Postoperative neural function real-time monitoring method and system for stroke patient

Abstract

The invention discloses a cerebral apoplexy patient postoperative neural function real-time monitoring method. The method comprises the steps that brain MRI image data and clinical information of a to-be-monitored patient and post-operation electroencephalogram data collected in real time are collected and preprocessed; respectively extracting neurophysiological features and radiomics features of the patient to be monitored based on the preprocessed data; performing significant feature screening on the clinical information, the neurophysiological features and the radiomics features, performing preprocessing on the screened data, and combining minimum absolute contraction and selection operator regression analysis to obtain quantitative electroencephalogram data feature indexes and radiomics scores; inputting the screened clinical information, the quantitative electroencephalogram data characteristic index and the radiomics score of the to-be-monitored patient into a trained prediction model, and predicting a risk index of early neurological deterioration of the to-be-monitored patient; the problem that the evaluation result is inaccurate due to the fact that a single clinical feature is adopted to evaluate the neural function in a traditional method is solved.

Description

Technical Field

[0001] This application relates to the field of medical device technology, and relates to, but is not limited to, a method and system for real-time monitoring of postoperative neurological function in stroke patients. Background Technology

[0002] According to the latest Global Burden of Disease Study 2021 (GBD2021) data released in 2024, the burden of stroke increased significantly in Eastern Europe, Central Asia, and my country and Southeast Asia from 1990 to 2021. Acute ischemic stroke (AIS) accounted for as much as 65.3% of all strokes, remaining a leading cause of high mortality and disability rates globally, even compared to 2019. my country's latest "China Stroke Prevention and Control Report 2021," released in 2023, indicates that stroke has become the leading cause of premature death in China. The high incidence, mortality, and disability rates of ischemic stroke seriously threaten the lives and health of the Chinese people, damage the socio-economic situation, and impose a heavy medical burden.

[0003] Improving the prognosis of stroke patients is a key challenge in current clinical practice. Accurately identifying risk factors associated with poor prognosis, especially modifiable ones, can effectively aid clinical decision-making and the development of personalized treatment plans. Currently, postoperative diagnostic methods mainly rely on CT, MRI, and physician assessment, lacking real-time monitoring methods. For patients with extremely poor prognoses, early detection and early intervention are of great value. Therefore, postoperative monitoring of neurological function needs to be more precise and timely. Traditional methods for assessing neurological function mainly rely on single clinical characteristics, which are prone to model underfitting. Furthermore, ischemic stroke patients experience unstable and rapidly progressing conditions during treatment, making real-time monitoring of patient status impossible with traditional methods. The National Institutes of Health Stroke Score (NIHSS) and CT / MRI imaging are the most common assessment methods, but these methods place a heavy burden on physicians and are highly subjective. This inconvenience in assessment, the inability to provide dynamic and continuous monitoring information, and the potential for delays can increase the risk of complications. Summary of the Invention

[0004] To overcome the shortcomings of the prior art, this application provides a method and system for real-time monitoring of postoperative neurological function in stroke patients.

[0005] The technical solution of this application embodiment is implemented as follows: In a first aspect, embodiments of this application provide a method for real-time monitoring of postoperative neurological function in stroke patients. The method includes: collecting brain MRI image data, clinical information, and postoperative real-time electroencephalogram (EEG) data of the patient to be monitored, and performing filtering, noise reduction, and artifact removal processing to obtain preprocessed corresponding data; extracting neurophysiological features of the patient to be monitored based on the preprocessed EEG data; obtaining radiomics features of the patient to be monitored based on the preprocessed brain MRI image data using the PyRadiomics toolkit on the Python platform; using a lasso regression screening algorithm to screen significant features based on clinical information, neurophysiological features, and radiomics features to obtain corresponding feature data after screening; performing preprocessing on the screened neurophysiological features and screened radiomics features, and combining minimum absolute contraction and selection operator regression analysis to obtain a quantitative EEG data feature index and radiomics score of the neurophysiological features; inputting the screened clinical information, quantitative EEG data feature index, and radiomics score of the patient to be monitored into a trained prediction model to predict the risk index of early neurological function deterioration in the patient to be monitored, and determining the risk level based on the risk index; wherein, the prediction model is trained based on a random forest model.

[0006] In some embodiments, the training process of the prediction model is as follows: Connect to the hospital database to obtain all EEG data, clinical information, brain MRI image data, quantitative EEG data characteristic index, and radiomics scores of all stroke patients admitted within a preset time period, and mark whether their early neurological function has deteriorated, generating training samples; divide the training samples into a training set and a test set in a 7:3 ratio using random stratified sampling; batch input the training set into the random forest model for training; then input the test set into the trained random forest model for performance testing until a preset standard is reached, thus obtaining the prediction model.

[0007] In some embodiments, the preprocessing of the selected neurophysiological features and the selected radiomics features, combined with minimum absolute contraction and selection operator regression analysis, to obtain the quantitative EEG data feature index and radiomics score of the neurophysiological features includes: performing variance analysis, one-way ANOVA, and redundancy analysis on the selected neurophysiological features and the selected radiomics features respectively to obtain the first feature dimensionality reduction data of the neurophysiological features and the second feature dimensionality reduction data of the radiomics features; and performing minimum absolute contraction and selection operator regression analysis on the first feature dimensionality reduction data and the second feature dimensionality reduction data respectively to obtain the quantitative EEG data feature index and radiomics score of the neurophysiological features.

