Intelligent anesthesia depth regulation and control system based on multi-mode biological signals

Abstract

The invention discloses an intelligent anesthesia depth regulation and control system based on a multi-modal biological signal, which relates to the technical field of anesthesia depth regulation and control and comprises a multi-modal biological signal acquisition module, a signal fusion processing module, an intelligent anesthesia depth analysis module, an intelligent regulation and control decision module and an intelligent control terminal. Multi-modal biological signals of a patient are collected in real time, multi-modal physiological signals such as electroencephalogram, electrocardio and blood oxygen are fused, intelligent analysis is performed in combination with a deep learning model and a fuzzy PID control algorithm, and in combination with a pharmacokinetic model, the individual difference of the patient can be considered, the anesthesia depth of the patient is accurately evaluated, and the accuracy of anesthesia is improved. Individual anesthetic dosage adjustment is completed, a basis is provided for accurate regulation and control of anesthetic, accurate closed-loop regulation and control of anesthesia depth are achieved, physiological changes of a patient can be responded in real time, the anesthetic dosage can be adjusted in time, and compared with a traditional manual regulation and control mode, the risk of knowing and excessive anesthesia in an operation can be reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of anesthesia depth regulation, and in particular to an intelligent anesthesia depth regulation system based on multimodal biological signals. Background Art

[0002] Biological signals are time-series data that reflect the physiological state of organisms. They are widely used in medical diagnosis, health monitoring, and scientific research. Their research covers signal acquisition, processing, analysis, and multi-scenario applications. Multimodal biosignals refer to multiple biological data signals from different sensors or devices, including physiological signals, imaging data, biochemical data, and motion data. Through fusion processing, they can provide more comprehensive and accurate biological information.

[0003] Anesthesia refers to the technology of suppressing the function of the nervous system through drugs or other methods, causing the patient to temporarily lose pain perception, consciousness or movement ability. It is widely used in surgery, pain management and critical care. The depth of anesthesia refers to the anesthetic management goal of ensuring intraoperative unconsciousness and stable physiological state by monitoring the patient's probability of no response to stimulation and brain function status. Its core is to balance the dosage of anesthetic drugs to prevent intraoperative awareness and avoid excessive inhibition.

[0004] Anesthesia depth regulation refers to the process of precisely controlling the dosage and type of anesthetic drugs during general anesthesia, and monitoring the patient's physiological indicators in real time to ensure that the patient is in an appropriate anesthetic state during surgery. This not only avoids risks such as respiratory depression and circulatory instability caused by excessively deep anesthesia, but also prevents intraoperative awareness and pain caused by excessively shallow anesthesia. By precisely controlling the depth of anesthesia, the patient's safety and comfort during surgery can be ensured, and rapid postoperative recovery can be promoted.

[0005] In the existing technology, the regulation and management of anesthesia depth mainly relies on the experience of the anesthesiologist, and lacks objective and real-time monitoring means, resulting in inaccurate anesthesia depth control. In addition, the traditional anesthesia regulation method cannot respond to the patient's physiological changes in real time, and it is difficult to adjust the anesthetic drug dosage in time, which increases the risk of excessive anesthesia during surgery. The timeliness needs to be improved. In addition, due to individual differences in patients' sensitivity to anesthetic drugs, the existing technology is difficult to implement individualized anesthesia management according to the patient's specific situation, and cannot effectively improve the patient's comfort and surgical safety. Therefore, the present invention proposes an intelligent anesthesia depth regulation system based on multimodal biosignals to solve the problems existing in the existing technology. Summary of the Invention

[0006] In response to the above problems, the purpose of the present invention is to propose an intelligent anesthesia depth control system based on multimodal biological signals to solve the problems that the existing anesthesia depth control and management methods are not accurate enough in controlling the anesthesia depth, it is difficult to adjust the anesthetic drug dosage in time, and it is impossible to implement individualized anesthesia management according to the patient's specific situation.

