Method, device and system for assisting in recognizing attention deficit hyperactivity disorder based on electrocardiogram signals

Abstract

The invention discloses a method, a device and a system for assisting in recognizing attention deficit hyperactivity disorder based on electrocardiogram signals. The method comprises the following steps: (1) collecting and processing electrocardiogram data of a testee of which the risk of suffering from attention deficit hyperactivity disorder needs to be evaluated; (2) performing deep feature extraction on the processed electrocardiogram data of the person to be detected by adopting a one-dimensional convolutional neural network; generating a classification thermodynamic diagram for the feature map of the convolutional neural network through Score-CAM, and extracting time domain, frequency domain and local statistical features from the generated classification thermodynamic diagram; and (3) inputting the time domain, the frequency domain and the local statistical features into a machine learning classifier for classification to obtain an assessment result of the attention deficit hyperactivity disorder risk of the person to be detected. According to the method, deep learning and various machine learning methods are combined, so that the classification performance is improved, and rich feature interpretation is provided.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a method, device and system for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals. Background Art

[0002] Attention deficit hyperactivity disorder (ADHD) is a common neurodevelopmental disorder in children, which is mainly manifested by inattention, impulsivity and hyperactivity. ADHD affects about 5% to 7% of children and adolescents, and its etiology involves multiple factors such as genetics, neurophysiology and environment. This disease not only has a significant impact on the patient's daily life, learning and social interaction, but may also continue into adulthood, resulting in reduced work efficiency and impaired social function. At the same time, ADHD is often accompanied by comorbidities such as anxiety, conduct disorder and oppositional defiant disorder, which further increase the physical and mental burden on patients.

[0003] Currently, the diagnosis of ADHD mainly relies on clinical questionnaires (such as DSM-5 assessment criteria) and behavioral observations. However, these methods are highly subjective, and the diagnostic results are easily affected by the evaluator's experience and fluctuations in patient performance. In addition, traditional methods are time-consuming and costly, and it is difficult to meet the clinical needs of large-scale screening and early intervention. Especially in areas with limited medical resources, delayed diagnosis or misdiagnosis often occurs. Therefore, there is an urgent need to develop an efficient and objective auxiliary diagnostic tool to improve diagnostic accuracy and shorten diagnostic time.

[0004] In recent years, electrocardiogram (ECG) signals have gradually become a hot topic in the study of the physiological characteristics of ADHD because of their close relationship with the activity of the autonomic nervous system. Existing studies have shown that the ECG signals of ADHD patients exhibit obvious abnormal characteristics, such as an overall decrease in heart rate variability (HRV) and irregular signal rhythm. However, most existing studies are mainly based on linear time domain and frequency domain analysis methods, which make it difficult to capture the complex nonlinear characteristics of the signal. At the same time, these methods have limited classification accuracy and lack robustness when faced with large-scale data.

[0005] In addition, the application of machine learning and deep learning technologies in medical diagnosis has gradually deepened in recent years, but it is still in the initial exploration stage in the study of ADHD. For example, some studies have attempted to use simple statistical features combined with traditional machine learning algorithms for classification, but due to insufficient feature extraction, the classification performance is limited. For example, the Chinese patent document with publication number CN118983076A discloses an auxiliary diagnosis method and system for children with attention deficit hyperactivity disorder. The method includes the following steps: setting up a cognitive ability test according to the cognitive characteristics and development of the child group; collecting cognitive ability test data of subjects in the attention deficit hyperactivity disorder diagnosis group and cognitive ability test data of subjects in the normal control group; preprocessing the collected cognitive ability test data; constructing a support vector machine model, and using the trained support vector machine model to implement auxiliary diagnosis of children with attention deficit hyperactivity disorder. On the other hand, although deep learning technology can extract deep-level features of signals, its "black box" problem and insufficient model interpretability also limit the promotion of its clinical application.

