An intelligent monitoring system for symptoms of palliative care based on electroencephalogram signal analysis

Abstract

The present application relates to the technical field of electroencephalogram signal analysis, and particularly relates to a kind of intelligent monitoring system of symptom of peace therapy based on electroencephalogram signal analysis.Data acquisition module obtains current and historical monitoring period electroencephalogram signal and brain region topological space diagram;Abnormal period screening module is based on numerical feature division period, through comparison with historical data and adjacent monitoring point horizontal comparison, adaptive tracking electroencephalogram baseline drift and screening abnormal period;Channel decoupling and path extraction module is based on frequency domain feature and decouples abnormal period signal into multi-band channel signal, combines adjacent monitoring point comparison and divides abnormal sub-period, and determines spatial conduction path along brain region topological graph;Early warning module analyzes the timing response synchronism between abnormal sub-period across channel, distinguishes true symptom abnormality from sporadic fluctuation, and determines early warning index for early warning in combination with the overlap of conduction path and historical data.

Description

Technical Field

[0001] This invention relates to the field of electroencephalogram (EEG) signal analysis technology, and specifically to an intelligent monitoring system for palliative care symptoms based on EEG signal analysis. Background Technology

[0002] With the development of medical informatization, intelligent monitoring technology based on physiological signal analysis has been widely used in palliative care. Palliative care primarily serves terminally ill patients who often gradually lose their ability to speak due to extreme weakness, confusion, or coma. Healthcare professionals often struggle to obtain information about a patient's pain perception or emotional state through direct communication. Therefore, relying on physiological signals such as electroencephalogram (EEG) monitoring systems to assess hidden paroxysmal pain, anxiety, or agitation symptoms, and subsequently adjusting analgesia and intervention plans in a timely manner, is crucial for improving patients' quality of life.

[0003] Current EEG monitoring typically relies on extracting abnormal fluctuations in EEG signals based on fixed energy thresholds to provide early warning of abnormal physiological states in patients. However, in actual palliative care monitoring scenarios, patients' EEG signals often exhibit multiple overlapping waveforms. Specifically, patients frequently exhibit broadband electromyography (EMG) artifacts caused by behaviors such as localized muscle tension, teeth clenching, or involuntary spasms. These EMG artifacts are highly susceptible to spatiotemporal overlap and signal coupling with real abnormal EEG fluctuations caused by actual pain (such as abnormal discharges of pain receptors). Therefore, directly determining abnormal fluctuations based on fixed energy thresholds can easily lead to misjudgments. Summary of the Invention

[0004] To address the technical problem of broadband electromyography (EMG) artifacts often observed in patients due to behaviors such as localized muscle tension, teeth clenching, or involuntary spasms, and the fact that these EMG artifacts easily overlap and couple with real abnormal EEG fluctuations caused by actual pain (such as abnormal discharges of pain nerves), this invention aims to provide an intelligent monitoring system for palliative care symptoms based on EEG signal analysis. The specific technical solution adopted is as follows: This invention proposes an intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis, the system comprising: The data acquisition module is used to acquire the electroencephalogram (EEG) signals of each monitoring point of a palliative care patient during the target monitoring period, which includes the current monitoring period and historical monitoring periods, and to acquire a brain region topological spatial map. The abnormal time period screening module is used to divide the time period based on the numerical characteristics of the EEG signal for any monitoring point in the current monitoring cycle; analyze the change characteristics of the EEG signal in each time period and the historical monitoring cycle, and compare it with adjacent monitoring points to screen out abnormal time periods; The channel decoupling and path extraction module is used to separate the EEG signal into multiple frequency bands for each abnormal time period of each monitoring point based on the frequency domain characteristics of the EEG signal of all monitoring points in the corresponding time period; to divide the abnormal sub-time periods based on the data change characteristics of each channel signal in each abnormal time period and to compare with adjacent monitoring points; to traverse the monitoring points along the brain region topology map and to determine the spatial conduction path of each abnormal sub-time period corresponding to each monitoring point based on the deviation of the spectral energy centroid between abnormal sub-time periods; The early warning module is used to analyze the temporal response synchronization between abnormal sub-periods under each abnormal period of each monitoring point across channels, and to determine early warning indicators and issue early warnings by combining the spatial transmission path of abnormal sub-periods of each monitoring point with the overlap between historical data and the data.

[0005] Furthermore, the method for obtaining the time period includes: Extract the envelope of the EEG signal for each monitoring point within the current monitoring period, and mark the maximum points on the envelope; The time interval between any two adjacent maxima is calculated and the average value is taken as the mean length. Based on the mean length, the EEG signal is truncated to obtain all time periods of each monitoring point within the current monitoring cycle.

[0006] Furthermore, the method for obtaining the abnormal time period includes: In the current monitoring period, the average energy of the EEG signal in each time period is calculated in the EEG signal of each monitoring point as a local average feature value. In the historical monitoring period, the average of all energy in the EEG signal of each monitoring point is calculated as a global average feature value. The absolute value of the difference between the local average feature value of each time period of each monitoring point and the global average feature value is used as the deviation value. Choose any monitoring point as the target point, and in the brain region topology map, use the monitoring points connected to the target point as reference points; Select any one of the time periods at the target point as the target segment. Among all the time periods at all the reference points, select the time periods that overlap with the target segment as the reference segments. Calculate the mean of the deviation values ​​of all reference segments as the comparison value. Normalize the absolute value of the difference between the deviation value of the target segment and the comparison value as the anomaly factor of the target segment. The abnormality factor of the target segment is multiplied by the deviation value and then normalized to obtain the abnormality value of the target segment. In all time periods of the target point, the time periods in which the abnormality level value is greater than the preset first abnormality threshold are defined as abnormal time periods.

