displays an icon or text label indicating "stable state," such as a green circular indicator light. If the inertia value exceeds the threshold, the software further analyzes the direction of the spectrum velocity vector. If both components of the velocity vector (corresponding to changes in MDF and TP) are positive, the system determines it to be in a "rapid recovery" state, potentially displaying a prominent red upward arrow on the screen and selectively triggering an audible alarm to alert the anesthesiologist to immediate attention. In other cases where the threshold is exceeded, a yellow marker indicating "slow recovery" or "changing status" may be displayed. In this way, the module presents the complex results of digital signal processing to clinicians in an intuitive and clear manner, achieving effective decision support.
[0032] Example 2
[0033] Reference Figure 2 This embodiment provides a method for predicting the depth of muscle relaxation based on electromyographic signals. This method can be executed by the system described in Embodiment 1. The various steps of the method will be described in detail below, combined with… Figure 3 The diagrams and numerical examples throughout are used to illustrate this.
[0034] Step S100: Acquire and preprocess electromyographic signals.
[0035] In this embodiment, S110: signal acquisition is performed first.
[0036] The system continuously acquires raw electromyographic (EMG) analog voltage signals at a sampling rate of 2048 Hz via Ag / AgCl surface electrodes attached to the patient's adductor pollicis muscle. At this sampling rate, 2048 voltage samples are generated per second, sufficient to capture frequency components up to 1024 Hz in the EMG signal, far exceeding the upper limit of the main energy range of EMG signals (approximately 500 Hz), ensuring signal integrity in the frequency domain. For example, at any given moment, the acquired raw signal might be a signal at... A weak signal that fluctuates within a certain range and is mixed with noise.
[0037] Next, execute S120: signal filtering.
[0038] The acquired raw signal is fed into a digital bandpass filter, which is a fourth-order IIR Butterworth filter with its passband precisely set between 20 Hz and 500 Hz. Components below 20 Hz (such as baseline drift caused by slight patient movement or breathing) are effectively filtered out; components above 500 Hz (such as high-frequency electromagnetic interference from surgical equipment or thermal noise from the circuit itself) are also suppressed. For example, after filtering, low-frequency baseline drift and high-frequency spike noise in the raw signal are significantly reduced, and the signal waveform becomes clearer, mainly consisting of bursts of electromyographic activity.
[0039] Then execute S130: Signal Amplification and Digitization.
[0040] The filtered signal is amplified by a programmable gain amplifier (PGA). The PGA automatically adjusts the gain based on the signal's root mean square (RMS) value, aiming to bring the signal's peak-to-peak value close to the full-scale range of the subsequent ADC, for example, adjusting it to... Left and right; subsequently, the amplified analog signal is converted by a 16-bit ADC at a sampling rate of 2048Hz. Here, 16-bit resolution means that the voltage range is divided into... Each level; and for a The range, its quantization accuracy is For example, in Instant arrival Within a time interval of seconds, obtain a sequence containing... A sequence of numbers from 32768 samples, where each value is a 16-bit integer between -32768 and 32767.
[0041] Step S200: Window and spectral transformation are performed on the preprocessed electromyographic signal to generate spectral characteristics for each time window.
[0042] In this embodiment, the time-domain signal is converted into a frequency-domain representation, and key features that can characterize the state of muscle relaxation are extracted, thereby facilitating subsequent processing.
[0043] Before this, a crucial step, S210, baseline calibration, needs to be performed. The specific process is as follows: After the anesthesiologist confirms that anesthesia induction is complete and the muscle relaxant (such as rocuronium bromide) has reached its maximum effect, the system will initiate a 30-second baseline acquisition cycle. During this period, the system calculates the MDF and TP values for each time window. After 30 seconds, the system will calculate the average of all MDF values and the average of all TP values during this period, and store these two averages as the patient's individualized baseline.
[0044] For example, after 30 seconds of calibration, the system calculates and stores the baseline median frequency. and baseline total power These two values will be used in all subsequent normalization calculations.
[0045] Then, execute S220: Time window division.
