Non-invasive multi-modal brain function monitoring method and device, electronic equipment and storage medium

Abstract

The application provides a non-invasive multi-modal brain function monitoring method and device, electronic equipment and storage medium, and relates to the technical field of multi-modal electric digital data processing. The method comprises the following steps: acquiring physiological state information of a measured target and electroencephalogram monitoring data, brain oxygen monitoring data and brain edema monitoring data of the measured target within a preset time length; determining electroencephalogram infection artifacts of the measured target according to the physiological state information, and removing the electroencephalogram infection artifacts from the electroencephalogram monitoring data to obtain real electroencephalogram data; correcting the brain oxygen monitoring data according to the physiological state information and the real electroencephalogram data to obtain real brain oxygen data; determining a brain function level of the measured target based on the physiological state information, the real electroencephalogram data, the real brain oxygen data and the brain edema monitoring data, and performing brain injury early warning according to the brain function level. The application can timely and accurately predict the change of the brain function of a child with severe infection through multi-modal monitoring data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-modal electric digital data processing, and in particular to a non-invasive multi-modal brain function monitoring method and device, an electronic device and a storage medium. BACKGROUND

[0002] Systemic inflammatory response caused by severe infection can involve all organ systems, among them, the central nervous system is affected by inflammation, oxidation and immune processes, and is prone to brain injury, which is a major cause of serious threat to children's health. Therefore, it is very important to timely and accurately determine the changes in brain function of children with severe infection and to early identify and warn brain injury.

[0003] In related technologies, the monitoring data of various monitors is mainly obtained, and the brain function of the child is determined by judging the monitoring data and grading the performance of the child's brain function, so as to determine whether brain injury occurs. However, children with severe infection may have different inflammatory responses, resulting in deviations in the obtained monitoring data, making it difficult to accurately identify the changes in brain function of children with severe infection in a timely manner. SUMMARY

[0004] The embodiments of the present application provide a non-invasive multi-modal brain function monitoring method, device, electronic device and storage medium, which timely and accurately predict the changes in brain function of children with severe infection and realize early identification and warning of brain injury.

[0005] In a first aspect, the embodiments of the present application provide a non-invasive multi-modal brain function monitoring method, comprising:

[0006] obtaining physiological state information of a measured target and electroencephalogram (EEG) monitoring data, brain oxygen monitoring data and brain edema monitoring data of the measured target within a preset time period;

[0007] According to the physiological state information, determine the EEG infection artifact of the measured target, and remove the EEG infection artifact from the EEG monitoring data to obtain true EEG data;

[0008] According to the physiological state information and the true EEG data, correct the brain oxygen monitoring data to obtain true brain oxygen data;

[0009] Based on the physiological state information, the true EEG data, the true brain oxygen data and the brain edema monitoring data, determine the brain function level of the measured target, and according to the brain function level, perform brain injury warning.

[0010] In a possible implementation manner, the physiological state information includes body temperature data and motion data of the measured target within a preset time period; the EEG infection artifact includes high temperature slow drift artifact and convulsion artifact.

[0011] Determining the EEG artifacts of the target according to the physiological state information, and removing the EEG artifacts from the EEG monitoring data to obtain real EEG data, including:

[0012] Decomposing the EEG monitoring data to obtain low-frequency band data, mid-frequency band data, and high-frequency band data;

[0013] determining a high-temperature slow-drift artifact of the measured target based on the body temperature data, and removing the high-temperature slow-drift artifact from the low-frequency band data to obtain low-frequency filtered data;

[0014] determining a convulsion artifact from the high-frequency band data according to the motion data, and attenuating the convulsion artifact to obtain high-frequency filtered data;

[0015] The low-frequency filtered data, the mid-frequency band data, and the high-frequency filtered data are reconstructed to obtain real EEG data.

[0016] In a possible implementation, the state information of the measured target further includes age; each age corresponds to a temperature sensitivity coefficient;

[0017] Determining a high temperature slow drift artifact of the measured target according to the body temperature data, and removing the high temperature slow drift artifact from the low frequency band data to obtain low frequency filtered data, including:

[0018] extracting slow drift frequency band data from the low frequency band data;

[0019] Determining a temperature compensation sequence according to the body temperature data and a temperature sensitivity coefficient corresponding to the measured target;

[0020] Using the temperature compensation sequence, the slow drift frequency band data is corrected to obtain temperature compensated frequency band data;

[0021] determining a high-temperature slow drift artifact of the measured target according to the temperature-compensated frequency band data and the slow-drift frequency band data;

[0022] The high-temperature slow drift artifact and the low-frequency band data are reconstructed to obtain low-frequency filtered data.

[0023] In a possible implementation, determining a seizure artifact from the high-frequency band data according to the motion data, and attenuating the seizure artifact to obtain high-frequency filtered data includes:

[0024] determining, based on the motion data, a convulsive period during which the detected target performs convulsive motion;

[0025] Perform blind source separation on high-frequency band data to obtain multiple first independent components;

[0026] The amplitude, zero-crossing rate and ridge verticality corresponding to the convulsion period in each first independent component were extracted respectively;

[0027] determining, based on the amplitude, the zero-crossing rate, and the ridge verticality, whether data corresponding to the convulsion period in each first independent component is a convulsion artifact;

[0028] Attenuate the part of each first independent component that is a convulsion artifact to obtain a second independent component;

[0029] All second independent components and first independent components of non-convulsive artifacts are reconstructed to obtain high-frequency filtered data.

[0030] In a possible implementation, the status information includes body temperature data and blood pressure data of the measured target within a preset time period;

[0031] Correcting the brain oxygen monitoring data according to the physiological state information and the real EEG data to obtain real brain oxygen data includes:

[0032] determining a first cerebral oxygen compensation sequence according to the body temperature data and a preset metabolic compensation coefficient;

[0033] determining a second cerebral oxygen compensation sequence according to the blood pressure data, a preset low-pressure compensation coefficient, and a preset high-pressure contraction coefficient;

[0034] Correcting the brain oxygen monitoring data according to the first brain oxygen compensation sequence and the second brain oxygen compensation sequence to obtain brain oxygen correction data;

[0035] detecting whether there is a sudden drop in brain oxygen saturation in the brain oxygen correction data;

[0036] If so, extracting slow wave data from the real EEG data and detecting whether the slow wave data corresponds to the sudden drop in brain oxygen saturation;

[0037] If they correspond, the brain oxygen correction data is determined as the real brain oxygen data.

[0038] In a possible implementation, the physiological state information includes age; each age group corresponds to a brain function monitoring model;

[0039] Determining the brain function level of the measured target based on the physiological state information, the real EEG data, the real brain oxygenation data, and the brain edema monitoring data, and providing a brain injury warning based on the brain function level, including:

[0040] determining a unit time according to a sampling frequency of the real EEG data, the real brain oxygenation data, and the brain edema monitoring data;

[0041] Performing feature extraction on the real EEG data, the real brain oxygenation data, and the brain edema monitoring data to obtain monitoring features within each unit time of the preset time length;

[0042] According to the preset time length, the monitoring features within each unit time are integrated to obtain a fusion matrix;

[0043] Obtaining a brain function level of the measured target according to the age, the monitoring characteristics, and a brain function monitoring model corresponding to the age of the measured target;

[0044] Based on the brain function level, a brain damage warning is performed.

[0045] In one possible implementation, before obtaining the brain function level of the measured target based on the age, the monitoring characteristics, and the brain function monitoring model corresponding to the age of the measured target, the method further includes:

[0046] Obtain the historical age, historical fusion matrix and corresponding brain function level of people in each age group;

[0047] Pre-training a preset machine learning model according to the historical age of each person, the historical fusion matrix, and the corresponding brain function level to obtain a pre-trained model;

[0048] The pre-training model is trained twice using the historical ages of people in each age group, the historical fusion matrix corresponding to the people, and the brain function level to obtain brain function monitoring models for each age group.

