ADHD auxiliary detection method and system based on image processing
By converting the EEG signal into an image and extracting features using the BP neural network, the problem of inaccurate ADHD detection in the prior art is solved, and a higher detection accuracy is achieved.
Patent Information
- Application Number
- CN202411707873.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-27
- Publication Date
- 2025-04-29
AI Technical Summary
The prior art is difficult to accurately judge the abnormal brain waves of ADHD patients, resulting in inaccurate ADHD assisted detection.
By converting the EEG signal into images and extracting image features using BP neural network-based image processing technology, the brain wave abnormalities are rated to improve the accuracy of ADHD-assisted detection.
The accuracy of ADHD assisted detection is improved, and more reliable detection results are obtained through the spatial and temporal feature extraction and fusion of dynamic thermal images, combined with the classification module of LightBGM algorithm.
Smart Images
Figure CN119896480A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of ADHD auxiliary detection, and in particular to an ADHD auxiliary detection method and system based on image processing. Background Art
[0002] Attention-deficit / hyperactivity disorder (ADHD) is a developmental disorder characterized by difficulty paying attention, hyperactivity, and impulsivity. It primarily affects children and adolescents, but can also persist into adulthood. People with ADHD may exhibit behavioral traits such as difficulty paying attention, impulsivity, and hyperactivity. It can be difficult for people with ADHD to stick to a fixed schedule and routine in their regular lives, as they often have difficulty controlling their behavior and attention.
[0003] Currently, ADHD requires repeated evaluations by professional psychiatrists, which may not be conducive to the early identification of ADHD in children, and detection based solely on clinical and ratings may be unreliable because it may be related to the clinician's own, culture, and country, etc. Therefore, there is a need for an objective ADHD detection method to assist clinicians in testing patients.
[0004] Electroencephalogram (EEG) is a method of recording brain activity using electrophysiological indicators. When the brain is active, the postsynaptic potentials of a large number of neurons occur synchronously and are summed up. It records the changes in electric waves during brain activity and is a general reflection of the electrophysiological activity of brain nerve cells in the cerebral cortex or scalp surface. Studies have found that there are differences between the brain waves of ADHD patients and those of ordinary people. Therefore, monitoring brain waves can assist in the detection of ADHD. However, the fluctuations of brain waves are nonlinear, and it is difficult to judge the abnormality of brain waves. Therefore, misdiagnosis is prone to occur by directly observing abnormalities in brain waves. Summary of the invention
[0005] In order to solve the problem that the existing technology for auxiliary detection of ADHD has difficulty in judging brain wave abnormalities and leads to inaccurate ADHD detection, the present invention provides an ADHD auxiliary detection method and device based on image processing, which converts EEG signals into images and uses BP neural network to extract image features, and then rates brain wave abnormalities, thereby improving the accuracy of ADHD auxiliary detection.
[0006] In a first aspect, the present invention provides an ADHD auxiliary detection method based on image processing, comprising:
[0007] Step 1: Obtain the brain wave signal of the target detection person and perform preprocessing;
[0008] Step 2: Perform time-frequency analysis on the preprocessed EEG signal to obtain EEG time-frequency data;
[0009] Step 3: Convert EEG time-frequency data into dynamic thermal images;
[0010] Step 4: Build and train an ADHD auxiliary assessment model;
[0011] The ADHD auxiliary assessment model includes a feature extraction module, a feature processing module and a classification module; the feature extraction module is used to extract spatial features and temporal features from the input dynamic thermal image; the feature fusion module is used to fuse and process the features extracted by the feature extraction module to generate a final feature map; the classification module performs classification based on the final feature map to generate an ADHD detection result;
[0012] Step 5: Input the thermal image corresponding to the EEG data of the target detection person into the trained ADHD auxiliary evaluation model to obtain the ADHD auxiliary detection result of the target detection person.
[0013] Furthermore, in step 1, the preprocessing includes filtering and re-referencing, and the specific process includes:
[0014] The Butterworth filter is used to filter the EEG signal;
[0015] Perform re-reference processing on the filtered signal.