[0046] The system divides the continuous digital signal stream output from the aforementioned steps into a series of overlapping time windows, each with a length of... There are 512 sampling points, and the corresponding time length is The overlap length between time windows With 256 sampling points, representing a 50% overlap, the system helps smooth the temporal evolution of the feature sequence and avoids abrupt feature changes caused by window boundary cutting effects. Therefore, the system samples every... This will generate a new analysis result. Optionally, in scenarios requiring faster response times, the overlap rate can be increased to 75%, i.e., an overlap length of 384 points, which will double the frequency of system result updates.
[0047] Next, execute S230: spectrum calculation.
[0048] For each time window containing 512 sampling points, the system first applies a Hamming window function. The Hamming window approaches zero at both ends and is 1 in the middle, effectively reducing the Gibbs phenomenon (spectral leakage) caused by signal truncation. Then, the system performs a 512-point Fast Fourier Transform (FFT) on the windowed 512 data points. The result of the FFT is a sequence of 512 complex numbers, where the nth complex number corresponds to a frequency of... The system then calculates the square of the modulus of each complex number to obtain the power spectral density (PSD) sequence for that time window. Of course, in addition to using FFT for batch processing, short-time Fourier transform (STFT) can also be used to treat the whole process as a time-frequency analysis of the signal, thereby obtaining the distribution map of electromyographic signal energy in both time and frequency dimensions.
[0049] Finally, execute S240: Spectral Feature Extraction.
[0050] Based on the obtained power spectral density sequence, the system calculates two core features for each time window: the first feature is the total power (TP), which is obtained by numerically integrating the power spectral density over the effective frequency range (20Hz to 500Hz).
[0051] in, This is the frequency resolution, here 4Hz. TP reflects the total energy of muscle activity within this time window.
[0052] The second characteristic is the median frequency (MDF), which is a frequency that divides the total power spectrum energy in half. It is calculated by accumulating the power spectrum from low to high frequencies until the sum equals half the total power at that frequency.
[0053] The median frequency of the MDF reflects the central trend of the power spectrum, which is related to the conduction velocity and recruitment patterns of muscle fibers. Specifically, spectral features may also include, but are not limited to: spectral entropy, spectral sideband 95 (SEF95), or the proportion of energy in a specific frequency band (e.g., 30-60 Hz) to the total energy. Spectral entropy measures the randomness and uncertainty of the power spectrum; it typically increases during muscle relaxation recovery. SEF95 refers to the lowest frequency containing 95% of the total energy and also reflects the shift of the spectrum towards higher frequencies. These features can also reflect changes in the spectral structure of electromyographic signals from different dimensions and can serve as a supplement or replacement for MDF and TP to construct a higher-dimensional spectral state vector.
[0054] S300: Based on a preset spectral inertia model, calculate the spectral inertia that reflects the dynamic changes in the spectral characteristics.
[0055] In this embodiment, the change in muscle relaxation state needs to be abstracted as a physical inertial process, specifically: First, execute S310: Construct the normalized spectral state vector.
[0056] For the k-th time window currently being processed, the system utilizes the previously extracted... and and baseline value and To construct a two-dimensional, dimensionless normalized spectral state vector. :
[0057] like Figure 3 As shown, this vector can be represented as a point in a two-dimensional state space with normalized MDF as the horizontal axis and normalized TP as the vertical axis. In the baseline state of deep muscle relaxation, this point is located near coordinate (1,1). As muscle relaxation recovers, this point will move away from the origin.
[0058] Next, execute S320: Calculate the spectral velocity vector.
[0059] The system calculates the current state vector. Compared to the state vector of the previous time window The first-order difference between them yields the spectral velocity vector. :
[0060] This vector represents the direction and distance (velocity) of the state point's movement within a unit time window interval, such as... Figure 3 As indicated by the arrows, its magnitude (norm) reflects the rate of state change, while its direction reflects whether MDF and TP are increasing or decreasing.
[0061] Finally, execute S330: calculate spectral inertia.
[0062] To smoothly measure the drasticness of state changes and avoid the influence of noise from single measurements, the system uses an exponentially weighted moving average (EWMA) to calculate the spectral inertia. :
[0063] in, It is the Euclidean norm of the spectral velocity vector, representing the magnitude of the velocity; It is a smoothing factor whose value is limited to the interval [0.9, 0.99].
