Autism child risk assessment early warning system and method

Abstract

The application discloses a system and method for autism risk assessment and early warning, and relates to the field of computer and electroencephalogram signal processing. The system comprises an electroencephalogram data acquisition module, a feature extraction module, a risk assessment module and a risk early warning module. The preset task state covers resting state, local motion state and social motion state. In the resting state, the user observes low-speed target objects in the display picture for a preset time length. In the local motion state, the user observes the first type of target person completing local limb motion. In the social motion state, the user observes the second type of target person waving hands to say hello. By collecting electroencephalogram data in each task state and extracting features, the autism risk probability and the corresponding risk level are determined, and the early warning is triggered according to the preset logic rules. The system does not need frequent language interaction with children, is suitable for young children, significantly reduces the screening manpower and time cost, and is convenient for popularization in areas with scarce medical resources.

Description

Technical Field

[0001] This application relates to the interdisciplinary field of computer technology and electroencephalogram (EEG) signal processing, specifically to an early risk assessment and warning system and method for children with autism, applicable to auxiliary screening scenarios for developmental risks in children. Background Technology

[0002] Autism spectrum disorder (ASD) is a group of developmental disorders that commonly occur in childhood. Affected children often exhibit poor social communication and interaction skills, insufficient responsiveness to external social stimuli, repetitive and stereotyped behaviors, and difficulty maintaining attention. In recent years, the detection rate of ASD among children worldwide has been on the rise, placing significant pressure on families and society in terms of parenting and intervention.

[0003] For children at risk of autism spectrum disorder, the earlier the intervention, the more significant the positive impact on their developmental trajectory. Therefore, early identification of autism-related risks in children is of great practical importance. Currently, early screening methods for autism-related risks in children mainly rely on scale assessments and manual observation. Preliminary screening results need to be obtained by professionals through multiple follow-up observations, combined with information such as family developmental history.

[0004] However, existing screening methods have many limitations: on the one hand, the screening process relies heavily on the experience of professionals, is time-consuming and costly, making it difficult to meet the growing screening demand; on the other hand, traditional screening methods require frequent verbal and behavioral interactions with children, making them less suitable for young children and those with language delays, and difficult to widely implement in economically underdeveloped areas with scarce medical resources, resulting in many potentially at-risk children missing the critical window for early intervention. Furthermore, existing screening methods lack objective physiological signals to support their results, making the results highly subjective and hindering large-scale early screening for developmental risks in children. Summary of the Invention

[0005] The technical objective of this invention is to address the technical problems of low efficiency and high subjectivity in existing autism assessment and screening methods by providing a risk assessment and early warning system and method for children with autism.

[0006] To achieve the above technical objectives, this application adopts the following technical solution.

[0007] In a first aspect, embodiments of this application provide a risk assessment and early warning system for children with autism, including:

[0008] The EEG data acquisition module is used to collect the user's EEG data in a preset task state;

[0009] The feature extraction module is used to extract EEG features based on the EEG data collected in the task state.

[0010] a risk assessment module configured to determine a probability of autism risk based on the extracted electroencephalogram features, and determine a risk level according to a preset threshold based on the probability;

[0011] a risk warning module configured to trigger a corresponding warning according to a preset logical rule based on the risk level;

[0012] The task state includes a resting state, a local motion state, and a social motion state.

[0013] In the resting state, the user observes a target object in a display screen at a speed less than a preset value for a preset time length, and a background of the display screen is black.

[0014] In the local motion state, the user observes a first type of target person completing a local limb motion in the display screen for a first sampling time length.

[0015] In the social motion state, the user observes a second type of target person waving an upper limb in the display screen for a second sampling time length.

[0016] In a second aspect, the embodiments of the present application provide a method for risk assessment and early warning of autism children, comprising:

[0017] collecting electroencephalogram data of the user in a preset task state;

[0018] extracting electroencephalogram features based on the electroencephalogram data collected in the task state;

[0019] determining a probability of autism risk based on the extracted electroencephalogram features, and determining a risk level according to a preset threshold based on the probability;

[0020] triggering a corresponding warning according to a preset logical rule based on the risk level;

[0021] The task state includes a resting state, a local motion state, and a social motion state.