[0007] Furthermore, the method for acquiring the channel signal includes: In the current monitoring cycle, for any monitoring point, select one of the abnormal time periods corresponding to the monitoring point as the test period, obtain the average power spectrum of the EEG signals of all monitoring points under the test period, and extract the maximum and minimum points on the envelope of the average power spectrum. Calculate the absolute value of the difference between the power spectral density of each minimum point and the mean of the power spectral densities of its two nearest maximum points on the frequency axis. Use this value as the valley feature value of each minimum point. Sort all the valley feature values ​​of the minimum points in descending order and select the minimum points corresponding to the first preset number of valley feature values ​​as the frequency band splitting points. The frequency values ​​corresponding to all frequency band segmentation points are arranged in ascending order to form the frequency band boundaries. The number of filter bank modes is determined based on the number of frequency band boundaries. Empirical wavelet transform is used to filter the EEG signals of all monitoring points during the test period to obtain the channel signals of each monitoring point in different frequency bands.

[0008] Furthermore, the method for obtaining the abnormal sub-time period includes: In the current monitoring cycle, select one monitoring point as the test point, and select one of the abnormal time periods of the test point as the test period. During the period to be measured, the maximum points of the signal envelope of each monitoring point in each frequency band are extracted, and the average time interval between adjacent maximum points is used as the length of the local window to divide the channel signal of each monitoring point in each frequency band into local time periods. The absolute value of the difference between the average energy value in each local time period and the average energy value of the channel signal as a whole is calculated and normalized to obtain the abnormal deviation parameter corresponding to each local time period. In the brain region topological space map, the monitoring points connected to the test point are used as analysis points; During the period to be tested, among the channel signals of the same frequency band at all monitoring points, one of the local time periods corresponding to the point to be tested is selected as a candidate time period. Among all local time periods at all analysis points, the local time periods that overlap with the candidate time periods are selected as analysis time periods. The mean of the abnormal deviation parameters of all analysis time periods is calculated as the benchmark value. The absolute value of the difference between the abnormal deviation parameters of the candidate time period and the benchmark value is normalized and used as the abnormal feature factor of the candidate time period. The abnormal feature factor of each local time period is multiplied by the abnormal deviation parameter and normalized to obtain the abnormal feature value of each local time period. Among all local time periods of the channel signal of each frequency band of the test point, the local time period with the abnormal feature value greater than the preset second abnormal threshold is regarded as the abnormal sub-time period.

[0009] Furthermore, the method for obtaining the spatial conduction path includes: In the channel signal of any frequency band during the test period, the test point is taken as the starting point, and multiple conduction branches are generated by traversing adjacent monitoring points level by level along the brain region topological space map. For any abnormal sub-period corresponding to the test point, when traversing the current conduction branch, the deviation direction of the spectral energy centroid of the current monitoring point and the next-level adjacent monitoring point relative to the normal baseline frequency in the corresponding time period of the abnormal sub-period is calculated respectively. If the deviation direction of the current monitoring point and the next-level adjacent monitoring point is the same, and the next-level adjacent monitoring point does not exist in the current conduction branch, then the next-level adjacent monitoring point is included in the current conduction branch and traversed as a new current monitoring point. If the deviation direction is different, or the next-level adjacent monitoring point already exists in the current conduction branch, then the traversal of the next-level adjacent monitoring point is terminated. The transmission branch with the most monitoring points among all the transmission branches generated by the traversal is taken as the spatial transmission path of the test point under the abnormal sub-period.

[0010] Furthermore, the method for obtaining the early warning indicators includes: Under any abnormal time period at any monitoring point, analyze the timing response synchronization of abnormal sub-time periods between channel signals across channels to determine the reasonable indicators for that abnormal time period; Within the current monitoring period, for any abnormal time period at any monitoring point, extract the monitoring points contained in the spatial propagation path of all abnormal sub-time periods under the channel signal of each frequency band, and construct a path set; The path set is matched with the historical path set corresponding to the historical abnormal sub-periods that overlap with the abnormal period in the historical monitoring cycle. The normalized value of the ratio of the number of intersection elements to the number of union elements is used as the deterioration weight of the channel signal of each frequency band. The proportion of channel signals with a degradation weight greater than a preset degradation threshold to the total number of channel signals is used as the degradation ratio. The deterioration ratio corresponding to each abnormal period is multiplied by the corresponding reasonable indicator and normalized to obtain the early warning factor for each abnormal period. The mean of the early warning factors for all abnormal periods is used as the early warning indicator for each monitoring point in the current monitoring cycle.

[0011] Furthermore, the method for obtaining the reasonable indicators includes: At any abnormal time period at any monitoring point, for any abnormal sub-time period in any channel signal, the abnormal sub-time periods in the other channel signals that have temporal overlap with that abnormal sub-time period are taken as reference segments. The number of overlapping moments between each reference segment and the abnormal sub-period is calculated as the synchronization factor, and the variance of all data values ​​in each reference segment is used as the volatility factor. The volatility factor is weighted and summed using the synchronization factor of the reference segment. The normalized weighted result is then multiplied by the number of reference segments to obtain a reasonable factor for the abnormal sub-period. The average of the reasonable factors for all abnormal sub-periods in all channel signals is used as the reasonable index for the corresponding abnormal period.

[0012] Furthermore, the provision of early warning includes: During the current monitoring period, if the warning indicator of a certain monitoring point is greater than or equal to the preset warning threshold, then the monitoring point is designated as a high-risk monitoring point, and a warning needs to be issued and the EEG signal of the monitoring point needs to be output. If the warning indicator of a certain monitoring point is lower than the preset warning threshold, then the monitoring point is regarded as a normal monitoring point and no warning is required.