[0008] In some embodiments, the neurophysiological features include power spectral density, relative power ratio, brain symmetry index, coherence, and small-world properties corresponding to different frequency bands of EEG signals in the non-lesion hemisphere and lesion hemisphere. The extraction of neurophysiological features of the patient under monitoring based on preprocessed EEG data includes: calculating the spectral power of each electrode in each region of the non-lesion hemisphere and lesion hemisphere using Fast Fourier Transform, with a frequency range of 1 to 45 Hz; dividing the spectral power into multiple frequency bands based on the frequency range, and calculating the average power spectral density corresponding to each frequency band using the Welch method; the multiple frequency bands include: delta band, theta band, alpha band, beta band, and gamma band; removing midline electrodes, and calculating the average power spectral density corresponding to each non-lesion hemisphere and lesion hemisphere using the Welch method; and calculating the average power spectral density corresponding to each frequency band. The average spectral density is used to calculate the relative power ratio of the frequency bands; the relative power ratio of the frequency bands includes DAR, TAR, and DTABR; electrode pairs are determined, consisting of electrodes in the non-lesion hemisphere and electrodes in the lesion hemisphere corresponding to the symmetrical positions of the electrodes. Based on the frequency band power of the electrode pairs, the brain symmetry index of different frequency bands is calculated using the brain symmetry index calculation formula for different frequency bands; electrode lead pairs consisting of any two electrodes in the non-lesion hemisphere or any two electrodes in the lesion hemisphere are determined, and the coherence of different frequency bands is calculated based on the power spectral density of the electrode lead pairs using the coherence calculation formula for different frequency bands; based on the frequency band power of each electrode in the whole brain, the weighted clustering coefficient and weighted feature path length of each frequency band are obtained using a brain imaging analysis toolkit combined with the constructed undirected weighted network, and the ratio of the standardized weighted clustering coefficient to the weighted feature path length is used as the small-world property.

[0009] In some embodiments, the formula for calculating the brain symmetry index of different frequency bands is as follows: in, It is a brain symmetry index; , These represent the band power of the i-th electrode pair in the non-lesion hemisphere and the band power in the lesion hemisphere, respectively; N is the number of effective electrode pairs.

[0010] In some embodiments, the coherence calculation formula for the different frequency bands is as follows: in, The coherence between the x and y electrodes within the cerebral hemisphere; The cross-power spectral density of the leads of the x and y electrodes; and denoted as the self-power spectral density of electrode x and electrode y, respectively; f represents the power corresponding to different frequency bands.

[0011] In some embodiments, the method further includes: connecting to a hospital database, acquiring multiple sets of brain MRI image data collected by two doctors for multiple identical patients, obtaining initial radiomics features corresponding to each brain MRI image data, obtaining the intra-group correlation coefficient according to the intra-group correlation coefficient calculation formula, deleting the corresponding radiomics feature value when the intra-group correlation coefficient is less than a first preset threshold, updating the content of the initial radiomics features, and optimizing the radiomics features of the patient to be monitored based on the updated content of the initial radiomics features.

[0012] In some embodiments, the radiomics features include: first-order statistical features, 2D morphological features, 3D morphological features, gray-level co-occurrence matrix features, gray-level size region matrix features, gray-level run length features, neighborhood gray-level difference matrix features, and gray-level correlation matrix features.

[0013] Secondly, embodiments of this application provide a real-time monitoring system for postoperative neurological function in stroke patients. The system includes: a data acquisition and preprocessing module for acquiring brain MRI image data, clinical information, and postoperative real-time electroencephalogram (EEG) data of the patient to be monitored, and performing filtering, noise reduction, and artifact removal to obtain preprocessed corresponding data; a neurophysiological feature extraction module for extracting neurophysiological features of the patient to be monitored based on the preprocessed EEG data; a radiomics feature extraction module for obtaining radiomics features of the patient to be monitored based on the preprocessed brain MRI image data using the PyRadiomics toolkit on the Python platform; and a salient feature screening module for selecting features based on clinical information and neurophysiological data. The system employs a lasso regression screening algorithm to select significant features from both neurophysiological and radiomic features, yielding corresponding feature data. A feature data preprocessing module preprocesses the selected neurophysiological and radiomic features, combining minimum absolute contraction and selection operator regression analysis to obtain quantitative EEG data feature indices and radiomic scores for the neurophysiological features. A feature data analysis module inputs the selected clinical information, quantitative EEG data feature indices, and radiomic scores of the patients to be monitored into a trained prediction model to predict the risk index of early neurological functional deterioration in the patients, and determines the risk level based on the risk index. The prediction model is trained using a random forest model.

[0014] In some embodiments, the system further includes a display module and an early warning module: the display module is used to display in real time the radiomics characteristics, clinical information, postoperative real-time electroencephalogram data, neurophysiological characteristics, and predicted risk level of early neurological function deterioration of the patient to be monitored; it is also used to display the number of times the risk level of early neurological function deterioration is high; the early warning module is used to issue an early warning based on the number of times the risk level of early neurological function deterioration is high: an early warning is issued when the number of times the risk level of early neurological function deterioration is high exceeds a second preset threshold.

[0015] The beneficial effects of the technical solutions provided in this application include at least the following: This application collects and preprocesses brain MRI images, clinical information, and postoperative real-time electroencephalogram (EEG) data from patients under monitoring. Based on the preprocessed EEG and brain MRI images, neurophysiological and radiomics features of the patients under monitoring are extracted. A lasso regression algorithm is used to screen for significant features in the clinical information, neurophysiological features, and radiomics features of the patients under monitoring. Based on the screened neurophysiological and radiomics features, preprocessing is performed, and combined with minimum absolute contraction and selection operator regression analysis, quantitative EEG feature indices and radiomics scores for neurophysiological features are obtained. The screened clinical information, quantitative EEG feature indices, and radiomics scores of the patients under monitoring are input into a trained prediction model to predict the risk index of early neurological function deterioration in the patients under monitoring, and the risk level is determined based on the risk index. This application avoids the problem of inaccurate assessment results caused by traditional methods that use a single clinical feature to assess neurological function, and enables real-time monitoring of postoperative neurological function in stroke patients. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 A flowchart illustrating a method for real-time monitoring of postoperative neurological function in stroke patients, provided in an embodiment of this application; Figure 2 This is a structural block diagram of the real-time monitoring system for postoperative neurological function in stroke patients provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0019] It should be noted that the terms "first, second, and third" used in the embodiments of this application are merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, and third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0020] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiments of this application pertain. It should also be understood that terms such as those defined in general dictionaries should be understood to have a meaning consistent with their meaning in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0021] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for real-time monitoring of postoperative neurological function in stroke patients provided in this application embodiment. The method includes: step S1, collecting brain MRI image data, clinical information, and postoperative electroencephalogram data of the patient to be monitored, and performing filtering, noise reduction, and artifact removal processing to obtain preprocessed corresponding data.