[0006] Therefore, how to combine deep learning and machine learning techniques to improve the model classification performance while enhancing its clinical interpretability has become an important technical challenge in current ADHD research. Summary of the invention

[0007] The present invention provides a device for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals, which combines deep learning with multiple machine learning methods, not only improving the classification performance but also providing rich feature interpretability.

[0008] The technical solution of the present invention is as follows: A method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals, comprising: (1) Collecting and processing electrocardiogram data of subjects who need to be assessed for the risk of attention deficit hyperactivity disorder; (2) A one-dimensional convolutional neural network is used to perform deep feature extraction on the processed ECG data of the subjects; a classification heat map is generated from the feature map of the convolutional neural network through Score-CAM, and time domain, frequency domain and local statistical features are extracted from the generated classification heat map; (3) The time domain, frequency domain, and local statistical features are input into the machine learning classifier for classification to obtain the assessment results of the subject's risk of developing attention deficit hyperactivity disorder.

[0009] In step (1), processing the electrocardiogram data includes: capturing a signal segment from the 3rd second to the 9th second of lead II in the electrocardiogram data as processed electrocardiogram data.

[0010] In step (2), the one-dimensional convolutional neural network includes four convolutional layers and a fully connected layer connected in sequence. Each convolution layer is followed by a batch normalization (BatchNorm) and ReLU activation function as well as a maximum pooling layer.

[0011] The first convolutional layer uses a convolution kernel of 3 filters with a kernel size of 27 to capture the basic local features of the ECG signal; The second convolutional layer uses a convolution kernel of 10 filters with a kernel size of 15 to extract the time domain related features of the ECG signal; The third convolutional layer uses a convolution kernel of 10 filters with a kernel size of 3 to further extract the mid-order features of the signal; The fourth convolutional layer uses a convolution kernel of 10 filters with a kernel size of 3, which is responsible for capturing the high-order nonlinear characteristics of the ECG signal; Each convolution layer is followed by batch normalization (BatchNorm) and ReLU activation function, and the feature dimension is reduced through the maximum pooling layer. The final feature is flattened through the fully connected layer to form a deep feature.

[0012] The training process of the one-dimensional convolutional neural network includes: collecting and processing the electrocardiogram data of several healthy control children (HC) and children with attention deficit hyperactivity disorder (ADHD) to construct a training data set; and using the training data set to train the one-dimensional convolutional neural network.

[0013] The construction of the training data set includes: collecting 12-lead ECG signal data of healthy control children and children with attention deficit hyperactivity disorder aged 6 to 12 years old from a real clinical environment, excluding drug intervention and other disease interference factors; intercepting the signal fragment of a specific time period (3rd to 9th seconds) of lead II to construct the training data set.

[0014] Furthermore, the one-dimensional convolutional neural network was trained using the Adam optimizer with a learning rate of 0.0002.

[0015] Furthermore, L2 regularization and Dropout techniques are combined in the one-dimensional convolutional neural network training process to improve the generalization performance of the model.

[0016] In step (2), Score-CAM generates a classification heat map by weighting the activation maps of each convolutional layer of the one-dimensional convolutional neural network.

[0017] The time domain features include the global activation mean, the activation value standard deviation, and the maximum activation value.

[0018] The global activation mean reflects the average level of the overall activation value of the heat map and is used to measure the overall activity of the signal. The activation mean of the ADHD group is usually higher than that of the healthy control group, indicating that their signal fluctuations are greater. The activation value standard deviation represents the distribution range of the heat map activation value. The standard deviation of the ADHD group is significantly higher than that of the HC group, revealing the instability of its signal characteristics. The maximum activation value, the location of the strongest activation value in the heat map, reflects the area that the model pays most attention to, usually concentrated near the QRS complex and T wave.

[0019] The frequency domain features include Fourier transform mean, spectrum peak, and power peak.

