KAN network-based infantile autism assessment scale numerical regression method

Abstract

The invention provides a KAN network-based autism assessment scale numerical regression method, which realizes high-precision prediction of autism patient scale scores by collecting multichannel electroencephalogram (EEG) data, combining with behavior scale scores and utilizing feature extraction and modeling capabilities of a KAN network, improves the interpretability of a model, and improves the prediction accuracy of the autism assessment scale numerical regression. And the influence of each feature on scale scoring can be visually displayed. According to the method, the accuracy and consistency of scale scoring are effectively improved, dependence on manual evaluation of experts is reduced, and the method has wide clinical application potential and can be used for early screening, diagnosis support and personalized intervention analysis of autism.

Description

Technical Field

[0001] The present invention relates to the technical field of electroencephalogram (EEG) signal processing, and in particular to a numerical regression method for an autism assessment scale based on a KAN network. Background Art

[0002] Autism Spectrum Disorder (ASD) is a common neurodevelopmental disease, whose main features include social interaction disorders, language communication difficulties and repetitive stereotyped behaviors. With the increase in the incidence of ASD in recent years, early screening and personalized diagnosis play a vital role in the rehabilitation and intervention of patients with autism. At present, the diagnosis and evaluation of ASD mostly rely on the observation and scale evaluation of clinical experts. Autism assessment scales (such as ADOS, CARS, SRS, etc.) are commonly used clinical diagnostic tools that can quantify patients' social, language and behavioral performance. Although scale evaluation has a high diagnostic value, its operation is complicated and relies on the subjective judgment of experts, making it difficult to adapt to large-scale screening and standardized diagnosis needs.

[0003] In recent years, with the development of artificial intelligence technology, numerical regression models have gradually been introduced into the field of prediction and analysis of autism scale scores to solve the limitation of existing scale evaluation relying on experts and improve diagnostic efficiency and consistency. Numerical regression models can automatically generate scale scores or perform predictive analysis by analyzing patients' multimodal data, such as behavioral characteristics, language characteristics, and EEG signals. However, traditional machine learning regression methods usually have a high dependence on data features and lack model interpretability, which makes it difficult to apply the model to the prediction of autism scale scores. Summary of the invention

[0004] The present invention provides a KAN network-based autism assessment scale numerical regression method to solve the defects in the prior art.

[0005] In a first aspect, the present invention provides a numerical regression method for an autism assessment scale based on a KAN network, comprising: Collecting EEG signal data of the target subject; Preprocessing the EEG signal data to obtain preprocessed EEG signal data; Extracting EEG features of the preprocessed EEG signal data; The EEG features are used to train and optimize the KAN network to generate a numerical regression model for the autism assessment scale; The regression results output by the numerical regression model of the autism assessment scale are analyzed for interpretability to obtain key EEG channel features of the scale score.

[0006] According to a KAN network-based autism assessment scale numerical regression method provided by the present invention, EEG signal data of a target subject is collected, including: Collect high-density 62-channel 1000Hz resting-state EEG signal data; The autism assessment scale was used to quantitatively score the behavioral characteristics of the subjects to form structured scale data.

[0007] According to a KAN network-based autism assessment scale numerical regression method provided by the present invention, the EEG signal data is preprocessed to obtain preprocessed EEG signal data, including: Downsampling, filtering and removing artifacts on the EEG signal data in sequence to obtain filtered EEG signal data; Using a fixed time window to segment the screened EEG signal data into a plurality of EEG segments; Performing discrete Fourier transform and bandpass filtering on the multiple EEG segments respectively to obtain EEG frequency domain signals, and retaining EEG time domain signals; The EEG frequency domain signal and the EEG time domain signal are randomly mixed according to the subjects and divided into a training set, a validation set and a test set.

[0008] According to a KAN network-based autism assessment scale numerical regression method provided by the present invention, the EEG features of the pre-processed EEG signal data are extracted, including: Based on the time domain and frequency domain methods, the neural activity features of the preprocessed EEG signal data are extracted, and the EEG features are constituted by the neural activity features.

[0009] According to a KAN network-based autism assessment scale numerical regression method provided by the present invention, the KAN network is trained and optimized using the EEG features to generate an autism assessment scale numerical regression model, including: Determining that the KAN network includes an input layer, a plurality of hidden layers and an output layer; Using the EEG features, the KAN network is trained with minimizing mean square error as a model objective; The parameters of the KAN network are optimized by back propagation and gradient descent method to obtain the numerical regression model of the autism assessment scale.

