Artificial intelligence applied system for evaluating influence of excessive anesthetic on brain function

Abstract

The invention relates to the technical field of anesthesia management, in particular to an anesthesia medicine excess brain function influence assessment system applying artificial intelligence, which comprises a data acquisition module, a data preprocessing module, a model training module, a real-time assessment module, a risk early warning module, a treatment suggestion module, a system interaction module and a remote monitoring module. Through the combination of a convolutional neural network and a long-short-term memory network, spatial-temporal characteristics of brain wave and functional magnetic resonance imaging data are extracted, a brain function thermodynamic diagram is generated in real time, and the influence of anesthetic drugs is analyzed; the system is combined with a reinforcement learning algorithm to dynamically adjust the anesthetic dosage, a risk early warning module dynamically adjusts a threshold value through physiological parameters and tolerance of a patient, an expert team can provide medical assistance in real time through a remote monitoring module, and brain function safety of the patient in the operation or anesthesia process is ensured; according to the invention, the safety, accuracy and individuation level of anesthesia management are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of anesthesia management, and in particular to an artificial intelligence-based system for evaluating the impact of an anesthetic drug overdose on brain function. Background Art

[0002] Anesthetics are widely used in modern medicine, particularly in surgical procedures and intensive care settings, to help patients relieve pain and maintain unconsciousness. However, while anesthetics exert their effects, they can also potentially affect a patient's brain function. In particular, an overdose of anesthetics can cause temporary or permanent damage to brain function. Monitoring and evaluating the effects of anesthetics on a patient's brain function, especially in real-time monitoring and analysis of anesthetic overdose, has become a critical issue in the medical field.

[0003] Existing anesthesia monitoring technologies primarily focus on monitoring vital signs, such as heart rate, blood pressure, and blood oxygen saturation, but remain deficient in assessing the specific effects of anesthetics on brain function. These traditional monitoring systems struggle to process and analyze brainwaves and functional brain imaging data in real time, resulting in delayed risk identification and response to anesthetic overdose. Furthermore, existing technologies are significantly limited in providing personalized anesthetic dosage adjustments and remote medical assistance. They are unable to effectively combine the patient's personalized physiological parameters with real-time data for dynamic adjustments, and lack intelligent optimization and early warning capabilities for anesthetic dosage. Summary of the Invention

[0004] The present invention provides an artificial intelligence-based system for evaluating the effects of an overdose of anesthetic drugs on brain function.

[0005] An artificial intelligence-based anesthetic drug overdose brain function impact assessment system includes a data acquisition module, a data preprocessing module, a model training module, a real-time assessment module, a risk warning module, a treatment recommendation module, a system interaction module, and a remote monitoring module, wherein;

[0006] The data acquisition module is used to collect multimodal physiological data of the patient in real time, including brain waves, functional magnetic resonance imaging, and brain blood oxygen saturation. The brain wave data is obtained through a multi-channel electroencephalogram device, the functional magnetic resonance imaging data represents the brain activity state through three-dimensional voxel data, and the brain blood oxygen saturation data is collected through an optical sensor to evaluate the effect of anesthetic drugs on brain function;

[0007] The data preprocessing module is used to perform denoising, normalization and feature extraction on the collected multimodal data;

[0008] The model training module is used to establish a prediction model for the effect of anesthetic overdose on brain function based on historical patient data by integrating multiple machine learning algorithms, specifically including:

[0009] A deep learning model combining convolutional neural networks and long short-term memory networks was used to extract and fuse the spatiotemporal features of EEG and functional magnetic resonance imaging data.

[0010] Convolutional neural networks are used to extract temporal features of EEG and spatial features of functional magnetic resonance imaging. The fused features are then input into a pre-trained support vector machine model for binary classification to determine whether anesthetic drugs have an excessive effect on brain function.

[0011] The real-time evaluation module is used to dynamically evaluate whether the patient's current anesthetic drug dose will have an excessive effect on brain function based on the real-time data of the data acquisition module and the prediction model generated by the model training module. Specifically:

[0012] The real-time assessment module uses a deep convolutional neural network to generate a brain function heat map, marking the brain areas affected by anesthetics. It also uses a long-short-term memory network to dynamically analyze the time series data of brain waves to assess the risk of brain function impairment at different time points.

[0013] The risk warning module determines whether the patient's brain function is at risk of damage based on the evaluation results of the real-time evaluation module, and triggers an early warning when the safety threshold is exceeded;

[0014] The treatment suggestion module is used to generate personalized anesthetic drug adjustment suggestions based on the analysis results of the risk warning module and in combination with the patient's personalized physiological parameters and historical anesthetic drug usage data;

[0015] The system interaction module is used for interactive operations between medical staff and the system, supporting real-time data query, review of prediction results and adjustment of treatment recommendations;

[0016] The remote monitoring module supports the expert team to remotely monitor the patient's brain function status through the Internet and provide remote assistance.

[0017] Optionally, the data acquisition module includes:

[0018] EEG data collection: Use multi-channel EEG equipment to collect EEG data in real time;

[0019] Functional MRI data acquisition: High-resolution functional MRI equipment using a gradient echo EPI sequence acquires three-dimensional voxel data to capture blood oxygen level-dependent signals under anesthesia;

[0020] Brain blood oxygen saturation data collection: Use near-infrared optical sensors to non-invasively collect local brain blood oxygen saturation data on the patient's scalp;

[0021] Data synchronization: By integrating a time synchronization unit, EEG data, functional magnetic resonance imaging data, and brain blood oxygen saturation data are collected synchronously on the same time axis;

[0022] Multimodal data transmission and storage: The collected multimodal data is encrypted and transmitted through a high-speed data interface and stored on a dedicated server.

