Hepatobiliary surgery nuclear magnetism guided ablation anesthesia depth real-time regulation and control system

Abstract

The invention relates to the technical field of medical monitoring, in particular to a hepatobiliary surgery nuclear magnetism guided ablation anesthesia depth real-time regulation and control system which comprises a signal synchronous acquisition module, an interval dislocation judgment module, an energy compensation modeling module, a continuity reconstruction module and an anesthesia criterion generation module. According to the method, the electroencephalogram signal and the nuclear magnetic gradient switching time sequence are synchronously collected, the interference occurrence moment is accurately positioned, the interference section and the normal physiological signal are separated and corrected, and it is ensured that the electroencephalogram signal can accurately reflect the real anesthesia depth. And through energy compensation and time continuity restoration, the phenomenon of signal instability caused by interference is eliminated, so that the change of anesthesia depth parameters is more continuous and smoother. The real-time performance and the accuracy of anesthetic regulation and control are further improved, the dependence of artificial experience judgment is reduced, the consistency and the safety of anesthesia control are improved, and the stable anesthesia state of a patient in an ablation operation is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of medical monitoring technology, and in particular to a real-time control system for the depth of anesthesia in hepatobiliary surgery guided by MRI. Background Technology

[0002] The field of medical monitoring technology is a comprehensive technology direction that revolves around the continuous acquisition, recording, and evaluation of patients' physiological status throughout the entire diagnosis and treatment process. Core aspects include vital sign acquisition and monitoring equipment, anesthesia-related physiological parameter measurement, monitoring coordination during medical imaging examinations, and data synchronization and medical decision support during surgery. It encompasses hardware sensor placement, parameter acquisition paths, monitoring process standardization, and collaborative methods with clinical operations, and is particularly important in minimally invasive surgical treatments and image-guided procedures.

[0003] Among them, the real-time control system for anesthesia depth in traditional hepatobiliary surgery MRI-guided ablation refers to the monitoring and adjustment system used to assist in anesthesia management during MRI-guided ablation treatment in hepatobiliary surgery. This type of system focuses on the technical aspect of the patient's anesthesia depth during the ablation procedure. It uses EEG index acquisition electrodes fixed to the patient's scalp to obtain anesthesia-related electrical signals. These signals are combined with routine vital signs such as heart rate, blood pressure, and blood oxygen saturation, which are read in real time by a monitor. The anesthesiologist manually adjusts the infusion rate or dosage of anesthetic drugs based on the changes in values ​​displayed at the MRI examination bedside, thus maintaining the anesthesia state during the continuous MRI-guided ablation procedure.

[0004] Traditional MRI-guided ablation primarily relies on bedside monitors to control anesthesia depth, combined with human experience to adjust the dosage and infusion rate of anesthetic drugs. However, during MRI gradient switching, EEG signals are susceptible to electromagnetic interference, leading to signal distortion or temporal misalignment, affecting monitoring accuracy. Disruptive regions are difficult to distinguish from normal physiological changes, resulting in reduced accuracy in assessing anesthesia depth. Anesthesia control carries the risk of lag or instability, increasing the risk of excessively deep or shallow anesthesia, impacting patient safety and the continuity of the surgical procedure. Summary of the Invention

[0005] To address the challenges of traditional MRI-guided ablation surgery where anesthesia depth control relies primarily on bedside monitoring data and human experience to adjust anesthetic drug dosage and infusion rate, the present invention provides a real-time anesthesia depth control system for MRI-guided ablation surgery. This system addresses the issue that traditional MRI-guided ablation surgery relies heavily on bedside monitoring for vital sign data, combined with human experience to adjust anesthetic drug dosage and infusion rate. However, during MRI gradient switching, EEG signals are susceptible to electromagnetic interference, leading to signal distortion or temporal misalignment and affecting monitoring accuracy. Furthermore, interference zones are difficult to distinguish from normal physiological changes, resulting in reduced accuracy in anesthesia depth assessment. This also introduces the risk of delayed or unstable anesthesia control, increasing the risk of excessively deep or shallow anesthesia and impacting patient safety and the continuity of surgical procedures.

[0006] On the one hand, a real-time control system for anesthesia depth in hepatobiliary surgery guided by MRI was provided. This system includes: The signal synchronization acquisition module acquires multi-channel EEG signals through non-magnetic electrodes, obtains gradient switching timing markers through the synchronization interface, maps them to a unified time reference, and generates gradient interference time nodes. The interval misalignment determination module calls the gradient interference time node to perform time window division on the EEG signal and calculate the frequency band energy, calculates the time offset between the time window and the gradient interference time node, compares it with the interference radius threshold, and generates an interference time window identifier. The energy compensation modeling module filters the time windows corresponding to the interference time window identifiers, extracts the frequency band energy and calculates the reference value of the non-interference time window, performs energy compensation mapping based on the differences, and generates a compensated EEG energy dataset. The continuity reconstruction module calculates the time change rate of the compensated EEG energy dataset, determines whether it exceeds the physiological continuity threshold, performs continuity interpolation reconstruction on abnormal locations, and generates a time-continuous EEG energy sequence. The anesthesia criterion generation module inputs the time-continuous EEG energy sequence into the anesthesia depth mapping model, calculates the anesthesia depth state parameter sequence, compares it with the ablation anesthesia control target range, and generates anesthesia drug administration regulation criterion data.

[0007] As a further embodiment of the present invention, the gradient interference time node includes a timestamp, a channel synchronization identifier, and a time reference offset; the interference time window identifier includes a time window number, an interference intensity level, and a window validity status; the compensated EEG energy dataset includes corrected frequency band energy, a compensation coefficient, and an energy difference; the time-continuous EEG energy sequence includes an energy change rate, a smoothing repair curve, and an abnormal repair marker; and the anesthesia drug administration regulation criterion data includes anesthesia depth value, control deviation amount, and drug administration regulation direction.