[0064] For example, It can be set to 0.95. The meaning of this formula is that the current inertia value is composed of 95% of the inertia from the previous moment and 5% of the current velocity. This makes... It can smoothly track the long-term trend of state changes; it should be noted that the smoothing factor... The selection of [a specific component] has a direct impact on system performance; a larger [specific component]... A value (such as 0.99) produces a very smooth inertia curve, effectively filtering out noise, but it responds slowly to abrupt changes in state; a smaller value... A value (such as 0.90) will make the inertia curve more closely resemble the instantaneous velocity change, resulting in a more sensitive response, but it is also more susceptible to noise interference. In practice, adjustments can be made based on different clinical requirements for response speed and stability. Alternatively, a simplified sliding window averaging model can be used for the spectral inertia model. Specifically, the system can calculate the arithmetic mean of the norms of the most recent N (e.g., N=10) spectral velocity vectors as the current spectral inertia. While this method results in a slight delay in response, it is simpler to calculate and is a viable alternative on some devices with extremely limited computing resources.
[0065] S400: Combines the current muscle relaxation depth with spectral inertia to generate a muscle relaxation trend prediction result.
[0066] In this embodiment, S410 is executed first: calculate the current muscle relaxation depth.
[0067] Current muscle relaxation depth It can be determined by the normalized total power. To define it. Since depth and power are inversely proportional, the following formula can be used for calculation:
[0068] Here we use The function ensures that the depth value is not negative; when equal When the depth is 0, it indicates that muscle relaxation has returned to the baseline level; when much smaller For example If the depth is 95, it indicates deep muscle relaxation; this value provides doctors with an intuitive quantitative assessment of the current state of muscle relaxation.
[0069] Next, execute S420: Determine the predicted trend.
[0070] The system calculates the spectral inertia based on the current data. With a preset inertia threshold Comparisons are made to determine trends; the inertia threshold is used here. It can be set to 0.1.
[0071] Rule 1: If The system determines that the current muscle relaxation state is "stable," which means that the spectral characteristics have not changed drastically in the recent period. This result informs the doctor that the current dose of anesthetic drugs can maintain a stable level of muscle relaxation and no immediate intervention is required.
[0072] Rule 2: If The system determines that the current muscle relaxation state is "changing," indicating that the balance of muscle relaxation levels has been disrupted and a significant change is about to occur or is already occurring. In this case, further analysis of the spectral velocity vector is needed. The direction is to provide more refined predictions.
[0073] Rule 2a: If and This means that both the median frequency and total power are increasing significantly, which is a strong signal of rapid recovery of muscle relaxation. The system judges the trend as "rapid recovery". This suggests to the doctor that the patient's muscle relaxation may suddenly disappear in a short period of time, with the risk of physical movement, and that additional muscle relaxant medication needs to be prepared immediately.
[0074] Rule 2b: In other cases (e.g., only one component is positive, or some components are negative), the system determines the trend as "slow recovery / change," which indicates that the muscle relaxation state is changing, but not as drastically as in Rule 2a; this provides the doctor with a buffer period to calmly assess the situation and decide whether to adjust the medication dosage.
[0075] Finally, execute S430: Output and Display.
[0076] The system sends the calculated muscle relaxation depth value (e.g., "95") and the determined trend (e.g., "stable state") to the display screen. The display interface can be designed to present information in a multi-level manner. For example, the main interface displays the depth value in large font, with a color-coded status indicator light next to it (green represents stable, yellow represents changing, and red represents rapid recovery). Users can also click to enter the details page to view the real-time change curves of parameters such as MDF, TP, and spectral inertia, providing data support for advanced clinical analysis. In addition, this system can also be linked with the Anesthesia Information Management System (AIMS) or the Target-Controlled Infusion (TCI) pump. When the system detects a "rapid recovery" trend, it can automatically send an alarm event to the AIMS or directly send a command to the TCI pump to suggest adjusting the drug infusion rate, thus forming a semi-automatic or fully automatic closed-loop system for muscle relaxation management.