[0013] Furthermore, the preset warning threshold ranges from (0.7, 1).

[0014] The present invention has the following beneficial effects: First, the data acquisition module obtains EEG signals and brain region topological maps for the current and historical monitoring periods. In the abnormal period screening module, the current monitoring period is divided into time periods based on the numerical characteristics of the EEG signals, improving analysis accuracy by utilizing local data. By comparing the change characteristics of EEG signals in each time period with those of historical monitoring periods, and by performing horizontal comparisons with EEG signals from adjacent monitoring points, the module can adaptively track the dynamic drift of the patient's EEG baseline and screen out abnormal periods. Because EEG signals may exhibit complex waveform overlaps, such as the overlap between electromyography artifacts and real EEG waves, and given that different physiological processes correspond to different dominant frequencies, the channel decoupling and path extraction module can separate EEG signals from abnormal periods into multiple frequency bands based on frequency domain characteristics. This decouples signals from different physiological sources in the frequency domain, avoiding feature ambiguity caused by frequency band mixing. Furthermore, within each channel signal, abnormal sub-periods are further segmented by combining data change characteristics with comparisons to adjacent monitoring points. The module then traverses the monitoring points along the brain region topology map and determines spatial conduction paths based on the deviation of the spectral energy centroid between abnormal sub-periods, providing direct evidence for brain region localization and mechanism analysis of symptoms. Finally, in the early warning module, the temporal response synchronicity between abnormal sub-periods is analyzed across channels to determine whether abnormal signals possess cross-frequency band collaborative response characteristics, effectively distinguishing between true symptom-related abnormalities and occasional background fluctuations. Simultaneously, early warning determination is made by combining the spatial conduction paths of abnormal sub-periods at each monitoring point with the overlap of historical data. Attached Figure Description

[0015] Figure 1This is a system block diagram of intelligent monitoring of palliative care symptoms based on electroencephalogram (EEG) signal analysis, provided by an embodiment of the present invention. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis provided by this invention.

[0017] Please see Figure 1 The diagram illustrates a system block diagram of an intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to an embodiment of the present invention. The system includes: a data acquisition module 101, an abnormal time period screening module 102, a channel decoupling and path extraction module 103, and an early warning module 104.

[0018] The data acquisition module 101 is used to acquire the electroencephalogram (EEG) signals of each monitoring point of the palliative care patient during the target monitoring period, which includes the current monitoring period and historical monitoring periods, and to acquire the brain region topological spatial map.

[0019] The medication rhythm and daily routine of palliative care patients are usually measured in days. Therefore, in this embodiment of the invention, a 24-hour monitoring cycle is used. In each monitoring cycle, a fixed sampling frequency (e.g., 250Hz) is set, and raw EEG signals from various monitoring points on the patient's head are collected through a wearable multi-channel EEG acquisition device. The raw EEG signals are amplified and converted by A / D. A high-pass filter is used for baseline drift correction, and a band-pass filter is used to remove power frequency interference and high-frequency noise to obtain preprocessed EEG signals.

[0020] In this embodiment of the invention, EEG signals are acquired under two target monitoring cycles, namely the current monitoring cycle and the historical monitoring cycle, wherein the time period corresponding to the historical monitoring cycle is the day before the current monitoring cycle.

[0021] Then, the three-dimensional physical coordinates of each monitoring point on the patient's head in the multi-channel EEG acquisition device are read, and the preset standard brain functional area mapping template is called. Based on the three-dimensional physical coordinates, each monitoring point is mapped to the brain functional area mapping template, thereby constructing a brain region topology map. The brain region topology map uses the monitoring points as nodes and the physical adjacency and neural conduction connectivity of the monitoring points on the scalp surface as edges.

[0022] In this embodiment of the invention, the collection and acquisition of patients' personal information data are all authorized by the relevant users, and the process does not violate relevant laws and regulations, nor does it violate public order and good morals.

[0023] The abnormal time period screening module 102 is used to divide the time period based on the numerical characteristics of the EEG signal for any monitoring point in the current monitoring cycle; analyze the change characteristics of the EEG signal in each time period and the historical monitoring cycle, and compare it with adjacent monitoring points, thereby screening out abnormal time periods.

[0024] Patients receiving palliative care have different disease courses and medication regimens, resulting in significant individual differences in their baseline EEG rhythms. Therefore, by analyzing the numerical characteristics of the EEG signals of the current patient in the current monitoring period to divide the time period, the pace of the time window can be made to match the patient's own physiological fluctuation baseline rhythm, avoiding feature truncation or overlap caused by a fixed time window.

[0025] Preferably, in one embodiment of the present invention, the method for obtaining the time period includes: EEG signals may contain a large number of high-frequency, complex, and subtle fluctuations. Therefore, extracting the envelope of the EEG signal at each monitoring point within the current monitoring period can filter out trivial spikes and reveal the overall fluctuation trend of EEG activity. Then, marking the maximum value points on the envelope, which represent the energy peaks of neural activity in local brain regions, helps to capture the macroscopic basic rhythm of the patient's EEG activity.

[0026] Then, the time interval between each two adjacent maxima is calculated and the average is taken as the mean length. Based on this mean length, the EEG signal is truncated to obtain all time periods of each monitoring point within the current monitoring cycle. At this time, the time scale of each time period can match the physiological fluctuation rhythm of the patient in the current monitoring cycle, which helps to capture its unique abnormal mutations more accurately in the subsequent process.