[0008] As a further aspect of the present invention, the signal synchronization acquisition module includes: The signal acquisition submodule acquires multi-channel EEG signals through non-magnetic electrodes, records the continuous voltage sampling values ​​corresponding to the electrode channels, marks the sampling time according to the sampling clock, performs amplitude difference calculation on the sampling points in the same channel and verifies the consistency of the time sequence, and generates a channel amplitude sequence. The timing marking submodule, based on the channel amplitude sequence, obtains the gradient switching trigger pulse signal output by the synchronization interface, extracts the pulse rise time, performs a comparison with the preset switching rhythm based on the time interval between adjacent pulses, filters out abnormal pulses whose time interval deviates from the rhythm, and obtains the gradient switching marking sequence. The time mapping submodule, based on the gradient switching marker sequence, calls the sampling time information corresponding to the channel amplitude sequence, performs offset calculation, adjusts the sampling index position according to the offset direction and offset duration interval, marks the set of sampling time points corresponding to the switching event, and generates gradient interference time nodes.

[0009] As a further aspect of the present invention, the preset switching rhythm is based on the gradient switching trigger pulse signal output by the synchronization interface. The time interval between the rising edges of adjacent pulses is counted within multiple consecutive gradient switching cycles, and the median value in the time interval set is used as the reference rhythm value.

[0010] As a further aspect of the present invention, the interval misalignment determination module includes: The time window segmentation submodule calls the gradient interference time nodes to obtain the time axis of continuous EEG signals, performs time window segmentation according to the time node positions, records the start and end times of each time window, performs continuity verification of adjacent time windows, and generates an EEG time window sequence. The frequency band energy calculation submodule, based on the EEG time window sequence, acquires multiple EEG signal sampling points within the time window, maps the EEG signals to the frequency dimension, selects corresponding frequency components according to the target frequency band range, performs accumulation and normalization processing on the amplitude of the selected frequency components, and generates the time window frequency band energy value. The offset determination submodule, based on the energy value of the time window frequency band, calls the time information of the EEG time window sequence, calculates the time difference between the center time of the time window and the gradient interference time node, obtains the offset, compares it with the interference radius threshold execution interval and marks the status, and generates an interference time window identifier.

[0011] As a further aspect of the present invention, the energy compensation modeling module includes: The time window filtering submodule obtains the interference time window identifier, reads the corresponding time window sequence index and time boundary, performs item-by-item judgment on the time window according to the identifier status, calls the time window identifier and time boundary to verify the correspondence, performs aggregation on the indexes that meet the interference conditions, and generates an interference time window set. The reference energy calculation submodule, based on the set of interference time windows, calls the index of the time window without interference to obtain the corresponding frequency band energy data, performs sequential alignment of energy values ​​according to the frequency band identifier, and performs energy summarization and benchmark calculation according to the time window order to obtain the reference energy value of the uninterrupted time window. The compensation mapping generation submodule, based on the reference energy value of the uninterrupted time window, calls the energy data of the corresponding frequency band of the interference time window set, performs window-by-window comparison on the energy of the same frequency band, calculates the difference between the interference energy and the reference energy, performs mapping adjustment on the energy according to the difference, and generates a compensated EEG energy dataset.

[0012] As a further aspect of the present invention, the continuity reconstruction module includes: The rate of change calculation submodule acquires the compensated EEG energy dataset, reads the energy values ​​of continuous sampling points in chronological order, calculates the energy difference based on the time identifier of adjacent sampling points, synchronously acquires the time interval attribute, performs normalization processing on the energy difference and time interval, and generates an EEG energy time rate of change sequence. The continuity discrimination submodule, based on the EEG energy time change rate sequence, calls the physiological continuity threshold, compares the relationship between the change rate and the upper and lower bounds of the threshold interval point by point, records the time index corresponding to the threshold interval, performs order verification and duplicate item removal on the index, and obtains the continuity abnormality location index set. The interpolation and reconstruction submodule, based on the continuous abnormal location index set, calls the complete time axis of the compensated EEG energy dataset, extracts the adjacent effective energy values ​​and corresponding time intervals before and after the abnormal index, performs linear interpolation according to the time interval ratio, and writes them into the abnormal index position to generate a time-continuous EEG energy sequence.

[0013] As a further aspect of the present invention, the physiological continuity threshold is jointly defined by the lower bound and the upper bound of the rate of change. The lower bound and the upper bound of the rate of change are determined statistically based on the EEG energy time change rate sequence within a preset time window in the compensated EEG energy data set. The lower bound of the rate of change is the value corresponding to the 5% position in the absolute value distribution of the EEG energy time change rate sequence, and the upper bound of the rate of change is the value corresponding to the 95% position in the absolute value distribution of the EEG energy time change rate sequence.

[0014] As a further aspect of the present invention, the anesthesia criterion generation module includes: The data sequence receiving submodule acquires the time-continuous EEG energy sequence, reads energy samples in the order of sampling time markers and judges the continuity of adjacent time markers, marks abnormal intervals and performs sample position rearrangement to generate a standard EEG energy input sequence. The state parameter calculation submodule, based on the standard input sequence of EEG energy, performs mapping operations point by point according to the relationship between energy amplitude distribution and time evolution, calculates the corresponding state parameters for each sampling point and keeps the time index consistent, integrates the parameter values ​​in time order, and obtains the anesthesia depth state parameter sequence. The criterion generation submodule obtains the boundary parameters of the ablation anesthesia control target interval based on the anesthesia depth state parameter sequence, compares the relationship between the state parameters and the upper and lower limits of the interval under multiple time indices, summarizes the location identifiers of the landing area and organizes them in a structured manner according to time order, and generates anesthesia drug administration control criterion data.

[0015] As a further aspect of the present invention, the step of performing mapping operation point by point based on the relationship between energy amplitude distribution and time evolution refers to jointly mapping the energy amplitude of each sampling point with the energy change trend of its adjacent sampling points in the standard input sequence of EEG energy to generate state parameters that correspond one-to-one with the sampling time identifier. The comparison of the relationship between the state parameters under multiple time indices and the upper and lower limits of the interval refers to comparing the state parameters corresponding to multiple consecutive sampling time markers in the anesthesia depth state parameter sequence with the upper and lower limits of the boundary parameters of the ablation anesthesia control target interval, and recording the location markers of the areas falling within the interval, exceeding the upper limit, and falling below the lower limit in the order of the time index.