[0023] Optionally, the data preprocessing module includes:

[0024] Denoising and decomposition of EEG data: Use wavelet transform to decompose and denoise EEG signals;

[0025] Feature extraction of EEG data: Use fast Fourier transform to perform frequency domain analysis on the denoised EEG signal and extract power spectral density features;

[0026] Denoising and normalization of functional MRI data: aligning the functional MRI data in time and space, and normalizing the functional MRI data to adjust the data intensity consistency between different voxels;

[0027] Independent Component Analysis of fMRI Data: Feature extraction from fMRI data using independent component analysis;

[0028] Temporal alignment of multimodal data: After the EEG and fMRI data are processed, a time synchronization algorithm is used to align the two types of data.

[0029] Optionally, the model training module includes:

[0030] Extraction of EEG timing features: Use convolutional neural networks to extract EEG timing features;

[0031] Extraction of functional magnetic resonance imaging spatial features: Perform three-dimensional convolution processing on functional magnetic resonance imaging data to extract spatial activity features of different brain regions;

[0032] Spatiotemporal feature fusion: The EEG temporal features extracted by convolutional neural networks and the spatial features of functional magnetic resonance imaging are fused to form fused spatiotemporal features.

[0033] Optionally, the model training module further includes:

[0034] Feature classification: The fused spatiotemporal features are input into the pre-trained support vector machine model for binary classification to determine whether the anesthetic drug has an excessive effect on brain function;

[0035] Adaptive Gradient Optimization Algorithm: Use the adaptive gradient optimization algorithm to optimize the prediction model. The adaptive gradient optimization algorithm updates the model parameters by calculating the adaptive learning rate of the gradient;

[0036] Cross-entropy loss function: The cross-entropy loss function is used for model training to measure the difference between the predicted results and the true labels.

[0037] Optionally, the real-time evaluation module specifically includes:

[0038] Real-time data input: The real-time evaluation module obtains multimodal data collected in real time from the data acquisition module, including brain waves, functional magnetic resonance imaging, and brain blood oxygen saturation data;

[0039] Deep convolutional neural network generates brain function heat maps: The real-time evaluation module processes functional magnetic resonance imaging data through a pre-trained deep convolutional neural network to generate brain function heat maps;

[0040] Long-short-term memory network analysis of EEG time series data: The real-time evaluation module uses a long-short-term memory network to perform dynamic analysis on EEG time series data, capturing the time dependency of EEG signals and assessing changes in brain function at different moments.

[0041] Dynamic evaluation result output: The real-time evaluation module comprehensively evaluates the brain function heat map generated by the convolutional neural network and the dynamic analysis results of the brain wave signals by the long short-term memory network. Based on the anesthetic drug dosage, the patient's current physiological state, the brain function heat map and the time series analysis results, a comprehensive evaluation report is generated to indicate whether the current anesthetic drug dosage exceeds the safety range and to assess whether there is a risk of brain function damage.

[0042] Optionally, the risk warning module specifically includes:

[0043] Setting and adjusting warning thresholds: The system initially sets a safety threshold range based on clinical data on the effects of anesthetics on brain function under normal circumstances. The threshold range is used to measure whether the brain function status is safe. The risk warning module automatically adjusts the warning threshold based on the patient's individual physiological data (including age, weight, gender, medical history, etc.) and the patient's tolerance to anesthetics. The adjustment process is carried out through an adaptive algorithm.

[0044] Real-time monitoring and evaluation: The risk warning module obtains the brain function heat map, brain wave timing analysis and anesthetic drug dosage from the real-time evaluation module to calculate the current brain function status value;

[0045] Risk warning triggering mechanism: When the calculated brain function status value exceeds the adjusted warning threshold, the risk warning module triggers the warning mechanism;

[0046] Display and output of warning information: The warning module reminds medical staff of the current risk status through visual and auditory means.

[0047] Optionally, the treatment suggestion module includes:

[0048] Personalized parameter collection: The treatment recommendation module extracts personalized physiological parameters from the patient's historical data;

[0049] Risk assessment and anesthesia dose adjustment: Based on the assessment results provided by the risk warning module, identify whether the patient is at risk of overdose or underdose of anesthetic drugs and generate an anesthesia dose adjustment plan;

[0050] Reinforcement learning algorithm optimization adjustment strategy: Use reinforcement learning algorithms (such as Q-Learning) to optimize the adjustment strategy of anesthesia dosage;

[0051] Real-time adjustment and feedback: The dosage of anesthetic drugs is adjusted in real time according to the results of reinforcement learning algorithm optimization; Safety and personalized recommendations: When generating an adjustment plan, the treatment recommendation module always ensures that the dosage adjustment does not exceed the safety threshold range.

[0052] Optionally, the system interaction module includes:

[0053] Graphical user interface design: The system interaction module provides medical staff with an intuitive and concise operation platform through a graphical user interface;

[0054] Voice recognition function: supports medical staff to interact with the system through voice commands to realize data query and operation execution;

[0055] Natural language processing and feedback: The system interaction module incorporates natural language processing technology, allowing medical staff to interact with the system using natural language;

[0056] Data query and review: Allows medical staff to query real-time and historical data of any time period, and perform visual display and comparative analysis;

[0057] Manual adjustment of treatment recommendations: Supports medical staff to manually adjust anesthesia dosage and parameters, and record the adjustment results for subsequent reference.

[0058] Optionally, the remote monitoring module includes:

[0059] Data transmission: The patient's multimodal physiological data is encrypted and transmitted to the expert team. At the same time, a time synchronization mechanism is used to ensure that the expert team obtains the latest patient multimodal physiological data;

[0060] Remote monitoring platform setup: The remote monitoring module provides access to the expert team through a dedicated remote monitoring platform with a graphical user interface, allowing experts to remotely view the patient's real-time brain function data, historical data records, and system-generated assessment results;

[0061] Real-time assistance and decision support: The remote monitoring module allows the expert team to provide immediate decision support based on real-time monitored patient data;

[0062] Multi-user collaborative monitoring: supports multiple experts to remotely monitor the patient's condition simultaneously and collaborate to generate the optimal anesthesia management and treatment plan;

[0063] Remote operation authority management and confirmation: Remote adjustment suggestions provided by the expert team will only take effect after confirmation by local medical staff.