[0049] In a second aspect, an embodiment of the present application provides a non-invasive multimodal brain function monitoring device, comprising:

[0050] An acquisition module is used to obtain physiological status information of the measured target and EEG monitoring data, brain oxygen monitoring data and brain edema monitoring data of the measured target within a preset time period;

[0051] A first correction module is used to determine the EEG artifacts of the measured target according to the physiological state information, and remove the EEG artifacts from the EEG monitoring data to obtain real EEG data;

[0052] A second correction module is used to correct the brain oxygen monitoring data according to the physiological state information and the real EEG data to obtain real brain oxygen data;

[0053] The prediction module is used to determine the brain function level of the target being measured based on the physiological state information, the real EEG data, the real brain oxygen data and the brain edema monitoring data, and to provide a brain injury warning according to the brain function level.

[0054] In a third aspect, an embodiment of the present invention provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method in the first aspect or any possible implementation of the first aspect is implemented.

[0055] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method in the first aspect or any possible implementation of the first aspect.

[0056] In a fifth aspect, an embodiment of the present invention provides a computer program product, including a computer program, which, when executed by a processor, implements the method in the first aspect or any possible implementation of the first aspect.

[0057] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0058] The embodiment of the present application performs multimodal comprehensive monitoring of the target under test by acquiring the state information of the target under test and the EEG monitoring data, brain oxygen monitoring data and brain edema monitoring data of the target under test within a preset time period; considering that the infection status of the seriously infected child, i.e., the target under test, will have an impact on the monitoring data, the EEG infection artifacts of the target under test are determined according to the physiological state information, and the EEG infection artifacts are removed from the EEG monitoring data to obtain real EEG data; and the brain oxygen monitoring data are corrected according to the physiological state information and the real EEG data to obtain real brain oxygen data, and the artifacts are removed. The impact of the target's own high temperature and convulsions on the monitoring data is measured, and real and reliable EEG data and brain oxygen data are extracted; finally, based on the physiological state information, real EEG data, real brain oxygen data and brain edema monitoring data, the brain function level of the target is determined, and according to the brain function level, brain damage warning is carried out. It can fully consider the infection status of the target and multimodal monitoring data, accurately predict the brain function level of the target, judge whether brain damage is possible, and then provide timely warning and early identification when brain damage is possible, thereby improving the prognosis of seriously infected targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0060] Figure 1This is a flowchart of the implementation of the non-invasive multimodal brain function monitoring method provided in the embodiment of the present application;

[0061] Figure 2 Schematic diagram of the structure of the non-invasive multimodal brain function monitoring device provided in an embodiment of the present application;

[0062] Figure 3 Schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0063] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0064] Severe infection-associated encephalopathy (SAE) is an acute, diffuse brain dysfunction secondary to severe infection after a primary central nervous system infection has been excluded. Clinically, it can manifest as delirium, coma, epilepsy, or focal neuropathy.

[0065] The inventors have discovered that when monitoring children with severe infections, multiple monitoring devices are typically used to obtain data to monitor brain function. For example, electroencephalograms (EEGs) can be obtained using EEG monitors, regional cerebral oxygen saturation (rSO2) can be obtained using cerebral oximetry, and perturbation coefficients and intracranial pressure (ICP) can be obtained using cerebral edema monitors. However, children with severe infections may experience inflammatory reactions during monitoring. For example, severe infection may be accompanied by high fever or seizures. High fever can affect the EEG and cerebral oxygen demand, resulting in interference with EEG and rSO2 data. Furthermore, seizures can produce artifacts that interfere with the acquisition of true EEG data. This makes it difficult to directly monitor brain function in children with severe infections based on the acquired monitoring data.

[0066] In order to accurately monitor the brain function of children with severe infections and provide early identification and warning of brain damage, in the implementation of this application, considering that the infection status of severely infected children, i.e., the target being tested, will affect the monitoring data, the EEG monitoring data and brain oxygen monitoring data are corrected based on the physiological status information of the target being tested, and the EEG infection artifacts in the EEG monitoring data are removed to obtain true and reliable EEG data and brain oxygen data. Finally, the brain function of the target being tested is predicted based on the physiological status information, true EEG data, true brain oxygen data, and brain edema monitoring data, so as to provide timely warning and early identification when brain damage may occur, and assist relevant personnel in judging and identifying the target being tested.

[0067] In order to make the purpose, technical solutions and advantages of this application clearer, specific embodiments will be described below with reference to the accompanying drawings.

[0068] Figure 1 The following is a flowchart of the implementation of the non-invasive multimodal brain function monitoring method provided by the embodiment of the present invention:

[0069] Step 101 : Acquire physiological status information of a measured target and EEG monitoring data, brain oxygen monitoring data, and brain edema monitoring data of the measured target within a preset time period.

[0070] In this embodiment, the subject being measured can be a child or an adult. Through multimodal monitoring technology, relevant data of the subject being measured is comprehensively collected to provide a basis for subsequent brain function assessment.

[0071] Here, the physiological status information can be key physiological indicators related to severe infection, including dynamic physiological parameters such as the body temperature, limb movement data, and blood pressure data of the measured target, and can also include static physiological characteristics such as age.

[0072] EEG data can record the electrical activity of brain neurons, reflecting the real-time state of brain function. Brain oxygenation data can monitor brain tissue oxygen saturation, reflecting brain metabolism and blood perfusion. Brain edema monitoring data can include disturbance coefficients and intracranial pressure, reflecting the degree of brain tissue edema and changes in intracranial pressure.

[0073] Among them, EEG monitoring data can be obtained through an EEG monitor, brain oxygen monitoring data can be obtained through a brain oxygen monitor, and brain edema monitoring data can be obtained through a non-invasive brain edema dynamic monitor, so as to obtain relevant data of the target being measured through non-invasive multimodal monitoring.

[0074] Step 102: Determine the EEG artifacts of the target subject based on the physiological state information, and remove the EEG artifacts from the EEG monitoring data to obtain real EEG data.

[0075] In this embodiment, the subject's physiological state can affect the accuracy of EEG monitoring data, generating EEG artifacts. For example, a high fever can cause low-frequency signal drift, generating a high-temperature slow-drift artifact. When a subject is experiencing a convulsion, the EMG signals from limb movement can generate convulsion artifacts, leading to high-frequency interference.

[0076] Therefore, physiological state information can be used to determine EEG infection artifacts such as high temperature slow drift artifacts and convulsion artifacts corresponding to the physiological state information, so as to correct the EEG monitoring data, eliminate the interference of reactions such as high fever and convulsion on the EEG signals, and obtain real and reliable neuronal electrical activity data that is conducive to subsequent brain function analysis.

[0077] Step 103: Correct the brain oxygen monitoring data based on the physiological state information and the real EEG data to obtain the real brain oxygen data.

[0078] In this embodiment, the subject's physiological status can also affect the accuracy of brain oxygen monitoring data, leading to inaccurate brain oxygen demand monitoring. For example, the subject's body temperature can affect brain metabolic rate. Hyperthermia increases brain metabolic rate, increasing brain oxygen demand and localized oxygen uptake, leading to a false decrease in brain oxygen saturation. Therefore, physiological status information can be used to correct brain oxygen monitoring data.

[0079] At the same time, considering that there is a corresponding relationship between brain oxygen saturation and electroencephalogram, when brain oxygen saturation decreases, it will affect neurons and generate slow-wave EEG data. Therefore, the real EEG data obtained in the above steps can also be used to verify and correct the brain oxygen monitoring data to ensure that the brain oxygen data can truly reflect the oxygen supply and demand status of brain tissue.