[0022] In the resting state, the user observes a target object in a display screen at a speed less than a preset value for a preset time length, and a background of the display screen is black.

[0023] In the local motion state, the user observes a first type of target person completing a local limb motion in the display screen for a first sampling time length.

[0024] In the social motion state, the user observes a second type of target person waving an upper limb in the display screen for a second sampling time length.

[0025] Compared with the prior art, the autism risk assessment and early warning system and method provided by the embodiment has the following beneficial technical effects: the present application avoids the subjectivity of traditional screening relying on artificial experience by integrating resting state, local motion and social motion multi-task state design, combining with the modularized collaborative implementation of electroencephalogram signal acquisition, feature extraction, risk probability and grade evaluation and grading early warning, taking objective electroencephalogram physiological signals as support; without frequent language interaction with children, it is suitable for low-age children and other groups, significantly reduces the screening labor and time cost, and is convenient for popularization in areas with scarce medical resources; through risk grade division and risk early warning output, a universal early childhood autism-related risk early screening scheme is provided, helping to grasp the key window period of early intervention.

[0026] It should be understood that the summary section is not intended to identify key or essential features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0027] The drawings described herein are for illustrative purposes only and are not intended to limit the scope of the present disclosure in any way. In addition, the shapes and scale sizes of the components in the figures are only illustrative, used to help understand the present application, and are not specific limitations on the shapes and scale sizes of the components of the present application. Those skilled in the art can select various possible shapes and scale sizes to implement the present application according to specific circumstances under the guidance of the present application. In the drawings:

[0028] Figure 1 Structure schematic diagram of the autism risk assessment and early warning system provided by the embodiment;

[0029] Figure 2 Flowchart of the autism risk assessment and early warning method provided by the embodiment;

[0030] Figure 3 Schematic diagram of the display screen in the resting state in the embodiment. DETAILED DESCRIPTION

[0031] In order for those skilled in the art to better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0032] The scheme is not directly aimed at obtaining a disease diagnosis result or a health condition, but is only a method for processing electroencephalogram data, and all steps are information processing methods implemented by devices such as computers.

[0033] It should be fully understood that the user information (including but not limited to user physiological sensor information, user personal information, etc.) involved in the present application is information and data authorized by the user or authorized by all parties, and the use of user information should comply with the privacy policies and practices generally considered to meet or exceed the industry for maintaining user privacy, the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards, and provide corresponding operation portal for user to choose authorization or refusal.

[0034] As shown in Figure 1 , the embodiment provides an autism risk assessment and early warning system, including an electroencephalogram data acquisition module, a feature extraction module, a risk assessment module, and a risk warning module.

[0035] The electroencephalogram data acquisition module is used to acquire electroencephalogram data of a user in a preset task state;

[0036] The feature extraction module is used to extract electroencephalogram features based on the electroencephalogram data acquired in the task state;

[0037] The risk assessment module is used to determine an autism risk probability based on the extracted electroencephalogram features, and divide a risk level according to a preset threshold based on the probability;

[0038] The risk warning module is used to trigger a corresponding warning according to a preset logic rule based on the risk level;

[0039] The task state includes a resting state, a local motion state, and a social motion state; in the resting state, the user observes a target object in a display picture with a speed less than a preset value for a preset duration, and the background of the display picture is black; in the local motion state, the user observes a first type of target person completing a local limb motion in a display picture for a first sampling duration; in the social motion state, the user observes a second type of target person waving an upper limb to say hello in a display picture for a second sampling duration.

[0040] Based on the same inventive concept as the autism risk assessment and early warning system provided in the above embodiment, the present embodiment also provides an autism risk assessment and early warning method, as shown in Figure 2 , which includes:

[0041] Step S1: acquiring electroencephalogram data of a user in a preset task state;

[0042] Step S2: extracting electroencephalogram features based on the electroencephalogram data acquired in the task state;

[0043] Step S3: Based on the extracted EEG features, determine the probability of autism risk, and classify the risk level according to a preset threshold based on the probability;

[0044] Step S4: Based on the risk level, trigger the corresponding warning according to the preset logical rules;

[0045] The task state includes the resting state, the local motion state, and the social motion state;

[0046] In a resting state, the user observes a target object on the display screen with a speed less than a preset value for a preset duration, and the background of the display screen is black.