[0022] Here, MRI (Magnetic Resonance Imaging) generally refers to magnetic resonance imaging. Head MRI is an imaging examination primarily used to check for abnormalities in the head's structures. Head MRI scans the head using the principles of magnetic resonance imaging, clearly revealing any structural abnormalities such as cerebral infarction, cerebral hemorrhage, or brain tumors. Head MRI is a non-invasive examination with no radiation exposure and minimal harm to the body, and can be used to detect intracranial vascular lesions and intracranial tumors.

[0023] Clinical information included: NIHSS score at admission, geographic region, and whether the patient was elderly, had hypertension, diabetes, consumed alcohol, smoked, or was overweight or obese. The NIHSS score (National Institutes of Health Stroke Scale) is a standardized neurological deficit assessment tool primarily used to quantify the degree of neurological impairment in stroke patients (especially ischemic stroke), guiding treatment decisions, assessing prognosis, and evaluating treatment effectiveness.

[0024] Electroencephalography (EEG) is a technique for recording and monitoring the electrical activity of the human brain. It involves placing electrodes on the scalp and measuring the electrical activity signals of neurons in the brain, thus creating an EEG image. The EEG recording is represented as an image, typically plotted with time on the x-axis and voltage on the y-axis. The waveforms on the image represent the electrical activity of neurons in the brain. EEG records the electrical activity of neurons in the brain. This electrical activity can be divided into different frequency bands, including the delta, theta, alpha, beta, and gamma bands.

[0025] Step S2: Extract neurophysiological characteristics of the patient to be monitored based on the preprocessed EEG data.

[0026] Here, extracting neurophysiological features based on electroencephalogram (EEG) data essentially involves converting the synchronized electrical activity of neurons in the cerebral cortex recorded by EEG data into quantifiable indicators. These indicators directly or indirectly reflect the neurophysiological state of the brain (such as arousal, cognitive processing, pathological abnormalities, etc.). These neurophysiological features include the power spectral density, relative power ratio, brain symmetry index, coherence, and small-world properties of different frequency bands of EEG signals from the non-lesion hemisphere and the lesion hemisphere. Step S3: Based on the preprocessed brain MRI image data, the radiomics characteristics of the patient to be monitored are obtained using the PyRadiomics toolkit on the Python platform.

[0027] Here, radiomics features are extracted from brain MRI image data. This involves quantitatively analyzing information such as grayscale distribution, spatial structure, and texture patterns of tissues in MRI images, transforming the image information into calculable digital features that reflect the microscopic pathology, structural changes, or functional state of brain tissue. These features provide objective evidence for disease diagnosis, prognostic assessment, and treatment response prediction. The radiomics features include first-order statistical eigenvalues, 2D morphological eigenvalues, 3D morphological eigenvalues, grayscale co-occurrence matrix eigenvalues, grayscale size region matrix eigenvalues, grayscale run length eigenvalues, neighborhood grayscale difference matrix eigenvalues, and grayscale correlation matrix eigenvalues. Step S4: Based on clinical information, neurophysiological characteristics, and radiomics characteristics, lasso regression screening algorithm is used to screen significant features to obtain the corresponding feature data after screening.

[0028] Here, a lasso regression screening algorithm was used to screen for significant features of the clinical information, neurophysiological characteristics, and radiomics characteristics of the patients to be monitored, thereby optimizing the feature data.

[0029] Step S5: Based on the screened neurophysiological features and the screened radiomics features, preprocessing is performed respectively. Combined with minimum absolute contraction and selection operator regression analysis, quantitative EEG data feature index and radiomics score of neurophysiological features are obtained.

[0030] Here, quantitative EEG data characteristic indices are a series of quantifiable feature indicators obtained through digital processing, statistical analysis, and algorithmic extraction of selected neurophysiological and radiomic features. These indices are used to objectively describe the rhythm, synchronicity, complexity, and other characteristics of EEG activity, and are widely applied in areas such as brain function assessment, diagnosis of neurological diseases (e.g., epilepsy, Alzheimer's disease), and cognitive state monitoring (e.g., arousal, attention). Quantitative EEG characteristic indices provide an objective tool for brain function assessment by transforming abstract selected neurophysiological and radiomic features into quantifiable indicators. Different indices analyze brain activity from dimensions such as frequency, power, synchronicity, and complexity; their combined analysis can more comprehensively reflect brain functional status and is an important auxiliary means for neuroscience research and clinical diagnosis.

[0031] Radiomics Score is a quantitative indicator built upon radiomics technology. It integrates high-throughput features extracted from medical images to provide a quantitative description of disease-related information. By standardizing and preprocessing medical images (such as CT, MRI, and PET) (denoising, lesion segmentation, etc.), subtle features that are difficult to discern with the naked eye (such as lesion morphology, density, texture, and grayscale distribution) are extracted. Then, key features are selected and models are built using statistical analysis or machine learning algorithms, ultimately transforming multi-dimensional features into one or more quantifiable "scores." The main functions of this score are to assist in disease diagnosis (e.g., differentiating between benign and malignant lesions), prognostic assessment (e.g., predicting patient survival), and treatment response prediction (e.g., assessing sensitivity to radiotherapy and chemotherapy), providing objective radiomics evidence for clinical decision-making.