[0080] Step 104, based on the physiological state information, real EEG data, real brain oxygen data and brain edema monitoring data, the brain function level of the target is determined, and brain damage warning is issued according to the brain function level.

[0081] In this embodiment, the brain function level of the target being measured is evaluated through the obtained real data, which can remove the interference of EEG artifacts and accurately predict and evaluate the brain function state of the target being measured.

[0082] Here, when the predicted brain function level is damaged or the degree of damage to the brain function level worsens, a corresponding warning can be triggered to prompt relevant personnel or medical staff to conduct clinical examinations and interventions to improve the prognosis of the target being tested.

[0083] The embodiment of the present application performs multimodal comprehensive monitoring of the target under test by acquiring the state information of the target under test and the EEG monitoring data, brain oxygen monitoring data and brain edema monitoring data of the target under test within a preset time period; considering that the infection status of the seriously infected child, i.e., the target under test, will have an impact on the monitoring data, the EEG infection artifacts of the target under test are determined according to the physiological state information, and the EEG infection artifacts are removed from the EEG monitoring data to obtain real EEG data; and the brain oxygen monitoring data are corrected according to the physiological state information and the real EEG data to obtain real brain oxygen data, and the artifacts are removed. The impact of the target's own high temperature and convulsions on the monitoring data is measured, and real and reliable EEG data and brain oxygen data are extracted; finally, based on the physiological state information, real EEG data, real brain oxygen data and brain edema monitoring data, the brain function level of the target is determined, and according to the brain function level, brain damage warning is carried out. It can fully consider the infection status of the target and multimodal monitoring data, accurately predict the brain function level of the target, judge whether brain damage is possible, and then provide timely warning and early identification when brain damage is possible, thereby improving the prognosis of seriously infected targets.

[0084] In some embodiments, the physiological state information includes body temperature data and movement data of the measured target within a preset time period; the EEG infection artifacts include high temperature slow drift artifacts and convulsion artifacts.

[0085] In this embodiment, when the subject experiences a high temperature or fever, a slow baseline drift occurs in the low-frequency band of the EEG monitoring data, i.e., a high-temperature slow drift artifact. When the subject experiences a convulsion, their limbs move rapidly and violently, introducing contagious data in the high-frequency band of the EEG monitoring data, i.e., a seizure artifact. Therefore, the required physiological state information, i.e., the subject's body temperature and movement data, can be determined based on the cause of the EEG contagious artifact.

[0086] Here, the temperature data can be obtained by a relevant temperature measuring device, such as a temperature sensor. The motion data can be obtained by an accelerometer. For example, an accelerometer can be set on each limb of the measured target to obtain the motion status of each limb of the measured target.

[0087] This embodiment determines the EEG infection artifacts of the target under test based on physiological state information, and removes the EEG infection artifacts from the EEG monitoring data to obtain real EEG data. The method can first decompose the EEG monitoring data to obtain low-frequency band data, medium-frequency band data and high-frequency band data; then determine the high-temperature slow drift artifacts of the target under test based on body temperature data, and remove the high-temperature slow drift artifacts from the low-frequency band data to obtain low-frequency filtered data; determine the convulsion artifacts from the high-frequency band data based on motion data, and attenuate the convulsion artifacts to obtain high-frequency filtered data; finally, reconstruct the low-frequency filtered data, medium-frequency band data and high-frequency filtered data to obtain real EEG data.

[0088] In the embodiment, considering that the frequency bands of the electroencephalogram infection artifacts such as high-temperature slow drift artifact and convulsion artifact are different, the electroencephalogram monitoring data can be first decomposed according to frequency bands to accurately eliminate the electroencephalogram infection artifacts and restore the real electroencephalogram signal.

[0089] Here, the electroencephalogram monitoring data can be decomposed into three parts according to the frequency range, including low-frequency band data, medium-frequency band data and high-frequency band data. The low-frequency band data is easily affected by the body temperature of the measured target and generates high-temperature slow drift artifact. The medium-frequency band data is relatively stable. The high-frequency band data is prone to contain electromyographic data of the violent movement of the measured target and generates convulsion artifact and the like, which covers the real epileptiform discharge.

[0090] The three parts of the electroencephalogram monitoring data can be divided according to the frequency bands of the electroencephalogram waves, for example, the frequency range of δ wave is 0 Hz-4 Hz, which generally exists in deep sleep stage or when the vital signs decrease; the frequency range of θ wave is 4 Hz-8 Hz, which generally exists in the initial stage of deep relaxation sleep; the frequency range of α wave is 8 Hz-12 Hz, which generally exists in the relaxed state of clear and quiet purpose or when the attention is concentrated; the frequency range of β wave is 12 Hz-30 Hz, which generally exists in mental tension, emotional excitement or excitement, and when waking up from sleep; the frequency range of γ wave is greater than 30 Hz. Therefore, δ wave can be taken as low-frequency band data, θ wave, α wave and β wave can be taken as medium-frequency band data, and γ wave can be taken as high-frequency band data.

[0091] Since high fever has a certain corresponding relationship with high-temperature slow drift artifact, the body temperature data of the measured target can be used to generate corresponding high-temperature slow drift artifact or electroencephalogram data equivalent to high-temperature slow drift artifact, so as to process the low-frequency band data, remove the high-temperature slow drift artifact therein, and obtain filtered real low-frequency band data, i.e., low-frequency filtered data.

[0092] Correspondingly, when the measured target has a convulsion attack, the corresponding electromyographic signal of the limb movement data of the measured target will interfere with the high-frequency band data and may cover the real epileptiform discharge data generated by the measured target. Therefore, by using the movement data of the measured target, the electroencephalogram signal corresponding to the convulsion attack in the high-frequency band data is identified, so as to attenuate the convulsion artifact, reduce the signal strength thereof, and reduce the interference, and obtain filtered real high-frequency band data, i.e., high-frequency filtered data.

[0093] Finally, by reconstructing the low-frequency filtered data and mid-frequency data after removing the high-temperature slow drift artifacts, and the high-frequency filtered data after removing the convulsion artifacts, they can be reintegrated into a complete EEG signal, that is, the real EEG data can be obtained, reducing the interference of the high fever and convulsion symptoms of the target itself on the EEG data, retaining the EEG signal that reflects the real activity of brain neurons, and providing a reliable data basis for subsequent brain function level judgment.

[0094] Optionally, the state information of the measured target also includes age; each age corresponds to a temperature sensitivity coefficient. Here, different ages of measured targets have different sensitivities to temperature, and accordingly, the degree of high-temperature slow-drift artifacts in the EEG signal will also vary. Therefore, the temperature sensitivity coefficient can be used to characterize the relationship between measured targets of different ages and the high-temperature slow-drift artifacts they produce. Age and temperature sensitivity coefficient are negatively correlated: the younger the age, the greater the temperature sensitivity coefficient, and the older the age, the smaller the temperature sensitivity coefficient.

[0095] This is primarily because subjects of different ages have different brain development and different brain water content, which results in different effects of high fever on EEG data. For example, a newborn's temperature regulation center is developed but immature at birth. Premature infants, in particular, have unstable body temperature regulation and a significantly narrower range of tolerance to environmental changes than adults. They are easily affected by environmental changes and develop abnormal body temperatures, which can affect metabolism and physiological function. Therefore, they are more sensitive to temperature and have the highest corresponding temperature sensitivity coefficient.

[0096] For example, the temperature sensitivity coefficient for newborns and premature infants can be 0.20 / °C, the temperature sensitivity coefficient for infants under 1 year old can be 0.18 / °C, the temperature sensitivity coefficient for toddlers between 1 and 3 years old can be 0.17 / °C, the temperature sensitivity coefficient for children between 3 and 5 years old can be 0.16 / °C, the temperature sensitivity coefficient for children between 6 and 12 years old can be 0.14 / °C, and the temperature sensitivity coefficient for subjects over 13 years old can be 0.12 / °C. It should be noted that the above is only an example to illustrate the relationship between age and temperature sensitivity coefficient and is not intended to be limiting.