[0007] To achieve the purpose of the present invention, the present invention is implemented through the following technical solutions: an intelligent anesthesia depth control system based on multimodal biosignals, including a multimodal biosignal acquisition module, a signal fusion processing module, an anesthesia depth intelligent analysis module, an intelligent control decision module and an intelligent control terminal, wherein the modal biosignal acquisition module is used to obtain the patient's electroencephalogram (EEG) signal, electrocardiogram (ECG) signal, photoplethysmogram (PEP) signal, blood pressure and respiratory rate signal, and body temperature data signal in real time;

[0008] The signal fusion processing module performs noise reduction, feature extraction and multimodal data fusion processing on each signal collected by the multimodal biological signal acquisition module to generate a standardized physiological feature vector of the patient;

[0009] The anesthesia depth intelligent analysis module analyzes the patient's standardized physiological feature vector based on a deep learning model and outputs an anesthesia depth assessment index and a predicted intraoperative risk level;

[0010] The intelligent control decision module generates an anesthetic drug infusion rate adjustment instruction based on the deviation between the anesthesia depth assessment index output by the anesthesia depth intelligent analysis module and the preset target range, combined with the patient's individual pharmacokinetic model;

[0011] The intelligent control terminal feeds back the adjustment instructions generated by the intelligent control decision module to the anesthesia infusion equipment in real time, dynamically adjusts the infusion rate of the anesthetic drug, and maintains a stable anesthesia depth.

[0012] A further improvement is that the multimodal biosignal acquisition module includes a wireless bioelectrode array for acquiring the patient's EEG signals and ECG signals, a PPG sensor for synchronously acquiring the patient's blood oxygen saturation and pulse waveform, a respiratory monitoring unit for obtaining the patient's respiratory rate, and a body temperature monitoring unit for acquiring the patient's temperature data.

[0013] Further improvements are: the wireless bioelectrode array integrates dry EEG electrodes and flexible ECG electrodes and uses a non-invasive method to collect the patient's EEG signals and ECG signals; the PPG sensor is integrated into the patient's earlobe or fingertip wearable device; the respiratory monitoring unit is an impedance respiratory sensor and calculates the respiratory rate through impedance changes; the body temperature monitoring unit uses a MEMS temperature sensor.

[0014] Further improvements are as follows: the specific steps of signal processing in the signal fusion processing module are: using wavelet transform to remove power frequency interference from EEG signals, and extracting the power spectral density of α waves and β waves as features; using adaptive filtering to eliminate motion artifacts from ECG signals, extracting time domain and frequency domain indicators of heart rate variability, fusing multimodal features through the attention mechanism and generating the patient's physiological feature vector.

[0015] Further improvements are as follows: the deep learning model is a hybrid architecture of a dual-channel convolutional neural network and a long short-term memory network, wherein: the dual-channel convolutional neural network branch is used to extract the spatiotemporal features of EEG signals, and the long short-term memory network branch is used to model the temporal dependencies of ECG signals and photoplethysmography signals. The fusion layer integrates multimodal features through a weighted attention mechanism to output the anesthesia depth assessment index value and the risk probability of intraoperative hypotension and delayed awakening.

[0016] Further improvements are as follows: the intelligent control decision module adopts fuzzy PID control algorithm to output decision, and the specific steps are: input anesthesia depth assessment index deviation and its rate of change, dynamically adjust the proportion, integral, and differential coefficients through the fuzzy rule library, combine the patient's weight, age, liver and kidney function data, call the pharmacokinetic model to calculate the drug plasma target concentration, and output the infusion rate adjustment amount of the anesthetic drug.