[0010] According to a KAN network-based autism assessment scale numerical regression method provided by the present invention, an interpretability analysis is performed on the regression results output by the autism assessment scale numerical regression model to obtain key EEG channel features of the scale score, including: Obtaining the regression result of the autism assessment scale numerical regression model; The regression results are used to analyze the contribution of different features to the scale score, and the key EEG channel features are identified by visualizing the KAN weights and function importance; Weighted visualization is used to highlight the key EEG channel features.

[0011] In a second aspect, the present invention further provides a KAN network-based autism assessment scale numerical regression system, comprising: An acquisition module, used to acquire EEG signal data of a target subject; A preprocessing module, used for preprocessing the EEG signal data to obtain preprocessed EEG signal data; An extraction module, used for extracting EEG features of the preprocessed EEG signal data; A training module, used to train and optimize the KAN network using the EEG features to generate a numerical regression model for an autism assessment scale; The regression module is used to perform interpretable analysis on the regression results output by the numerical regression model of the autism assessment scale to obtain key EEG channel features of the scale score.

[0012] In a third aspect, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the numerical regression method for an autism assessment scale based on a KAN network as described above is implemented.

[0013] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a numerical regression method for an autism assessment scale based on a KAN network as described in any one of the above.

[0014] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-described KAN network-based autism assessment scale numerical regression methods.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This paper uses the structured design and nonlinear decomposition capabilities of the KAN network to enable the model to effectively capture the complex features in the EEG data of autistic patients and achieve high-precision scale score prediction. Through feature selection and optimization of the regression model, the model can maintain excellent prediction performance even with fewer data samples, improving the adaptability and practicality of the algorithm; This invention not only improves the prediction accuracy of the scale score, but also significantly enhances the interpretability of the model. By analyzing the feature weights and importance in the KAN network, it can intuitively display the impact of different EEG features on the scale score, making it easier for clinicians to understand the decision-making process of the model, and providing a more transparent and feasible reference for personalized intervention and diagnosis; The lightweight design of the present invention reduces the demand for computing resources and consumes fewer computing resources than traditional neural networks, which allows the model to be deployed on lower-configuration hardware and is suitable for practical application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of a numerical regression method for autism assessment scale based on KAN network provided by the present invention; Figure 2 It is the overall structural principle diagram provided by the present invention; Figure 3 It is a structural schematic diagram of the KAN network-based autism assessment scale numerical regression system provided by the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] In view of the problems existing in the prior art, the present invention proposes a numerical regression method for autism assessment scale based on Kolmogorov-Arnold Networks (KAN). The method uses the KAN network to effectively model the patient's EEG data and scale characteristics. The method can not only automatically extract and aggregate the key features of the EEG data, but also provide good model interpretability, thereby providing a scientific basis for the assessment and personalized intervention of autistic patients. The present invention effectively solves the application limitations of traditional models in large-scale screening through the KAN network, improves the efficiency and accuracy of autism assessment, and provides an innovative and practical solution for the early diagnosis and intervention of autism.

[0020] Figure 1 is a flow chart of a numerical regression method for autism assessment scale based on KAN network provided by an embodiment of the present invention, such as Figure 1 As shown, including: Step 100: Collecting EEG signal data of a target subject; Step 200: preprocessing the EEG signal data to obtain preprocessed EEG signal data; Step 300: extracting EEG features of the preprocessed EEG signal data; Step 400: Using the EEG features, training and optimizing the KAN network to generate a numerical regression model for an autism assessment scale; Step 500: Perform interpretability analysis on the regression results output by the numerical regression model of the autism assessment scale to obtain key EEG channel features of the scale score.

[0021] Specifically, the numerical regression method for autism assessment scale based on KAN network proposed in the embodiment of the present invention comprises the following steps: Step 1: Data Collection This method uses high-density 62-channel EEG data and autism assessment scale data as the data source for model training and prediction. The EEG data acquisition frequency is set to 1000Hz to ensure high temporal resolution. The scale data contains multiple sub-scores related to ASD symptoms to quantify the patient's social, language and behavioral performance.

[0022] Step 2: Data Preprocessing For EEG signals and scale data, the preprocessing steps include the following parts: EEG signal processing: Use downsampling, bandpass filtering and artifact removal algorithms to remove artifact noise such as electrooculography and electromyography to improve signal quality.

[0023] Split time window: Split the preprocessed EEG signal into several time segments according to a fixed time window for feature extraction.

[0024] Frequency domain conversion: Perform discrete Fourier transform on the time segment data to highlight the frequency domain features.