[0097] This embodiment determines the high-temperature slow-drift artifact of the measured target based on the body temperature data, and removes the high-temperature slow-drift artifact from the low-frequency band data to obtain low-frequency filtered data. The method can be as follows: first, the slow-drift frequency band data in the low-frequency band data is extracted; then, a temperature compensation sequence is determined based on the body temperature data and the temperature sensitivity coefficient corresponding to the measured target; the slow-drift frequency band data is corrected using the temperature compensation sequence to obtain temperature-compensated frequency band data; then, the high-temperature slow-drift artifact of the measured target is determined based on the temperature-compensated frequency band data and the slow-drift frequency band data; finally, the high-temperature slow-drift artifact and the low-frequency band data are reconstructed to obtain low-frequency filtered data.

[0098] In this embodiment, high-temperature slow-drift artifacts are primarily concentrated in the lower frequency band, typically 0Hz-1Hz, manifesting as a slow, continuous baseline shift. The low-frequency data decomposed from EEG monitoring data typically ranges from 0Hz to 4Hz. Therefore, it is necessary to precisely locate the slow-drift components susceptible to high-temperature interference, narrow the scope of targeted processing, and avoid misprocessing other non-drifting signals in the low-frequency band (such as normal low-frequency EEG activity).

[0099] Extracting slow-drift frequency band data directly from raw EEG data requires processing the full frequency band, making it susceptible to interference from mid- and high-frequency signals. For example, the harmonics of high-frequency seizure artifacts can mix into the low-frequency range, blurring the boundary between artifacts and valid signals and increasing the risk of false extraction. Therefore, we first separate the low-frequency data into frequency bands, then extract the narrower slow-drift frequency band data from it, focusing on the low-frequency range where artifacts are concentrated and reducing cross-band interference.

[0100] This embodiment constructs a temperature compensation sequence for the target duration using the temperature sensitivity coefficient corresponding to the body temperature data and the age of the measured subject. This temperature compensation sequence can indicate the temperature compensation of the slow drift frequency band based on the age of the measured subject and the real-time body temperature data, thereby eliminating baseline drift caused by high temperature.

[0101] For example, the value at each moment in the temperature compensation sequence can be calculated using the following formula: Where, Indicates the time in the temperature compensation sequence of the measured target The value of Indicates the temperature sensitivity coefficient corresponding to the age of the measured target, Indicates that the target is at time Body temperature data, Indicates the standard body temperature corresponding to the age of the measured target. The standard body temperature can be 37°C, or can be determined based on the age of each measured target, or based on the average body temperature of the measured target in the recent normal state.

[0102] Then, the obtained temperature compensation sequence is applied to the slow drift frequency band data, which can remove the influence of high temperature slow drift from the original slow drift frequency band data and obtain temperature compensated frequency band data without baseline drift caused by high temperature. = the value of the temperature compensation sequence at the moment The value of × the time in the slow drift frequency band data The value of .

[0103] Based on this, by comparing the temperature compensation frequency band data and the slow drift frequency band data, the high temperature slow drift artifact of the target being measured can be determined. By reconstructing the low frequency band data to remove the high temperature slow drift artifact, low frequency filtered data can be obtained to ensure that the low frequency band data only reflects the physiological electrical activity of the brain itself.

[0104] Alternatively, reconstruction can be performed directly based on the temperature-compensated frequency band data, the slow-drift frequency band data, and the low-frequency band data. First, the slow-drift frequency band data is removed from the low-frequency band data, that is, the low-frequency portion containing high-temperature slow drift is removed. The temperature-compensated frequency band data is then added to the data. This allows only the corrected low-frequency portion to be included, thereby removing the high-temperature slow-drift artifact.

[0105] Furthermore, before using the temperature compensation sequence to correct the slow drift frequency band data and obtain the temperature compensated frequency band data, the temperature compensation sequence can be subjected to sliding average filtering to avoid sudden changes in the corrected EEG signal due to fluctuations in body temperature data.

[0106] Optionally, based on the motion data, the convulsion artifact is determined from the high-frequency band data, and the convulsion artifact is attenuated to obtain high-frequency filtered data, which can be: first, based on the motion data, the convulsion period of the target under test performing convulsive motion is determined; then, blind source separation is performed on the high-frequency band data to obtain multiple first independent components; the amplitude, zero-crossing rate and ridge verticality corresponding to the convulsion period in each first independent component are extracted respectively; based on the amplitude, zero-crossing rate and ridge verticality, it is determined whether the data corresponding to the convulsion period in each first independent component is a convulsion artifact; finally, the part of each first independent component that is a convulsion artifact is attenuated to obtain a second independent component; all second independent components and first independent components that are not convulsion artifacts are reconstructed to obtain high-frequency filtered data.

[0107] In this embodiment, considering the violent limb twitching during a seizure, the electrical signals generated by limb movement can be mixed into the high-frequency EEG data, such as 30-100 Hz, creating seizure artifacts that interfere with the assessment of true high-frequency EEG activity (such as rapid neuronal discharges). To accurately distinguish between seizure artifacts and high-frequency EEG activity, the motion data of the target is used to locate the seizure period. The artifacts are then identified and attenuated based on signal characteristics. Ultimately, a purified high-frequency EEG signal is obtained, ensuring its authenticity and providing a reliable basis for subsequent multimodal data-based assessment of brain function (such as identifying epileptic discharges and determining neuronal excitability).

[0108] Here, motion data can reflect the limb activity state of the target being measured. During a convulsion, rapid and violent limb movements, such as limb rigidity and convulsions, can cause significant abnormal fluctuations in motion data, such as a sudden increase in amplitude or a sudden change in frequency. At the same time, during a convulsion, there are symmetrical tonic-clonic movements with strong rhythmicity. During an epileptic seizure, the movements can be symmetrical or asymmetrical, with irregular rhythms. By analyzing this abnormal feature of the motion data, the time interval during which the convulsion occurred can be accurately located, that is, the convulsion period can be obtained. The high-frequency data is divided into a time range, and only the high-frequency data during the convulsion period is processed in a targeted manner to avoid erroneous interference with the normal high-frequency signals during the non-convulsion period.

[0109] High-frequency data is a mixture of true EEG high-frequency signals and seizure artifacts. By performing blind source separation on this data, the mixed signal can be decomposed into multiple independent signals, namely first independent components. Each first independent component represents an independent signal source, which may be the EEG itself, myoelectric artifacts, or myoelectric interference. Blind source separation can be achieved through independent component analysis.

[0110] Because the waveform characteristics of seizure artifacts differ significantly from those of true EEG high-frequency signals, they can be distinguished by amplitude, zero-crossing rate, and ridge verticality. Seizure artifacts are generated by limb movements, and their signal strength and amplitude are usually much greater than true EEG high-frequency signals. The zero-crossing rate refers to the number of times a signal crosses the baseline or zero value per unit time. Seizure artifact waveforms are generally more chaotic, and their zero-crossing rate is significantly higher than the regular true EEG high-frequency signals. Ridge verticality refers to the steepness of the rising and falling edges of the signal waveform. Because seizures are characterized by sudden movements, the baseline of the seizure artifact waveform is steeper and more vertical, while the waveform of the true EEG high-frequency signal changes relatively smoothly.

[0111] By comparing the amplitude, zero-crossing rate and ridge verticality corresponding to the convulsion period in each independent component, the type of each independent component can be determined, that is, whether it is a convulsion artifact.