[0016] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By simultaneously acquiring EEG signals and switching the gradient of MRI, the timing of interference is precisely located. The interfering segment is separated from and corrected to ensure that the EEG signal accurately reflects the true depth of anesthesia. Through energy compensation and temporal continuity restoration, signal instability caused by interference is eliminated, making changes in anesthesia depth parameters more continuous and smooth. This further improves the real-time performance and precision of anesthetic drug control, reduces reliance on human experience, enhances the consistency and safety of anesthesia control, and ensures a stable anesthetic state for patients during ablation procedures. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the accompanying drawings without creative effort.

[0018] Figure 1 This is a system schematic diagram of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the signal synchronization acquisition module in this invention; Figure 4 This is a flowchart of the interval misalignment determination module in this invention; Figure 5 This is a flowchart of the energy compensation modeling module in this invention; Figure 6 This is a flowchart of the continuous reconstruction module in this invention; Figure 7 This is a flowchart of the anesthesia criterion generation module in this invention. Detailed Implementation

[0019] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0020] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0021] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0022] In this embodiment of the invention, sometimes the subscript such as W1 is written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0023] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0024] This invention provides a real-time anesthesia depth control system for magnetic resonance-guided ablation in hepatobiliary surgery, such as... Figure 1-2 The diagram shown illustrates a real-time anesthesia depth control system for hepatobiliary surgery guided by magnetic resonance imaging (MRI). This system includes: The signal synchronization acquisition module acquires multi-channel EEG signals through non-magnetic electrodes, obtains gradient switching timing markers through the synchronization interface, maps them to a unified time reference, and generates gradient interference time nodes. The interval misalignment determination module calls the gradient interference time node, performs time window division on the EEG signal and calculates the frequency band energy, calculates the time offset between the time window and the gradient interference time node, compares it with the interference radius threshold, and generates an interference time window identifier. The energy compensation modeling module filters the time windows corresponding to the interference time windows, extracts the frequency band energy and calculates the reference value of the uninterrupted time window, performs energy compensation mapping based on the differences, and generates a compensated EEG energy dataset. The continuity reconstruction module calculates the time change rate of the compensated EEG energy dataset, determines whether it exceeds the physiological continuity threshold, performs continuity interpolation reconstruction on abnormal locations, and generates a time-continuous EEG energy sequence. The anesthesia criterion generation module inputs the time-continuous EEG energy sequence into the anesthesia depth mapping model, calculates the anesthesia depth state parameter sequence, compares it with the target range for ablation anesthesia control, and generates anesthesia drug administration regulation criterion data.

[0025] The gradient interference time nodes include timestamps, channel synchronization identifiers, and time reference offsets; the interference time window identifiers include time window numbers, interference intensity levels, and window validity status; the compensated EEG energy dataset includes corrected frequency band energy, compensation coefficients, and energy differences; the time-continuous EEG energy sequence includes energy change rate, smoothing repair curves, and abnormal repair markers; and the anesthesia drug administration regulation criterion data includes anesthesia depth values, control deviations, and drug administration regulation directions.

[0026] Specifically, such as Figure 2 , 3 As shown, the signal synchronization acquisition module includes: The signal acquisition submodule acquires multi-channel EEG signals through non-magnetic electrodes, records the continuous voltage sampling values ​​corresponding to the electrode channels, marks the sampling time according to the sampling clock, performs amplitude difference calculation on the sampling points in the same channel and verifies the consistency of the time sequence, and generates a channel amplitude sequence. The signal acquisition submodule is equipped with a high-sensitivity non-magnetic carbon fiber composite electrode array, specifically designed for a 3.0 Tesla MRI environment. This array is used to resist strong magnetic field interference and capture weak electroencephalographic signals during hepatobiliary ablation procedures. Internally, it integrates a multi-channel 24-bit high-precision analog-to-digital converter, with a sampling frequency of 2048 Hz, meaning it performs 2048 voltage value reads per second for each EEG channel. During amplitude difference calculation, it first retrieves two adjacent sampling points within the same channel. For example, if the voltage value at the current sampling point is 150 microvolts and the voltage value at the previous sampling point is 148 microvolts, the arithmetic unit performs a subtraction operation to obtain the difference of 2 microvolts. This difference is used to eliminate common-mode interference and highlight the dynamic changes in the EEG signal. Subsequently, through a built-in high-precision clock synchronization unit, each sampling point is assigned a timestamp accurate to the microsecond level, for example, marking the first sampling point as 10 minutes 00 seconds 000 milliseconds, ensuring that subsequent processing has a strict timing reference. Next, a time sequence consistency check is performed. By comparing the clock tags of consecutive sampling points, it is ensured that the time tags are strictly monotonically increasing. Once an out-of-order condition is detected (e.g., the timestamp of a later sampling point is less than that of a previous sampling point), an anomaly marker is immediately triggered and the bad data is discarded. Finally, the voltage data, after differential processing and time sequence verification, is arranged in an orderly manner to form a standardized channel amplitude sequence. This sequence intuitively reflects the continuous voltage fluctuation of the EEG signal over time, providing the original data foundation for the subsequent identification of gradient interference.