[0032] Step S6: Input the clinical information, quantitative EEG data feature index and radiomics score of the patients to be monitored after screening into the trained prediction model to predict the risk index of early neurological function deterioration of the patients to be monitored, and determine the risk level based on the risk index; wherein, the prediction model is trained based on the random forest model.

[0033] Here, the random forest training algorithm can efficiently process multi-dimensional quantitative EEG data feature indices, clinical information, and radiomics scores, revealing complex potential relationships between features and more accurately monitoring the patient's neurological function status. The risk level is determined based on the risk index. After the prediction model outputs a risk index predicting early neurological deterioration in the patient to be monitored, the relationship between this risk index and a third preset threshold is compared. If the risk index is greater than the third preset threshold, the risk level corresponding to this risk index is assessed as high risk; otherwise, it is considered low risk. The value of the third preset threshold can be set according to actual conditions and is not limited here.

[0034] This method collects and preprocesses brain MRI images, clinical information, and postoperative real-time EEG data from patients under monitoring. Based on the preprocessed EEG and MRI data, neurophysiological and radiomics features are extracted. A lasso regression algorithm is used to screen for significant features in the clinical information, neurophysiological characteristics, and radiomics features of the patients. Based on the screened neurophysiological and radiomics features, preprocessing is performed, and combined with minimum absolute contraction and selection operator regression analysis, quantitative EEG feature indices and radiomics scores for neurophysiological features are obtained. The screened clinical information, quantitative EEG feature indices, and radiomics scores of the patients are input into a trained prediction model to predict the risk index of early neurological function deterioration. The risk level is determined based on the risk index. This method avoids the inaccurate assessment results caused by traditional methods that rely on a single clinical feature to evaluate neurological function, and enables real-time monitoring of postoperative neurological function in stroke patients.

[0035] In some possible embodiments, the training process of the prediction model in step S5 is as follows: S61, connect to the hospital database, obtain all EEG data, clinical information, brain MRI image data, quantitative EEG data characteristic index and radiomics score of all stroke patients after admission within a preset time, and mark whether their early neurological function has deteriorated, and generate training samples; divide the training samples into training set and test set in a 7:3 ratio by random stratified sampling.

[0036] Here, the preset time period can be set according to the actual situation and is not limited here. For example, it could be all EEG data, clinical information, and brain MRI images of all stroke patients admitted to the hospital between May 2023 and January 2024. Feature data corresponding to each of the EEG data, clinical information, and brain MRI images are extracted. Quantitative EEG feature indices are obtained from the EEG data, and radiomics scores are obtained from the MRI images. Doctors use a voting system based on clinical information, quantitative EEG feature indices, and radiomics scores to determine whether the patient's early neurological function has deteriorated and to label these indices. The patient's clinical information, quantitative EEG feature indices, and radiomics scores, along with their corresponding labels indicating early neurological function deterioration, are used as training samples. These training samples are randomly stratified and divided into training and test sets in a 7:3 ratio. In machine learning, the training and test sets are two core subsets of the dataset; their roles are independent yet closely coordinated, jointly supporting the construction, evaluation, and optimization of the model. The training set is the "textbook" for model learning; it's the core data used by the model to fit parameters and learn data patterns. The test set is the model's "final exam," used to evaluate its generalization ability on unseen data. The training set is responsible for "teaching" the model, and the test set is responsible for "testing" the model. The ultimate goal is to obtain a model with strong generalization ability that can work stably on new data.

[0037] S62, the training set is batch-input into the random forest model for training; then the test set is input into the trained random forest model for performance testing until the preset standard is reached, and the prediction model is obtained.

[0038] Here, this step involves batch-feeding the training set into the random forest model for training. The training set is responsible for "teaching" the random forest model, enabling it to predict the risk index of early neurological deterioration in patients. The test set is responsible for "testing" the random forest model, ensuring it can stably predict the risk index of early neurological deterioration on new data. Finally, the stable random forest model is used as the prediction model.

[0039] The training set is batch-fed into the random forest model for training. The extracted feature data and its corresponding labels indicating whether early neurological function is deteriorating are used to "teach" the random forest model, enabling it to predict the risk index of early neurological function deterioration in patients. The test set is then fed into the random forest model for performance testing. Feature data from the test set is fed into the random forest model to predict the risk index of early neurological function deterioration in corresponding patients. The prediction results are compared with the labels indicating early neurological function deterioration in corresponding patients. Based on the comparison results, the performance of the random forest model is judged. When the comparison results meet the preset standard, the random forest model is adopted as the prediction model; when the comparison results do not meet the preset standard, the random forest model is optimized based on the comparison results until it meets the preset standard.

[0040] In some possible embodiments, the preprocessing of the selected neurophysiological features and the selected radiomics features, combined with minimum absolute contraction and selection operator regression analysis, yields quantitative EEG data feature indices and radiomics scores for the neurophysiological features, including: Step S51: Perform variance analysis, one-way ANOVA, and redundancy analysis on the screened neurophysiological features and the screened radiomics features respectively to obtain the first feature dimensionality reduction data of the neurophysiological features and the second feature dimensionality reduction data of the radiomics features.