[0112] Specifically, this can be determined using preset thresholds. For example, if the signal amplitude exceeds a preset amplitude, it indicates that the signal strength and amplitude of the independent component are abnormal, possibly indicating a seizure artifact. If the zero-crossing rate exceeds a preset zero-crossing rate, it indicates that the independent component frequently crosses the baseline, possibly indicating a seizure artifact. If the ridge verticality exceeds a preset verticality, it indicates that the waveform of the independent component is steep, possibly indicating a seizure artifact. If the independent component is determined to be a seizure artifact in all three cases, the independent component can be determined to be a seizure artifact.

[0113] Alternatively, by extracting features of amplitude, zero-crossing rate, and ridge verticality, and using these features and a convulsion artifact classification model, it can be determined whether each independent component is a convulsion artifact. The convulsion artifact classification model can be trained using features of convulsion artifacts and real high-frequency EEG data, as well as the corresponding amplitude, zero-crossing rate, and ridge verticality of the convulsion artifacts and real high-frequency EEG data.

[0114] In this embodiment, attenuation processing is performed on the signal during the convulsion period of the first independent component determined to be the convulsion artifact, which can reduce the amplitude of the convulsion artifact, eliminate artifact interference, avoid destroying the real EEG signal that may be contained in this component during the non-convulsion period, and accurately obtain the second independent component.

[0115] Finally, by reconstructing the second independent component and the first independent component of non-convulsive artifacts, the attenuated components of convulsive artifacts and the real EEG components that are not judged as convulsive artifacts can be recombined to restore the complete high-frequency signal, that is, high-frequency filtered data, removing the interference of convulsive artifacts, and obtaining a signal reflecting the real EEG high-frequency activity.

[0116] In some embodiments, the status information includes the body temperature data and blood pressure data of the target being measured within a preset time period. Here, brain oxygen monitoring data, such as brain oxygen saturation, reflects the balance of brain oxygen supply and demand. The body temperature of the target being measured will affect the brain metabolic rate. For example, when the brain oxygen consumption is high, it will cause a false decrease in brain oxygen saturation, thereby interfering with brain oxygen saturation. Blood pressure fluctuations will affect perfusion through automatic regulation of cerebral blood flow. For example, hypotension will cause insufficient brain perfusion and a true decrease in brain oxygen saturation; hypertension will cause cerebral vasoconstriction and a false increase in brain oxygen saturation. Therefore, the required physiological state information, that is, the body temperature data and blood pressure data of the target being measured, can be determined based on the cause of the impact of brain oxygen saturation.

[0117] In this embodiment, the brain oxygen monitoring data is corrected based on the physiological state information and the real EEG data to obtain the real brain oxygen data, which can be:

[0118] Step 1: Determine the first brain oxygen compensation sequence based on body temperature data and a preset metabolic compensation coefficient.

[0119] In this embodiment, the metabolic compensation coefficient can also be determined based on the age of the subject being measured. Subjects of different ages correspond to different metabolic compensation coefficients. This is primarily due to the fact that the development of brain metabolism and vascular regulation varies among subjects of different ages, leading to different effects of body temperature on brain oxygen saturation. Similarly, the metabolic compensation coefficient is negatively correlated with the subject's age: the younger the subject, the greater the metabolic compensation coefficient; the older the subject, the smaller the metabolic compensation coefficient.

[0120] When a subject experiences a high fever, the metabolic rate of brain tissue increases, leading to increased oxygen consumption and potentially causing deviations in cerebral oxygen saturation measurements. For example, under the same blood flow, high fever can increase actual oxygen demand, resulting in the appearance of false hypoxia. Using body temperature data and a metabolic compensation coefficient, we can generate a metabolic interference correction value that matches the subject's body temperature. This is known as the first cerebral oxygen compensation sequence, which corrects for the effects of body temperature on cerebral oxygen monitoring data.

[0121] Step 2: Determine the second cerebral oxygen compensation sequence based on the blood pressure data, the preset low-pressure compensation coefficient and the high-pressure contraction coefficient.

[0122] Here, the low-pressure compensation coefficient and the high-pressure contraction coefficient are determined for low-pressure and high-pressure blood pressure conditions, respectively, to evaluate the degree of influence of low and high pressure on cerebral oxygen saturation.

[0123] Blood pressure data is used to determine whether each moment within a preset time period is within the low-pressure compensation or high-pressure contraction range. If within the low-pressure compensation range, the corresponding brain oxygen compensation value is determined using the blood pressure data and the low-pressure compensation coefficient; if within the high-pressure contraction range, the corresponding brain oxygen compensation value is determined using the blood pressure data and the high-pressure contraction coefficient. Combining the brain oxygen compensation values ​​at each moment, a corresponding second brain oxygen compensation sequence can be generated to correct for the influence of blood pressure on the brain oxygen detection data.

[0124] Step three: correcting the brain oxygen monitoring data according to the first brain oxygen compensation sequence and the second brain oxygen compensation sequence to obtain brain oxygen correction data.

[0125] In this embodiment, since the original brain oxygen monitoring data is interfered with by both body temperature-driven metabolism and blood pressure-driven blood perfusion, it is necessary to perform comprehensive correction using the first brain oxygen compensation sequence and the second brain oxygen compensation sequence to remove the physiological interference caused by body temperature and blood pressure and obtain brain oxygen saturation data that is close to the actual value.

[0126] Optionally, the first brain oxygen compensation sequence and the second brain oxygen compensation sequence can be directly multiplied with the brain oxygen monitoring data in chronological order to obtain brain oxygen correction data. = value of brain oxygen monitoring data at the moment The value of × the time in the first brain oxygen compensation sequence The value of × the moment in the second brain oxygen compensation sequence The value of .

[0127] Step 4: Detect whether there is a sudden drop in brain oxygen saturation in the brain oxygen correction data.

[0128] Step 5: If it exists, extract slow wave data from the real EEG data and detect whether the slow wave data corresponds to the sudden drop in brain oxygen saturation; if so, determine the brain oxygen correction data as the real brain oxygen data.

[0129] In this embodiment, a sudden drop in brain oxygen saturation may be caused by severe brain dysfunction such as cerebral ischemia or hypoxia, or by a probe falling off. Therefore, when a sudden drop in brain oxygen saturation is present in the corrected brain oxygen data, real EEG data is also used for verification and judgment.

[0130] Slow waves are diffuse, high-amplitude delta waves in EEG data. They typically appear after true hypoxia. For example, slow waves will appear in EEG data 8 seconds after a sudden drop in brain oxygen saturation. Therefore, slow waves in EEG data can be used to verify a sudden drop in brain oxygen saturation.

[0131] Here, if the slow wave data in the EEG data corresponds to a sudden drop in brain oxygen saturation, it means that the sudden drop in brain oxygen saturation is real brain oxygen data and no correction is required.

[0132] If the slow wave data in the EEG data does not correspond to a sudden drop in brain oxygen saturation, for example, brain oxygen saturation drops suddenly but there is no corresponding slow wave data, or the time interval between the slow wave data and the sudden drop in brain oxygen saturation is greater than the preset time threshold, it indicates that there is a problem with the sudden drop in brain oxygen saturation, which may be caused by the probe falling off or other interference. An alarm can be issued so that relevant personnel can conduct an inspection.

[0133] The above article mainly describes how to correct EEG and cerebral oxygen monitoring data. Considering that critically ill individuals of different age groups have different data characteristics, the following will continue to explain how to determine the brain function of the measured subject using real EEG data, real cerebral oxygen data, and brain edema monitoring data.