[0064] Beneficial effects of the present invention:

[0065] This invention, through the highly integrated data acquisition and real-time assessment modules, can accurately collect and analyze patients' multimodal physiological data, including EEG, functional magnetic resonance imaging, and brain oxygen saturation, to assess the effects of anesthetic drugs on patients' brain function in real time. The deep learning model, which combines convolutional neural networks with long-short-term memory networks, accurately extracts spatiotemporal features and dynamically generates brain function heat maps, ensuring high-precision monitoring and assessment during anesthesia, providing medical staff with detailed feedback on brain function status, and improving patient safety.

[0066] This invention, through the application of a treatment recommendation module and a reinforcement learning algorithm, enables the system to generate an optimal anesthetic dosage adjustment plan based on the patient's personalized physiological parameters, historical anesthesia records, and real-time monitoring results. The system not only automatically adjusts the anesthetic dosage but also, through the risk warning module, provides timely notifications when the anesthetic dosage is too high or too low, dynamically adjusting the warning threshold to ensure the patient's treatment safety and optimal results. The personalized control function of the treatment recommendation module significantly improves the accuracy and flexibility of anesthesia management.

[0067] This invention, through a remote monitoring module and a system interaction module, supports a team of experts in remotely monitoring and collaborating on a patient's brain function status. Through encrypted transmission, rights management, and an operation confirmation mechanism, the expert team can provide real-time remote assistance, helping local medical staff develop optimal anesthesia strategies and ensuring that all remote adjustments are implemented after local confirmation. This collaborative model not only enhances the convenience of remote collaboration but also ensures the security and traceability of all operations, making it suitable for remote medical assistance in complex or emergency situations. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0069] Figure 1 A schematic diagram of a system flow diagram of an embodiment of the present invention;

[0070] Figure 2 Schematic diagram of the model training module flow in an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0072] It should be noted that references in the specification to "one embodiment," "an embodiment," "an exemplary embodiment," "some embodiments," etc. indicate that the described embodiments may include specific features, structures, or characteristics, but not every embodiment necessarily includes such specific features, structures, or characteristics. In addition, when specific features, structures, or characteristics are described in conjunction with an embodiment, it is within the knowledge of persons skilled in the relevant art to implement such features, structures, or characteristics in conjunction with other embodiments (whether or not explicitly described).

[0073] In general, terms can be understood, at least in part, from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending at least in part on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0074] like Figure 1-Figure 2 As shown, an artificial intelligence-based anesthetic drug overdose brain function impact assessment system includes a data acquisition module, a data preprocessing module, a model training module, a real-time assessment module, a risk warning module, a treatment suggestion module, a system interaction module, and a remote monitoring module, wherein;

[0075] The data acquisition module is used to collect multimodal physiological data of patients in real time, including electroencephalogram (EEG), functional magnetic resonance imaging (fMRI), and brain blood oxygen saturation. EEG data is acquired through a multi-channel EEG device, fMRI data represents brain activity status through three-dimensional voxel data, and brain blood oxygen saturation data is collected through optical sensors to evaluate the effects of anesthetic drugs on brain function.

[0076] The data preprocessing module is used to perform denoising, normalization and feature extraction on the collected multimodal data;

[0077] The model training module is used to establish a predictive model for the effects of anesthetic overdose on brain function based on historical patient data by integrating multiple machine learning algorithms. Specifically, it includes:

[0078] A deep learning model combining convolutional neural networks and long short-term memory networks was used to extract and fuse the spatiotemporal features of EEG and functional magnetic resonance imaging data.

[0079] Convolutional neural networks are used to extract temporal features of EEG and spatial features of functional magnetic resonance imaging. The fused features are then input into a pre-trained support vector machine model for binary classification to determine whether anesthetic drugs have an excessive effect on brain function.

[0080] The real-time evaluation module is used to dynamically evaluate whether the patient's current anesthetic drug dose will have an excessive effect on brain function based on the real-time data of the data acquisition module and the prediction model generated by the model training module. Specifically:

[0081] The real-time assessment module uses a deep convolutional neural network to generate a brain function heat map, marking the brain areas affected by anesthetics. It also uses a long-short-term memory network to dynamically analyze the time series data of brain waves to assess the risk of brain function impairment at different time points.

[0082] The risk warning module determines whether the patient's brain function is at risk of damage based on the evaluation results of the real-time evaluation module, and triggers an early warning when the safety threshold is exceeded;

[0083] The treatment recommendation module is used to generate personalized anesthetic drug adjustment recommendations based on the analysis results of the risk warning module, combined with the patient's personalized physiological parameters and historical anesthetic drug usage data;

[0084] The system interaction module is used for interactive operations between medical staff and the system, supporting real-time data query, review of prediction results, and adjustment of treatment recommendations;

[0085] The remote monitoring module supports the expert team to remotely monitor the patient's brain function status through the Internet and provide remote assistance.

[0086] The data acquisition module includes:

[0087] EEG data collection: Using a multi-channel EEG device (32 or 64 channels), with electrodes arranged according to the international 10-20 system, real-time EEG data collection is performed at a sampling frequency of 500 Hz or higher, ensuring coverage of electrical activity in different brain regions. The signals are filtered through a pre-processing filter to remove noise and transmitted to the data pre-processing module in real time.