[0041] Here, feature dimensionality reduction is performed on both the screened neurophysiological and radiomics features. First, features with zero variance are excluded. Then, one-way ANOVA is used to screen for differential features with p-values ​​less than 0.05. The p-value is a statistical indicator used to determine whether a result is statistically significant. Next, redundant features are eliminated by calculating the Pearson or Spearman correlation coefficients between features, with a threshold of 0.9 or higher. This threshold means that if the correlation is too high, the feature is considered redundant and is excluded; that is, a Pearson or Spearman correlation coefficient greater than 0.9 is used to exclude redundant features. Finally, the first-dimensionality-reduced data for neurophysiological features and the second-dimensionality-reduced data for radiomics features are obtained. Dimensionality reduction of the feature data improves computational efficiency, reduces resource consumption, decreases the risk of overfitting, improves data quality, and achieves computational efficiency.

[0042] Step S52: The first feature dimensionality reduction data and the second feature dimensionality reduction data are respectively subjected to minimum absolute contraction and selection operator regression analysis to obtain the quantitative EEG data feature index and radiomics score of neurophysiological characteristics.

[0043] Here, quantitative EEG data characteristic indices are a series of quantifiable feature indicators obtained through digital processing, statistical analysis, and algorithmic extraction of selected neurophysiological and radiomic features. These indices are used to objectively describe the rhythm, synchronicity, complexity, and other characteristics of EEG activity, and are widely applied in areas such as brain function assessment, diagnosis of neurological diseases (e.g., epilepsy, Alzheimer's disease), and cognitive state monitoring (e.g., arousal, attention). Quantitative EEG characteristic indices provide an objective tool for brain function assessment by transforming abstract selected neurophysiological and radiomic features into quantifiable indicators. Different indices analyze brain activity from dimensions such as frequency, power, synchronicity, and complexity; their combined analysis can more comprehensively reflect brain functional status and is an important auxiliary means for neuroscience research and clinical diagnosis.

[0044] Radiomics Score is a quantitative indicator built upon radiomics technology. It integrates high-throughput features extracted from medical images to provide a quantitative description of disease-related information. By standardizing and preprocessing medical images (such as CT, MRI, and PET) (denoising, lesion segmentation, etc.), subtle features that are difficult to discern with the naked eye (such as lesion morphology, density, texture, and grayscale distribution) are extracted. Then, key features are selected and models are built using statistical analysis or machine learning algorithms, ultimately transforming multi-dimensional features into one or more quantifiable "scores." The main functions of this score are to assist in disease diagnosis (e.g., differentiating between benign and malignant lesions), prognostic assessment (e.g., predicting patient survival), and treatment response prediction (e.g., assessing sensitivity to radiotherapy and chemotherapy), providing objective radiomics evidence for clinical decision-making.

[0045] In some possible embodiments, the neurophysiological features include power spectral density, relative power ratio, brain symmetry index, coherence, and small-world properties corresponding to different frequency bands of EEG signals in the non-lesion hemisphere and lesion hemisphere. The extraction of neurophysiological features of the patient to be monitored based on preprocessed EEG data includes: For each electrode within the non-lesion hemisphere and the lesion hemisphere, the spectral power of the corresponding electrode is calculated by fast Fourier transform, with a frequency range of 1 to 45 Hz.

[0046] Here, the non-lesion hemisphere refers to the other hemisphere of the brain (as opposed to the lesion hemisphere) where no definite lesion or abnormality directly related to the current disease is present. The lesion hemisphere refers to the one hemisphere (left or right) of the brain where a definite lesion (such as a lesion, injury, or abnormal area) is present. When collecting EEG data, electrodes are placed on the scalp to measure the electrical activity signals of neurons in the brain, thus obtaining an EEG. EEG can be divided into different frequency bands, each reflecting a specific type of brain electrical activity. The calculation of EEG frequency uses the Fourier transform method to convert the time-domain electrical signal into a frequency-domain energy spectrum.

[0047] The spectral power is divided into multiple frequency bands based on the frequency range, and the average power spectral density of each frequency band is calculated using the Welch method; the multiple frequency bands include: delta band, theta band, alpha band, beta band and gamma band.

[0048] Here, an electroencephalogram (EEG) records the electrical activity of neurons in the brain. This electrical activity can be divided into different frequency bands, including: the delta band (1 to 4 Hz), the theta band (4 to 8 Hz), the alpha band (8 to 13 Hz), the beta band (13 to 30 Hz), and the gamma band (30 to 45 Hz). Power spectral density (PSD) is an indicator used to quantify the energy distribution of EEG signals across different frequency components. The PSD of different frequency bands refers to the energy value within each band after dividing the PSD according to the classic frequency range (bands) of the EEG signal. The PSD of these frequency bands reflects different physiological states and functional activities of the brain and is a core indicator in EEG analysis.

[0049] The midline electrode was removed, and the average power spectral density of the non-lesion hemisphere and the lesion hemisphere was calculated using the Welch method.

[0050] Here, the midline electrodes include: FZ, FCZ, CZ, CPZ, and PZ. FZ is the electrode for the frontal midline; FCZ is the electrode for the central frontal midline; CZ is the electrode for the central midline; CPZ is the electrode for the central parietal midline; and PZ is the electrode for the parietal midline.

[0051] The relative power ratio of a frequency band is calculated based on the average power spectral density corresponding to each frequency band; the relative power ratio of a frequency band includes: DAR, TAR and DTABR.

[0052] Here, DAR is the ratio of the power of the delta band to the power of the alpha band; TAR is the ratio of the power of the theta band to the power of the alpha band; and DTABR is the ratio of the sum of the power of the delta and theta bands to the sum of the power of the alpha and beta bands.

[0053] Electrode pairs are formed by electrodes located in the non-lesion hemisphere and electrodes located symmetrically to the electrodes in the lesion hemisphere. Based on the frequency band power of the electrode pairs, the brain symmetry index for different frequency bands is calculated using the brain symmetry index calculation formula for different frequency bands.