[0134] Here, brain edema monitoring data can include the disturbance coefficient of brain edema and intracranial pressure. A high disturbance coefficient indicates the presence of hematoma or tumors, while a low disturbance coefficient indicates the presence of edema and fluid accumulation. Intracranial pressure is also an important indicator of brain function; increased intracranial pressure can, to a certain extent, indicate worsening brain damage. Therefore, real EEG data, real brain oxygenation data, and brain edema monitoring data can be used simultaneously to determine brain function.

[0135] In some embodiments, the physiological state information includes age; each age group corresponds to a brain function monitoring model. Here, considering that the brain development maturity and pathological response to severe infection of the subjects in different age groups are significantly different, the performance of the EEG data, brain oxygen data and brain edema monitoring data of the subjects in different age groups is also different. Therefore, a brain function monitoring model is set up for each age group to ensure that the model is targeted at the brain function changes of targets of a specific age, avoid assessment bias caused by age differences, and lay the foundation for accurate judgment of brain function level.

[0136] In the EEG data, 2-month-old to 12-month-old infants, in a state of wakefulness, the widespread is 2-3.5 Hz at 2 months, 4 Hz at 3 months, 5 Hz at 5 months, and the occipital region frequency reaches 6-7 Hz, and occasionally reaches 8 Hz at 12 months. 1-year-old to 3-year-old children, the posterior head basic rhythm gradually evolves from faster theta activity to slow alpha rhythm at 1 year old, the occipital region rhythm is about 6-7 Hz at 2 years old, and the occipital region rhythm is about 7-8 Hz at 3 years old. 3-year-old to 5-year-old children, the occipital region rhythm is about 8 Hz, the voltage is high, and the posterior head alpha rhythm often mixes with varying amounts of slow wave activity. 6-year-old to 12-year-old children, the occipital region alpha rhythm continues to gradually increase, the occipital region rhythm is about 9 Hz at 7 years old, and the occipital region rhythm is stable at 10 Hz at 10 years old and above. 13-year-old to 20-year-old adolescents, the occipital region rhythm is between 8-12 Hz, and the average is about 10 Hz.

[0137] In the cerebral oxygen data, the cerebral oxygen saturation of 0-year-old to 1-year-old infants is about 70.41%±4.66%, the cerebral oxygen saturation of 1-year-old to 3-year-old children is about 72.43%±3.81%, the cerebral oxygen saturation of 3-year-old to 5-year-old children is about 70.77%±3.27%, the cerebral oxygen saturation of 6-year-old to 18-year-old children is about 70.62%±2.20%, and the cerebral oxygen saturation of 18-year-old to 65-year-old adults is about 69.76%±6.02%.

[0138] In the perturbation coefficient, the perturbation coefficient of 0-year-old to 1-year-old infants is 60±14, the perturbation coefficient of 1-year-old to 3-year-old children is 92±18, the perturbation coefficient of 3-year-old to 5-year-old children is 112±18, and the perturbation coefficient of 6-year-old to 18-year-old children is 135±18.

[0139] Based on this, it can be considered to set a brain function monitoring model for 0-year-old to 1-year-old infants, 1-year-old to 3-year-old children, 3-year-old to 5-year-old children, 6-year-old to 12-year-old children, 13-year-old to 20-year-old adolescents, adults over 18 years old, and the elderly over 65 years old. The age division here is only for illustration, and the upper and lower limits of each interval can be adjusted as needed, and new age intervals can be added to make each brain function detection model more suitable for the corresponding age interval of the measured target.

[0140] This embodiment determines the brain function level of the target being measured based on physiological state information, real EEG data, real brain oxygen data, and brain edema monitoring data, and issues a brain injury warning based on the brain function level. The method may be as follows: first, the unit time is determined based on the sampling frequency of the real EEG data, real brain oxygen data, and brain edema monitoring data; then, feature extraction is performed on the real EEG data, real brain oxygen data, and brain edema monitoring data respectively to obtain monitoring features within each unit time within a preset time length; according to the preset time length, the monitoring features within each unit time are fused to obtain a fusion matrix; the brain function level of the target being measured is obtained based on a brain function monitoring model corresponding to age, monitoring features, and the age of the target being measured; finally, a brain injury warning is issued based on the brain function level.

[0141] In this embodiment, considering the differences in sampling frequencies of EEG data, brain oxygen data, and brain edema monitoring data, such as the fact that EEG sampling frequencies are generally higher than those of brain oxygen and brain edema data, a unit time is selected to align the three types of data in the time dimension, for example, a unit time of 1 minute. Alternatively, the sampling frequencies of the three types of data can be determined to ensure that all three types of data are fully recorded in each unit time, avoiding feature misalignment due to inconsistent time dimensions and ensuring temporal synchronization of multimodal data.

[0142] The monitoring features of each unit time are extracted separately to reflect the brain function state within that unit time. Among them, the characteristics of real EEG data can include the power proportion of low-frequency, medium-frequency and high-frequency signals, the frequency of occurrence of EEG rhythms, amplitude changes, etc., to reflect the state of neuronal electrical activity. The characteristics of real brain oxygen data can include the mean value of brain oxygen saturation, fluctuation amplitude, number of sudden drops, etc., to reflect the balance of oxygen supply and demand in brain tissue. The characteristics of brain edema monitoring data can include the peak value of intracranial pressure, the mean value of disturbance coefficient, the rate of progression of brain edema, etc., to reflect the degree of brain tissue edema and changes in intracranial pressure.

[0143] In addition, features can also be extracted from real EEG data, real brain oxygen data, and brain edema monitoring data through the feature extraction module for subsequent corresponding evaluation.

[0144] Here, the monitoring features of all unit times within the preset time length are integrated in chronological order to form a fusion matrix, which can integrate the multimodal information of EEG data, brain oxygen data and brain edema data to comprehensively reflect different aspects of brain function. At the same time, it retains the time series change information and reflects the dynamic trend of brain function status, such as the continuous decline of brain oxygen and the gradual slowing of EEG rhythm, providing more comprehensive input for the model.

[0145] Among them, considering that each age group corresponds to a brain function monitoring model, an age group includes multiple ages, and the situations of people of each age are different, therefore, by inputting the age and fusion matrix into the brain function monitoring model corresponding to the age of the target being measured, the age of the target being measured can be further considered to accurately obtain the brain function level of the target being measured, such as normal, mild abnormality, moderate damage, severe damage, damage maintenance and damage aggravation, etc.

[0146] In addition, when the brain function level is mild abnormality, moderate damage, severe damage and worsening damage, brain damage warning will be issued to promptly remind relevant personnel and assist medical staff to promptly identify the risk of brain damage in children with severe infections, achieve early intervention and improve prognosis.

[0147] Optionally, before obtaining the brain function level of the target being measured based on the age, monitoring characteristics and the brain function monitoring model corresponding to the age of the target being measured, this embodiment may also: obtain the historical age, historical fusion matrix and corresponding brain function level of people in each age group; then pre-train the preset machine learning model based on the historical age, historical fusion matrix and corresponding brain function level of each person to obtain a pre-trained model; finally, use the historical age of people in each age group, the historical fusion matrix corresponding to the people and the brain function level to perform secondary training on the pre-trained model to obtain brain function monitoring models for each age group.

[0148] In this embodiment, the historical fusion matrix is ​​derived from multimodal EEG, cerebral oxygenation, and cerebral edema monitoring data collected from individuals of various age groups over a preset period of time. Brain function levels can be determined based on clinical diagnosis or standardized assessments (such as brain function rating scales) and serve as labels for model training.

[0149] Here, a machine model is first pre-trained based on the historical age, historical fusion matrix and corresponding brain function level of people in each age group to learn the common correlation between multimodal brain function characteristics and brain function levels across age groups.