[0077] It should be noted that, in order to avoid inertia threshold Determining the optimal setting, which requires subjective input or excessive experimentation by clinicians, can be achieved through a systematic offline calibration method. This method can be used before the device leaves the factory or in the early stages of clinical use, by analyzing an existing, labeled electromyography (EMG) signal database to determine an optimal and universally applicable setting. value.
[0078] Specifically, the calibration method includes the following steps: Data preparation: Prepare a dataset containing electromyographic signals from multiple patients during anesthesia recovery. The dataset should include expert annotations on the state of muscle relaxation, for example, clearly marking the precise time point at which "muscle relaxation begins to recover."
[0079] Feature calculation: For each signal segment in the dataset, calculate the continuous spectral inertia. Time series.
[0080] ROC curve analysis: Set a candidate threshold range, for example, from 0.01 to 0.5, with a step size of 0.005. For each candidate within this range... The system iterates through the entire dataset to determine whether it can successfully issue an alert within a short time window (e.g., within 30 seconds) after the expert-annotated "recovery point" (i.e.,...). Based on this, each candidate can be calculated. The values correspond to the true positive rate (sensitivity) and false positive rate (1-specificity); by plotting these (false positive rate, true positive rate) pairs on a two-dimensional coordinate system, the receiver operating characteristic (ROC) curve can be obtained.
[0081] Optimal threshold selection: Calculate the Youden index for each point on the ROC curve: Sensitivity + Specificity - 1, and its corresponding candidate threshold. The value is considered to be the threshold that best balances sensitivity and specificity.
[0082] For example, by analyzing a database containing data from 50 patients, it was found that... At this point, the Youden index reaches its maximum value of 0.85 (at which point the sensitivity is 92% and the specificity is 93%); therefore, 0.115 can be used as the default preset inertia threshold for this device, and so on.
[0083] The following examples illustrate this.
[0084] Assuming the system has completed baseline calibration: , .
[0085] Initial state (k=0): , Smoothing factor Inertia threshold .
[0086] Time k=52 (rapid recovery of muscle relaxation): S240: Measured , .
[0087] S310: .
[0088] S320: .
[0089] S330: .
[0090] S410: .
[0091] S420: The assessment result is that the state is stable.
[0092] Time k=54 (inertia exceeds threshold): S240: Measured , .
[0093] S310: .
[0094] S320: .
[0095] S330: ; .
[0096] S410: .
[0097] S420: Triggering Rule 2; Further Analysis If both components are positive, rule 2a is triggered.
[0098] S430: Display: Depth: 0, Trend: Fast Recovery (and trigger an audible or visual alarm).
[0099] Example 3
[0100] In actual clinical applications, the original electromyographic signals may be contaminated by non-physiological artifacts such as electrosurgery interference, patient movement, and poor electrode contact, which can lead to serious deviations in the predicted results of muscle relaxation depth and trend based on artifact signals, misleading clinical decisions. In this regard, unlike Example 2, this example evaluates the data quality and dynamically adjusts the final output based on the evaluation results.
[0101] like Figure 4 As shown, in Figure 2 Based on the process shown, a step S500 is added that is executed in parallel with S400.
[0102] S500: Determines the reliability of muscle relaxation trend prediction results based on signal quality metrics.
[0103] This step is performed synchronously with S400 (generating muscle relaxation trend prediction results). It receives raw time-domain and frequency-domain data from the same time window from S200 and outputs a reliability level for the data quality of that time window. This step can be further broken down into the following sub-steps: S510: Calculate the time-domain saturation index.
[0104] The system iterates through each 16-bit digital sample value within the current time window (e.g., 512 sampling points), counting the number of sampling points whose absolute value is equal to or exceeds 99% of the ADC's maximum range, and records this as the saturation point count. .
[0105] For example, for a 16-bit ADC (range -32768 to 32767), its maximum range is 32767, and the threshold can be set to... The system will then count the number of points with an absolute value greater than 32440 out of the current 512 sampling points. Assuming that within a severely disturbed time window, 15 points are detected with values of 32767 or -32768, the saturation point count for that time window will be determined. .
[0106] S520: Calculate the spectral morphology index.