[0027] The timing marking submodule, based on the channel amplitude sequence, obtains the gradient switching trigger pulse signal output by the synchronization interface, extracts the pulse rise time, compares the time interval between adjacent pulses with the preset switching rhythm, filters out abnormal pulses whose time interval deviates from the rhythm, and obtains the gradient switching marking sequence. The timing marker submodule is directly connected to the gradient control system of the MRI scanner via an opto-isolated synchronization interface, receiving the synchronization trigger pulse signal generated during gradient coil switching in real time. The module is internally equipped with a pulse edge detection unit, with a voltage rising edge detection threshold set to 3.3 volts. When the input signal voltage jumps from 0 volts to above 3.3 volts within 1 microsecond, the detection unit identifies the pulse rising edge and records the precise time point of its occurrence. To determine the preset switching rhythm, after continuously monitoring 100 gradient switching cycles, the module calculates the time difference between two adjacent pulse rising edges. For example, it obtains a set of 100 time intervals containing values ​​such as 50 milliseconds, 50 milliseconds, 52 milliseconds, and 49 milliseconds. The module calls a median statistical algorithm to sort the values ​​in this set and selects the value in the middle, 50 milliseconds, as the baseline rhythm value. Subsequently, the module performs an abnormal pulse screening operation, setting a tolerance range of ±5% of the baseline rhythm value, i.e., the interval between 47.5 milliseconds and 52.5 milliseconds. Pulse signals whose calculated time intervals fall outside this interval are considered random interference or artifacts and are not marked. The pulse time points after cleaning and screening are confirmed as valid gradient switching events. These time points are arranged chronologically to construct a high-confidence gradient switching marker sequence. This sequence accurately depicts the rhythm of gradient field changes during MRI scanning, providing precise temporal targets for subsequent removal of gradient artifacts from EEG signals.

[0028] The time mapping submodule, based on the gradient switching marker sequence, calls the sampling time information corresponding to the channel amplitude sequence, performs offset calculation, adjusts the sampling index position according to the offset direction and offset duration interval, marks the set of sampling time points corresponding to the switching event, and generates gradient interference time nodes. The time mapping submodule is primarily responsible for resolving minor asynchrony issues between the clocks of the EEG signal acquisition system and the MRI system. The module first reads a specific gradient event time point from the gradient switching marker sequence, for example, 10 minutes 05 seconds 200 milliseconds, and simultaneously retrieves the corresponding sampling point time information from the channel amplitude sequence. It then performs offset calculations, setting a fixed delay parameter of 15 milliseconds to account for signal transmission delay and hardware response time. Adding this delay parameter to the gradient event time point, it calculates the start time of gradient interference in the EEG signal as 10 minutes 05 seconds 215 milliseconds. Based on the pre-determined characteristics of gradient artifact persistence, the offset duration interval is determined to be 30 milliseconds, meaning the interference will last until 10 minutes 05 seconds 245 milliseconds. Using the calculated start and end times, it searches the index table of the channel amplitude sequence to extract all sampling point indices whose time tags fall within the closed interval of 10 minutes 05 seconds 215 milliseconds to 10 minutes 05 seconds 245 milliseconds. The selected set of sampling point indices represents the EEG data segments directly affected by gradient field switching. The set of time points is defined as the gradient interference time nodes and used as key reference coordinates for subsequent signal segmentation and energy correction, ensuring that the accuracy of interference localization reaches the millisecond level.

[0029] Specifically, such as Figure 2 , 4 As shown, the interval misalignment determination module includes: The time window division sub-module calls the gradient interference time node to obtain the time axis of continuous EEG signals, performs time window segmentation according to the time node position, records the start and end time of each time window, performs continuity verification of adjacent time windows, and generates EEG time window sequence. The time window segmentation submodule performs fine segmentation on the time axis of continuous EEG signals based on gradient interference time nodes. A standard time window length of 200 milliseconds is set, but the module dynamically adjusts the segmentation strategy when encountering gradient interference time nodes. Specifically, it first identifies the time interval where the interference node is located, using the moment before the interval as the end time of the previous effective time window and the moment after the interval as the start time of the next effective time window, thus physically separating the interfered data segment from the clean EEG data segment. For example, if the interference node is located between 10:05.215 and 10:05.245, and the original time window should cover 10:05.200 to 10:05.400, the window will be truncated, generating a short window of 10:05.200 to 10:05.215, and a new window starting from 10:05.245. The absolute start and end times of each segmented time window are recorded, and the continuity of adjacent time windows is checked. It is determined whether there are unmarked gaps between adjacent time windows; if a gap larger than the sampling interval exists, it is automatically filled with a marker or the windows are merged. After this process, a set of EEG time window sequences with contiguous beginnings and endings or clearly marked intervals is generated. This sequence provides a structured data carrier for subsequent segmented frequency domain analysis.

[0030] The frequency band energy calculation submodule, based on the EEG time window sequence, acquires EEG signal sampling points within multiple time windows, maps the EEG signals to the frequency dimension, selects corresponding frequency components according to the target frequency band range, performs accumulation and normalization processing on the amplitude of the selected frequency components, and generates time window frequency band energy values. The frequency band energy calculation submodule performs spectral analysis for each independent EEG time window. First, it acquires the voltage values ​​of, for example, 400 EEG signal sampling points within the time window. A Fast Fourier Transform (FFT) algorithm is used to map the time-domain voltage signal to the frequency dimension, generating corresponding spectral data. Based on the requirements for anesthesia depth monitoring, the module identifies the target frequency band as the Alpha wave band (8 Hz to 13 Hz) and the Beta wave band (13 Hz to 30 Hz). Frequency components within these ranges are selected from the spectral data, and their corresponding amplitude data is extracted. Subsequently, a sum-of-squares operation is performed on the amplitudes of the selected frequency components, i.e., the sum of the squares of the amplitudes at each frequency point is calculated to characterize the total power of that frequency band. To eliminate energy magnitude differences caused by different time window lengths, a normalization process is performed, dividing the cumulative power value by the number of sampling points in the time window to obtain the average energy density per sampling point. For example, if the cumulative power in the Alpha band for a certain time window is 5000 microvolts squared and contains 400 sampling points, the normalized time window frequency band energy value is 12.5. This value objectively reflects the activity level of a specific EEG rhythm within that time segment, and the generated time window frequency band energy value sequence is directly used for subsequent interference determination and state assessment.