[0016] Furthermore, in step 2, wavelet transform is used to perform time-frequency analysis on the image to obtain brain wave time-frequency data; the wavelet transform calculation formula is as follows:
[0017]
[0018] Among them, W f (a,b) represents the transformed EEG time-frequency data, a represents the scale factor, b represents the displacement factor, and f(t) represents the preprocessed EEG signal. represents the wavelet basis function.
[0019] Furthermore, the step 3 specifically includes:
[0020] According to the spatial position of the electrode points corresponding to the EEG time-frequency data, the EEG topographic map is obtained by mapping it to the plane coordinate system;
[0021] The EEG time-frequency data corresponding to each pole in the EEG topographic map are normalized, and different colors are used to map the EEG data of different frequencies to generate a dynamic EEG thermal image.
[0022] Furthermore, the spatial positions of the electrode points corresponding to the EEG time-frequency data are mapped into a plane coordinate system to obtain an EEG topographic map, which specifically includes:
[0023] Polar coordinate mapping is used to map the spatial position of the electrode points into a plane coordinate system to obtain an EEG topographic map.
[0024] Furthermore, the feature extraction module includes a spatial feature extraction module and a temporal feature extraction module processed in parallel; wherein the spatial feature extraction module includes three 3D convolutional layers and a fully connected layer connected in sequence, and each of the 3D convolutional layers is connected to a splicing layer, a BN layer and a maximum pooling layer; wherein the 3D convolutional layer uses three convolution kernels of different sizes for feature extraction, and the sizes of the three convolution kernels are 3×3×3, 5×5×5 and 7×7×7 respectively; the temporal feature extraction module includes multiple LSTM units, a BN layer and a fully connected layer.
[0025] Furthermore, the feature fusion module includes a multi-layer perceptron, a flatten layer and a normalization layer.
[0026] Furthermore, the classification module introduces the LightBGM algorithm to classify the feature graph output by the feature fusion module.
[0027] Furthermore, when training the ADHD auxiliary assessment model, a cross-loss entropy function is used to adjust the model parameters. The cross-loss entropy function is as follows:
[0028] L(y,p)=-[ylog(p)+(1-y)log(1-p)]
[0029] Among them, p represents the probability of model output, y represents the sign function, and the ADHD auxiliary assessment model predicts that the accurate output is 1 and the error is 0.
[0030] In a second aspect, the present invention provides an ADHD auxiliary detection system based on image processing, comprising:
[0031] The EEG signal acquisition module is used to acquire the EEG signal of the person to be tested and perform preprocessing;
[0032] The time-frequency data conversion module is used to perform time-frequency analysis on the preprocessed EEG signal to obtain EEG time-frequency data;
[0033] The thermal map generation module is used to map the spatial position of the electrode points of the test EEG data into a plane coordinate system, normalize the EEG time-frequency data and use different colors to map EEG data of different frequencies to generate a dynamic EEG thermal image;
[0034] Model building module, used to build and train ADHD auxiliary assessment model;
[0035] The result output module is used to input the thermal image corresponding to the EEG data of the target detection person into the trained ADHD auxiliary evaluation model to obtain the ADHD auxiliary detection result of the target detection person.
[0036] Beneficial effects of the present invention:
[0037] The present invention proposes an ADHD auxiliary detection method and system based on image processing. By detecting the brain wave signals of the target person and converting them into dynamic thermal images, the changes in the intensity of the brain wave signals of the target person can be directly observed; and an ADHD auxiliary detection model is proposed. By extracting the spatial and temporal features of the dynamic thermal images and fusing them, and classifying them based on the fused features, and introducing the LightBGM algorithm in the classification module for classification, the results obtained by the ADHD auxiliary detection model can improve the accuracy of ADHD auxiliary detection. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 A flowchart of an ADHD auxiliary detection method based on image processing provided by an embodiment of the present invention;
[0039] Figure 2 A schematic diagram of the ADHD auxiliary detection model framework provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0040] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all 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.
[0041] like Figure 1 As shown, an ADHD auxiliary detection method based on image processing provided by an embodiment of the present invention includes:
[0042] Step 1: Obtain the brain wave signal of the target detection person and perform preprocessing;
[0043] Specifically, EEG detection equipment is used to detect the brain wave signals of the target detection personnel, and pre-processing is performed to remove external and internal interference of the human body.