[0054] Here, the brain symmetry index is an indicator used to quantify the degree of symmetry between the two hemispheres (left and right) of the brain in terms of structure, function, or physiological activity. It is widely used in fields such as neuroimaging and electroencephalography (EEG), and is especially important in assessing hemispheric asymmetry caused by brain diseases.

[0055] Determine the electrode lead pair consisting of any two electrodes within the non-lesion hemisphere or the lesion hemisphere. Based on the power spectral density of the electrode lead pair, calculate the coherence of different frequency bands using the coherence calculation formula for different frequency bands.

[0056] Here, the correlation is a quantitative indicator used to describe the degree of association between electroencephalogram (EEG) signals (including different brain regions, different time points, or different frequency components). By analyzing the synchronicity, consistency, or coordinated change patterns of the signals, the functional connectivity or activity patterns of different brain regions can be revealed.

[0057] Based on the frequency band power of each electrode in the whole brain, the weighted clustering coefficient and weighted feature path length of each frequency band are obtained by using a brain imaging analysis toolkit and an undirected weighted network. The ratio of the standardized weighted clustering coefficient to the weighted feature path length is used as the small-world property.

[0058] Here, the small-world property is based on complex network theory and describes an efficient organizational pattern of brain functional networks (composed of brain region functional connections recorded by electroencephalography). It maintains tight connections in local areas while enabling rapid information transmission over a global scale, which is an important characteristic of efficient normal brain function.

[0059] In some possible embodiments, the formula for calculating the brain symmetry index of different frequency bands is as follows: in, It is a brain symmetry index; , These represent the band power of the i-th electrode pair in the non-lesion hemisphere and the band power in the lesion hemisphere, respectively; N is the number of effective electrode pairs.

[0060] Here, the number of effective electrode pairs is generally 11.

[0061] In some possible embodiments, the coherence calculation formula for the different frequency bands is as follows: in, The coherence between the x and y electrodes within the cerebral hemisphere; The cross-power spectral density of the leads of the x and y electrodes; and denoted as the self-power spectral density of electrode x and electrode y, respectively; f represents the power corresponding to different frequency bands.

[0062] In some possible embodiments, the method further includes: connecting to a hospital database, obtaining multiple sets of brain MRI image data collected by two doctors for multiple identical patients, obtaining initial radiomics features corresponding to each brain MRI image data, obtaining the intra-group correlation coefficient according to the intra-group correlation coefficient calculation formula, and deleting the corresponding radiomics feature value and updating the content of the initial radiomics features when the intra-group correlation coefficient is less than a first preset threshold.

[0063] Here, multiple sets of brain MRI images collected by two doctors from multiple identical patients are acquired. For example, Doctor A and Doctor B collect brain MRI images from each of the same 10 patients. Each set of brain MRI images has its corresponding initial radiomics features. Within-group correlation coefficients are calculated, where "within-group" refers to the relationship between the initial radiomics features of all patients collected by one doctor and these multiple sets of brain MRI images. Between-group correlation coefficients are the relationship between the initial radiomics features of all patients collected by two doctors, for example, the relationship between the initial radiomics features of multiple sets of brain MRI images collected by Doctor A and Doctor B.

[0064] The formula for calculating the intragroup correlation coefficient is as follows: in, The correlation coefficient within the group; The variance between groups is due to inconsistencies in the eigenvalues ​​of different experts in different groups. The variance is within the group; the eigenvalues ​​of different experts in the same group are inconsistent.

[0065] By analyzing the relationship between the intragroup correlation coefficient and a first preset threshold, inconsistent feature values ​​are filtered out, thereby updating the initial image omics features. The first preset threshold is typically set to 0.8, but can be changed according to actual circumstances; no specific limit is imposed here.

[0066] The radiomics features of the patients to be monitored are optimized based on the updated initial radiomics features.

[0067] Here, the content of the radiomics features of the patients to be monitored is optimized by comparing the updated initial radiomics features with the content of the radiomics features.

[0068] In some possible embodiments, the radiomics features include: first-order statistical features, 2D morphological features, 3D morphological features, gray-level co-occurrence matrix features, gray-level size region matrix features, gray-level run length features, neighborhood gray-level difference matrix features, and gray-level correlation matrix features.

[0069] Here, radiomics features refer to the quantitative analysis of information such as grayscale distribution, spatial structure, and texture patterns of tissues in MRI images. This transforms image information into calculable digital features, reflecting the microscopic pathology, structural changes, or functional state of brain tissue. These features provide objective evidence for disease diagnosis, prognostic assessment, and treatment response prediction.

[0070] Example 2 Based on the foregoing embodiments, this application further provides a real-time monitoring system for postoperative neurological function in stroke patients. The system includes various modules and units included in each module, which can be implemented by a processor in an electronic device; of course, it can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), etc.

[0071] Please see Figure 2 , Figure 2This is a structural block diagram of the real-time postoperative neurological function monitoring system for stroke patients provided in this embodiment. The system 200 includes: a data acquisition and preprocessing module 201, used to acquire brain MRI image data, clinical information, and postoperative real-time electroencephalogram (EEG) data of the patient to be monitored, and to perform filtering, noise reduction, and artifact removal to obtain preprocessed corresponding data; a neurophysiological feature extraction module 202, used to extract neurophysiological features of the patient to be monitored based on the preprocessed EEG data; a radiomics feature extraction module 203, used to obtain radiomics features of the patient to be monitored based on the preprocessed brain MRI image data using the PyRadiomics toolkit on the Python platform; and a significant feature screening module 204. 4. A feature data preprocessing module 205 is used to preprocess the selected neurophysiological features and radiomics features using a lasso regression screening algorithm to obtain the corresponding feature data after screening. The feature data preprocessing module 205 is used to preprocess the selected neurophysiological features and radiomics features respectively, and combine minimum absolute contraction and selection operator regression analysis to obtain the quantitative EEG data feature index and radiomics score of the neurophysiological features. The feature data analysis module 206 is used to input the selected clinical information, quantitative EEG data feature index and radiomics score of the patient to be monitored into the trained prediction model to predict the risk index of early neurological function deterioration of the patient to be monitored, and determine the risk level according to the risk index. The prediction model is trained based on a random forest model.