[0027] Since paroxysmal pain and other abnormal neurological discharges can cause a sudden increase in energy in local brain regions, it is possible to analyze the changes in EEG signals in each time period of the current monitoring cycle and in historical monitoring cycles. This effectively isolates the patient's inherent background baseline and accurately captures high-energy, variable-frequency bursts triggered by abnormal physiological events. At the same time, because real abnormal neurological discharges (or strong electromyographic interference) in the brain have the physical characteristic of spreading to the surrounding neural topology rather than existing in isolation, comparisons were also made with adjacent monitoring points to screen out abnormal time periods among all time periods corresponding to each monitoring point.

[0028] Preferably, in one embodiment of the present invention, the method for obtaining abnormal time periods includes: In the current monitoring period, the average energy of the EEG signal at each monitoring point within each time period is calculated as a local mean feature value. The local mean feature value characterizes the instantaneous firing intensity and activity of neuronal clusters within a short period of time. In the historical monitoring period, the average of all energy in the EEG signal at each monitoring point is calculated as a global mean feature value. The global mean feature value characterizes the patient's EEG background baseline level over an entire historical monitoring period. Then, the absolute value of the difference between the local mean feature value and the global mean feature value at each time period of each monitoring point is used as the deviation value. The larger the deviation value, the more serious the deviation of the patient's EEG signal from steady state within that time period.

[0029] Choose any monitoring point as the target point. In the brain region topology map, use the monitoring points connected to the target point as reference points. Select any time period corresponding to the target point as the target segment. Among all time periods of all reference points, select the time periods that overlap with the target segment as reference segments. Since the reference segments are topologically connected and temporally overlapping with the target segment, they have higher reference value. Therefore, calculate the mean of the deviation values ​​of all reference segments as a comparison value. This comparison value quantifies the spatial common-mode fluctuation level around the target segment. Then, normalize the absolute value of the difference between the deviation value of the target segment and the comparison value to obtain the abnormality factor of the target segment. The smaller the abnormality factor, the more synchronously the target segment and the reference segments in the surrounding monitoring points change, thus increasing the likelihood of common-mode interference artifacts caused by whole-body movements. Conversely, the larger the abnormality factor, the stronger the local independent mutation characteristics of the data in the target segment, and the higher the probability of anomalies. The normalization here can use maximum-minimum normalization, where the maximum and minimum values ​​can be obtained from a database of all reference segments corresponding to the target segment.

[0030] The anomaly severity factor corresponding to the target segment is multiplied by the deviation severity value and then normalized to obtain the anomaly severity value of the target segment. Based on the previous analysis, the larger the anomaly severity value, the higher the probability of the target segment being abnormal. Therefore, among all time periods of the target point, the time periods with an anomaly severity value greater than the preset first anomaly threshold are finally identified as abnormal time periods. The normalization here can be achieved using maximum and minimum value normalization, where the maximum and minimum values ​​can be obtained from the database of all time periods of the target point.

[0031] It should be noted that the preset first abnormal threshold is 0.75, and the specific value can be adjusted according to the implementation scenario, without limitation here.

[0032] The channel decoupling and path extraction module 103 is used to separate the EEG signal into multiple frequency bands for each abnormal time period of each monitoring point based on the frequency domain characteristics of the EEG signal of all monitoring points in the corresponding time period; to divide the abnormal sub-time periods based on the data change characteristics of each channel signal in each abnormal time period and to compare with adjacent monitoring points; to traverse the monitoring points along the brain region topology map and to determine the spatial conduction path of each abnormal sub-time period corresponding to each monitoring point based on the deviation of the spectral energy centroid between abnormal sub-time periods.

[0033] In actual palliative care settings, genuine pathological abnormalities (such as nerve discharges causing paroxysmal pain or anxiety waves) typically manifest as dominant waves within a specific frequency band, while electromyography (EMG) artifacts generated by the patient's tensing muscles and clenching teeth present as strong broadband interference. When the two overlap, they can mask the true pathological features. Therefore, for each abnormal time period at each monitoring point, the frequency domain characteristics of the EEG signals at all monitoring points during the corresponding time period can be used to separate the EEG signals into multiple frequency band channels. This can effectively remove broadband artifacts and assign different physiological or interference components to different frequency domain channels, thus revealing the hidden true pathological discharges.

[0034] Preferably, in one embodiment of the present invention, the method for acquiring the channel signal includes: In the current monitoring cycle, for any monitoring point, one of the abnormal time periods corresponding to that monitoring point is selected as the test period. Since different physiological events have different frequency characteristics, the average power spectrum of the EEG signals of all monitoring points under the test period is obtained. The average power spectrum integrates the frequency distribution of the whole brain in the test period. By extracting the maximum and minimum points on the envelope of the average power spectrum, the boundary between the dominant frequency and the non-dominant frequency can be clearly defined.

[0035] Minimum points may be slight fluctuations or true breakpoints between two dominant frequency bands. Therefore, the absolute value of the difference between the power spectral density of each minimum point and the mean of the power spectral densities of its two nearest maxima on the frequency axis is calculated as the valley feature value for each minimum point. The larger the valley feature value, the more significant the difference in power spectral density between the minimum point and its two nearest maxima, indicating a greater gap between the minimum point and surrounding peak energies, and thus a more likely clear breakpoint. Therefore, all valley feature values ​​are sorted in descending order, and the minimum points corresponding to the first preset number of valley feature values ​​are selected as frequency band breakpoints. It should be noted that the preset number is set to one-fifth of the total number of minimum points, rounded up. The specific value can be adjusted according to the implementation scenario and is not limited here.

[0036] Finally, the frequency values ​​corresponding to all frequency band segmentation points are arranged in ascending order to serve as the frequency band boundaries. The number of filter bank modes is determined based on the number of frequency band boundaries. Empirical wavelet transform is used to filter the EEG signals of all monitoring points during the test period to obtain the channel signals of each monitoring point in different frequency bands.