[0047] In the local motion state, the user observes the first type of target person in the display screen complete a local limb movement for the first sampling duration;

[0048] In the social movement state, the user observes the second type of target person in the display screen moving their upper limbs to greet others, for a continuous second sampling time.

[0049] In this embodiment, during the EEG data acquisition preparation phase, the user wears an EEG acquisition device to collect brain activity signals, including those from the left and right motor brain regions (typically C3 and C4) and the frontal lobe region. The EEG acquisition device, such as an EEG cap, can be a dry electrode, a wet electrode, or a gel electrode.

[0050] The significance of brain region segmentation lies in reducing individual electrode noise, reducing the number of features to a certain extent, and improving the effectiveness of the system in small sample situations.

[0051] In one implementation, it may include not only the frontal lobe, left brain region, and right brain region, but also more brain regions, such as the central region and occipital region. Alternatively, it may calculate the features of each electrode separately without distinguishing brain regions, so as to increase the number of features and improve the system performance in conjunction with the increase in the number of samples.

[0052] During the resting state, the user needs to keep their eyes open and observe the display screen (such as a screen). The display screen includes multiple target objects (such as moving light spots), and the preset duration is 1 minute.

[0053] In some embodiments, in the resting state, the displayed screen is as follows: Figure 3 As shown, the overall background of the screen is black, each of the target objects is a solid color, and there are multiple target objects of different shapes (such as color blocks, dots, etc.) in the screen. The target objects have different colors and can also be of different sizes, and they move slowly (the movement speed is less than the preset value).

[0054] Slowly moving objects add a sense of dynamism to the image, which is attractive to children, especially those with autistic tendencies, and helps them stay calm.

[0055] EEG data collected in the resting state will be used as a benchmark for comparison with EEG data from subsequent stages (local motor state and social motor state). The slowly moving target object induces a series of EEG responses related to the observed movement, which are different from the EEG responses generated by mirror movement in subsequent stages. These responses will be used as baselines to better identify the movement of mirror neurons.

[0056] In some embodiments, during the local motion state acquisition phase, the user observes the display screen, which shows the palm and wrist of an adult (aged 18 or older) as the first type of target person. In the video, the palm repeatedly clenches and unclenches, with each clenching and unclenching cycle lasting for a first interval duration (which can be set to 5 seconds), and the entire cycle lasting for a first sampling duration (e.g., 1 minute).

[0057] The inclusion of a localized motion state acquisition stage allows children to observe hand movements in a controlled environment. Typical children, when observing others' movements, especially hand movements, instinctively and spontaneously understand the intention behind the action. Simultaneously, specific brain regions (such as the inferior frontal gyrus and inferior parietal lobule) are significantly activated, with mirror neurons firing, providing a foundation for imitation, empathy, and social interaction. Children with autism often lack this type of feedback. Related research indicates that mirror motor neurons in children with autism are weaker than those in typical children.

[0058] Furthermore, hand movements are more varied than those of the upper limbs, lower limbs, and hands, making them more conducive to eliciting mirror neuron feedback.

[0059] In the embodiments, in addition to the above actions, other hand movements can also be used as mirror stimuli, such as waving to greet someone, extending one or more fingers, and then making a fist, or putting both hands together.

[0060] In the social activity data collection phase, users observe a display screen showing children of the same age as the user waving and greeting the screen with one or both hands. The children appear in batches, with one or more children in each batch. Optionally, some batches may contain only one child, while others may contain multiple children. Optionally, each segment lasts 10 seconds. Six segments are provided, for a total second sampling duration of one minute.

[0061] The social activity data collection phase allows users to observe the actions of children of the same age, such as waving, during complete social interactions. Emphasizing motor observation in social situations allows for deeper testing of children's mirror neuron system and mental network feedback. Compared to typically developing peers, autistic children tend to avoid social interaction; social scenarios provide a more comprehensive way to detect EEG differences between typically developing and autistic children.

[0062] In the example, children of the same age waved at the screen, emphasizing hand movements, in order to better induce mirror neurons that are sensitive to movement.