[0072] This system uses a data acquisition and preprocessing module 201 to collect and preprocess brain MRI images, clinical information, and postoperative real-time electroencephalogram (EEG) data from the patient to be monitored. The preprocessed data is then obtained through these preprocessing steps. Next, a neurophysiological feature extraction module 202 and a radiomics feature extraction module 203 extract the neurophysiological and radiomics features of the patient, respectively. A salient feature screening module 204 then filters for salient features, resulting in the filtered neurophysiological and radiomics features. A feature data preprocessing module 205 then obtains the quantitative EEG data feature index and radiomics score for the neurophysiological features. Finally, a feature data analysis module 206 predicts the risk index of early neurological function deterioration in the patient and determines the risk level based on the risk index. This system avoids the inaccurate assessment results caused by traditional methods that rely on a single clinical feature to evaluate neurological function, and enables real-time monitoring of postoperative neurological function in stroke patients.

[0073] In some possible embodiments, the system further includes a display module 207 and an early warning module 208: the display module 207 is used to display in real time the radiomics characteristics, clinical information, postoperative real-time electroencephalogram data, neurophysiological characteristics, and predicted risk level of early neurological function deterioration of the patient to be monitored; at the same time, it is also used to display the number of times the risk level of early neurological function deterioration is high.

[0074] Here, display module 207 is used to display the radiomics characteristics, clinical information, postoperative real-time EEG data, neurophysiological characteristics, and predicted risk level of early neurological deterioration of the patient under monitoring in real time; it also displays the number of times the risk level of early neurological deterioration is high. Feature values ​​can be displayed via pie charts; the display method is not limited here. Display module 207 allows medical staff to quickly grasp the patient's condition by viewing changes in the feature pie chart at any time. Clicking the "Enter Patient Information" button updates the patient's clinical information and radiomics characteristics; the system will then re-analyze the patient's status based on the newly added information.

[0075] The early warning module 208 is used to issue an early warning based on the number of times the risk level of early neurological deterioration is high: when the number of times the risk level of early neurological deterioration is high exceeds a second preset threshold, an early warning is issued.

[0076] Here, the early warning module 208 is used to issue an early warning based on the number of times the risk level of early neurological function deterioration is classified as high: an early warning is issued when the number of times the risk level of early neurological function deterioration is classified as high exceeds a second preset threshold. The second preset threshold is generally set to 5 times, but can be changed according to actual circumstances, and is not limited here. When the number of times the risk level of early neurological function deterioration is classified as high exceeds the second preset threshold, the warning will be displayed on the display module and the warning information will be sent to clinicians and nurses simultaneously. The warning information can be sent to clinicians and nurses via SMS or a medical APP, and the sending method is not limited here.

[0077] It should be noted that the description of the above system embodiments is similar to the description of the above method embodiments, and has similar beneficial effects. For technical details not disclosed in the system embodiments of this application, please refer to the description of the method embodiments of this application for understanding.

[0078] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential 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. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0079] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0080] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0081] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0082] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0083] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause the device automatic test line to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0084] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0085] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0086] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for real-time monitoring of postoperative neurological function in stroke patients, characterized in that, The method includes: collecting brain MRI image data, clinical information, and postoperative real-time electroencephalogram (EEG) data from the patient to be monitored, and performing filtering, noise reduction, and artifact removal to obtain preprocessed corresponding data; extracting neurophysiological characteristics of the patient to be monitored based on the preprocessed EEG data; obtaining radiomics characteristics of the patient to be monitored based on the preprocessed brain MRI image data using the PyRadiomics toolkit on the Python platform; using lasso regression to screen significant features based on clinical information, neurophysiological characteristics, and radiomics characteristics, respectively, to obtain the corresponding feature data after screening; and performing preprocessing on the screened neurophysiological characteristics and screened radiomics characteristics, combined with minimum absolute contraction and selection operator regression analysis, to obtain the quantitative EEG data feature index and radiomics score of the neurophysiological characteristics. Clinical information, quantitative EEG data characteristic index, and radiomics score of the patients to be monitored are input into a trained prediction model to predict the risk index of early neurological deterioration in the patients to be monitored, and the risk level is determined based on the risk index; wherein, the prediction model is trained based on a random forest model.

2. The method for real-time monitoring of postoperative neurological function in stroke patients according to claim 1, characterized in that, The training process of the prediction model is as follows: Connect to the hospital database to obtain all EEG data, clinical information, brain MRI image data, quantitative EEG data characteristic index and radiomics score of all stroke patients after admission within a preset time period, and mark whether their early neurological function has deteriorated to generate training samples; the training samples are divided into training set and test set in a 7:3 ratio by random stratified sampling. The training set is batch-input into the random forest model for training; then the test set is input into the trained random forest model for performance testing until a preset standard is reached, thus obtaining the prediction model.

3. The method for real-time monitoring of postoperative neurological function in stroke patients according to claim 1, characterized in that, The selected neurophysiological features and radiomics features are preprocessed separately, and combined with minimum absolute contraction and selection operator regression analysis to obtain quantitative EEG data feature indices and radiomics scores for neurophysiological features, including: The selected neurophysiological features and radiomics features were subjected to variance analysis, one-way ANOVA, and redundancy analysis to obtain the first feature dimensionality reduction data of neurophysiological features and the second feature dimensionality reduction data of radiomics features. The first feature dimensionality reduction data and the second feature dimensionality reduction data were then subjected to minimum absolute contraction and selection operator regression analysis to obtain the quantitative EEG data feature index and radiomics score of neurophysiological features.