[0037] It should be noted that wavelet transform is a well-known technique, and the specific process will not be elaborated here.

[0038] During palliative care, signal fluctuations caused by broadband electromyography artifacts such as local muscle tension and teeth clenching usually manifest as in situ bursts with indiscriminate interference across all frequency bands. In contrast, real pathological fluctuations such as pain and anxiety often have specific dominant frequency bands and exhibit spatial conduction and diffusion patterns along neural topology. Therefore, we can analyze the change characteristics of data in each channel signal during each abnormal time period and compare them with adjacent monitoring points to identify abnormal sub-time periods.

[0039] Preferably, in one embodiment of the present invention, the method for obtaining abnormal sub-time periods includes: First, in the current monitoring period, select one monitoring point as the test point, and then select one of the abnormal time periods of the test point as the test period.

[0040] After frequency domain decoupling, the channel signals of different frequency bands have completely different oscillation frequencies and rhythms. Therefore, during the period to be measured, the maximum points of the signal envelope of each monitoring point in each frequency band are extracted, and the average time interval between adjacent maximum points is used as the length of the local window to divide the channel signals of each monitoring point in each frequency band into local time periods. At this time, the method of dividing local time periods can better conform to the physiological cycle of the corresponding frequency components.

[0041] Within a single frequency band's corresponding channel signal, neural activity also exhibits normal, stable fluctuations. Therefore, the absolute value of the difference between the average energy value within each local time segment and the overall average energy value of the channel signal is calculated and normalized to obtain the abnormal deviation parameter for each local time segment. The larger the abnormal deviation parameter, the greater the likelihood of a sudden change in transient energy within that local time segment. Normalization can be performed using maximum and minimum values, where the maximum and minimum values ​​can be extracted from a database of all local time segments within the test period.

[0042] In the brain region topological spatial map, monitoring points connected to the test point are used as analysis points. Then, during the test period, among the channel signals of the same frequency band of all monitoring points, one local time period corresponding to the test point is randomly selected as a candidate time period. Among all local time periods of all analysis points, local time periods that overlap with the candidate time periods are used as analysis time periods. The mean of the abnormal deviation parameters of all analysis time periods is calculated as the baseline value. The baseline value represents the average abnormal fluctuation background of monitoring points around the test point in the same frequency band. The normalized value of the absolute value of the difference between the abnormal deviation parameters of the candidate time period and the baseline value is used as the abnormal feature factor of the candidate time period. The larger the abnormal feature factor, the greater the possibility of spatially independent mutation of the test point within the local spatial range. The normalization here can be achieved by using maximum and minimum value normalization, where the maximum and minimum values ​​can be selected from all local time periods corresponding to the test point as the database.

[0043] The abnormal characteristic factor of each local time period is multiplied by the abnormal deviation parameter and then normalized to obtain the abnormal characteristic value of each local time period. Based on the aforementioned logical analysis, it is known that the larger the abnormal characteristic value, the higher the probability of an anomaly in that local time period. Finally, among all local time periods of the channel signal in each frequency band of the test point, the local time periods with abnormal characteristic values ​​greater than a preset second abnormal threshold are regarded as abnormal sub-time periods. The normalization here can be achieved using maximum and minimum value normalization, where the maximum and minimum values ​​can be selected from the database of all local time periods corresponding to the test point.

[0044] It should be noted that the preset second abnormal threshold is 0.75, and the specific value can be adjusted according to the implementation scenario, without limitation here.

[0045] The propagation of bioelectrical signals in the brain is limited by the real neural network and anatomical structure, and will not produce instantaneous displacement across regions. Therefore, by traversing the brain region topology map step by step, it can be ensured that the searched abnormal signals are spatially continuous and physically accessible, thereby effectively filtering out spatially isolated random electrode noise. At the same time, when a real pathological abnormality spreads from one brain region to an adjacent brain region, not only will the energy fluctuate, but the distribution characteristics of its frequency components will also usually change. Therefore, when traversing the monitoring points along the brain region topology map, the spatial conduction path of each abnormal sub-time period corresponding to each monitoring point can also be determined based on the spectral energy centroid shift between abnormal sub-time periods.

[0046] Preferably, in one embodiment of the present invention, the method for obtaining the spatial conduction path includes: The brain's neural network is highly interconnected. Abnormal discharges (such as pain stimuli) will radiate and spread in multiple directions along the nerve fibers from the source, forming multiple possible branches. Therefore, by traversing the brain region topology map step by step based on a pre-constructed topological spatial map, it can be ensured that the signal propagation path being analyzed is physically coherent and reachable, thereby eliminating random electromagnetic noise that is spatially isolated or jumps across boundaries. In the channel signal of any frequency band during the test period, the test point is taken as the starting point, and multiple conduction branches are generated by traversing the brain region topology map step by step along the adjacent monitoring points.

[0047] For any abnormal sub-segment corresponding to the test point, when traversing the current conduction branch, the deviation direction of the spectral energy centroid of the current monitoring point and the next-level adjacent monitoring point relative to the normal baseline frequency within the corresponding time period of the abnormal sub-segment is calculated. Since the migration trend of frequency components of real neural signals is usually consistent during conduction in a specific neural pathway, if the deviation direction of the current monitoring point and the next-level adjacent monitoring point is the same, and the next-level adjacent monitoring point does not exist in the current conduction branch, then the next-level adjacent monitoring point is included in the current conduction branch and used as a new current monitoring point for continued traversal; conversely, if the deviation directions are different, or the next-level adjacent monitoring point already exists in the current conduction branch, then the traversal of the next-level adjacent monitoring point is terminated. It should be noted that the normal baseline frequency is the average frequency of the patient's EEG signal during the historical monitoring period, and can also be set as an empirical constant according to the implementation scenario, which is not limited here.