[0031] Table 1: Energy Monitoring Table for Time Window Bands in EEG Time window number Start time (seconds) Number of sampling points Alpha band cumulative power Beta band cumulative power Alpha normalized energy value Beta normalized energy value 001 10.000 400 5200 2100 13.00 5.25 002 10.200 380 4940 2000 13.00 5.26 003 10.400 410 18500 6500 45.12 15.85 As shown in Table 1, this table presents examples of energy calculation results for different time windows in a specific frequency band. The energy value of time window number 003 is significantly higher than that of the previous two windows, indicating that there is an anomaly or interference, which needs to be further determined by subsequent modules.

[0032] The offset determination submodule calls the EEG time window sequence time information based on the energy value of the time window frequency band, calculates the time difference between the center time of the time window and the gradient interference time node, obtains the offset, compares it with the interference radius threshold execution interval and marks the status, and generates an interference time window identifier. The offset determination submodule aims to quantify the potential degree of gradient interference in each EEG time window. First, it retrieves the time information of the EEG time window sequence and calculates the center time of each window. For example, for a time window starting at 10:00:000 and ending at 10:00:200, its center time is 10:00:100. Simultaneously, it retrieves the gradient interference time node data and calculates the absolute value of the time difference between this center time and the center position of the most recent gradient interference node; this difference is the offset. An interference radius threshold of 100 milliseconds is set, based on the decay time of the gradient coil afterglow effect. The calculated offset is compared with the 100-millisecond threshold: if the offset is less than or equal to 100 milliseconds, the time window is determined to be in an "interferenced" state, marked with a status code of 1, indicating that the window data contains residual gradient artifacts; if the offset is greater than 100 milliseconds, the time window is determined to be in a "clean" state, marked with a status code of 0, indicating that the window data mainly consists of physiological EEG data. This status code is appended to the metadata of each time window to generate an interference time window identifier, which clearly distinguishes which data segments need to be corrected and which can be used as a reference benchmark.

[0033] Specifically, such as Figure 2 , 5 As shown, the energy compensation modeling module includes: The time window filtering submodule obtains the interference time window identifier, reads the corresponding time window sequence index and time boundary, performs item-by-item judgment on the time window according to the identifier status, calls the time window identifier and time boundary to verify the correspondence, performs aggregation on the index that meets the interference conditions, and generates a set of interference time windows. The time window filtering submodule splits the data stream based on the interference time window identifiers. It iterates through the identifier states of the time windows; when a status code of 1 is read, it is confirmed as a window that meets the interference conditions. The sequence index number and its precise time boundaries (start and end times) of that window are immediately read. A collection operation is performed on the indexes that meet the conditions, constructing a list containing the indices of the interfered windows, i.e., the interference time window set. Simultaneously, a correspondence verification is performed to ensure that each collected index can be correctly mapped back to the original energy data, preventing erroneous corrections due to data alignment errors. For example, if the window status codes corresponding to indices 5, 6, and 7 are all 1, these three indices are stored in the set, meaning that subsequent compensation algorithms will specifically process the data from these three time periods. The role of this module is to accurately locate the lesions, physically isolating the data segments that need "treatment" from the healthy baseline data segments, preparing for subsequent targeted compensation.

[0034] The reference energy calculation submodule, based on the set of interference time windows, calls the index of the time window without interference to obtain the energy data of the corresponding frequency band, performs sequential alignment of energy values ​​according to the frequency band identifier, and performs energy summarization and benchmark calculation according to the time window order to obtain the reference energy value of the uninterrupted time window. The reference energy calculation submodule is responsible for constructing an energy baseline for repairing damaged data. First, it calls the index of time windows not marked as interference (i.e., windows with a status code of 0) to extract energy data for the corresponding frequency band (e.g., alpha waves) from clean windows. Considering the non-stationarity of EEG signals, the module uses a locally weighted average method. For each interfered time window, it selects the five nearest adjacent uninterrupted time windows as reference samples. Based on the frequency band identifier, the module aligns the alpha band energy values ​​of these five samples sequentially and performs summarization and baseline calculations. The calculation logic is as follows: remove the maximum and minimum values ​​from the five samples, and calculate the arithmetic mean of the energy values ​​of the remaining three samples. This average is established as the reference energy value for the uninterrupted time window. For example, if the energy values ​​of the five adjacent clean windows are 12, 13, 12, 45 (abnormally high), and 11, the module removes 45 and 11, calculates (12+13+12) / 3, and obtains the reference value 12.33. This reference value represents the theoretical level of brain electrical energy that should exist during this time period under the current physiological state, without gradient interference.

[0035] The compensation mapping generation submodule calls the frequency band energy data corresponding to the interference time window set based on the reference energy value of the uninterrupted time window, performs window-by-window comparison of the energy in the same frequency band, calculates the difference between the interference energy and the reference energy, performs mapping adjustment on the energy according to the difference, and generates a compensated EEG energy dataset. The compensation mapping generation submodule uses a reference energy value to repair the disturbed data. It calls the measured energy data of a specific window in the set of disturbance time windows, such as time window number 003 in Table 1, where the measured Alpha energy is as high as 45.12. Simultaneously, it calls the reference energy value of 12.33 calculated for this window. The module performs a difference calculation, subtracting the reference energy from the measured energy to obtain a difference of 32.79, which is considered as the superimposed gradient noise energy. Based on the difference, it performs energy mapping adjustment. Considering the nonlinearity of noise superposition, the module uses the attenuation coefficient method, setting the attenuation coefficient to 0.95 (this coefficient was measured through phantom experiments). It subtracts (difference × attenuation coefficient) from the measured energy, i.e., 45.12 - (32.79 × 0.95) = 13.97. This calculation process pulls the abnormally high disturbed energy back to a range consistent with normal physiological conditions; the generated 13.97 is the corrected energy value. This process was performed on each of the disturbed windows in the dataset, ultimately generating a complete dataset of compensated EEG energy. This dataset eliminated the specific interference from the MRI environment and restored the true characteristics of EEG energy.