[0088] Functional MRI data acquisition: High-resolution functional MRI equipment using a gradient echo EPI sequence acquires three-dimensional voxel data to capture blood oxygen level-dependent signals under anesthesia;

[0089] Brain blood oxygen saturation data collection: Use near-infrared optical sensors to non-invasively collect local brain blood oxygen saturation data on the patient's scalp;

[0090] Data synchronization: By integrating a time synchronization unit, EEG data, functional magnetic resonance imaging data, and brain oxygen saturation data are collected synchronously on the same timeline. Key event markers, such as changes in anesthesia dose and significant changes in the patient's physiological state, are set to ensure temporal consistency in multimodal data analysis.

[0091] Multimodal data transmission and storage: The collected multimodal data is encrypted and transmitted through a high-speed data interface and stored on a dedicated server. Each data type is stored in a different format and timestamped to ensure data integrity and security.

[0092] The data acquisition module provides comprehensive and accurate real-time monitoring of the patient's brain function status by synchronously collecting multimodal data. It also ensures the security and accuracy of the data through data synchronization and encrypted transmission, supporting subsequent analysis and evaluation.

[0093] The data preprocessing module includes:

[0094] Denoising and decomposition of EEG data: Use wavelet transform to decompose and denoise EEG signals. Select appropriate wavelet basis functions (such as Daubechies or Coiflet) to perform multi-scale decomposition of EEG data. Use soft thresholding or hard thresholding to filter out high-frequency noise. Finally, reconstruct the decomposed data to retain the useful components of the original data and ensure accurate capture of different frequency bands.

[0095] Feature extraction of EEG data: Use fast Fourier transform to perform frequency domain analysis on the denoised EEG signal and extract the power spectral density features. The specific steps are as follows:

[0096] The time domain EEG signal is converted into a frequency domain signal through fast Fourier transform, and the power spectrum density value of each frequency is calculated. The calculation formula of power spectrum density is:

[0097]

[0098] Where P(f) is the power spectral density at frequency f, X(f) is the signal amplitude at frequency f, and N is the number of sampling points. The power spectral density is used to evaluate the activity intensity of different frequency bands of the brain and extract the main frequency band features related to brain activity.

[0099] Denoising and normalization of functional MRI data: Temporally and spatially align the functional MRI data to ensure that images acquired at different time points are accurately spatially matched, eliminating errors caused by patient movement or other external interference. Image registration methods based on mutual information or correlation are used to align the images to a standard brain template. The functional MRI data are then normalized to adjust the data intensity consistency between different voxels and reduce individual differences. The normalization steps include intensity normalization and spatial normalization, adjusting all data to the same scale and intensity range.

[0100] Independent component analysis of functional magnetic resonance imaging data: Independent component analysis is used to extract features from functional magnetic resonance imaging data. First, all collected signals are decorrelated to ensure data independence. Then, independent components corresponding to brain functional areas are decomposed. Each independent component represents a different area of ​​brain functional activity. The matrix decomposition of independent component analysis is expressed as:

[0101] X = A·S;

[0102] Where X is the observed functional magnetic resonance imaging data matrix, A is the mixing matrix, and S is the independent source signal matrix. By separating the independent source signals, the brain functional activity components related to the effects of anesthetic drugs are extracted, and the noise and irrelevant components are removed;

[0103] Temporal alignment of multimodal data: After processing EEG and fMRI data, a time synchronization algorithm is used to align the two types of data. The dynamic time warping algorithm addresses the different sampling frequencies of EEG and fMRI data, ensuring that the two data types can be synchronized and matched on the same timeline for subsequent temporal analysis and evaluation.

[0104] The data preprocessing module accurately denoises, extracts features, and synchronizes EEG and functional magnetic resonance imaging data through methods such as wavelet transform, fast Fourier transform, spatiotemporal alignment, normalization, and independent component analysis, ensuring data consistency, accuracy, and effectiveness in subsequent model analysis. This provides high-quality input data for accurately evaluating the effects of anesthetic drugs on brain function.

[0105] The model training module includes:

[0106] Extraction of brainwave timing features: Convolutional neural networks are used to extract brainwave timing features. Brainwave data is input in the form of time series and processed through multiple convolutional layers. The convolution operation formula is:

[0107]

[0108] Among them, O(i,j) is the convolution output, I(i,j) is the input EEG signal, W(m,n) is the convolution kernel, k is the size of the convolution kernel, and each layer of convolution extracts temporal features at different levels, such as the dynamic changes of EEG frequency bands such as α waves and β waves;

[0109] A pooling layer is used after each convolutional layer to reduce the dimension of the feature map, improving computational efficiency while preserving the main features;

[0110] Extraction of functional magnetic resonance imaging spatial features: 3D convolution processing is performed on the functional magnetic resonance imaging data to extract the spatial activity features of different brain regions. The convolution operation moves the 3D convolution kernel on the 3D voxels of the functional magnetic resonance imaging data, which is expressed as:

[0111]

[0112] Among them, O(x,y,z) is the output feature map, I(x,y,z) is the input voxel, and W(i,j,k) is the three-dimensional convolution kernel k x ,k y ,k z are the sizes of the convolution kernels, respectively. The spatial features of different brain regions are extracted through the three-dimensional convolution layer;

[0113] Spatiotemporal feature fusion: The EEG temporal features extracted by the convolutional neural network and the spatial features of the functional magnetic resonance imaging are fused to form fused spatiotemporal features. The long short-term memory network is used to capture the temporal dependency in the EEG data. The state update formula of the long short-term memory network is expressed as:

[0114] h t =o t tanh(c t );

[0115]

[0116] Among them, h t is the current hidden state, c t is the cell state, i t 、f t 、o t They are input gate, forget gate and output gate respectively. It is a candidate memory state. The long-short-term memory network combines the spatiotemporal characteristics to establish a comprehensive temporal dependency relationship on brain functional activities;

[0117] Through the combination of convolutional neural networks and long short-term memory networks, the model training module can effectively extract the temporal characteristics of brain waves and the spatial characteristics of functional magnetic resonance imaging data, and fuse the two to achieve a comprehensive analysis of the patient's brain functional activities, providing accurate data support for predicting the impact of anesthetic overdose on brain function.