[0048] Finally, the conduction branch with the most monitoring points among all the conduction branches generated by the traversal is taken as the spatial conduction path of the test point in the abnormal sub-period. The spatial conduction path with the most monitoring points will also have a higher radiation breadth and invasion depth of the abnormal electrophysiological activity in the cerebral cortex.

[0049] The early warning module 104 is used to analyze the temporal response synchronization between abnormal sub-periods under each abnormal period of each monitoring point across channels, and to determine early warning indicators and issue early warnings by combining the spatial transmission path of abnormal sub-periods of each monitoring point with the overlap between historical data and the data.

[0050] In palliative care settings, patients are often in a terminal state with multiple intertwined symptoms (e.g., severe pain is often accompanied by intense anxiety). Different physiological symptoms are controlled by different dominant frequency bands. When complex symptoms occur concurrently, multiple originally independent frequency channels will exhibit highly overlapping abnormal responses on the timeline. By analyzing the temporal response synchronicity of abnormal sub-periods across channels, it is possible to effectively determine whether the current EEG fluctuation is an isolated event of a single frequency or a serious complication that has triggered a multi-frequency chain reaction across the entire brain. Furthermore, the patient's true pathological abnormalities (such as paroxysmal discharges of specific neurons) are constrained by the inherent neuroanatomical structure of the human body. The spatial conduction path of the same true physiological abnormality on the brain region topological map has extremely high historical stability and reproducibility. In contrast, the conduction paths of occasional random noise or random muscle twitches are often chaotic and non-repeating. Therefore, the currently extracted conduction path is compared with the paths of historical data within the historical monitoring period to overlap nodes, filtering out occasional artifacts and accurately pinpointing the true symptom. Based on the aforementioned two characteristics, early warning indicators can be determined for early warning purposes.

[0051] Preferably, in one embodiment of the present invention, the method for obtaining early warning indicators includes: At any abnormal time period at any monitoring point, the temporal response synchronicity of abnormal sub-segments between channel signals is analyzed across channels to determine reasonable indicators for the abnormal time period. At any abnormal time period at any monitoring point, any abnormal sub-segment in any channel signal is designated as a specific sub-segment, and abnormal sub-segments in other channel signals that temporally overlap with the specific sub-segment are designated as reference segments. Abnormal sub-segments in different frequency bands may have time differences, but the smaller the temporal interval, the more likely the neuronal clusters in different frequency bands are resonating and co-triggered in response to the same physiological event. Therefore, the number of overlapping moments between each reference segment and the specific sub-segment is calculated as a synchronization factor. A larger synchronization factor indicates a higher probability of co-triggered responses. The variance of all data values ​​in each reference segment is used as a fluctuation factor; a larger fluctuation factor indicates more drastic changes in the data values ​​within the reference segment. The negative correlation mapping here can be achieved using… in, This represents an exponential function with the natural constant e as its base, where x represents the independent variable, t represents the dimensionless conversion coefficient, and in this case, the time unit.

[0052] Then, the fluctuation factor is weighted and summed using the synchronization factors of all reference segments corresponding to the specific sub-segment. The normalized value of the weighted sum is multiplied by the number of reference segments to serve as the rational factor for the specific sub-segment. The larger the weighted result, the more co-responsive other frequency bands are to the specific sub-segment in time, and the more intense the energy oscillations of the response. The larger the number of reference segments, the more simultaneous resonances of different frequency bands the specific sub-segment triggers, and the wider the scope involved. Therefore, the larger the rational factor, the more it indicates that the physiological abnormality is not an isolated interference of a single frequency band, but rather triggers a strong and synchronous joint response of multiple frequency bands in the whole brain. Its pathological components are more complex and realistic, and its clinical and pathological significance is higher. Therefore, the mean of the rational factors of all abnormal sub-segments in all channel signals is used as the rational indicator of the corresponding abnormal segment. The normalization here can use maximum and minimum value normalization, where the maximum and minimum values ​​can be extracted from the database of all abnormal sub-segments in the channel signals.

[0053] Within the current monitoring period, for any abnormal time period at any monitoring point, extract the monitoring points contained in the spatial propagation path of all abnormal sub-time periods under the channel signal of each frequency band, and construct a path set.

[0054] The path set is matched with the historical path set corresponding to historical abnormal sub-periods that overlap with the current abnormal period in the historical monitoring cycle. The ratio of the number of elements in the intersection to the number of elements in the union of the two is used as the path overlap degree. The higher the path overlap degree, the more historical recurrence characteristics the path set in the current monitoring cycle has. Therefore, the path overlap degree is normalized to obtain the degradation weight of the channel signal in each frequency band. The proportion of channel signals with degradation weights greater than a preset degradation threshold is used as the degradation ratio of each abnormal period. Normalization can be performed using maximum and minimum value normalization, where the maximum and minimum values ​​can be obtained from a database of all abnormal periods in the current monitoring cycle. It should be noted that the preset degradation threshold is set to 0.75 in this embodiment of the invention. The specific value can be adjusted according to the implementation scenario and is not limited here.

[0055] Finally, the deterioration ratio corresponding to each abnormal period is multiplied by the corresponding reasonable indicator and normalized to obtain the warning factor for each abnormal period. Based on the previous analysis, the larger the warning factor, the more complex the pathological components, and the more necessary further analysis is required. Therefore, the mean of the warning factors for all abnormal periods is used as the warning indicator for each monitoring point in the current monitoring period. Normalization can be performed using maximum and minimum value normalization, where the maximum and minimum values ​​can be obtained from a database of all abnormal periods in the current monitoring period.