[0025] Scale data matrixization: The autism assessment scale data is converted into matrix form and standardized so that it can be combined with EEG data features to facilitate subsequent model input.

[0026] Step 3: EEG feature extraction In the preprocessed EEG data, multiple features in the time domain and frequency domain are extracted, including neural activity features such as power spectrum, entropy, and phase coupling, to construct a rich feature set. This feature set and the scale data matrix together constitute the input feature.

[0027] Step 4: KAN network model construction and training Network architecture initialization: Based on the multi-layer structure of the KAN network, the initial model includes an input layer, multiple hidden layers, and an output layer, which aims to decompose and regress complex nonlinear features.

[0028] Training and optimization: The KAN network is trained using input feature data. The model objective is to minimize the mean squared error (MSE) to improve the accuracy of scale score prediction.

[0029] Parameter optimization: The parameters of the KAN network are optimized through back propagation and gradient descent method, and finally a stable regression model is generated.

[0030] Step 5: Interpretability Analysis Based on the regression results of the KAN network, the contribution of each feature to the scale score is analyzed, and the symptom representations corresponding to different features are identified. By analyzing the weights and feature importance of the KAN network, the key EEG features that affect the scale score are identified, providing diagnostic support and personalized intervention reference for clinicians.

[0031] The present invention uses the KAN network to perform deep feature extraction and modeling of multi-channel EEG signals and scale data to efficiently predict the scale scores of autistic patients, thereby achieving accurate evaluation and providing interpretable analysis for clinical practice.

[0032] Based on the above embodiment, step 1 includes: 1.1 EEG signal acquisition process EEG signal collection is usually carried out in a professional EEG collection room. Sensors are attached to the surface of the subject's scalp to record the brain's electrophysiological activity data in resting or task states, and the real-time collected signals are transmitted to a computer for storage.

[0033] 1.2 Preliminary preparation and environment setup To ensure the quality of signal collection, the subjects are required to remove items that may affect signal collection, such as glasses, earrings, etc. At the same time, during the collection process, the subjects will be given a headgear made of special materials to ensure that the electrodes can fit the scalp tightly and reduce external interference.

[0034] 1.3 Electrode arrangement and conductive treatment The electrodes corresponding to the 62 channels, a reference electrode and a ground wire were passed through the headgear respectively, so that the electrodes were closely attached to the scalp surface of the subject. In order to further reduce the resistance and improve the stability of signal acquisition, conductive paste was applied to the contact position.

[0035] 1.4 Signal acquisition and recording The collection process requires the subject to maintain steady breathing, sit quietly, and keep their eyes open for 3-5 minutes. The collected EEG signals will be stored together with the subject's basic information for subsequent analysis.

[0036] 1.5 Scale data collection Autism assessment scales (such as ADOS, CARS, SRS, etc.) are used to quantify the behavioral characteristics of the subjects to form structured scale data. The scale data includes scores in multiple dimensions such as social skills, speech development, and behavioral characteristics.

[0037] Based on the above embodiment, step 2 includes: 2.1 Data downsampling and filtering First, the collected EEG data is downsampled to 256Hz to enhance its anti-noise ability and reduce storage and computing resource consumption. Next, a bandpass filter (0.5-45Hz) is used to remove interference signals in non-EEG frequency bands, effectively retaining the main components of EEG activity. In addition, the artifact removal algorithm is used to remove artifact signals such as electrooculography and electromyography in the data to ensure the purity of the data.

[0038] 2.2 Data fragmentation processing The complete EEG recording of each subject was divided into 4-second segments as the basic unit for subsequent processing. In the discrimination process, the voting results of multiple segments were used to generate the final classification judgment of the subject to improve the robustness of individual discrimination.

[0039] 2.3 Feature extraction and matrixization The segmented EEG signal segments were subjected to discrete Fourier transform (DFT) to enhance the frequency domain features. The scale data was matrixed and the scoring data was standardized for subsequent model input.

[0040] 2.4 Dataset Division All samples are randomly mixed according to people and divided into training set, validation set and test set (ratio is 3:1:1). The training set is used for model training, the validation set is used for parameter optimization, and the test set is used to evaluate the final model performance.

[0041] Based on the above embodiment, step 3 includes: Through time and frequency domain methods, neural activity features such as power spectrum, entropy, phase coupling, etc. are extracted. After feature selection, a complete set of EEG features is formed for the prediction of scale scores.

[0042] Based on the above embodiment, step 4 includes: 4.1 Network Architecture Initialization According to the Kolmogorov-Arnold representation theorem, the KAN network architecture is designed to enable regression by decomposing multivariate complex features. The model input layer contains EEG features, and the output layer is used for regression prediction of scale scores.