[0118] The model training module also includes:

[0119] Feature classification: The fused spatiotemporal features are input into the pre-trained support vector machine model for binary classification to determine whether the anesthetic drug has an excessive effect on brain function. The support vector machine classifies samples into different categories by defining a hyperplane. The decision function of the support vector machine model is:

[0120] f(x)=w·x+b;

[0121] Among them, f(x) is the decision function, w is the weight vector, x is the input feature vector, b is the bias, and the support vector machine finds the optimal hyperplane through training to maximize the classification interval;

[0122] Adaptive gradient optimization algorithm: Use the adaptive gradient optimization algorithm to optimize the prediction model. The adaptive gradient optimization algorithm updates the model parameters by calculating the adaptive learning rate of the gradient. The optimization formula is expressed as:

[0123]

[0124] Among them, θ t is the current model parameter, η is the learning rate, m t and v t are the first-order and second-order momentum estimates of the gradient, ∈ is a constant to prevent the denominator from being zero. The adaptive gradient optimization algorithm improves the convergence speed and stability of the model by dynamically adjusting the learning rate;

[0125] Cross-entropy loss function: The cross-entropy loss function is used for model training to measure the difference between the predicted results and the true labels. The formula for cross-entropy loss is:

[0126]

[0127] Among them, L is the loss value, y i is the true label, p i is the probability predicted by the model, N is the number of samples, and the classification accuracy and generalization ability of the model are ensured by minimizing the cross entropy loss function;

[0128] By integrating support vector machines for binary classification, it is determined whether anesthetic drugs have excessive effects on brain function. The model is optimized and trained using an adaptive gradient optimization algorithm and a cross-entropy loss function, ensuring the accuracy and stability of the model in predicting the effects of anesthetic drug overdose.

[0129] The real-time evaluation module specifically includes:

[0130] Real-time data input: The real-time evaluation module obtains multimodal data collected in real time from the data acquisition module, including EEG, functional magnetic resonance imaging, and brain blood oxygen saturation data. The time synchronization unit ensures the temporal consistency of different modal data so that they can be analyzed on the same timeline.

[0131] Deep convolutional neural network generates brain function heatmaps: The real-time evaluation module processes functional magnetic resonance imaging data through a pre-trained deep convolutional neural network to generate brain function heatmaps. The specific steps are as follows:

[0132] Functional magnetic resonance imaging data is used as input and processed through multiple layers of two-dimensional convolutional layers to extract the spatial features of different brain regions. The formula for the convolution operation is:

[0133]

[0134] Where O(i,j) is the convolution output, I(i,j) is the input image, W(m,n) is the convolution kernel, and k is the convolution kernel size. The output of the convolution process is passed through an activation function (such as ReLU) and a pooling layer to generate a two-dimensional brain function heat map. The heat map shows the changes in activity of different brain regions under the influence of anesthetics, and color coding identifies areas that are more affected by anesthetics.

[0135] Long-short-term memory network analysis of EEG time series data: The real-time evaluation module uses a long-short-term memory network to perform dynamic analysis on EEG time series data, capturing the time dependency of EEG signals and assessing changes in brain function at different moments.

[0136] Dynamic evaluation result output: The real-time evaluation module comprehensively evaluates the brain function heat map generated by the convolutional neural network and the dynamic analysis results of the long-short-term memory network on the brain wave signal. Based on the anesthetic drug dosage, the patient's current physiological state, the brain function heat map and the time series analysis results, a comprehensive evaluation report is generated to indicate whether the current anesthetic drug dosage exceeds the safe range and assess whether there is a risk of brain function damage;

[0137] The real-time assessment module combines the brain function heat map generated by the deep convolutional neural network and the dynamic analysis of brain wave time series data by the long short-term memory network. It can evaluate the patient's brain function status during anesthesia in real time, and promptly determine whether the current anesthetic drug dose will have an excessive effect on brain function, providing comprehensive and accurate dynamic monitoring and analysis.

[0138] The risk warning module specifically includes:

[0139] Setting and adjusting the warning threshold: The system initially sets a safety threshold range based on clinical data on the effects of anesthetics on brain function under normal circumstances. The threshold range is used to measure whether the brain function state is safe. The risk warning module automatically adjusts the warning threshold based on the patient's individual physiological data (including age, weight, gender, medical history, etc.) and the patient's tolerance to anesthetics. The adjustment process is performed through an adaptive algorithm, which is expressed as:

[0140] T adjusted =T initial ·(1+α·R);

[0141] Among them, T adjusted is the adjusted warning threshold, T initial is the initial safety threshold, α is the adjustment coefficient, and R is the patient's tolerance index. The adjustment coefficient is updated in real time based on the patient's physiological parameters to ensure that the threshold is dynamically adjusted according to the patient's individual condition;

[0142] Real-time monitoring and evaluation: The risk warning module obtains the brain function heat map, brain wave timing analysis and anesthetic drug dosage from the real-time evaluation module and calculates the current brain function status value. The formula is:

[0143]

[0144] Among them, S current is the current brain function state value, W i is the weight of different brain regions or temporal features, F i is the functional status value of these areas or features, and N is the number of monitored features;

[0145] Triggering mechanism of risk warning: When the calculated brain function state value S current Exceeding the adjusted warning threshold T adjusted When the risk warning module triggers the warning mechanism, the warning mechanism is divided into different levels according to the degree of deviation between the status value and the safety threshold. The warning levels are divided as follows:

[0146] Low risk (green): The current status is close to the threshold but within a safe range, and the system alerts medical staff with a green signal;

[0147] Moderate Risk (Yellow): The status value is slightly higher than the safety threshold, indicating that there may be a slight risk of damage. The system issues a warning with a yellow signal and an audible tone.