[0150] Considering the varying characteristics of multimodal monitoring data across age groups, and to improve the model's predictive accuracy, the pre-trained model was retrained separately using the historical age, historical fusion matrix, and corresponding brain function levels of individuals from different age groups. This further studied the associations between age-specific characteristics and brain function levels, forming a brain function monitoring model applicable to each age group. This approach allowed the model to focus on age-specific characteristics based on common patterns, eliminating interference from age-specific brain development differences and ensuring that each model accurately judged the brain function level of its respective age group.

[0151] Among them, the machine learning model can be a deep learning model, a random forest model, or other models that have the ability to process multimodal time series features.

[0152] Optionally, considering that subjects with severe infections generally require long-term monitoring, the age of the target being measured, the historical fusion matrix and the corresponding brain function level can be used to train a brain function monitoring model corresponding to the age group of the target being measured, so as to form a brain function monitoring model corresponding to the target being measured, so as to make the obtained brain function monitoring model more targeted and improve the precision and accuracy of the model prediction.

[0153] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0154] The following are device embodiments of the present application. For details not fully described therein, please refer to the corresponding method embodiments described above.

[0155] Figure 2 The following is a schematic diagram of the structure of a non-invasive multimodal brain function monitoring device provided in an embodiment of the present application. For ease of explanation, only the parts related to the embodiment of the present application are shown, which are detailed as follows:

[0156] like Figure 2 As shown, the non-invasive multimodal brain function monitoring device 20 includes:

[0157] An acquisition module 21 is used to acquire physiological status information of the measured target and EEG monitoring data, brain oxygen monitoring data, and brain edema monitoring data of the measured target within a preset time period;

[0158] The first correction module 22 is used to determine the EEG artifacts of the target according to the physiological state information, and remove the EEG artifacts from the EEG monitoring data to obtain the real EEG data;

[0159] The second correction module 23 is used to correct the brain oxygen monitoring data according to the physiological state information and the real EEG data to obtain the real brain oxygen data;

[0160] The prediction module 24 is used to determine the brain function level of the target being tested based on physiological state information, real EEG data, real brain oxygen data and brain edema monitoring data, and to provide brain injury warning according to the brain function level.

[0161] In a possible implementation, the physiological state information includes body temperature data and movement data of the measured target within a preset time period; the EEG infection artifacts include high temperature slow drift artifacts and convulsion artifacts;

[0162] The first correction module 22 is specifically used for:

[0163] decompose the brain electrical monitoring data to obtain low-frequency data, medium-frequency data and high-frequency data;

[0164] According to the body temperature data, determine the high temperature slow drift artifact of the measured target, and remove the high temperature slow drift artifact from the low-frequency data to obtain low-frequency filtered data;

[0165] According to the motion data, determine the convulsion artifact from the high-frequency data, and attenuate the convulsion artifact to obtain high-frequency filtered data;

[0166] Reconstruct the low-frequency filtered data, the medium-frequency data and the high-frequency filtered data to obtain real brain electrical data.

[0167] In a possible implementation, the state information of the measured target further includes age; each age corresponds to a temperature sensitivity coefficient;

[0168] The first correction module 22 is specifically configured to:

[0169] Extract slow drift frequency band data in the low-frequency data;

[0170] According to the body temperature data and the temperature sensitivity coefficient corresponding to the measured target, determine a temperature compensation sequence;

[0171] Correct the slow drift frequency band data by using the temperature compensation sequence to obtain temperature compensation frequency band data;

[0172] According to the temperature compensation frequency band data and the slow drift frequency band data, determine the high temperature slow drift artifact of the measured target;

[0173] Reconstruct the high temperature slow drift artifact and the low-frequency data to obtain low-frequency filtered data.

[0174] In a possible implementation, the first correction module 22 is specifically configured to:

[0175] According to the motion data, determine a convulsion period in which the measured target performs convulsion motion;

[0176] Perform blind source separation on the high-frequency data to obtain a plurality of first independent components;

[0177] Respectively extract the amplitude, zero-crossing rate and spine perpendicularity of the convulsion period in each first independent component;

[0178] According to the amplitude, zero-crossing rate and spine perpendicularity, determine whether the data corresponding to the convulsion period in each first independent component is a convulsion artifact;

[0179] Attenuate the part of each first independent component that is a convulsion artifact to obtain a second independent component;

[0180] All second independent components and first independent components of non-convulsive artifacts are reconstructed to obtain high-frequency filtered data.

[0181] In one possible implementation, the status information includes body temperature data and blood pressure data of the measured target within a preset time period;

[0182] The second correction module 23 is specifically used for:

[0183] Determine the first brain oxygen compensation sequence based on body temperature data and a preset metabolic compensation coefficient;

[0184] Determine the second cerebral oxygen compensation sequence based on blood pressure data, preset low-pressure compensation coefficient and high-pressure contraction coefficient;

[0185] Correcting the brain oxygen monitoring data according to the first brain oxygen compensation sequence and the second brain oxygen compensation sequence to obtain brain oxygen correction data;

[0186] Detect whether there is a sudden drop in brain oxygen saturation in brain oxygen correction data;

[0187] If it exists, slow wave data is extracted from the real EEG data, and whether the slow wave data corresponds to a sudden drop in brain oxygen saturation is detected;

[0188] If they correspond, the brain oxygen correction data is determined as the real brain oxygen data.

[0189] In a possible implementation, the physiological state information includes age; each age group corresponds to a brain function monitoring model;

[0190] The prediction module 24 is specifically used for:

[0191] Determine the unit time based on the sampling frequency of real EEG data, real brain oxygen data, and brain edema monitoring data;

[0192] Feature extraction is performed on real EEG data, real brain oxygenation data, and brain edema monitoring data to obtain monitoring features for each unit time within the preset duration;

[0193] According to the preset time length, the monitoring features in each unit time are integrated to obtain the fusion matrix;

[0194] Obtaining the brain function level of the target according to the age, monitoring characteristics and the brain function monitoring model corresponding to the target's age;

[0195] Provide early warning of brain damage based on brain function level.

[0196] In a possible implementation, the prediction module 24 is further configured to:

[0197] Obtain the historical age, historical fusion matrix and corresponding brain function level of people in each age group;

[0198] According to the historical age, the historical fusion matrix and the corresponding brain function level of each personnel, a preset machine learning model is pre-trained to obtain a pre-trained model;

[0199] According to the historical age, the historical fusion matrix and the corresponding brain function level of each personnel, a preset machine learning model is pre-trained to obtain a pre-trained model;

[0200] Figure 3 is a schematic diagram of an electronic device provided by an embodiment of the present application. As shown in Figure 3 The electronic device 30 of this embodiment includes a processor 31 and a memory 32. The memory 32 stores a computer program 33. The processor 31 implements the steps in each of the method embodiments described above when executing the computer program 33. Alternatively, the processor 31 implements the functions of each module / unit in each of the device embodiments described above when executing the computer program 33.

[0201] For example, the computer program 33 can be divided into one or more modules / units, which are stored in the memory 32 and executed by the processor 31 to complete the present application. The one or more modules / units can be a series of computer program instruction segments that can complete a specific function, which are used to describe the execution process of the computer program 33 in the electronic device 30.

[0202] The electronic device 30 can include, but is not limited to, the processor 31 and the memory 32. Those skilled in the art can understand that Figure 3 The electronic device 30 is only an example and does not constitute a limitation on the electronic device 30, and can include more or fewer components than the diagram, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus, etc.

[0203] The processor 31 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0204] The memory 32 can be an internal storage unit of the electronic device 30, such as a hard disk or a memory of the electronic device 30. The memory 32 can also be an external storage device of the electronic device 30, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device 30. Further, the memory 32 can include both the internal storage unit and the external storage device of the electronic device 30. The memory 32 is used to store computer programs and other programs and data required by the electronic device. The memory 32 can also be used to temporarily store data that has been output or will be output.