[0056] After obtaining the early warning indicators, early warnings can be issued based on the early warning indicators for each monitoring point within the current monitoring period. Preferably, issuing early warnings includes: During the current monitoring period, if the warning indicator of a certain monitoring point is greater than or equal to the preset warning threshold, the monitoring point will be designated as a high-risk monitoring point, and a warning will be issued and the EEG signal of the monitoring point will be output; if the warning indicator of a certain monitoring point is less than the preset warning threshold, the monitoring point will be designated as a normal monitoring point and no warning is required.

[0057] It should be noted that the preset warning threshold is set to a range of (0.7, 1). In this embodiment of the invention, it can be set to 0.75. The specific value can be adjusted according to the implementation scenario and is not limited here.

[0058] The historical indicators, such as historical abnormal periods, historical abnormal sub-periods, and historical path sets, involved in this embodiment are obtained and calculated in the same way as the aforementioned processing method for EEG signals in the current monitoring period. That is, the same data acquisition, abnormal screening, and channel decoupling path extraction logic are used to process the historical EEG signals, which will not be repeated here.

[0059] In summary, the data acquisition module first obtained EEG signals and brain region topological maps for the current and historical monitoring periods. In the abnormal period screening module, the current monitoring period was divided into time segments based on the numerical characteristics of the EEG signals, improving analysis accuracy by utilizing local data. By comparing the change characteristics of each time segment with those of historical monitoring periods, and by performing horizontal comparisons with EEG signals from adjacent monitoring points, the system can adaptively track the dynamic drift of the patient's EEG baseline and screen out abnormal time segments. Because EEG signals may exhibit complex waveform overlaps, such as the overlap between electromyography artifacts and real EEG waves, and given that different physiological processes correspond to different dominant frequencies, the channel decoupling and path extraction module can separate EEG signals from abnormal periods into multiple frequency bands based on frequency domain characteristics. This decouples signals from different physiological sources in the frequency domain, avoiding feature ambiguity caused by frequency band mixing. Furthermore, within each channel signal, abnormal sub-periods are further segmented by combining data change characteristics with comparisons to adjacent monitoring points. The module then traverses the monitoring points along the brain region topology map and determines spatial conduction paths based on the deviation of the spectral energy centroid between abnormal sub-periods, providing direct evidence for brain region localization and mechanism analysis of symptoms. Finally, in the early warning module, the temporal response synchronicity between abnormal sub-periods is analyzed across channels to determine whether abnormal signals possess cross-frequency band collaborative response characteristics, effectively distinguishing between true symptom-related abnormalities and occasional background fluctuations. Simultaneously, early warning determination is made by combining the spatial conduction paths of abnormal sub-periods at each monitoring point with the overlap of historical data.

[0060] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis, characterized in that, The system includes: The data acquisition module is used to acquire the electroencephalogram (EEG) signals of each monitoring point of a palliative care patient during the target monitoring period, which includes the current monitoring period and historical monitoring periods, and to acquire a brain region topological spatial map. The abnormal time period screening module is used to divide the time period based on the numerical characteristics of the EEG signal for any monitoring point in the current monitoring cycle; analyze the change characteristics of the EEG signal in each time period and the historical monitoring cycle, and compare it with adjacent monitoring points to screen out abnormal time periods; The channel decoupling and path extraction module is used to separate the EEG signal into multiple frequency bands for each abnormal time period of each monitoring point based on the frequency domain characteristics of the EEG signal of all monitoring points in the corresponding time period; to divide the abnormal sub-time periods based on the data change characteristics of each channel signal in each abnormal time period and to compare with adjacent monitoring points; to traverse the monitoring points along the brain region topology map and to determine the spatial conduction path of each abnormal sub-time period corresponding to each monitoring point based on the deviation of the spectral energy centroid between abnormal sub-time periods; The early warning module is used to analyze the temporal response synchronization between abnormal sub-periods under each abnormal period of each monitoring point across channels, and to determine early warning indicators and issue early warnings by combining the spatial transmission path of abnormal sub-periods of each monitoring point with the overlap between historical data and the data.

2. The intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to claim 1, characterized in that, The methods for obtaining the time period include: Extract the envelope of the EEG signal for each monitoring point within the current monitoring period, and mark the maximum points on the envelope; The time interval between any two adjacent maxima is calculated and the average value is taken as the mean length. Based on the mean length, the EEG signal is truncated to obtain all time periods of each monitoring point within the current monitoring cycle.

3. The intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to claim 1, characterized in that, The method for obtaining the abnormal time period includes: In the current monitoring period, the average energy of the EEG signal in each time period is calculated in the EEG signal of each monitoring point as a local average feature value. In the historical monitoring period, the average of all energy in the EEG signal of each monitoring point is calculated as a global average feature value. The absolute value of the difference between the local average feature value of each time period of each monitoring point and the global average feature value is used as the deviation value. Choose any monitoring point as the target point, and in the brain region topology map, use the monitoring points connected to the target point as reference points; Select any one of the time periods at the target point as the target segment. Among all the time periods at all the reference points, select the time periods that overlap with the target segment as the reference segments. Calculate the mean of the deviation values ​​of all reference segments as the comparison value. Normalize the absolute value of the difference between the deviation value of the target segment and the comparison value as the anomaly factor of the target segment. The abnormality factor of the target segment is multiplied by the deviation value and then normalized to obtain the abnormality value of the target segment. In all time periods of the target point, the time periods in which the abnormality value is greater than the preset first abnormality threshold are defined as abnormal time periods.