[0044] It should be noted that the EEG detection device used in the embodiment of the present invention is a detection device of the 10-20 international standard lead system, including 75 electrode points. Similarly, it is also possible to use a device containing different numbers of electrode points to collect brain wave signals, which will not be elaborated in detail here.
[0045] Specifically, preprocessing includes:
[0046] The Butterworth filter is used to filter the EEG signal;
[0047] It can be understood that the original brain wave signal contains a lot of high-frequency noise and interference, and needs to be filtered. The present invention uses a Butterworth filter to filter the original brain wave signal, and filters the signal with a frequency other than 0.25Hz-40Hz in the original brain wave signal to obtain a filtered brain wave signal;
[0048] Perform re-reference processing on the filtered signal.
[0049] Specifically, a single electrode point may introduce noise, and the whole-brain average reference is used to re-reference the filtered signal. The mean of all data in the whole brain is used as the reference data, and the potential difference of each pole is recorded.
[0050] It can be understood that the data collected by the EEG signal is the potential difference between the location of each electrode and the reference electrode. Generally, the reference electrode used in EEG recording is the electrode in the center of the top of the head. Therefore, the embodiment of the present invention needs to perform re-reference processing on the EEG data.
[0051] Furthermore, in order to avoid measurement data errors caused by eye movements that may affect the test results, the data recorded by the electrode points recording the eye points are deleted before the re-reference process.
[0052] Step 2: Perform time-frequency analysis on the preprocessed EEG signal to obtain EEG time-frequency data;
[0053] Specifically, the preprocessed EEG signal is subjected to time-frequency analysis using wavelet transform to obtain EEG time-frequency data. The wavelet transform calculation formula is as follows:
[0054]
[0055] Among them, W f (a,b) represents the transformed EEG time-frequency data, a represents the scale factor, b represents the displacement factor, and f(t) represents the preprocessed EEG signal. represents the wavelet basis function.
[0056] Step 3: Convert EEG time-frequency data into dynamic thermal images, including:
[0057] According to the spatial position of the electrode points corresponding to the EEG time-frequency data, the EEG topographic map is obtained by mapping it to the plane coordinate system;
[0058] Specifically, the spatial position of the electrode point is mapped to the plane coordinate system through polar coordinates to obtain the EEG topography map. The specific process is as follows:
[0059] The spatial position of the electrode point E1 is (x1, y1, z1), which is first mapped to the three-dimensional polar coordinate system and calculated as follows:
[0060]
[0061] Wherein, r represents the polar diameter length of the electrode point E1 in the polar coordinate system, θ represents the inclination angle of the polar diameter corresponding to the electrode point E1, and μ represents the rotation angle of the polar diameter.
[0062] The electrode point E1 is mapped to the two-dimensional plane coordinate system according to the above tilt angle and rotation angle. The formula is as follows:
[0063]
[0064] Wherein, (x′1, y′1) represents the coordinate data of the electrode point E1 mapped to the two-dimensional plane coordinate system.
[0065] Similarly, the spatial positions corresponding to all EEG time-frequency data are mapped to a two-dimensional plane coordinate system in the same way to obtain an EEG topographic map;
[0066] The EEG time-frequency data corresponding to each electrode point in the EEG topographic map are normalized, and different colors are used to map the EEG data of different frequencies to generate a dynamic EEG thermal image.
[0067] It can be understood that by mapping the electrode points in the EEG topography to different colors, different colors correspond to different frequencies, and by observing the changes in thermal images at different times, the changes in the intensity of EEG activity of the target detection personnel can be intuitively observed.
[0068] Step 4: Build and train an ADHD auxiliary assessment model; Figure 2 As shown, the ADHD auxiliary assessment model includes a feature extraction module, a feature processing module and a classification module; the feature extraction module is used to extract spatial features and temporal features from the input dynamic thermal image; the feature fusion module is used to fuse and process the features extracted by the feature extraction module to generate a final feature map; the classification module performs classification based on the final feature map to generate an ADHD detection result;
[0069] Specifically, the feature extraction module includes a spatial feature extraction module and a temporal feature extraction module processed in parallel; the spatial feature extraction module includes three sequentially connected 3D convolutional layers and a fully connected layer, and each 3D convolutional layer is connected with a splicing layer, a BN layer and a maximum pooling layer; the 3D convolutional layer uses three convolution kernels of different sizes for feature extraction, and the sizes of the three convolution kernels are 3×3×3, 5×5×5 and 7×7×7 respectively.