[0020] The Fourier transform mean is used to analyze the spectral distribution of the signal corresponding to the heat map. The ADHD group usually shows a higher spectral mean, indicating the complexity of the signal frequency components. The spectral peak is the most significant frequency component captured in the signal. The ADHD group shows greater fluctuations in this feature than the HC group. The power peak indicates the concentration of signal energy in the frequency domain. The power peak of the ADHD group is higher than that of the HC group, reflecting that the energy fluctuation of its signal is more intense.

[0021] Local statistical features include sliding window mean, sliding window standard deviation, and local maximum.

[0022] The sliding window mean is the average level of local activation values ​​calculated by sliding the window. The sliding mean of the ADHD group has a wider fluctuation range, reflecting the strong changes in the signal in different time periods. The sliding window standard deviation is used to quantify the volatility of local activation values. The ADHD group is significantly higher than the HC group, indicating the instability of its signal on a small time scale. The local maximum is the maximum activation value within the sliding window, which identifies the key area of ​​the local signal. The ADHD group usually has a higher local maximum.

[0023] Through the statistical analysis of the heat map features, it was found that there were significant differences in signal characteristics between the ADHD group and the HC group: the time domain analysis showed that the overall activation level and volatility of the ADHD group's signals were higher, reflecting the disordered characteristics of the patient's autonomic nervous system activity. The frequency domain analysis revealed the diversity and complexity of the ADHD group in the spectral components, indicating that its signals had stronger irregularities. The local statistical characteristics further reflected that the dynamic changes of the ECG signals of ADHD patients were more obvious in the time dimension.

[0024] In step (3), the machine learning classifier is support vector machine (SVM), random forest (RF), logistic regression (LR), K nearest neighbor algorithm (KNN), decision tree (DT) or XGBoost.

[0025] Training a machine learning classifier involves: Collect and process the electrocardiogram data of several healthy control children and children with attention deficit hyperactivity disorder, and divide the processed electrocardiogram data into a training set and a test set; The training set is input into the trained one-dimensional convolutional neural network, and the classification heat map is generated from the feature map of the convolutional neural network through Score-CAM. The time domain, frequency domain and local statistical features are extracted from the generated classification heat map. The extracted time domain, frequency domain and local statistical features are input into the machine learning classifier to train the machine learning classifier.

[0026] The present invention also provides a device for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals, comprising: A data collection and processing module collects and processes the electrocardiogram data of the subjects whose risk of suffering from attention deficit hyperactivity disorder needs to be assessed; The heat map feature extraction module uses a one-dimensional convolutional neural network to perform deep feature extraction on the processed ECG data of the subject; Score-CAM is used to generate a classification heat map from the feature map of the convolutional neural network, and the time domain, frequency domain and local statistical features are extracted from the generated classification heat map; The classification judgment module inputs the time domain, frequency domain and local statistical features into the machine learning classifier for classification, and obtains the assessment results of the risk of the subject suffering from attention deficit hyperactivity disorder.

[0027] The present invention also provides a system for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals are implemented.

[0028] Compared with the prior art, the present invention has the following beneficial effects: (1) High precision: Experimental results show that the method of the present invention achieves a classification accuracy of over 95% on clinical datasets.

[0029] (2) Interpretability: The heat map generated by Score-CAM technology improves the interpretability of the model, helping doctors understand the decision-making basis of the model and making the model's diagnostic results more transparent and credible.

[0030] (3) Universality: A variety of machine learning classifiers are used to verify the extracted features, which improves the robustness and reliability of the model.

[0031] (4) Low cost and non-invasive: Compared with traditional imaging and behavioral assessment methods, ECG signal-based methods have the advantages of low cost and convenience for clinical outpatient screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The present invention is a flow chart of assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals.

[0033] Figure 2 Schematic diagram of the 1D-CNN model structure.