4. The intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to claim 1, characterized in that, The method for acquiring the channel signal includes: In the current monitoring cycle, for any monitoring point, one of the abnormal time periods corresponding to the monitoring point is selected as the test period, the average power spectrum of the EEG signals of all monitoring points under the test period is obtained, and the maximum and minimum points on the envelope of the average power spectrum are extracted. Calculate the absolute value of the difference between the power spectral density of each minimum point and the mean of the power spectral densities of its two nearest maximum points on the frequency axis. Use this value as the valley feature value of each minimum point. Sort all the valley feature values ​​of the minimum points in descending order and select the minimum points corresponding to the first preset number of valley feature values ​​as the frequency band splitting points. The frequency values ​​corresponding to all frequency band segmentation points are arranged in ascending order to form the frequency band boundaries. The number of filter bank modes is determined based on the number of frequency band boundaries. Empirical wavelet transform is used to filter the EEG signals of all monitoring points during the test period to obtain the channel signals of each monitoring point in different frequency bands.

5. The intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to claim 1, characterized in that, The method for obtaining the abnormal sub-time period includes: In the current monitoring cycle, select one monitoring point as the test point, and select one of the abnormal time periods of the test point as the test period. During the period to be measured, the maximum points of the signal envelope of each monitoring point in each frequency band are extracted, and the average time interval between adjacent maximum points is used as the length of the local window to divide the channel signal of each monitoring point in each frequency band into local time periods. The absolute value of the difference between the average energy value in each local time period and the average energy value of the channel signal as a whole is calculated and normalized to obtain the abnormal deviation parameter corresponding to each local time period. In the brain region topological space map, the monitoring points connected to the test point are used as analysis points; During the period to be tested, among the channel signals of the same frequency band at all monitoring points, one of the local time periods corresponding to the point to be tested is selected as a candidate time period. Among all local time periods at all analysis points, the local time periods that overlap with the candidate time periods are selected as analysis time periods. The mean of the abnormal deviation parameters of all analysis time periods is calculated as the benchmark value. The absolute value of the difference between the abnormal deviation parameters of the candidate time period and the benchmark value is normalized and used as the abnormal feature factor of the candidate time period. The abnormal feature factor of each local time period is multiplied by the abnormal deviation parameter and normalized to obtain the abnormal feature value of each local time period. Among all local time periods of the channel signal of each frequency band of the test point, the local time period with the abnormal feature value greater than the preset second abnormal threshold is regarded as the abnormal sub-time period.

6. The intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to claim 5, characterized in that, The method for obtaining the spatial transmission path includes: In the channel signal of any frequency band during the test period, the test point is taken as the starting point, and multiple conduction branches are generated by traversing adjacent monitoring points level by level along the brain region topological space map. For any abnormal sub-period corresponding to the test point, when traversing the current conduction branch, the deviation direction of the spectral energy centroid of the current monitoring point and the next-level adjacent monitoring point relative to the normal baseline frequency in the corresponding time period of the abnormal sub-period is calculated respectively. If the deviation direction of the current monitoring point and the next-level adjacent monitoring point is the same, and the next-level adjacent monitoring point does not exist in the current conduction branch, then the next-level adjacent monitoring point is included in the current conduction branch and traversed as a new current monitoring point. If the deviation direction is different, or the next-level adjacent monitoring point already exists in the current conduction branch, then the traversal of the next-level adjacent monitoring point is terminated. The transmission branch with the most monitoring points among all the transmission branches generated by the traversal is taken as the spatial transmission path of the test point under the abnormal sub-period.

7. The intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to claim 1, characterized in that, The methods for obtaining the early warning indicators include: Under any abnormal time period at any monitoring point, analyze the timing response synchronization of abnormal sub-time periods between channel signals across channels to determine the reasonable indicators for that abnormal time period; Within the current monitoring period, for any abnormal time period at any monitoring point, extract the monitoring points contained in the spatial propagation path of all abnormal sub-time periods under the channel signal of each frequency band, and construct a path set; The path set is matched with the historical path set corresponding to the historical abnormal sub-periods that overlap with the abnormal period in the historical monitoring cycle. The normalized value of the ratio of the number of intersection elements to the number of union elements is used as the deterioration weight of the channel signal of each frequency band. The proportion of channel signals with a degradation weight greater than a preset degradation threshold to the total number of channel signals is used as the degradation ratio. The deterioration ratio corresponding to each abnormal period is multiplied by the corresponding reasonable indicator and normalized to obtain the early warning factor for each abnormal period. The mean of the early warning factors for all abnormal periods is used as the early warning indicator for each monitoring point in the current monitoring cycle.

8. The intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to claim 7, characterized in that, The methods for obtaining the reasonable indicators include: At any abnormal time period at any monitoring point, for any abnormal sub-time period in any channel signal, the abnormal sub-time periods in the other channel signals that have temporal overlap with that abnormal sub-time period are taken as reference segments. The number of overlapping moments between each reference segment and the abnormal sub-period is calculated as the synchronization factor, and the variance of all data values ​​in each reference segment is used as the volatility factor. The volatility factor is weighted and summed using the synchronization factor of the reference segment. The normalized weighted result is then multiplied by the number of reference segments to obtain a reasonable factor for the abnormal sub-period. The average of the reasonable factors for all abnormal sub-periods in all channel signals is used as the reasonable index for the corresponding abnormal period.

9. The intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to claim 8, characterized in that, The aforementioned early warning includes: During the current monitoring period, if the warning indicator of a certain monitoring point is greater than or equal to the preset warning threshold, then the monitoring point is designated as a high-risk monitoring point, and a warning needs to be issued and the EEG signal of the monitoring point needs to be output. If the warning indicator of a certain monitoring point is lower than the preset warning threshold, then the monitoring point is regarded as a normal monitoring point and no warning is required.

10. The intelligent monitoring system for palliative care symptoms based on electroencephalogram (EEG) signal analysis according to claim 9, characterized in that, The preset warning threshold ranges from (0.7, 1).

Images