[0107] The power spectrum of physiological electromyographic signals has specific morphological characteristics (e.g., single-peak, Gaussian-like distribution), while the spectra of various artifact signals (such as electrosurgical noise) exhibit significantly different morphologies. In this case, the source of the signal can be distinguished from the frequency domain by calculating at least two spectral morphology indices.
[0108] S521: Calculate the spectral flatness.
[0109] Spectral flatness (SFM) measures the uniformity of power spectrum energy distribution within a frequency band. A value close to 1 indicates a uniform spectral energy distribution, similar to white noise; a value much less than 1 indicates that the energy is concentrated at a few frequency points, similar to pure tone. Physiological electromyographic signals exhibit relatively concentrated energy, while the spectrum of high-frequency noise such as that from electrosurgical procedures is typically wider and flatter.
[0110] Spectral flatness The ratio of the geometric mean to the arithmetic mean of the power spectrum is obtained by calculating:
[0111] in, It is a power spectral density sequence, and N is the number of frequency points involved in the calculation.
[0112] For example, for a time window dominated by physiological signals, its power spectrum energy is mainly concentrated in the range of 40-150Hz, and the calculated... It could be 0.15. However, for a signal contaminated by electrosurgical noise, its power spectrum may maintain a high energy level within the 20-500Hz range, and the calculated value... It could be as high as 0.70, and so on.
[0113] S522: Calculate the high-frequency energy ratio.
[0114] This indicator is used to quantify the high-frequency noise component in a signal; the energy of physiological electromyography signals is mostly below 300Hz, while artifacts such as those from electrosurgical excision typically contain very rich high-frequency components; and the high-frequency energy ratio The ratio of energy set to the high-frequency band (e.g., 300-500Hz) to the total frequency band (e.g., 20-500Hz):
[0115] For example, in a normal electromyography (EMG) signal time window, the proportion of high-frequency energy is very small. The calculated value might be 0.08; while a signal interfered with by an electrosurgical unit, its... The value could rise sharply to 0.65, and so on.
[0116] S530: Synthetic signal reliability index and adjusts the prediction results.
[0117] Based on multiple calculated quality indicators, discrete signal reliability levels (SRLs) are synthesized using preset clinical rules, and these levels are used to adjust the display method of the original prediction results generated by the S400.
[0118] The preset rules and reliability levels are defined as follows, where , , It is a preset threshold: adjustable. (indivual), , ).
[0119] Rule 1: If or or If the signal reliability level SRL for the current time window is low, then the signal reliability level for that time window is determined to be low.
[0120] Rule 2: If Rule 1 is not satisfied, but or In the middle range (e.g., [0.3, 0.5]) or If it falls within the medium range (e.g., [0.2, 0.4]), then SRL is determined to be of medium reliability.
[0121] Rule 3: If none of the above rules are met, then SRL is determined to be highly reliable.
[0122] Based on the judgment-based SRL, the system dynamically adjusts the display of the muscle relaxation depth and trend prediction results output by S400 as follows: When SRL=High Reliability: All results calculated by S400 are displayed normally. For example, "Depth: 55, Trend: Slow Recovery" is displayed on the screen in standard colors (such as white or green font).
[0123] When SRL = Medium Reliability: The result calculated by S400 is displayed, but with explicit warning labels. For example, the displayed font color will change to yellow, and a question mark icon (?) will be shown next to it, indicating to clinicians that the current reading may be slightly affected and should be interpreted with caution.
[0124] When SRL = Low Reliability: The display of numerical and trend results calculated by S400 is completely suppressed, and instead a clear warning message covering the entire display area is displayed, such as "Poor signal quality, unreliable data" or "Electrical knife interference detected"; at the same time, the system will suspend the calculation of spectral inertia. The update maintains the value from the previous high-reliability moment to avoid artifact data contaminating the inertia accumulation calculation. Normal calculation and display will resume once the signal quality recovers to medium or high reliability.
[0125] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0126] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.
[0127] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.
[0128] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.
[0129] Embodiments of the present invention also provide an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.
[0130] In one exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.
[0131] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.