[0017] Further improvements are as follows: the intelligent control terminal includes an instruction execution module, a risk warning unit and a human-computer interaction interface. The instruction execution module controls the anesthesia infusion equipment to execute the adjustment instructions generated by the intelligent control decision module. When the predicted intraoperative risk probability exceeds the threshold, the risk warning unit triggers an audible and visual alarm and recommends intervention measures. The human-computer interaction interface displays the anesthesia depth assessment index trend chart, drug infusion curve and vital signs dashboard in real time, supporting the anesthesiologist to manually modify the control instructions.

[0018] A further improvement is that it also includes a cloud platform docking module, which is used to upload intraoperative data to the medical cloud database in real time for cloud storage backup.

[0019] The beneficial effects of the present invention are as follows: the present invention collects multimodal biological signals of patients in real time, integrates multimodal physiological signals such as EEG, ECG, and blood oxygen, combines deep learning models with fuzzy PID control algorithms for intelligent analysis, and combines pharmacokinetic models. It can take into account individual differences in patients, accurately evaluate the depth of anesthesia of patients, and complete individualized anesthetic drug dosage adjustment, providing a basis for precise regulation of anesthetic drugs, realizing precise closed-loop regulation of anesthesia depth, and can respond to physiological changes of patients in real time and adjust the dosage of anesthetic drugs in time. Compared with traditional manual control methods, it can reduce the risk of intraoperative awareness and excessive anesthesia, improve the safety of complex operations, reduce the incidence of postoperative complications, and enhance the patient's surgical experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a structural schematic diagram of the intelligent anesthesia depth control system based on multimodal biological signals of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0022] Biosignals refer to physical or chemical change signals generated inside or outside an organism that carry information about the biological system. They are widely present at all levels of life activities. Their main types are electrophysiological signals, biochemical signals, mechanical signals, and optical signals. Electrophysiological signals include electrocardiographic signals (reflecting cardiac electrical activity and used to diagnose arrhythmias, myocardial ischemia, etc.), electroencephalographic signals (monitoring brain neuronal discharges and used for epilepsy diagnosis and brain-computer interfaces), electromyographic signals (recording muscle electrical activity, assessing motor function or controlling prosthetic limbs), and electrooculographic signals (tracking eye movements and used for sleep research or interactive devices). Biochemical signals include hormones (insulin, adrenaline), neurotransmitters (dopamine, serotonin), and cytokines (interleukins). Mechanical signals include blood pressure / pulse waves (reflecting cardiovascular status) and respiratory signals (monitoring respiratory rate and depth and used for sleep disorders or intensive care). Optical signals include blood oxygen saturation (measured by photoplethysmography) and fluorescently labeled molecules (used for cell imaging or gene expression research).

[0023] Anesthesia is to keep the patient in a painless, unconscious and stable state during surgery to ensure the safety and success of the operation. The depth of anesthesia refers to the patient's response to external stimuli during anesthesia, and is an important indicator for evaluating the effectiveness of anesthesia. Anesthesia depth regulation is the process of monitoring and regulating the patient's anesthetic state during anesthesia. Its purpose is to ensure that the patient is in an appropriate depth of anesthesia during surgery, without experiencing pain or recovery of consciousness, and without complications caused by excessive anesthesia.

[0024] according to Figure 1 As shown, this embodiment provides an intelligent anesthesia depth control system based on multimodal biosignals, which consists of a multimodal biosignal acquisition module for collecting various biosignals of patients to be anesthetized, a signal fusion processing module for fusing the collected biosignals, an anesthesia depth intelligent analysis module for analyzing the patient's anesthesia depth according to the signal fusion results, an intelligent control decision module for generating anesthesia control instructions, and an intelligent control terminal for human-computer interaction and executing anesthesia control instructions. The multimodal biosignal acquisition module, the signal fusion processing module, the anesthesia depth intelligent analysis module, and the intelligent control decision module are wirelessly connected in sequence, and the intelligent control terminal is wirelessly connected to the multimodal biosignal acquisition module, the signal fusion processing module, the anesthesia depth intelligent analysis module, and the intelligent control decision module respectively. The wireless transmission protocol adopts Bluetooth 5.0 and Wi-Fi 6 dual-mode transmission to ensure signal stability in the operating room (packet loss rate <0.1%) and delay ≤50ms;