[0063] In other embodiments, in addition to waving gestures, related actions include passing items between multiple children, a series of behaviors with obvious hand gestures and social attributes.

[0064] like Figure 1 and Figure 2 As shown, step S2 extracts EEG features based on EEG data collected in the task state, including: step S21, obtaining the first feature set: obtaining the first feature set based on EEG data collected in the resting state, local motion state and social motion state.

[0065] In this embodiment, the data preprocessing of the collected raw EEG data is also included, including filtering the EEG data, removing artifacts, and grouping the EEG data of different tasks to form a first feature set.

[0066] As an example, EEG data acquired under different task states were preprocessed to form a first feature set. Preprocessing included: filtering: applying a 3rd-order Butterworth filter to the data from 1-30Hz; time window segmentation: removing the first 5 seconds of signal and segmenting the data into time windows. The time window length was 5 seconds, sliding at 1-second intervals, resulting in a total of 51 data windows; artifact removal: discarding windows with absolute amplitude values ​​exceeding 100μV.

[0067] Step S22: Obtain the second feature set: For each time window, calculate the spectral energy curve of each lead based on the first feature set, and average the spectral energy curves of the left motor brain region, right motor brain region and frontal lobe region under each task state to obtain the average spectral energy curve of each brain region under each task state for each time window.

[0068] The average curves of the spectral energy of the sub-brain regions in all effective time windows of the resting state are then globally averaged to obtain the global average curves of the spectral energy of the sub-brain regions corresponding to the left motor brain region, the right motor brain region, and the frontal lobe region.

[0069] The average spectral energy curve of each brain region in the local motion state is subtracted from the global average spectral energy curve of the corresponding brain region in the resting state. The average spectral energy curve of each brain region in the social motion state is subtracted from the global average spectral energy curve of the corresponding brain region in the resting state to obtain two types of brain region difference curves. The two types of brain region difference curves together constitute the second feature set.

[0070] In this embodiment, the energy curves represent the intensity of EEG energy at different frequencies. When averaging the lead spectral energy curves of the left motor brain region, right motor brain region, and frontal lobe region under each task state, if a certain EEG region has only a single electrode, then no averaging process is required.

[0071] In some embodiments, after obtaining the average curve of the spectral energy of the brain region for each time window under each task state, the method further includes smoothing the average curve of the spectral energy of the brain region.

[0072] In this embodiment, the average curve (energy / frequency) of the averaged brain region spectrum energy is smoothed. The smoothing process increases the robustness of the algorithm by filtering out some interference and spectrum leakage.

[0073] After processing, each time window contains only three sets of spectral data, representing the left motion region, right motion region, and forehead region, respectively.

[0074] The average spectral energy curves of different brain regions in the resting state for all effective time windows were then globally averaged to obtain three baselines, each representing a different brain region.

[0075] Step S23: Based on the average spectral energy curve of the resting-state frontal lobe region, calculate the ratio of α-rhythm energy to β-rhythm energy for each time window. Based on the brain region sub-region difference curve of the second feature set, calculate the ratio of α-rhythm energy to β-rhythm energy for each time window in the local motor state and social motor state, respectively. Based on the average spectral energy curve of the resting-state brain region and the brain region difference curve of the second feature set, perform nonlinear fitting on the curves of the left motor region and right motor region for each time window, calculate the μ-rhythm peak value, and finally obtain the initial feature set.

[0076] The α rhythm energy is the energy area in the 8-12Hz frequency range of the frontal lobe region spectral data, and the β rhythm energy is the energy area in the 14-28Hz frequency range of the frontal lobe region spectral data.

[0077] Step S24: Obtain the final third feature set: For each task state, calculate the average value and standard deviation of the feature values ​​corresponding to the effective time window in the initial feature set, and filter out abnormal feature values ​​that exceed the range of ±k times (±3 times) the average value and standard deviation. This step can be performed a total of 3 times. Finally, the average ratio of α rhythm energy to β rhythm energy, the average value of μ rhythm peak value in the left motion region, and the average value of μ rhythm peak value in the right motion region are obtained under different task states, forming the final third feature set.