[0036] Specifically, such as Figure 2 , 6 As shown, the continuity reconstruction module includes: The rate of change calculation submodule acquires the compensated EEG energy dataset, reads the energy values ​​of continuous sampling points in chronological order, calculates the energy difference based on the time identifier of adjacent sampling points, synchronously acquires the time interval attribute, performs normalization processing on the energy difference and time interval, and generates the EEG energy time rate of change sequence. The rate of change calculation submodule is used to capture the dynamic fluctuation trend of EEG energy over time. It reads the energy values ​​of consecutive sampling points in the compensated EEG energy dataset in chronological order, and sets the energy of the current sampling point as... The energy of the previous sampling point is First, calculate the energy difference, that is, use... minus At the same time, the time interval between these two sampling points is obtained. (This is the length of the time window, such as 0.2 seconds). The processing unit performs normalization, dividing the energy difference by the time interval, which is the calculation. The rate of energy change per unit time is obtained. For example, if the current energy is 14, the previous energy was 13, and the time interval is 0.2 seconds, then the rate of change is (14-13) / 0.2 = 5 units / second. This sequence reflects the rate of change in anesthesia depth; extremely high rates of change in mutations indicate artifact residue or physiological mutations. The calculated rate of change values ​​are arranged sequentially to generate a sequence of EEG energy temporal change rates, providing a quantitative indicator for subsequent identification of signal smoothness and continuity abnormalities.

[0037] The continuity discrimination submodule, based on the EEG energy time change rate sequence, calls the physiological continuity threshold, compares the relationship between the change rate and the upper and lower bounds of the threshold interval point by point, records the time index corresponding to the threshold interval, performs order verification and duplicate item removal on the index, and obtains the continuity abnormality location index set. The continuity discrimination submodule defines the reasonable fluctuation range of physiological signals using statistical methods. First, it takes the absolute value of the EEG energy temporal change rate sequence within a preset time window (e.g., the past minute) and sorts the entire sequence from smallest to largest. Based on statistical distribution patterns, the module determines the lower and upper bounds of the change rate. The lower bound is selected from the value at the 5th percentile of the sorted sequence, and the upper bound is selected from the value at the 95th percentile. For example, if the sorted change rate absolute value sequence contains 100 data points, with the 5th data point having a value of 0.1 and the 95th data point having a value of 20, then the physiological continuity threshold interval is limited to [0.1, 20]. Subsequently, the current change rate data is scanned point by point. If the absolute value of the change rate at a certain point exceeds 20 or is lower than 0.1, it is determined that this point violates the physiological continuity law and is considered an abnormal mutation. The time indices exceeding the threshold interval are recorded, and a sequence check is performed to remove duplicate indices, ultimately generating a continuous abnormality location index set. This step ensures that even after energy compensation, residual minute jumps can be identified, preventing them from misleading the calculation of anesthesia depth.

[0038] The interpolation and reconstruction submodule, based on the continuous abnormal location index set, calls the complete time axis of the compensated EEG energy dataset, extracts the adjacent effective energy values ​​and corresponding time intervals before and after the abnormal index, performs linear interpolation according to the time interval ratio, and writes them into the abnormal index position to generate a time-continuous EEG energy sequence. The interpolation and reconstruction submodule is responsible for smoothing the identified continuous anomalies. Based on the continuous anomaly location index set, the module locates the time position of a specific anomaly; for example, energy point index 50 is identified as an anomaly. It calls the compensated EEG energy dataset to extract the energy values ​​of the preceding valid point (index 49) and the following valid point (index 51), along with their corresponding time markers. Linear interpolation is then performed based on the time interval ratio. For example, if index 49 has an energy of 10 and a time of 10 seconds, index 51 has an energy of 12 and a time of 10.4 seconds, and the anomaly index 50 has a time of 10.2 seconds, since 10.2 seconds is exactly in the middle, the module calculates (10+12) / 2=11 and writes 11 as the new energy value for index 50, replacing the original anomaly value. For cases with multiple consecutive anomalies, the module uses piecewise linear interpolation logic to ensure a smooth transition of the reconstructed data on the time axis. After scanning and reconstructing the entire sequence, the module generates a time-continuous EEG energy sequence, eliminating abrupt transition points and providing high-quality, smooth input data.

[0039] Specifically, such as Figure 2 , 7 As shown, the anesthesia criterion generation module includes: The data sequence receiving submodule acquires the time-continuous EEG energy sequence, reads energy samples in the order of sampling time markers and judges the continuity of adjacent time markers, marks abnormal intervals and performs sample position rearrangement to generate the standard input sequence of EEG energy. The data sequence receiving submodule is the final checkpoint before entering state computation. It acquires a continuous EEG energy sequence, reading energy samples one by one according to the sampling time markers. It has built-in interval detection logic, setting the standard time step to 200 milliseconds. It checks whether the difference between two adjacent time markers is strictly equal to the standard time step. If an interval of 400 milliseconds is found, it indicates a missing sample; if the interval is 100 milliseconds, it indicates overlapping samples. For missing samples, it performs filler with the previous values; for overlapping samples, it performs mean merging. Simultaneously, the module rearranges the sample positions to ensure data contiguous addressing in memory. After this series of integrity checks and rectifications, the module outputs a standardized EEG energy input sequence with a uniform format and strictly equidistant timing. This sequence fully meets the input format requirements of subsequent complex mapping operations, preventing computational crashes caused by data structure errors.