[0148] High risk (red): The status value significantly exceeds the safety threshold, indicating that the patient's brain function is at risk of severe damage. The system notifies medical staff through a red signal and an emergency sound alarm, requiring immediate intervention.

[0149] Display and output of warning information: The warning module alerts medical staff to the current risk status through visual and auditory means. Visual warnings are displayed on a color-coded display screen, including green (safe), yellow (medium risk) and red (high risk). Audio warnings are presented through progressively stronger audio signals. As the risk level increases, the frequency and intensity of the sound will increase accordingly.

[0150] The risk warning module monitors the patient's brain function status in real time and combines it with dynamically adjusted personalized warning thresholds to quickly trigger the warning mechanism when the patient's brain function status exceeds the safe range. It also provides color coding, sound alarms and visual displays for different risk levels to ensure that medical staff can promptly identify and address the risk of brain function impairment in patients.

[0151] The treatment recommendation module includes:

[0152] Personalized parameter collection: The treatment recommendation module extracts personalized physiological parameters from the patient's historical data, such as age, weight, gender, medical history, and anesthetic tolerance. These parameters are combined with the patient's real-time physiological data (such as brain waves, brain oxygen saturation, heart rate, blood pressure, etc.) as the basis for generating personalized adjustment recommendations. The module also accesses the patient's historical anesthetic drug usage data, recording the past anesthetic drug type, dosage, duration, and corresponding reactions, to establish a personalized anesthetic drug usage profile.

[0153] Risk assessment and anesthesia dose adjustment: Based on the assessment results provided by the risk warning module, identify whether the patient is at risk of overdose or underdose of anesthetic drugs and generate an anesthesia dose adjustment plan;

[0154] When generating an adjustment plan, the treatment recommendation module considers the following factors: current anesthetic dose, risk level, patient-specific parameters, and the patient's historical anesthetic response. The adjustment plan includes reducing, increasing, or maintaining the anesthetic dose. The specific adjustment plan is calculated using the following formula:

[0155] D new =D current +ΔD;

[0156] Among them, D new is the adjusted anesthetic dose, D currentis the current anesthetic dose, and ΔD is the dose change, which is dynamically calculated based on the patient's current status and historical data;

[0157] Reinforcement learning algorithm optimization adjustment strategy: Use reinforcement learning algorithms (such as Q-Learning) to optimize the anesthetic dose adjustment strategy. The Q-Learning algorithm gradually optimizes the adjustment plan by learning the patient's response to different anesthetic doses to ensure the best anesthetic effect at the minimum dose. The update formula of the Q-Learning algorithm is:

[0158] Q(s,a)=Q(s,a)+α[r+γmaxQ(s′,a′)-Q(s,a)];

[0159] Where Q(s,a) is the value of performing action a (adjusting the dose) in state s, α is the learning rate, r is the current reward, γ is the discount factor, and s ′ and a ′ are the next state and the next action respectively. By continuously optimizing the Q function, the system can learn the optimal anesthesia dosage adjustment strategy;

[0160] Real-time adjustment and feedback: Based on the results of the reinforcement learning algorithm optimization, the anesthetic drug dosage is adjusted in real time. After each adjustment, the system monitors the patient's physiological response in real time and feeds these responses back to the reinforcement learning algorithm to continuously optimize future adjustment strategies.

[0161] When the adjustment is completed, the system generates an anesthetic dose adjustment report, which records in detail the dose changes before and after the adjustment, the patient's physiological response, and the corresponding evaluation results;

[0162] Safety and personalized recommendations: When generating an adjustment plan, the treatment recommendation module always ensures that the dosage adjustment does not exceed the safety threshold range. By dynamically calculating the patient's tolerance and combining personalized historical data, it ensures that all adjustments are made within the patient's safety range. If necessary, the system will also generate recommendations for alternative anesthetics or auxiliary drugs.

[0163] The treatment recommendation module combines the patient's personalized physiological parameters and historical anesthesia data, uses a reinforcement learning algorithm to generate a personalized anesthetic drug adjustment plan, and dynamically optimizes the adjustment strategy. This module ensures that the patient receives precise dosage control during anesthesia, minimizes risks, and ensures patient safety.

[0164] The system interaction modules include:

[0165] Graphical User Interface Design: The system interaction module provides medical staff with an intuitive and concise operating platform through a graphical user interface. The interface design complies with the ergonomic requirements of medical software and adopts a partitioned layout to ensure that medical staff can quickly access key functional modules;

[0166] The main interface includes the following core functional areas:

[0167] Real-time data monitoring area: displays real-time physiological data (such as brain waves, brain blood oxygen saturation, etc.) from the data acquisition module in the form of charts and numbers;

[0168] Evaluation result review area: provides the brain function heat map, anesthetic drug dosage and brain function damage risk assessment results generated by the real-time evaluation module. Medical staff can select data from different time periods for review.

[0169] Treatment suggestions and adjustment area: displays the anesthetic drug adjustment suggestions generated by the current system. Medical staff can view the suggestions and make manual adjustments.

[0170] Voice recognition function: supports medical staff to interact with the system through voice commands to realize data query and operation execution;

[0171] Natural Language Processing and Feedback: The system's interaction module incorporates natural language processing technology, allowing medical staff to interact with the system using natural language, rather than just pre-set voice commands. The system can understand complex sentences and parse key information. For example, when a medical staff asks, "What is the patient's current brain function status?" the system provides detailed feedback based on the assessment results, such as: "The patient's current EEG waves are within the normal range, but the anesthetic dose is recommended to be reduced by 5% to reduce the risk of brain damage."