[0078] Through steps S21-S24, a total of 9 feature values ​​are obtained for each user in the three sampling stages of resting state, local motion state, and social motion state, including 3 average α / β values ​​in the forehead region and 6 average values ​​of motion-related potential peaks in the left and right motion regions. These are normalized to obtain the third feature set.

[0079] As an example, EEG signals are collected without distinguishing brain regions, and EEG features are extracted based on EEG data collected in a task state, including: obtaining a first feature set: obtaining a first feature set based on EEG data collected in a resting state, a local motion state, and a social motion state;

[0080] Obtain the second feature set: For each time window, calculate the spectral energy curve of each lead based on the first feature set, and perform a global average of the spectral energy curves of all leads under each task state to obtain the average spectral energy curve of all leads for each time window under each task state; perform a global average of the average spectral energy curves of all leads for all effective time windows in the resting state to obtain the global average spectral energy curve of the resting state; subtract the global average spectral energy curve of the resting state from the average spectral energy curve of the resting state for each time window under the local motion state; subtract the global average spectral energy curve of the resting state from the average spectral energy curve of the resting state for each time window under the social motion state to obtain two types of spectral energy difference curves; the two types of spectral energy difference curves together constitute the second feature set;

[0081] Obtaining the initial feature set: Based on the average energy curve of the resting-state full-lead spectrum, calculate the ratio of α-rhythm energy to β-rhythm energy for each time window; based on the full-lead difference curve of the second feature set, calculate the ratio of α-rhythm energy to β-rhythm energy for each time window under local motion and social motion states respectively; based on the average energy curve of the resting-state full-lead spectrum and the full-lead difference curve of the second feature set, perform nonlinear fitting on the full-lead curve of each time window (the fitting object for resting-state is the average energy curve of the full-lead spectrum, and the fitting object for local motion / social motion is the full-lead difference curve), calculate the μ-rhythm peak value, and finally obtain the initial feature set; Obtaining the final third feature set: For each task state, calculate the mean and standard deviation of the feature values ​​corresponding to the effective time windows in the initial feature set; filter out abnormal feature values ​​that exceed the range of mean ± 3 times the standard deviation; finally obtain the average ratio of α-rhythm energy to β-rhythm energy and the average value of the full-lead μ-rhythm peak value under different task states, forming the final third feature set.

[0082] In some embodiments, the feature values ​​of the three sampling stages—resting state, local motion state, and social motion state—may also include the α rhythm energy and / or β rhythm energy of the corresponding stages.

[0083] In the embodiment, step S3 involves inputting the features of the third feature set into the pre-trained risk assessment module to obtain the autism risk probability.

[0084] In this embodiment, the risk assessment module predicts the probability of a user having autism.

[0085] In this embodiment, users can be divided into 6 groups based on their age and gender. A corresponding classifier and a corresponding autism prediction probability table are selected based on the user's gender and age. When the probability is higher than 50%, the probability is output.

[0086] As an example, if the probability of a user having autism is less than 50%, it means that the user has a low risk of having autism. The corresponding classifier and probability table are selected according to the user's gender and age, and the third feature set is sent to the risk assessment module. The probability of the user having autism in the future is obtained according to the autism prediction probability table, and the corresponding autism risk level is assessed.

[0087] In this embodiment, the spectral energy curve of the second feature set is modeled using nonlinear fitting to calculate the μ-rhythm peak value at different acquisition stages. The nonlinear fitting model used is as follows:

[0088] Y = f1(x) + f2(x) + c0;

[0089] Where Y is the fitted spectrum curve, c0 is the parameter term, the baseline model f1(x) = -a1log2(a2x) and the normal distribution model f2(x), and a1 and a2 are the fitting parameters to be sought;

[0090] ;

[0091] a3 and a4 are the fitting parameters we need to find.

[0092] In this embodiment, the spectral energy curve is fitted, with a1 ranging from (0, 10) with a value interval of 0.05; a2 ranging from (0, 1) with a value interval of 0.01; a3 ranging from (0, 20) with a value interval of 0.1; a4 ranging from (7, 13) with a value interval of 0.1; and c0 ranging from (0, 10) with a value interval of 0.1. The root mean square error (RMSE) of the evaluation function is used, employing an optimization algorithm to find the optimal parameters that minimize the fit between the model-fitted curve and the actual spectral curve, thus minimizing the RMSE.