[0040] The state parameter calculation submodule, based on the standard input sequence of EEG energy, performs mapping operations point by point according to the relationship between energy amplitude distribution and time evolution, calculates the corresponding state parameters for each sampling point and keeps the time index consistent, integrates the parameter values ​​in time order, and obtains the anesthesia depth state parameter sequence. The core function of the state parameter calculation submodule is to map physical-level EEG energy into clinically readable anesthesia depth state parameters. Internally, this module constructs a mapping computation network based on the principle of a multilayer perceptron (MLP). This network structure has been defined in a white-box manner: the input layer contains 5 neurons, receiving the current alpha wave energy, beta wave energy, alpha / beta energy ratio, and the raw energy values ​​from the previous two time steps; the hidden layer consists of two layers: the first hidden layer contains 10 neurons, using ReLU (Rectified Linear Unit) as the activation function to extract the nonlinear combination relationship of energy features; the second hidden layer contains 5 neurons, also using the ReLU activation function for feature compression; the output layer contains 1 neuron, using the sigmoid activation function to restrict the output to between 0 and 1, and then multiplying it by 100 to match the 0-100 anesthesia depth index scale. Data from the standard EEG energy input sequence are input into this network point by point. For example, when inputting a set of data such as an Alpha energy of 12, a Beta energy of 5, and a ratio of 2.4, the neurons in the first hidden layer perform weighted summation (weight parameters established through offline clinical big data training) and ReLU activation, outputting an intermediate feature vector. After the signal flows through the second hidden layer, the final output layer calculates a value of 0.55, which, after scaling, outputs a state parameter of 55. This process is repeated for each sampling point, keeping the time index consistent with the input sequence, and the calculated parameter values ​​are integrated in chronological order to obtain the anesthesia depth state parameter sequence. The lower the value in this sequence, the deeper the anesthesia; the higher the value, the more awake the patient. For example, 55 corresponds to an appropriate surgical anesthesia depth.

[0041] The criteria generation submodule obtains the boundary parameters of the ablation anesthesia control target interval based on the anesthesia depth state parameter sequence, compares the relationship between the state parameters and the upper and lower limits of the interval under multiple time indices, summarizes the location identifiers of the landing area and organizes them in a structured manner according to time order, and generates anesthesia drug administration control criteria data. The criterion generation submodule transforms continuous anesthesia depth state parameters into specific drug administration control instructions. First, it acquires the boundary parameters of the target interval for ablation anesthesia control, setting the lower limit of the interval at 40 and the upper limit at 60. This interval represents the ideal range for maintaining the patient's unconsciousness and stable vital signs. It then compares the values ​​in the anesthesia depth state parameter sequence with this interval. If the state parameter is 65 at a certain moment, higher than the upper limit of 60, the module determines that the patient is at risk of awakening and generates a "deepen anesthesia" location marker; if the parameter is 35, lower than the lower limit of 40, the module determines that the anesthesia is too deep and generates a "lighten anesthesia" marker; if the parameter is 50, within the interval, a "maintain" marker is generated. These markers are then structured and organized in chronological order to generate anesthesia drug administration control criterion data.

[0042] Table 2: Data Table of Criteria for Anesthetic Drug Administration Regulation Time index (minutes:seconds) State parameter values Target range Landing location marker Regulation Recommendations 15:00 55 [40，60] Within the range Maintain current pumping rate 15:05 62 [40，60] Above the upper limit Increase propofol infusion rate 15:10 38 [40，60] Below the lower limit Reduce propofol infusion rate As shown in Table 2, the data clearly indicates the control strategy at different time points. This criterion data is sent to the anesthesia infusion pump workstation in real time through a standard communication protocol, realizing full-process automation from signal acquisition to closed-loop control. This ensures that the patient's anesthesia depth is always precisely locked within a safe range in the complex environment of MRI-guided ablation surgery.

[0043] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the described technical solutions.

Claims

1. A real-time anesthesia depth control system for magnetic resonance-guided ablation in hepatobiliary surgery, characterized in that, The system includes: The signal synchronization acquisition module acquires multi-channel EEG signals through non-magnetic electrodes, obtains gradient switching timing markers through the synchronization interface, maps them to a unified time reference, and generates gradient interference time nodes. The interval misalignment determination module calls the gradient interference time node to perform time window division on the EEG signal and calculate the frequency band energy, calculates the time offset between the time window and the gradient interference time node, compares it with the interference radius threshold, and generates an interference time window identifier. The energy compensation modeling module filters the time windows corresponding to the interference time window identifiers, extracts the frequency band energy and calculates the reference value of the non-interference time window, performs energy compensation mapping based on the differences, and generates a compensated EEG energy dataset. The continuity reconstruction module calculates the time change rate of the compensated EEG energy dataset, determines whether it exceeds the physiological continuity threshold, performs continuity interpolation reconstruction on abnormal locations, and generates a time-continuous EEG energy sequence. The anesthesia criterion generation module inputs the time-continuous EEG energy sequence into the anesthesia depth mapping model, calculates the anesthesia depth state parameter sequence, compares it with the ablation anesthesia control target range, and generates anesthesia drug administration regulation criterion data.

2. The real-time anesthesia depth control system for hepatobiliary surgery guided by MRI as described in claim 1, characterized in that, The gradient interference time node includes a timestamp, channel synchronization identifier, and time reference offset; the interference time window identifier includes a time window number, interference intensity level, and window validity status; the compensated EEG energy dataset includes corrected frequency band energy, compensation coefficient, and energy difference; the time-continuous EEG energy sequence includes energy change rate, smoothing repair curve, and abnormal repair marker; and the anesthesia drug administration regulation criterion data includes anesthesia depth value, control deviation amount, and drug administration regulation direction.

3. The real-time anesthesia depth control system for hepatobiliary surgery guided by MRI as described in claim 1, characterized in that, The signal synchronization acquisition module includes: The signal acquisition submodule acquires multi-channel EEG signals through non-magnetic electrodes, records the continuous voltage sampling values ​​corresponding to the electrode channels, marks the sampling time according to the sampling clock, performs amplitude difference calculation on the sampling points in the same channel and verifies the consistency of the time sequence, and generates a channel amplitude sequence. The timing marking submodule, based on the channel amplitude sequence, obtains the gradient switching trigger pulse signal output by the synchronization interface, extracts the pulse rise time, performs a comparison with the preset switching rhythm based on the time interval between adjacent pulses, filters out abnormal pulses whose time interval deviates from the rhythm, and obtains the gradient switching marking sequence. The time mapping submodule, based on the gradient switching marker sequence, calls the sampling time information corresponding to the channel amplitude sequence, performs offset calculation, adjusts the sampling index position according to the offset direction and offset duration interval, marks the set of sampling time points corresponding to the switching event, and generates gradient interference time nodes.