[0172] Data query and review: Allows medical staff to query real-time and historical data of any time period, and perform visual display and comparative analysis;

[0173] Manual adjustment of treatment recommendations: Supports medical staff to manually adjust anesthesia dosage and parameters, and record the adjustment results for subsequent reference.

[0174] The remote monitoring module includes:

[0175] Data transmission: The patient's multimodal physiological data is encrypted and transmitted to the expert team. At the same time, a time synchronization mechanism is used to ensure that the expert team obtains the latest patient multimodal physiological data;

[0176] Remote Monitoring Platform Setup: The remote monitoring module provides access to the expert team through a dedicated remote monitoring platform. The platform features a graphical user interface, allowing experts to remotely view patients' real-time brain function data, historical data records, and system-generated assessment results. Experts can select data from specific time periods for detailed analysis, including real-time brain function heat maps, EEG time series analysis, and current anesthetic drug dosages.

[0177] Real-time assistance and decision support: The remote monitoring module allows the expert team to provide immediate decision support based on real-time monitored patient data. Experts can make treatment recommendations, adjust anesthesia dosages, or modify treatment plans based on the system's assessment results and the patient's current physiological condition. Experts communicate and collaborate directly with local medical staff through the platform and submit adjustment recommendations remotely.

[0178] Multi-user collaborative monitoring: supports multiple experts to remotely monitor the patient's condition simultaneously and collaborate to generate the optimal anesthesia management and treatment plan;

[0179] Remote operation authority management and confirmation: Remote adjustment suggestions provided by the expert team will only take effect after confirmation by local medical staff. All operation records will be archived to ensure the security and traceability of remote operations.

[0180] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0181] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A system for evaluating the effects of anesthetic overdose on brain function using artificial intelligence, characterized in that: It includes data acquisition module, data preprocessing module, model training module, real-time evaluation module, risk warning module, treatment suggestion module, system interaction module and remote monitoring module, among which; The data acquisition module is used to collect multimodal physiological data of the patient in real time, including electroencephalogram (EEG), functional magnetic resonance imaging and brain blood oxygen saturation. The electroencephalogram data is acquired through a multi-channel electroencephalogram device, the functional magnetic resonance imaging data represents the brain activity state through three-dimensional voxel data, and the brain blood oxygen saturation data is acquired through an optical sensor to evaluate the effect of anesthetic drugs on brain function; The data preprocessing module is used to perform denoising, normalization and feature extraction on the collected multimodal data; The model training module is used to establish a prediction model of the effect of anesthetic overdose on brain function based on historical patient data by integrating multiple machine learning algorithms, specifically including: A deep learning model combining convolutional neural networks and long short-term memory networks is used to extract and fuse the spatiotemporal features of EEG and functional magnetic resonance imaging data. Convolutional neural networks are used to extract the temporal features of EEG and the spatial features of functional magnetic resonance imaging. The fused features are input into the pre-trained support vector machine model for binary classification to determine whether anesthetic drugs have excessive effects on brain function. The real-time evaluation module is used to dynamically evaluate whether the patient's current anesthetic drug dose will have an excessive effect on brain function based on the real-time data of the data acquisition module and the prediction model generated by the model training module, specifically: The real-time assessment module uses a deep convolutional neural network to generate a brain function heat map, marking the brain areas affected by anesthetics, and dynamically analyzes the time series data of brain waves through a long short-term memory network to assess the risk of brain function impairment at different time points; The risk warning module determines whether the patient's brain function is at risk of damage based on the evaluation results of the real-time evaluation module, and triggers a warning when it exceeds the safety threshold; The treatment suggestion module is used to generate personalized anesthetic drug adjustment suggestions based on the analysis results of the risk warning module, combined with the patient's personalized physiological parameters and historical anesthetic drug use data; The system interaction module is used for interactive operations between medical staff and the system, supporting real-time data query, review of prediction results, and adjustment of treatment recommendations; The remote monitoring module supports the expert team to remotely monitor the patient's brain function status through the network and provide remote assistance.

2. According to the artificial intelligence-based anesthetic drug overdose brain function impact assessment system of claim 1, it is characterized in that: The data acquisition module comprises: EEG data collection: Use multi-channel EEG equipment to collect EEG data in real time; Acquisition of functional magnetic resonance imaging data: A functional magnetic resonance imaging device using a gradient echo EPI sequence is used to collect three-dimensional voxel data and capture blood oxygen level-dependent signals under anesthesia; Brain blood oxygen saturation data collection: Use near-infrared optical sensors to non-invasively collect local brain blood oxygen saturation data on the patient's scalp; Data synchronization: By integrating a time synchronization unit, it ensures that EEG data, functional magnetic resonance imaging data, and brain blood oxygen saturation data are collected synchronously on the same time axis; Multimodal data transmission and storage: The collected multimodal data is encrypted and transmitted through a high-speed data interface and stored on a dedicated server.

3. The system for evaluating the effect of anesthetic overdose on brain function using artificial intelligence according to claim 2, characterized in that: The data preprocessing module comprises: Denoising and decomposition of EEG data: Use wavelet transform to decompose and denoise EEG signals; Feature extraction of EEG data: Use fast Fourier transform to perform frequency domain analysis on the denoised EEG signal and extract power spectral density features; Denoising and normalization of functional MRI data: perform spatiotemporal alignment of functional MRI data, and normalize the functional MRI data to adjust the consistency of data intensity between different voxels; Independent Component Analysis of fMRI Data: Feature extraction of fMRI data using independent component analysis; Temporal alignment of multimodal data: After the EEG and fMRI data are processed, the two types of data are aligned using a time synchronization algorithm.