[0093] In this embodiment, the formula for calculating the peak value of motion-related μ-rhythm potentials is as follows:

[0094] v = f2(a4) - f1(a4);

[0095] Where v is the peak potential, and a4 is the fitted value, representing the frequency center of the motion-related potential, which is also the frequency point where the peak value of the motion-related potential unit is located. Normally v>0; if not, the window data is invalid, and all feature data in that window are discarded.

[0096] In this embodiment, the test is invalid if any of the following conditions are met:

[0097] During the preprocessing stage, after discarding data due to artifacts, the remaining windows are less than 2 / 3 of the total number of windows in this task phase.

[0098] During the preprocessing stage, after discarding data due to peak value v≤0, the remaining window is less than 2 / 3 of the total number of windows in this task state stage;

[0099] When filtering feature values, the number of remaining feature values ​​after filtering is less than 2 / 3 of the total number of windows in the task state.

[0100] In this embodiment, the training method for the risk assessment module is as follows: healthy children with autism are invited to participate in the data collection process, a third feature set is obtained according to the above method, and the children's age, gender and autism diagnosis data are collected to obtain the first training set.

[0101] The first training set was divided into three stages based on age (0-2, 2-6, 6-10) and into six sub-training sets based on gender. Each training set was labeled with whether the user had autism.

[0102] Train the classifier on different training subsets separately, and perform the following steps:

[0103] Using the above data, each classifier was trained. The training process was based on existing technology. The model input features were 6, and the output dimension was 2, representing the scores y1 for autism and y2 for health, respectively. y1 and y2 were used as the first classification output.

[0104] The method for converting scores into disease probabilities is as follows:

[0105] 1) Feed all samples into the trained classifier to obtain the first classification output for each sample;

[0106] 2) Calculate log(y1 / y2) as the second category output;

[0107] 3) Divide the second-class output into 10 intervals from largest to smallest. If the second-class output of a sample exceeds the minimum threshold of an interval, the sample is judged to have autism. Calculate the classification accuracy p for each interval. The formula for classification accuracy is as follows:

[0108] ;

[0109] Wherein, TP represents the number of samples with autism that were detected by classification, and FP represents the number of healthy samples that were detected by classification.

[0110] 4) Continuously fine-tune the start and end values ​​of each interval to ensure that the difference in classification accuracy p between adjacent intervals is around 10%, thus obtaining the autism inference probability table. In the autism inference probability table, the probability of having autism in the first interval is 0-10%, the probability of having autism in the second interval is 11%-20%, and so on.

[0111] In addition, for children with autism, data before and after the onset of the condition were collected, and a third feature set was obtained using the method described above. This feature set was then combined with the child's age, gender, and autism diagnosis data to obtain a second training set. Through the above training method, self-classifiers and autism prediction probability tables for six subsets of the second training set were obtained.

[0112] In this embodiment, the risk warning module triggers corresponding warnings through preset logical rules, as follows:

[0113] Low risk: No active warning is triggered; only the risk probability and a "recommendation for routine child health follow-up" message are presented in the output results.

[0114] Medium risk: Triggers a Level 1 warning, presented in the form of visual prompts and text descriptions, emphasizing "It is recommended to complete a professional child development assessment within 3 months";

[0115] High risk: Triggers a Level 2 warning, presented in the form of visual prompts + text descriptions + voice broadcast (optional), clearly stating "It is recommended to go to a medical institution with autism diagnosis and treatment qualifications for a comprehensive evaluation within 1 month".

[0116] In the embodiments, visual output, text output, or text output can be used. Visual output may include color differentiation (low risk green, medium risk yellow, high risk orange), or a pie chart to display the probability of risk.

[0117] The risk assessment and early warning system and method for children with autism provided in this application do not require frequent verbal interaction with children, are suitable for young children, significantly reduce the manpower and time costs of screening, are easy to promote in areas with scarce medical resources, and provide a universally applicable auxiliary screening program through risk warning output. It is only used as an auxiliary screening tool to help grasp the critical window period for early intervention.