4. The method for real-time monitoring of postoperative neurological function in stroke patients according to claim 1, characterized in that, The neurophysiological features include power spectral density, relative power ratio, brain symmetry index, coherence, and small-world properties corresponding to different frequency bands of EEG signals in the non-lesion hemisphere and lesion hemisphere. The extraction of neurophysiological features of the patient to be monitored based on preprocessed EEG data includes: For each electrode within the non-lesion hemisphere and the lesion hemisphere, the spectral power of the corresponding electrode is calculated by fast Fourier transform, with a frequency range of 1 to 45 Hz. The spectral power is divided into multiple frequency bands based on the frequency range, and the average power spectral density of each frequency band is calculated using the Welch method; the multiple frequency bands include: delta band, theta band, alpha band, beta band, and gamma band; The midline electrode was removed, and the average power spectral density of the non-lesion hemisphere and the lesion hemisphere was calculated using the Welch method. The relative power ratio of the frequency bands is calculated based on the average power spectral density corresponding to each frequency band; the relative power ratio of the frequency bands includes: DAR, TAR and DTABR; Electrode pairs are formed by electrodes in the non-lesion hemisphere and electrodes in the lesion hemisphere that are symmetrical to the electrodes in the non-lesion hemisphere. Based on the frequency band power of the electrode pairs, the brain symmetry index of different frequency bands is calculated using the brain symmetry index calculation formula for different frequency bands. Determine the electrode lead pair consisting of any two electrodes in the non-lesion hemisphere or any two electrodes in the lesion hemisphere. Based on the power spectral density of the electrode lead pair, calculate the coherence of different frequency bands using the coherence calculation formula for different frequency bands. Based on the frequency band power of each electrode in the whole brain, the weighted clustering coefficient and weighted feature path length of each frequency band are obtained by using a brain imaging analysis toolkit and an undirected weighted network. The ratio of the standardized weighted clustering coefficient to the weighted feature path length is used as the small-world property.

5. The method for real-time monitoring of postoperative neurological function in stroke patients according to claim 4, characterized in that, The formulas for calculating the brain symmetry index for different frequency bands are as follows: in, It is a brain symmetry index; , These represent the band power of the i-th electrode pair in the non-lesion hemisphere and the band power in the lesion hemisphere, respectively; N is the number of effective electrode pairs.

6. The method for real-time monitoring of postoperative neurological function in stroke patients according to claim 4, characterized in that, The formulas for calculating the coherence of different frequency bands are as follows: in, The coherence between the x and y electrodes within the cerebral hemisphere; The cross-power spectral density of the x-electrode and the y-electrode leads; and denoted as the self-power spectral density of electrode x and electrode y, respectively; f represents the power corresponding to different frequency bands.

7. The method for real-time monitoring of postoperative neurological function in stroke patients according to claim 1, characterized in that, The method further includes: The system connects to the hospital database to obtain multiple sets of brain MRI images collected by two doctors for multiple identical patients, and obtains the initial radiomics features corresponding to each brain MRI image. Based on the formula for calculating the intragroup correlation coefficient, the intragroup correlation coefficient is obtained. When the intragroup correlation coefficient is less than the first preset threshold, the corresponding radiomics feature value is deleted and the content of the initial radiomics features is updated. The radiomics features of the patients to be monitored are optimized based on the updated initial radiomics features.

8. The method for real-time monitoring of postoperative neurological function in stroke patients according to any one of claims 1 to 7, characterized in that, The image omics features include: first-order statistical features, 2D morphological features, 3D morphological features, gray-level co-occurrence matrix features, gray-level size region matrix features, gray-level run length features, neighborhood gray-level difference matrix features, and gray-level correlation matrix features.

9. A real-time monitoring system for postoperative neurological function in stroke patients, characterized in that, The system includes: The data acquisition and preprocessing module is used to acquire brain MRI image data, clinical information and postoperative real-time electroencephalogram data of the patients to be monitored, and to perform filtering, noise reduction and artifact removal to obtain the corresponding preprocessed data. The neurophysiological feature extraction module is used to extract the neurophysiological features of the patient to be monitored based on the preprocessed electroencephalogram data. The radiomics feature extraction module is used to obtain the radiomics features of the patients to be monitored based on preprocessed brain MRI image data using the PyRadiomics toolkit on the Python platform. The salient feature screening module is used to screen salient features based on clinical information, neurophysiological features and radiomics features using lasso regression screening algorithms to obtain the corresponding feature data after screening. The feature data preprocessing module is used to preprocess the selected neurophysiological features and the selected radiomics features respectively, and combine the minimum absolute contraction and selection operator regression analysis to obtain the quantitative EEG data feature index and radiomics score of the neurophysiological features. The feature data analysis module is used to input the clinical information, quantitative EEG data feature index and radiomics score of the patients to be monitored after screening into the trained prediction model to predict the risk index of early neurological function deterioration of the patients to be monitored, and determine the risk level based on the risk index; wherein, the prediction model is trained based on the random forest model.

10. The real-time monitoring system for postoperative neurological function in stroke patients according to claim 9, characterized in that, The system also includes a display module and an early warning module: The display module is used to display the radiomics characteristics, clinical information, postoperative electroencephalogram data, neurophysiological characteristics, and predicted risk level of early neurological deterioration of the patient to be monitored in real time; it is also used to display the number of times the risk level of early neurological deterioration is high. The early warning module is used to issue warnings based on the number of times the risk level of early neurological function deterioration is high: when the number of times the risk level of early neurological function deterioration is high exceeds a second preset threshold, an early warning is issued.

Images