[0043] 4.2 Model Training and Optimization Based on the mean square error (MSE) optimization objective, the model parameters are adjusted through back propagation to achieve high-precision prediction of the scale score. The formula used is:

[0044] in, Represents a univariate function in KAN, which is parameterized by B-spline or other learnable functions, enabling the model to learn complex nonlinear relationships.

[0045] Based on the above embodiment, step 5 includes: 5.1 Feature Importance Analysis Based on the KAN network regression results, the contribution of different features to the scale score is analyzed. By visualizing the KAN weights and function importance, key EEG channels or features are identified to provide explanatory support for autism diagnosis.

[0046] 5.2 Feature Visualization and Interpretation Use weighted visualization techniques, such as transparency mapping, to highlight important features, making it easier for clinicians to identify important signals.

[0047] The scheme of the present invention is described below with specific embodiments. Figure 2 As shown, including: The first is data collection. The collection of EEG signals is usually carried out in a professional EEG collection room. To ensure signal quality, the collection environment needs to be kept quiet and the power supply voltage needs to be stable. Subjects with sparse or soft hair are preferred. Before wearing the EEG collection headgear, the scalp needs to be cleaned and head ornaments need to be removed to ensure that the electrodes fit well. Accurately attach the electrodes to specific locations on the subject's scalp, and apply special conductive paste to the electrode contact points to reduce contact resistance and ensure that the impedance of each electrode is less than 100kΩ (less than 30kΩ in the best case). During the data collection process, the subjects are required to keep their eyes open, not speak, try to relax, breathe steadily, and use air conditioning and other means to prevent signal drift caused by sweating. Finally, 3 to 5 minutes of EEG data are collected and the data is saved together with the basic information of the subjects.

[0048] In this embodiment, the default frequency of EEG data collected is 1000Hz, and this resolution far exceeds the basic requirements of deep neural networks for data accuracy. To enhance noise resistance, downsampling is used to reduce the frequency to 256Hz. Subsequently, a bandpass filter is used to remove noise interference above 0-0.5Hz and 45Hz. The EEG recording of each subject is divided into 4-second segments as processing units for subsequent analysis; for tail segments less than 4 seconds, they are discarded. During classification and discrimination, the discrimination results of more than half of all individual segments are used as the overall discrimination results. After manual screening, segments with heavy noise or obvious signal drift are eliminated. Then a discrete Fourier transform is performed on each EEG segment, and the signal is converted from the time domain to the frequency domain to make the features more significant, and the effective frequency band of 0.5-45Hz is retained, while retaining the original time domain features. The autism assessment scale data is converted into a matrix form and standardized to ensure consistency with the EEG data features, so as to facilitate the combination and unified processing of the model input. Finally, the samples were randomly mixed by individual (not fragment) and divided into training set, validation set and test set in a ratio of 3:1:1. The first two were used for model training and optimization, and the latter was used for model testing.

[0049] A variety of time domain and frequency domain features are extracted from the preprocessed EEG segment data to establish a rich feature set. The extracted features include: (1) Power spectrum: Analyze the power distribution of each frequency band to reflect the activity state of a specific brain area. (2) Entropy feature: Used to quantify the complexity of EEG signals and help identify EEG changes related to autism. (3) Phase coupling feature: Calculate the phase consistency between different frequency bands and channels to capture the synergistic information of cross-regional EEG activity. These features, together with the scale data matrix, constitute the input feature set of the model, which aims to provide a more comprehensive EEG data representation for accurate regression prediction of scale scores.

[0050] The goal of the model in the embodiment of the present invention is to achieve high-precision prediction of autism scale scores. During the model training process, based on the EEG feature data and scale score labels of the training set, the KAN network parameters are optimized through the back propagation algorithm to reduce the mean square error (MSE). During the training process, the validation set is used to evaluate and adjust the hyperparameters (such as learning rate, batch size, dropout rate, etc.) to further optimize the regression performance of the model on the test set. Finally, through multiple rounds of iterations and parameter adjustments, it is ensured that the model achieves a balance between generalization ability, accuracy and stability, providing reliable guarantee for subsequent scale score predictions.

[0051] Based on the regression results of the KAN network, the interpretability analysis of the features was performed to identify the EEG features that contribute most to the scale score. By analyzing the feature weights and importance in the KAN network, the association between different EEG features and autism symptoms can be identified. Visualizing the contribution of key features makes it easier for clinicians to intuitively understand the prediction basis of the model, providing a scientific reference for diagnostic support and personalized intervention.