[0118] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, or a tablet computer, or any combination of these devices.

[0119] The above provides a detailed description of the risk assessment and early warning system and method for children with autism provided in this application. Specific examples have been used in this paper to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the concept of this application and should not be construed as a limitation on the scope of protection of this application.

Claims

1. A risk assessment and early warning system for children with autism, characterized in that, include: The EEG data acquisition module is used to collect the user's EEG data in a preset task state; The feature extraction module is used to extract EEG features based on the EEG data collected in the task state. The risk assessment module is used to determine the probability of autism risk based on the extracted EEG features, combined with the user's age and gender, and classify the risk level according to the probability based on a preset threshold. The risk warning module is used to trigger corresponding warnings based on the risk level and according to preset logical rules. The task states mentioned include resting state, localized movement state, and social-movement state, and each task state is adapted to the cognitive characteristics of children aged 3-10 years. In the resting state, the user observes a target object on the display screen with a speed less than a preset value for a preset duration. The display screen includes multiple target objects, and each target object has a different color. In the local motion state, the user observes the first type of target person in the display screen complete a local limb movement for a first sampling time, where the first type of target person is an adult. In the social movement state, the user observes the screen displaying the second type of target person moving their upper limbs to greet others for a second sampling period. The second type of target person is a child of the same age as the user. Based on the EEG data collected in the task state, EEG features are extracted, including: Obtaining the first feature set: Based on EEG data collected in the resting state, local motion state, and social motion state, the first feature set is obtained; Obtain the second feature set: For each time window, calculate the spectral energy curve of each lead based on the first feature set, and average the spectral energy curves of the left motor brain region, right motor brain region and frontal lobe region under each task state to obtain the average spectral energy curve of each brain region under each task state for each time window. The average curves of the spectral energy of the sub-brain regions in all effective time windows of the resting state are then globally averaged to obtain the global average curves of the spectral energy of the sub-brain regions corresponding to the left motor brain region, the right motor brain region, and the frontal lobe region. The average spectral energy curve of each brain region in the local motion state is subtracted from the global average spectral energy curve of the corresponding brain region in the resting state. The average spectral energy curve of each brain region in the social motion state is subtracted from the global average spectral energy curve of the corresponding brain region in the resting state to obtain two types of brain region difference curves. The two types of brain region difference curves together constitute the second feature set. Obtain the initial feature set: Based on the average spectral energy curve of the resting frontal lobe region, calculate the ratio of α rhythm energy to β rhythm energy for each time window. Based on the brain region difference curve of the frontal lobe region of the second feature set, calculate the ratio of α rhythm energy to β rhythm energy for each time window under local motor state and social motor state respectively. Based on the average energy curve of the resting state brain region spectrum and the brain region difference curve of the second feature set, nonlinear fitting is performed on the curves of the left motor region and the right motor region of each time window to calculate the μ rhythm peak and finally obtain the initial feature set. Obtain the final third feature set: For each task state, calculate the average value and standard deviation of the feature values ​​corresponding to the effective time window in the initial feature set, and filter out abnormal feature values ​​that exceed the range of plus or minus k times the standard deviation of the average value; finally, obtain the average ratio of α rhythm energy to β rhythm energy, the average value of μ rhythm peak value in the left motion region and the average value of μ rhythm peak value in the right motion region under different task states, forming the final third feature set.

2. The autism risk assessment and early warning system according to claim 1, characterized in that, In the resting state, the background of the displayed screen is black, and each of the target objects is a solid color.

3. The autism risk assessment and early warning system according to claim 1, characterized in that, In the local motion state, the first type of target person in the display screen completes local limb movement: the palm and wrist of the first type of target person are displayed. The palm is repeatedly clenched and unclenched, with each clenching and unclenching cycle lasting the first interval duration, and the entire cycle lasting the first sampling duration.

4. The autism risk assessment and early warning system according to claim 1, characterized in that, In the social movement state, the scene of the second type of target person waving with their upper limbs includes: children appearing in batches in the display screen, with one or more children in each batch, and each batch of children raising one or both hands to wave and greet.