[0034] Figure 3 The following are the distribution histograms of the heat map features of the ADHD and HC groups; (a) is the distribution histogram of the average activation value; (b) is the distribution histogram of the maximum activation value; (c) is the distribution histogram of the standard deviation of the activation value; (d) is the distribution histogram of the Fourier transform mean; (e) is the distribution histogram of the Fourier transform standard deviation; (f) is the distribution histogram of the maximum Fourier transform value; (g) is the distribution histogram of the power mean; (h) is the distribution histogram of the power maximum; (i) is the distribution histogram of the local mean; (j) is the distribution histogram of the local standard deviation; (k) is the distribution histogram of the local maximum.

[0035] Figure 4 The performance comparison chart of 1D-CNN and classifier cross-validation; (a) is the accuracy comparison line chart of ten-fold cross-validation; (b) is the precision comparison line chart of ten-fold cross-validation; (c) is the recall comparison line chart of ten-fold cross-validation; (d) is the F1 score comparison line chart of ten-fold cross-validation.

[0036] Figure 5 Examples of 1D-Score-CAM heatmaps for ADHD samples (a) and HC samples (b). DETAILED DESCRIPTION

[0037] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be pointed out that the embodiments described below are intended to facilitate the understanding of the present invention and do not have any limiting effect on the present invention.

[0038] The present invention designs an automatic detection technology for attention deficit hyperactivity disorder (ADHD) based on ECG signals by deeply analyzing children's electrocardiogram (ECG) signals, combining one-dimensional convolutional neural network (1D-CNN) and multiple machine learning classifiers (the flowchart is shown in the figure). Figure 1 As shown in Figure 1). This technology uses the heat map generated by Score-CAM as the core, and constructs a feature extraction method that combines time domain, frequency domain and local statistical features. The classifier accurately classifies the ECG features of ADHD and healthy individuals. By combining 1D-CNN deep feature extraction with interpretability technology, this invention significantly improves the transparency and clinical applicability of the model classification performance, and ultimately forms an efficient and interpretable ADHD detection process, allowing clinicians to achieve rapid screening and preliminary diagnosis without relying on behavioral assessment.

[0039] The specific technical solutions of the present invention are further illustrated by examples below.

[0040] Devices for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals include: (1) ECG data acquisition and processing module In this example, ECG data of children aged 6 to 12 years old without drug intervention were collected from real clinical environments, including 2368 children diagnosed with ADHD and 2368 healthy control children (HC). All participants were matched 1:1 by age, and patients with cardiovascular diseases that may affect ECG characteristics were excluded to ensure the authenticity and representativeness of the data.

[0041] The collected ECG signal is a 12-lead record. To improve signal quality and consistency, only the II lead signal is used as the research data. The sampling frequency is 300 Hz, and the signal segment from the 3rd to the 9th second is intercepted to form an analysis sample of 3000 sampling points per segment. The signal is not filtered or denoised, and all characteristic information in the original data is retained.

[0042] (2) Deep learning feature extraction module A one-dimensional convolutional neural network (1D-CNN) was constructed to gradually extract signal features through four layers of convolution and pooling (the model structure is as follows Figure 2 The specific process is as follows: The first convolutional layer uses a convolution kernel with 3 filters and a kernel size of 27 to capture the basic local features of the ECG signal. Batch normalization (BatchNorm) and ReLU activation function are used to enhance the stability and nonlinear expression ability of the model.

[0043] The second convolution layer uses a convolution kernel with 10 filters and a kernel size of 15 to extract the time domain related features of the ECG signal. The pooling layer uses a 2×2 maximum pooling operation to reduce the feature dimension and retain key information.

[0044] The third convolutional layer: uses a convolution kernel of 10 filters with a kernel size of 3 to further extract the mid-order pattern of the signal and reduce the dimensionality through pooling operations.

[0045] The fourth convolutional layer uses a convolution kernel of 10 filters and a kernel size of 3, which is responsible for capturing the high-order nonlinear features of the ECG signal. Combined with batch normalization and pooling operations, it improves the compactness and expressiveness of the features.