[0052] The following is a description of the KAN network-based autism assessment scale numerical regression system provided by the present invention. The KAN network-based autism assessment scale numerical regression system described below and the KAN network-based autism assessment scale numerical regression method described above can correspond to each other.

[0053] Figure 3 is a schematic diagram of the structure of a KAN network-based autism assessment scale numerical regression system provided by an embodiment of the present invention, such as Figure 3 As shown, it includes: a collection module 31, a preprocessing module 32, an extraction module 33, a training module 34 and a regression module 35, wherein: The acquisition module 31 is used to collect the EEG signal data of the target subject; the preprocessing module 32 is used to preprocess the EEG signal data to obtain the preprocessed EEG signal data; the extraction module 33 is used to extract the EEG features of the preprocessed EEG signal data; the training module 34 is used to use the EEG features to train and optimize the KAN network to generate a numerical regression model for the autism assessment scale; the regression module 35 is used to perform an interpretable analysis on the regression results output by the numerical regression model of the autism assessment scale to obtain the key EEG channel features of the scale score.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A numerical regression method for autism assessment scale based on KAN network, characterized in that: include: Collecting EEG signal data of the target subject; Preprocessing the EEG signal data to obtain preprocessed EEG signal data; Extracting EEG features of the preprocessed EEG signal data; The EEG features are used to train and optimize the KAN network to generate a numerical regression model for the autism assessment scale; The regression results output by the numerical regression model of the autism assessment scale are analyzed for interpretability to obtain key EEG channel features of the scale score.

2. The KAN network-based autism assessment scale numerical regression method according to claim 1, characterized in that: Collect EEG signal data of the target subject, including: Collect high-density 62-channel 1000Hz resting-state EEG signal data; The autism assessment scale was used to quantitatively score the behavioral characteristics of the subjects to form structured scale data.

3. The KAN network-based autism assessment scale numerical regression method according to claim 1, characterized in that: Preprocessing the EEG signal data to obtain preprocessed EEG signal data includes: Downsampling, filtering and removing artifacts on the EEG signal data in sequence to obtain filtered EEG signal data; Using a fixed time window to segment the screened EEG signal data into a plurality of EEG segments; Performing discrete Fourier transform and bandpass filtering on the multiple EEG segments respectively to obtain EEG frequency domain signals, and retaining EEG time domain signals; The EEG frequency domain signal and the EEG time domain signal are randomly mixed according to the subjects and divided into a training set, a validation set and a test set.

4. The KAN network-based autism assessment scale numerical regression method according to claim 1, characterized in that: Extracting the EEG features of the preprocessed EEG signal data includes: Based on the time domain and frequency domain methods, the neural activity features of the preprocessed EEG signal data are extracted, and the EEG features are constituted by the neural activity features.

5. The KAN network-based autism assessment scale numerical regression method according to claim 1, characterized in that: The EEG features are used to train and optimize the KAN network to generate a numerical regression model for the autism assessment scale, including: Determining that the KAN network includes an input layer, a plurality of hidden layers and an output layer; Using the EEG features, the KAN network is trained with minimizing mean square error as a model objective; The parameters of the KAN network are optimized by back propagation and gradient descent method to obtain the numerical regression model of the autism assessment scale.

6. The KAN network-based autism assessment scale numerical regression method according to claim 1, characterized in that: The regression results output by the numerical regression model of the autism assessment scale were analyzed for interpretability, and the key EEG channel features of the scale score were obtained, including: Obtaining the regression result of the autism assessment scale numerical regression model; The regression results are used to analyze the contribution of different features to the scale score, and the key EEG channel features are identified by visualizing the KAN weights and function importance; Weighted visualization is used to highlight the key EEG channel features.

7. A numerical regression system for autism assessment scale based on KAN network, characterized in that: include: An acquisition module, used to acquire EEG signal data of a target subject; A preprocessing module, used for preprocessing the EEG signal data to obtain preprocessed EEG signal data; An extraction module, used for extracting EEG features of the preprocessed EEG signal data; A training module, used to train and optimize the KAN network using the EEG features to generate a numerical regression model for an autism assessment scale; The regression module is used to perform interpretable analysis on the regression results output by the numerical regression model of the autism assessment scale to obtain key EEG channel features of the scale score.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the KAN network-based autism assessment scale numerical regression method as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the KAN network-based autism assessment scale numerical regression method as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the KAN network-based autism assessment scale numerical regression method as described in any one of claims 1 to 6 is implemented.

Images