5. The autism risk assessment and early warning system according to claim 1, characterized in that, The process of obtaining the first feature set includes: sequentially performing 1-30Hz and 3rd order Butterworth filtering on the EEG data, dividing it into time windows, removing artifacts, and grouping it according to different task states.

6. The autism risk assessment and early warning system according to claim 1, characterized in that, After obtaining the average curve of the spectral energy of the brain region for each time window under each task state, the method further includes smoothing the average curve of the spectral energy of the brain region.

7. The autism risk assessment and early warning system according to claim 1, characterized in that, The α rhythm energy is the energy area in the 8-12Hz frequency range of the frontal lobe region spectral data, and the β rhythm energy is the energy area in the 14-28Hz frequency range of the frontal lobe region spectral data.

8. The autism risk assessment and early warning system according to claim 1, characterized in that, The nonlinear fitting model is as follows: Y = f1(x) + f2(x) + c0; Where Y is the fitted spectrum curve, x is the frequency, c0 is the parameter, the baseline model f1(x) = -a1log2(a2x) and the normal distribution model f2(x), and a1 and a2 are the fitting parameters to be sought; ； a3 and a4 are the fitting parameters we need to find.

9. A computer-implemented method for processing electroencephalogram (EEG) information, characterized in that, include: Collect the user's EEG data under a preset task state; EEG features are extracted based on the EEG data collected under the aforementioned task state; Based on the extracted EEG features, combined with the user's age and gender, the probability of autism risk is determined, and the risk level is divided according to a preset threshold based on the probability. Based on the risk level, a corresponding warning is triggered according to preset logic rules; The task states mentioned include resting state, localized movement state, and social-movement state, and each task state is adapted to the cognitive characteristics of children aged 3-10 years. In the resting state, the user observes a target object on the display screen with a speed less than a preset value for a preset duration. The display screen includes multiple target objects, and each target object has a different color. In the local motion state, the user observes the first type of target person in the display screen complete a local limb movement for a first sampling time, where the first type of target person is an adult. In the social movement state, the user observes the screen displaying the second type of target person moving their upper limbs to greet others for a second sampling period. The second type of target person is a child of the same age as the user. Based on the EEG data collected in the task state, EEG features are extracted, including: Obtaining the first feature set: Based on EEG data collected in the resting state, local motion state, and social motion state, the first feature set is obtained; Obtain the second feature set: For each time window, calculate the spectral energy curve of each lead based on the first feature set, and average the spectral energy curves of the left motor brain region, right motor brain region and frontal lobe region under each task state to obtain the average spectral energy curve of each brain region under each task state for each time window. The average curves of the spectral energy of the sub-brain regions in all effective time windows of the resting state are then globally averaged to obtain the global average curves of the spectral energy of the sub-brain regions corresponding to the left motor brain region, the right motor brain region, and the frontal lobe region. The average spectral energy curve of each brain region in the local motion state is subtracted from the global average spectral energy curve of the corresponding brain region in the resting state. The average spectral energy curve of each brain region in the social motion state is subtracted from the global average spectral energy curve of the corresponding brain region in the resting state to obtain two types of brain region difference curves. The two types of brain region difference curves together constitute the second feature set. Obtain the initial feature set: Based on the average spectral energy curve of the resting frontal lobe region, calculate the ratio of α rhythm energy to β rhythm energy for each time window. Based on the brain region difference curve of the frontal lobe region of the second feature set, calculate the ratio of α rhythm energy to β rhythm energy for each time window under local motor state and social motor state respectively. Based on the average energy curve of the resting state brain region spectrum and the brain region difference curve of the second feature set, nonlinear fitting is performed on the curves of the left motor region and the right motor region of each time window to calculate the μ rhythm peak and finally obtain the initial feature set. Obtain the final third feature set: For each task state, calculate the average value and standard deviation of the feature values ​​corresponding to the effective time window in the initial feature set, and filter out abnormal feature values ​​that exceed the range of plus or minus k times the standard deviation of the average value; finally, obtain the average ratio of α rhythm energy to β rhythm energy, the average value of μ rhythm peak value in the left motion region and the average value of μ rhythm peak value in the right motion region under different task states, forming the final third feature set.

Images