[0025] The modal biosignal acquisition module of this embodiment is composed of a wireless bioelectrode array, a PPG sensor, a respiratory monitoring unit and a body temperature monitoring unit. The wireless bioelectrode array integrates dry EEG electrodes (international standard lead configuration with an electrode spacing of 10 to 20, a sampling rate of ≥256Hz, and an impedance of ≤5kΩ) and flexible ECG electrodes (supporting single-lead / three-lead mode, a dynamic range of ±5mV, and a common-mode rejection ratio of ≥100dB) and uses a non-invasive method to collect brain electroencephalogram (EEG) and electrocardiogram (ECG) signals. The PPG sensor (using a dual-wavelength design, a sampling frequency of 100Hz, and a blood oxygen saturation measurement error) is used to collect brain electroencephalogram (EEG) and electrocardiogram (ECG) signals. The device is integrated into the patient's earlobe or fingertip wearable device to synchronously collect the patient's blood oxygen saturation and pulse waveform signal (PPG). The respiratory monitoring unit is an impedance respiratory sensor (excitation frequency 50kHz, chest impedance measurement range 0.1-5Ω, respiratory rate resolution 0.1 times / min), embedded in the patient's chest strap, and calculates the patient's respiratory rate through impedance changes. The body temperature monitoring unit uses a MEMS temperature sensor (measurement range 25-45℃, accuracy ±0.05℃, response time <1 second), attached to the patient's armpit or tympanic membrane position, to collect the patient's body temperature data signal;

[0026] The signal fusion processing module of this embodiment uses a signal preprocessing unit to perform noise reduction and feature extraction on each collected biological signal, and uses a multimodal fusion unit to perform multimodal data fusion to generate a standardized physiological feature vector of the patient. The specific steps are: using wavelet transform to remove power frequency interference from the EEG signal, and extracting the power spectral density of α waves (8-13Hz) and β waves (14-30Hz) as features; using adaptive filtering to eliminate motion artifacts from the ECG signal, extracting time domain (SDNN) and frequency domain (LF / HF) indicators of heart rate variability (HRV); using Savitzky-Golay filter to smooth the waveform of the PPG signal, calculating the pulse wave propagation time (PTT) for blood pressure estimation, and fusing multimodal features through the attention mechanism to generate a physiological feature vector with a dimension of 128;

[0027] The anesthesia depth intelligent analysis module of this embodiment analyzes the patient's standardized physiological feature vectors based on a deep learning model, and outputs the ADAI and predicted intraoperative risk level. The deep learning model is a hybrid architecture of a dual-channel convolutional neural network (CNN) and a long short-term memory network (LSTM). The CNN branch in the model is used to extract the spatiotemporal features of the EEG signal, and the LSTM branch is used to model the temporal dependencies of the ECG and PPG signals. The fusion layer in the deep learning model integrates multimodal features through a weighted attention mechanism and outputs the ADAI value and the risk probability of intraoperative hypotension and delayed awakening. The system integrates multiple physiological signals such as EEG, ECG, and blood oxygen, and performs comprehensive analysis through the deep learning model, thereby improving the accuracy of anesthesia depth assessment.