4. The system for evaluating the effect of anesthetic overdose on brain function using artificial intelligence according to claim 3, characterized in that: The model training module includes: Extraction of EEG timing features: Use convolutional neural network to extract EEG timing features; Extraction of spatial features of functional magnetic resonance imaging: Perform three-dimensional convolution processing on functional magnetic resonance imaging data to extract spatial activity features of different brain regions; Spatiotemporal feature fusion: The EEG temporal features extracted by the convolutional neural network and the spatial features of functional magnetic resonance imaging are fused to form fused spatiotemporal features.

5. The system for evaluating the effect of anesthetic overdose on brain function using artificial intelligence according to claim 4, characterized in that: The model training module also includes: Feature classification: The fused spatiotemporal features are input into the pre-trained support vector machine model for binary classification to determine whether the anesthetic drugs have an excessive effect on brain function; Adaptive gradient optimization algorithm: Use the adaptive gradient optimization algorithm to optimize the prediction model. The adaptive gradient optimization algorithm updates the model parameters by calculating the adaptive learning rate of the gradient. Cross-entropy loss function: The cross-entropy loss function is used for model training to measure the difference between the predicted results and the true labels.

6. The system for evaluating the effects of anesthetic overdose on brain function using artificial intelligence according to claim 5, characterized in that: The real-time evaluation module specifically includes: Real-time data input: The real-time evaluation module obtains multimodal data collected in real time from the data acquisition module, including brain waves, functional magnetic resonance imaging, and brain blood oxygen saturation data; Deep convolutional neural network generates brain function heat map: The real-time evaluation module processes functional magnetic resonance imaging data through a pre-trained deep convolutional neural network to generate a brain function heat map; Analyze brain wave timing data using long short-term memory network: The real-time evaluation module uses long short-term memory network to dynamically analyze brain wave timing data, capture the time dependency of brain wave signals, and evaluate changes in brain function at different times; Dynamic evaluation result output: The real-time evaluation module conducts a comprehensive evaluation on the brain function heat map generated by the convolutional neural network and the dynamic analysis results of the brain wave signals by the long short-term memory network. According to the anesthetic drug dosage, the patient's current physiological state, the brain function heat map and the timing analysis results, a comprehensive evaluation report is generated to indicate whether the current anesthetic drug dosage exceeds the safety range and whether there is a risk of brain function damage.

7. The system for evaluating the effects of anesthetic overdose on brain function using artificial intelligence according to claim 6, characterized in that: The risk warning module specifically includes: Setting and adjusting the warning threshold: The system initially sets a safety threshold range, which is used to measure whether the brain function state is safe. The risk warning module automatically adjusts the warning threshold based on the patient's individual physiological data and the patient's tolerance to anesthetic drugs. The adjustment process is carried out through an adaptive algorithm. Real-time monitoring and evaluation: The risk warning module obtains the brain function heat map, brain wave timing analysis and anesthetic drug dosage from the real-time evaluation module to calculate the current state value of brain function; Triggering mechanism of risk warning: When the calculated brain function status value exceeds the adjusted warning threshold, the risk warning module triggers the warning mechanism; Display and output of warning information: The warning module reminds medical staff of the current risk status through visual and auditory means.

8. The system for evaluating the effect of anesthetic overdose on brain function using artificial intelligence according to claim 7, characterized in that: The treatment suggestion module includes: Personalized parameter collection: The treatment suggestion module extracts personalized physiological parameters from the patient's historical data; Risk assessment and anesthesia dose adjustment: Based on the assessment results provided by the risk warning module, identify whether the patient is at risk of overdose or underdose of anesthetic drugs, and generate an anesthesia dose adjustment plan; Reinforcement learning algorithm optimizes adjustment strategy: Use reinforcement learning algorithm to optimize the adjustment strategy of anesthesia dosage; Real-time adjustment and feedback: According to the optimization results of the reinforcement learning algorithm, the dosage of anesthetic drugs is adjusted in real time; Safety and personalized recommendations: When generating adjustment plans, the treatment recommendation module always ensures that the dosage adjustment does not exceed the safety threshold range.

9. The system for evaluating the effect of anesthetic overdose on brain function using artificial intelligence according to claim 8, characterized in that: The system interaction module includes: Graphical user interface design: The system interaction module provides an operating platform for medical staff through a graphical user interface; Voice recognition function: supports medical staff to interact with the system through voice commands to achieve data query and operation execution; Natural language processing and feedback: The system interaction module combines natural language processing technology to enable medical staff to interact with the system using natural language; Data query and review: Allow medical staff to query real-time and historical data of any time period, and perform visual display and comparative analysis; Manual adjustment of treatment recommendations: Supports medical staff to manually adjust anesthesia doses and parameters, and record the adjustment results for subsequent reference.

10. The system for evaluating the effect of anesthetic overdose on brain function using artificial intelligence according to claim 9, characterized in that: The remote monitoring module comprises: Data transmission: The patient's multimodal physiological data is encrypted and transmitted to the expert team. At the same time, a time synchronization mechanism is used to ensure that the expert team obtains the latest patient multimodal physiological data; Remote monitoring platform setup: The remote monitoring module provides access to the expert team through a dedicated remote monitoring platform with a graphical user interface that allows experts to remotely view the patient's real-time brain function data, historical data records, and system-generated assessment results; Real-time assistance and decision support: The remote monitoring module allows the expert team to provide immediate decision support based on real-time monitored patient data; Multi-user collaborative monitoring: Supports multiple experts to remotely monitor the patient's status at the same time and collaborate to generate the best anesthesia management and treatment plan; Remote operation authority management and confirmation: Remote adjustment suggestions provided by the expert team will only take effect after confirmation by local medical staff.

Images