4. The real-time anesthesia depth control system for hepatobiliary surgery guided by MRI as described in claim 3, characterized in that, The preset switching rhythm is based on the gradient switching trigger pulse signal output by the synchronization interface. The time interval between the rising edges of adjacent pulses is counted within multiple consecutive gradient switching cycles, and the median value in the time interval set is used as the reference rhythm value.

5. The real-time anesthesia depth control system for hepatobiliary surgery guided by MRI as described in claim 1, characterized in that, The interval misalignment determination module includes: The time window segmentation submodule calls the gradient interference time nodes to obtain the time axis of continuous EEG signals, performs time window segmentation according to the time node positions, records the start and end times of each time window, performs continuity verification of adjacent time windows, and generates an EEG time window sequence. The frequency band energy calculation submodule, based on the EEG time window sequence, acquires multiple EEG signal sampling points within the time window, maps the EEG signals to the frequency dimension, selects corresponding frequency components according to the target frequency band range, performs accumulation and normalization processing on the amplitude of the selected frequency components, and generates the time window frequency band energy value. The offset determination submodule, based on the energy value of the time window frequency band, calls the time information of the EEG time window sequence, calculates the time difference between the center time of the time window and the gradient interference time node, obtains the offset, compares it with the interference radius threshold execution interval and marks the status, and generates an interference time window identifier.

6. The real-time anesthesia depth control system for hepatobiliary surgery guided by MRI as described in claim 1, characterized in that, The energy compensation modeling module includes: The time window filtering submodule obtains the interference time window identifier, reads the corresponding time window sequence index and time boundary, performs item-by-item judgment on the time window according to the identifier status, calls the time window identifier and time boundary to verify the correspondence, performs aggregation on the indexes that meet the interference conditions, and generates an interference time window set. The reference energy calculation submodule, based on the set of interference time windows, calls the index of the time window without interference to obtain the corresponding frequency band energy data, performs sequential alignment of energy values ​​according to the frequency band identifier, and performs energy summarization and benchmark calculation according to the time window order to obtain the reference energy value of the uninterrupted time window. The compensation mapping generation submodule, based on the reference energy value of the uninterrupted time window, calls the energy data of the corresponding frequency band of the interference time window set, performs window-by-window comparison on the energy of the same frequency band, calculates the difference between the interference energy and the reference energy, performs mapping adjustment on the energy according to the difference, and generates a compensated EEG energy dataset.

7. The real-time anesthesia depth control system for hepatobiliary surgery guided by MRI as described in claim 1, characterized in that, The continuity reconstruction module includes: The rate of change calculation submodule acquires the compensated EEG energy dataset, reads the energy values ​​of continuous sampling points in chronological order, calculates the energy difference based on the time identifier of adjacent sampling points, synchronously acquires the time interval attribute, performs normalization processing on the energy difference and time interval, and generates an EEG energy time rate of change sequence. The continuity discrimination submodule, based on the EEG energy time change rate sequence, calls the physiological continuity threshold, compares the relationship between the change rate and the upper and lower bounds of the threshold interval point by point, records the time index corresponding to the threshold interval, performs order verification and duplicate item removal on the index, and obtains the continuity abnormality location index set. The interpolation and reconstruction submodule, based on the continuous abnormal location index set, calls the complete time axis of the compensated EEG energy dataset, extracts the adjacent effective energy values ​​and corresponding time intervals before and after the abnormal index, performs linear interpolation according to the time interval ratio, and writes them into the abnormal index position to generate a time-continuous EEG energy sequence.

8. The real-time anesthesia depth control system for hepatobiliary surgery guided by magnetic resonance imaging (MRI) ablation according to claim 7, characterized in that, The physiological continuity threshold is defined by a lower bound and an upper bound of the rate of change. The lower bound and the upper bound of the rate of change are determined by statistical analysis of the brain energy time change rate sequence within a preset time window in the compensated brain energy energy dataset. The lower bound of the rate of change is the value corresponding to the 5% position in the absolute value distribution of the brain energy time change rate sequence, and the upper bound of the rate of change is the value corresponding to the 95% position in the absolute value distribution of the brain energy time change rate sequence.

9. The real-time anesthesia depth control system for hepatobiliary surgery guided by MRI as described in claim 1, characterized in that, The anesthesia criterion generation module includes: The data sequence receiving submodule acquires the time-continuous EEG energy sequence, reads energy samples in the order of sampling time markers and judges the continuity of adjacent time markers, marks abnormal intervals and performs sample position rearrangement to generate a standard EEG energy input sequence. The state parameter calculation submodule, based on the standard input sequence of EEG energy, performs mapping operations point by point according to the relationship between energy amplitude distribution and time evolution, calculates the corresponding state parameters for each sampling point and keeps the time index consistent, integrates the parameter values ​​in time order, and obtains the anesthesia depth state parameter sequence. The criterion generation submodule obtains the boundary parameters of the ablation anesthesia control target interval based on the anesthesia depth state parameter sequence, compares the relationship between the state parameters and the upper and lower limits of the interval under multiple time indices, summarizes the location identifiers of the landing area and organizes them in a structured manner according to time order, and generates anesthesia drug administration control criterion data.

10. The real-time anesthesia depth control system for hepatobiliary surgery guided by magnetic resonance imaging (MRI) ablation according to claim 9, characterized in that, The point-by-point mapping operation based on the relationship between energy amplitude distribution and time evolution refers to jointly mapping the energy amplitude of each sampling point with the energy change trend of adjacent sampling points in the standard EEG energy input sequence to generate state parameters that correspond one-to-one with the sampling time identifier. The comparison of the relationship between the state parameters under multiple time indices and the upper and lower limits of the interval refers to comparing the state parameters corresponding to multiple consecutive sampling time markers in the anesthesia depth state parameter sequence with the upper and lower limits of the boundary parameters of the ablation anesthesia control target interval, and recording the location markers of the areas falling within the interval, exceeding the upper limit, and falling below the lower limit in the order of the time index.

Images