[0205] For the convenience and brevity of description, only the above-mentioned division of functional modules / units is exemplified, and in actual application, the above-mentioned functions can be completed by different functional modules / units according to needs. The above-mentioned modules / units can be realized in the form of hardware, software or a combination of hardware and software.

[0206] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program. When the computer program is executed by a processor, the method in each method embodiment described above is realized.

[0207] The embodiment of the present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the method in each method embodiment described above is realized.

[0208] The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms. The computer readable medium can include any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, Read-Only Memory (ROM), Random Access Memory (RAM), electric carrier wave signal, telecommunication signal and software distribution medium, etc.

[0209] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments. If there is no special description and no logical conflict, the terms and / or descriptions of different embodiments are consistent and can be mutually referenced. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0210] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A non-invasive multimodal brain function monitoring method, characterized in that: include: Acquiring physiological status information of the measured target and EEG monitoring data, brain oxygen monitoring data, and brain edema monitoring data of the measured target within a preset time period; Determining the EEG artifacts of the target subject based on the physiological state information, and removing the EEG artifacts from the EEG monitoring data to obtain true EEG data; Correcting the brain oxygen monitoring data according to the physiological state information and the real EEG data to obtain real brain oxygen data; Determining the brain function level of the measured target based on the physiological state information, the real EEG data, the real brain oxygenation data, and the brain edema monitoring data, and providing a brain injury warning according to the brain function level; The physiological state information includes the body temperature data and movement data of the measured target within a preset time period; the EEG infection artifacts include high temperature slow drift artifacts and convulsion artifacts; Determining the EEG artifacts of the target according to the physiological state information, and removing the EEG artifacts from the EEG monitoring data to obtain real EEG data, including: Decomposing the EEG monitoring data to obtain low-frequency band data, mid-frequency band data, and high-frequency band data; determining a high-temperature slow-drift artifact of the measured target based on the body temperature data, and removing the high-temperature slow-drift artifact from the low-frequency band data to obtain low-frequency filtered data; determining a convulsion artifact from the high-frequency band data according to the motion data, and attenuating the convulsion artifact to obtain high-frequency filtered data; The low-frequency filtered data, the mid-frequency band data, and the high-frequency filtered data are reconstructed to obtain real EEG data.

2. The non-invasive multimodal brain function monitoring method according to claim 1, characterized in that: The state information of the measured target also includes age; each age corresponds to a temperature sensitivity coefficient; Determining a high temperature slow drift artifact of the measured target according to the body temperature data, and removing the high temperature slow drift artifact from the low frequency band data to obtain low frequency filtered data, including: extracting slow drift frequency band data from the low frequency band data; Determining a temperature compensation sequence according to the body temperature data and a temperature sensitivity coefficient corresponding to the measured target; Using the temperature compensation sequence, the slow drift frequency band data is corrected to obtain temperature compensated frequency band data; determining a high-temperature slow drift artifact of the measured target according to the temperature-compensated frequency band data and the slow-drift frequency band data; The high-temperature slow drift artifact and the low-frequency band data are reconstructed to obtain low-frequency filtered data.

3. The non-invasive multimodal brain function monitoring method according to claim 1, characterized in that: Determining a convulsion artifact from the high-frequency band data according to the motion data, and attenuating the convulsion artifact to obtain high-frequency filtered data, including: determining, based on the motion data, a convulsive period during which the detected target performs convulsive motion; Perform blind source separation on high-frequency band data to obtain multiple first independent components; The amplitude, zero-crossing rate and ridge verticality corresponding to the convulsion period in each first independent component were extracted respectively; determining, based on the amplitude, the zero-crossing rate, and the ridge verticality, whether data corresponding to the convulsion period in each first independent component is a convulsion artifact; Attenuate the part of each first independent component that is a convulsion artifact to obtain a second independent component; All second independent components and first independent components of non-convulsive artifacts are reconstructed to obtain high-frequency filtered data.

4. The non-invasive multimodal brain function monitoring method according to claim 1, characterized in that: The status information includes the body temperature data and blood pressure data of the measured target within a preset time period; Correcting the brain oxygen monitoring data according to the physiological state information and the real EEG data to obtain real brain oxygen data includes: determining a first cerebral oxygen compensation sequence according to the body temperature data and a preset metabolic compensation coefficient; determining a second cerebral oxygen compensation sequence according to the blood pressure data, a preset low-pressure compensation coefficient, and a preset high-pressure contraction coefficient; Correcting the brain oxygen monitoring data according to the first brain oxygen compensation sequence and the second brain oxygen compensation sequence to obtain brain oxygen correction data; detecting whether there is a sudden drop in brain oxygen saturation in the brain oxygen correction data; If so, extracting slow wave data from the real EEG data and detecting whether the slow wave data corresponds to the sudden drop in brain oxygen saturation; If they correspond, the brain oxygen correction data is determined as the real brain oxygen data.

5. The non-invasive multimodal brain function monitoring method according to any one of claims 1 to 4, characterized in that: The physiological status information includes age; each age group corresponds to a brain function monitoring model; Determining the brain function level of the measured target based on the physiological state information, the real EEG data, the real brain oxygenation data, and the brain edema monitoring data, and providing a brain injury warning based on the brain function level, including: determining a unit time according to a sampling frequency of the real EEG data, the real brain oxygenation data, and the brain edema monitoring data; Performing feature extraction on the real EEG data, the real brain oxygenation data, and the brain edema monitoring data to obtain monitoring features within each unit time of the preset time length; According to the preset time length, the monitoring features within each unit time are integrated to obtain a fusion matrix; Obtaining a brain function level of the measured target according to the age, the monitoring characteristics, and a brain function monitoring model corresponding to the age of the measured target; Based on the brain function level, a brain damage warning is performed.

6. The non-invasive multimodal brain function monitoring method according to claim 5, characterized in that: Before obtaining the brain function level of the measured target according to the age, the monitoring characteristics, and the brain function monitoring model corresponding to the age of the measured target, the method further includes: Obtain the historical age, historical fusion matrix and corresponding brain function level of people in each age group; Pre-training a preset machine learning model according to the historical age of each person, the historical fusion matrix, and the corresponding brain function level to obtain a pre-trained model; The pre-training model is trained twice using the historical age of people in each age group, the historical fusion matrix corresponding to the people, and the brain function level to obtain brain function monitoring models for each age group.

7. A non-invasive multimodal brain function monitoring device, characterized in that: include: An acquisition module is used to obtain physiological status information of the measured target and EEG monitoring data, brain oxygen monitoring data and brain edema monitoring data of the measured target within a preset time period; A first correction module is used to determine the EEG artifacts of the measured target according to the physiological state information, and remove the EEG artifacts from the EEG monitoring data to obtain real EEG data; A second correction module is used to correct the brain oxygen monitoring data according to the physiological state information and the real EEG data to obtain real brain oxygen data; a prediction module, configured to determine the brain function level of the measured target based on the physiological state information, the real EEG data, the real brain oxygenation data, and the brain edema monitoring data, and to provide a brain injury warning based on the brain function level; The physiological state information includes the body temperature data and movement data of the measured target within a preset time period; the EEG infection artifacts include high temperature slow drift artifacts and convulsion artifacts; The first correction module is specifically used to: Decomposing the EEG monitoring data to obtain low-frequency band data, mid-frequency band data, and high-frequency band data; determining a high-temperature slow-drift artifact of the measured target based on the body temperature data, and removing the high-temperature slow-drift artifact from the low-frequency band data to obtain low-frequency filtered data; determining a convulsion artifact from the high-frequency band data according to the motion data, and attenuating the convulsion artifact to obtain high-frequency filtered data; The low-frequency filtered data, the mid-frequency band data, and the high-frequency filtered data are reconstructed to obtain real EEG data.

8. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images