[0028] The intelligent control decision module of this embodiment generates an anesthetic drug infusion rate adjustment instruction based on the deviation of the anesthesia depth assessment index output by the anesthesia depth intelligent analysis module and the preset target range (40-60) in combination with the patient's individual pharmacokinetic model using a fuzzy PID control algorithm. The specific decision-making steps are as follows: inputting the ADAI deviation (e) and its rate of change (Δe), dynamically adjusting the proportional (Kp), integral (Ki), and differential (Kd) coefficients through a fuzzy rule base, combining the patient's weight, age, and liver and kidney function data, calling a pharmacokinetic model (such as the Marsh or Schnider model) to calculate the drug plasma target concentration, and outputting the anesthetic drug infusion rate adjustment amount, including the propofol infusion rate adjustment amount (ΔVp) and the remifentanil infusion rate adjustment amount (ΔVr), with a control accuracy of ±0.1 μg / kg / min. Using the fuzzy PID control algorithm, the anesthetic drug infusion rate is dynamically adjusted according to the ADAI, realizing real-time closed-loop control of the anesthesia depth. The Marsh model is used for propofol, and the Schnider model is used for remifentanil, and the target concentration is adjusted in combination with age and BMI.

[0029] The intelligent control terminal of this embodiment is composed of a command execution module, a risk warning unit and a human-computer interaction interface. The command execution module feeds back the adjustment instructions generated by the intelligent control decision module to the anesthesia infusion equipment in real time to dynamically adjust the infusion rate of anesthetic drugs and maintain a stable anesthesia depth. When the predicted intraoperative risk probability exceeds the threshold (>30%), the risk warning unit triggers an audible and visual alarm and recommends intervention measures. The human-computer interaction interface displays the ADAI trend chart, drug infusion curve and vital signs dashboard in real time, supporting the anesthesiologist to manually correct the control instructions, so that the system has a risk warning function, can predict and prompt intraoperative risks, and at the same time provide a friendly human-computer interaction interface to support the anesthesiologist's decision-making.

[0030] The intelligent anesthesia depth control system based on multimodal biosignals provided in this embodiment also includes a cloud platform docking module, which is used to upload intraoperative data to the medical cloud database in real time via the HTTPS protocol for cloud storage backup to prepare for subsequent retrospective query of historical data. It supports query based on multiple conditions such as patient ID, operation date, anesthesiologist, etc., and can generate PDF reports including anesthesia depth fluctuation charts, drug dosage statistics and risk event records.

[0031] The effectiveness of the intelligent anesthesia depth control system based on multimodal biosignals provided in this embodiment is verified by the following methods:

[0032] During the operation, BIS monitoring values ​​were recorded simultaneously, and the correlation coefficient of ADAI was calculated with BIS as the gold standard (required R 2 ≥0.85);

[0033] Comparison of anesthesia depth deviation time (T out-range ), requiring the deviation time of the intelligent control group to be reduced by ≥40%.

[0034] When the intelligent anesthesia depth control system based on multimodal biosignals is actually used, the multimodal biosignal acquisition module is first used to collect the patient's multimodal biosignals in real time, and then the signal fusion processing module is used to perform signal preprocessing and multimodal fusion on the collected multimodal biosignals. Then, based on the multimodal fusion results of the signals, the anesthesia depth intelligent analysis module is used to perform a deep intelligent evaluation of the patient's anesthesia. Then, according to the anesthesia depth evaluation results, the intelligent control decision module is used to generate the corresponding intelligent control decision. Finally, the intelligent control terminal is used to execute the generated control decision, thereby realizing intelligent anesthesia depth control based on multimodal biosignals.

[0035] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the foregoing embodiments. The foregoing embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent anesthesia depth control system based on multimodal biosignals, comprising a multimodal biosignal acquisition module, a signal fusion processing module, an anesthesia depth intelligent analysis module, an intelligent control decision module, and an intelligent control terminal, characterized by: The modal biosignal acquisition module is used to acquire the patient's electroencephalogram (EEG) signal, electrocardiogram (ECG) signal, photoplethysmography (PEP) signal, blood pressure and respiratory rate signal, and body temperature data signal in real time; The signal fusion processing module performs noise reduction, feature extraction and multimodal data fusion processing on each signal collected by the multimodal biological signal acquisition module to generate a standardized physiological feature vector of the patient; The anesthesia depth intelligent analysis module analyzes the patient's standardized physiological feature vector based on a deep learning model and outputs an anesthesia depth assessment index and a predicted intraoperative risk level; The intelligent control decision module generates an anesthetic drug infusion rate adjustment instruction based on the deviation between the anesthesia depth assessment index output by the anesthesia depth intelligent analysis module and the preset target range, combined with the patient's individual pharmacokinetic model; The intelligent control terminal feeds back the adjustment instructions generated by the intelligent control decision module to the anesthesia infusion equipment in real time, dynamically adjusts the infusion rate of the anesthetic drug, and maintains a stable anesthesia depth.