[0132] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0133] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0134] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0135] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0136] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. An anesthesia system for predicting muscle relaxation depth based on electromyographic signals, characterized in that, include: Electromyography (EMG) signal acquisition module, used to acquire EMG signals; A signal preprocessing module is used to preprocess the electromyographic signals; The spectrum state analysis module is used to perform windowing and spectrum transformation on the preprocessed electromyographic signal to generate spectrum characteristics for each time window, and to calculate the spectrum inertia reflecting the dynamic changes of the spectrum characteristics based on a preset spectrum inertia model; wherein, the spectrum characteristics include median frequency and total power. The trend prediction and display module is used to combine the current muscle relaxation depth with the spectral inertia to generate and display the muscle relaxation trend prediction result.
2. The system according to claim 1, characterized in that, Also includes: Before the step of windowing and spectral transformation of the preprocessed electromyographic signal to generate the spectral characteristics of each time window, the baseline electromyographic signal is acquired when the muscle relaxant reaches its peak effect. Based on the baseline electromyographic signal, the baseline median frequency and baseline total power are determined.
3. The system according to claim 2, characterized in that, Also includes: For the current time window, a normalized spectral state vector is constructed based on the median frequency, the total power, the baseline median frequency, and the baseline total power. The spectral velocity vector is calculated based on the normalized spectral state vector of the current time window and the normalized spectral state vector of the previous time window. The spectral inertia is updated by applying an exponentially weighted moving average to the norm of the spectral velocity vector.
4. The system according to claim 1, characterized in that, The process of combining the current muscle relaxation depth with the spectral inertia to generate a muscle relaxation trend prediction result includes: The spectral inertia is compared with a preset inertia threshold. If the spectral inertia is greater than or equal to the inertia threshold, the muscle relaxation state is determined to be changing, and a corresponding muscle relaxation trend prediction result is generated; or, based on the component directions of the spectral velocity vector, the type of muscle relaxation recovery is distinguished; wherein, if all components of the spectral velocity vector are positive, the muscle relaxation trend prediction result is determined to be rapid recovery. If the spectral inertia is less than the inertia threshold, the muscle relaxation state is determined to be stable, and a corresponding muscle relaxation trend prediction result is generated.
5. A method for predicting the depth of muscle relaxation in anesthesia based on electromyographic signals, characterized in that, include: Acquire and preprocess electromyographic signals; The preprocessed electromyographic signal is windowed and subjected to spectral transformation to generate spectral features for each time window, wherein the spectral features include median frequency and total power. Based on a preset spectral inertia model, the spectral inertia that reflects the dynamic changes of the spectral characteristics is calculated. By combining the current muscle relaxation depth with the spectral inertia, a muscle relaxation trend prediction result is generated.
6. The method according to claim 5, characterized in that, Before performing windowing and spectral transformation on the preprocessed electromyographic signal to generate the spectral characteristics of each time window, the method further includes: Baseline electromyography (EMG) signals were acquired when the muscle relaxant reached its peak effect. Based on the baseline electromyographic signal, the baseline median frequency and baseline total power are determined.
7. The method according to claim 6, characterized in that, The step of calculating the spectral inertia reflecting the dynamic changes of the spectral characteristics based on a preset spectral inertia model includes: For the current time window, a normalized spectral state vector is constructed based on the median frequency, the total power, the baseline median frequency, and the baseline total power. The spectral velocity vector is calculated based on the normalized spectral state vector of the current time window and the normalized spectral state vector of the previous time window. The spectral inertia is updated based on the spectral velocity vector.
8. The method according to claim 5, characterized in that, The process of combining the current muscle relaxation depth with the spectral inertia to generate a muscle relaxation trend prediction result includes: The spectral inertia is compared with a preset inertia threshold. If the spectral inertia is greater than or equal to the inertia threshold, the muscle relaxation state is determined to be changing, and a corresponding muscle relaxation trend prediction result is generated; or, based on the component directions of the spectral velocity vector, the type of muscle relaxation recovery is distinguished; wherein, if all components of the spectral velocity vector are positive, the muscle relaxation trend prediction result is determined to be rapid recovery. If the spectral inertia is less than the inertia threshold, the muscle relaxation state is determined to be stable, and a corresponding muscle relaxation trend prediction result is generated.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is configured to perform the method described in any one of claims 5 to 8 when executed.
10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method as described in any one of claims 5 to 8.