[0046] Each convolution layer is followed by batch normalization (BatchNorm) and ReLU activation function, and the feature dimension is reduced through the maximum pooling layer to reduce the risk of overfitting. The final feature is flattened through the fully connected layer to form a deep feature.

[0047] The one-dimensional convolutional neural network is trained using the Adam optimizer with a learning rate of 0.0002. During the training process, L2 regularization and Dropout techniques are combined to improve the generalization performance of the model.

[0048] (3) Heat map feature extraction and interpretability module To enhance the clinical interpretability of the model, the present invention uses Score-CAM technology to generate classification heat maps for the feature maps of convolutional neural networks. Score-CAM calculates the contribution of each feature area to the classification result, and intuitively presents the decision basis of the deep learning model in the form of a heat map, helping to understand the signal area that the model focuses on.

[0049] Extract time domain, frequency domain and local statistical features from the generated heat map (feature histogram such as Figure 3 ), as follows: Time domain features: 1) Global activation mean: reflects the average level of the overall activation value of the heat map, which is used to measure the overall activity of the signal. The activation mean of the ADHD group is usually higher than that of the healthy control group (HC), indicating that its signal fluctuations are greater. 2) Activation value standard deviation: represents the distribution range of the heat map activation value. The standard deviation of the ADHD group is significantly higher than that of the HC group, revealing the instability of its signal characteristics. 3) Maximum activation value: The location of the strongest activation value in the heat map reflects the area that the model pays most attention to, which is usually concentrated near the QRS complex wave and T wave.

[0050] Frequency domain features: 1) Fourier transform mean: used to analyze the spectrum distribution of the signal corresponding to the heat map. The ADHD group usually shows a higher spectrum mean, indicating the complexity of the signal frequency components. 2) Spectral peak: captures the most significant frequency component of the signal. The ADHD group shows greater fluctuations in this feature than the HC group. 3) Power peak: indicates the concentration of signal energy in the frequency domain. The power peak of the ADHD group is higher than that of the HC group, reflecting that the energy fluctuation of its signal is more intense.

[0051] Local statistical features: 1) Sliding window mean: The average level of local activation values ​​is calculated by sliding window. The sliding mean of the ADHD group has a larger fluctuation range, reflecting the strong changes in the signal in different time periods. 2) Sliding window standard deviation: used to quantify the volatility of local activation values. The ADHD group is significantly higher than the HC group, indicating the instability of its signal on a small time scale. 3) Local maximum: The maximum activation value within the sliding window identifies the key area of ​​the local signal. The ADHD group usually has a higher local maximum.

[0052] Through the statistical analysis of the heat map features, it was found that there were significant differences in signal characteristics between the ADHD group and the HC group: the time domain analysis showed that the overall activation level and volatility of the ADHD group's signals were higher, reflecting the disordered characteristics of the patient's autonomic nervous system activity. The frequency domain analysis revealed the diversity and complexity of the ADHD group in the spectral components, indicating that its signals had stronger irregularities. The local statistical characteristics further reflected that the dynamic changes of the ECG signals of ADHD patients were more obvious in the time dimension.

[0053] The statistical results of the heat map features significantly improved the accuracy and robustness of ADHD classification, laying a solid foundation for the clinical promotion of the model.

[0054] (4) ADHD classification and performance verification module The extracted heat map features were input into the following six machine learning classifiers for ADHD and HC classification: support vector machine (SVM); random forest (RF); logistic regression (LR); K nearest neighbor algorithm (KNN); decision tree (DT); XGBoost. The performance of the classifier was evaluated using the 10-fold cross-validation method, which comprehensively evaluated the stability and robustness of the model by dividing the data set and repeating training and testing.

[0055] The results show that the combination of 1D-CNN and random forest classifier performs best (the difference in classifier performance is shown in Figure 4 As shown in the figure, its performance indicators are: classification accuracy: 95.97%; precision: 0.96; recall rate: 0.96; F1 score: 0.96.