2. The intelligent anesthesia depth control system based on multimodal biosignals according to claim 1, characterized in that: The multimodal biosignal acquisition module includes a wireless bioelectrode array for acquiring the patient's EEG signals and ECG signals, a PPG sensor for synchronously acquiring the patient's blood oxygen saturation and pulse waveform, a respiratory monitoring unit for obtaining the patient's respiratory rate, and a body temperature monitoring unit for acquiring the patient's temperature data.

3. The intelligent anesthesia depth control system based on multimodal biosignals according to claim 2, characterized in that: The wireless bioelectrode array integrates dry EEG electrodes and flexible ECG electrodes and collects the patient's EEG and ECG signals in a non-invasive manner. The PPG sensor is integrated into the patient's earlobe or fingertip wearable device. The respiratory monitoring unit is an impedance respiratory sensor and calculates the respiratory rate through impedance changes. The body temperature monitoring unit uses a MEMS temperature sensor.

4. The intelligent anesthesia depth control system based on multimodal biosignals according to claim 1, characterized in that: The specific steps of signal processing in the signal fusion processing module are as follows: using wavelet transform to remove power frequency interference from EEG signals and extracting the power spectral density of α waves and β waves as features; using adaptive filtering to eliminate motion artifacts from ECG signals and extracting time domain and frequency domain indicators of heart rate variability; and fusing multimodal features through an attention mechanism to generate a patient's physiological feature vector.

5. The intelligent anesthesia depth control system based on multimodal biological signals according to claim 1, characterized in that: The deep learning model is a hybrid architecture of a dual-channel convolutional neural network and a long-short-term memory network, wherein: the dual-channel convolutional neural network branch is used to extract the spatiotemporal features of EEG signals, and the long-short-term memory network branch is used to model the temporal dependencies of ECG signals and photoplethysmography signals. The fusion layer integrates multimodal features through a weighted attention mechanism to output an anesthesia depth assessment index value and the risk probability of intraoperative hypotension and delayed awakening.

6. The intelligent anesthesia depth control system based on multimodal biosignals according to claim 1, characterized in that: The intelligent control decision module uses a fuzzy PID control algorithm to output decisions. The specific steps are: input the anesthesia depth assessment index deviation and its rate of change, dynamically adjust the proportional, integral, and differential coefficients through the fuzzy rule library, combine the patient's weight, age, and liver and kidney function data, call the pharmacokinetic model to calculate the drug plasma target concentration, and output the infusion rate adjustment amount of the anesthetic drug.

7. The intelligent anesthesia depth control system based on multimodal biosignals according to claim 1, characterized in that: The intelligent control terminal includes an instruction execution module, a risk warning unit and a human-computer interaction interface. The instruction execution module controls the anesthesia infusion equipment to execute the adjustment instructions generated by the intelligent control decision module. When the predicted intraoperative risk probability exceeds a threshold, the risk warning unit triggers an audible and visual alarm and recommends intervention measures. The human-computer interaction interface displays an anesthesia depth assessment index trend chart, a drug infusion curve and a vital signs dashboard in real time, supporting the anesthesiologist to manually modify the control instructions.

8. The intelligent anesthesia depth control system based on multimodal biological signals according to claim 1, characterized in that: It also includes a cloud platform docking module, which is used to upload intraoperative data to the medical cloud database in real time for cloud storage backup.

Images