[0056] To further verify the reliability of the classification results and the interpretability of the model, we randomly selected samples of one ADHD subject and one HC subject for heat map visualization analysis (the sample ECG heat maps of ADHD and HC are shown in Figure 5 The results show that in the ADHD group, the heat map activation pattern distribution is more dispersed, and the model mainly focuses on the QRS complex wave and T wave area. This indicates that the ECG signals of ADHD patients have significant abnormal fluctuation characteristics in these areas. In the HC group, the activation values ​​are concentrated near the R wave, and the signal distribution is relatively stable, reflecting the regularity and consistency of the ECG signals of healthy individuals.

[0057] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals, characterized in that: include: (1) Collecting and processing electrocardiogram data of subjects who need to be assessed for the risk of attention deficit hyperactivity disorder; (2) A one-dimensional convolutional neural network is used to perform deep feature extraction on the processed ECG data of the subjects; a classification heat map is generated from the feature map of the convolutional neural network through Score-CAM, and time domain, frequency domain and local statistical features are extracted from the generated classification heat map; (3) The time domain, frequency domain, and local statistical features are input into the machine learning classifier for classification to obtain the assessment results of the subject's risk of developing attention deficit hyperactivity disorder.

2. The method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals according to claim 1, characterized in that: In step (1), processing the electrocardiogram data includes: capturing a signal segment from the 3rd second to the 9th second of lead II in the electrocardiogram data as processed electrocardiogram data.

3. The method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals according to claim 1, characterized in that: In step (2), the one-dimensional convolutional neural network includes four convolutional layers and a fully connected layer connected in sequence, and each convolution layer is followed by batch normalization, ReLU activation function and maximum pooling layer.

4. The method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals according to claim 1, characterized in that: The training process of the one-dimensional convolutional neural network includes: collecting and processing the electrocardiogram data of several healthy control children and children with attention deficit hyperactivity disorder to construct a training data set; and using the training data set to train the one-dimensional convolutional neural network.

5. The method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals according to claim 1, characterized in that: In step (2), Score-CAM generates a classification heat map by weighting the activation maps of each convolutional layer of the one-dimensional convolutional neural network.

6. The method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals according to claim 1, characterized in that: In step (2), the time domain features include global activation mean, activation value standard deviation, and maximum activation value; The frequency domain features include Fourier transform mean, spectrum peak, and power peak; Local statistical features include sliding window mean, sliding window standard deviation, and local maximum.

7. The method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals according to claim 1, characterized in that: In step (3), the machine learning classifier is a support vector machine, random forest, logistic regression, K nearest neighbor algorithm, decision tree or XGBoost.

8. The method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals according to claim 1, characterized in that: Training a machine learning classifier involves: Collect and process the electrocardiogram data of several healthy control children and children with attention deficit hyperactivity disorder, and divide the processed electrocardiogram data into a training set and a test set; The training set is input into the trained one-dimensional convolutional neural network, and the classification heat map is generated from the feature map of the convolutional neural network through Score-CAM. The time domain, frequency domain and local statistical features are extracted from the generated classification heat map. The extracted time domain, frequency domain and local statistical features are input into the machine learning classifier to train the machine learning classifier.

9. A device for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals, comprising: A data collection and processing module collects and processes the electrocardiogram data of the subjects whose risk of suffering from attention deficit hyperactivity disorder needs to be assessed; The heat map feature extraction module uses a one-dimensional convolutional neural network to perform deep feature extraction on the processed ECG data of the subject; Score-CAM is used to generate a classification heat map from the feature map of the convolutional neural network, and the time domain, frequency domain and local statistical features are extracted from the generated classification heat map; The classification judgment module inputs the time domain, frequency domain and local statistical features into the machine learning classifier for classification, and obtains the assessment results of the risk of the subject suffering from attention deficit hyperactivity disorder.

10. A system for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for assisting in identifying attention deficit hyperactivity disorder based on electrocardiogram signals as described in any one of claims 1 to 8 are implemented.

Images