[0070] It can be understood that the 3D convolution layer uses convolution kernels of different sizes to extract features from the input thermal image, which can effectively extract features of different granularities. Each 3D convolution layer is followed by a batch normalization (BN) layer and a maximum pooling layer, where the BN layer is used for data normalization, and the maximum pooling layer helps improve computational efficiency and the overall performance of the ADHD detection model. Using multiple layers of 3D convolution layers ensures the extraction of deeper feature information, and the fully connected layer is used to output the extracted feature information.
[0071] The temporal feature extraction module includes a neural network of multiple LSTM units, a BN layer and a fully connected layer.
[0072] In this embodiment, a neural network with 256 LSTM units is used to extract time features from the input thermal image, so as to better focus on the implicit relationship between thermal images of different time series.
[0073] Specifically, the feature fusion module includes a multilayer perceptron, a flatten layer and a normalization layer; wherein the multilayer perceptron (MLP) is used to perform weighted processing on the input spatial feature map and temporal feature map, and fuse the weighted results to generate a spatiotemporal feature map; the flatten layer is used to reduce the dimension of the spatiotemporal feature map.
[0074] It can be understood that fusing the temporal and spatial features of thermal images through MLP, thereby extracting the temporal and spatial features of brain waves, is conducive to increasing the reliability of ADHD detection.
[0075] Specifically, the classification module introduces the LightBGM algorithm to classify the feature maps output by the feature fusion module.
[0076] Furthermore, when training the ADHD auxiliary assessment model, the cross-loss entropy function is used to adjust the model parameters. The cross-loss entropy function is as follows:
[0077] L(y,p)=-[ylog(p)+(1-y)log(1-p)]
[0078] Among them, p represents the probability of model output, y represents the sign function, and the ADHD auxiliary assessment model predicts that the accurate output is 1 and the error is 0.
[0079] Step 5: Input the thermal image corresponding to the EEG data of the target detection person into the trained ADHD auxiliary evaluation model to obtain the ADHD auxiliary detection result of the target detection person.
[0080] An ADHD auxiliary detection method based on image processing provided by an embodiment of the present invention converts the data of EEG image detection through electrodes at different positions into thermal images through preprocessing and other operations, and proposes an ADHD auxiliary detection model. By extracting the spatial and temporal features of the dynamic thermal image and fusing them, and classifying them based on the fused features, and introducing the LightBGM algorithm in the classification module for classification, the results obtained by the ADHD auxiliary detection model can improve the accuracy of ADHD auxiliary detection.
[0081] The embodiment of the present invention further provides an ADHD auxiliary detection system based on image processing, comprising:
[0082] The EEG signal acquisition module is used to acquire the EEG signal of the person to be tested and perform preprocessing;
[0083] The time-frequency data conversion module is used to perform time-frequency analysis on the preprocessed EEG signal to obtain EEG time-frequency data;
[0084] The thermal map generation module is used to map the spatial position of the electrode points of the test EEG data into a plane coordinate system, normalize the EEG time-frequency data and use different colors to map EEG data of different frequencies to generate a dynamic thermal image;
[0085] A model building module is used to build and train an ADHD auxiliary assessment model; wherein the ADHD auxiliary assessment model includes a feature extraction module, a feature fusion module and a classification module; the feature extraction module is used to extract spatial features and temporal features from the input dynamic thermal image; the feature fusion module is used to fuse and process the features extracted by the feature extraction module to generate a final feature map; the classification module performs classification based on the final feature map to generate an ADHD detection result;
[0086] The result output module is used to input the thermal image corresponding to the EEG data of the target detection person into the trained ADHD auxiliary evaluation model to obtain the ADHD auxiliary detection results and dynamic thermal images of the target detection person.
[0087] 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. An ADHD auxiliary detection method based on image processing, characterized in that: include: Step 1: Obtain the brain wave signal of the target detection person and perform preprocessing; Step 2: Perform time-frequency analysis on the preprocessed EEG signal to obtain EEG time-frequency data; Step 3: Convert EEG time-frequency data into dynamic thermal images; Step 4: Build and train an ADHD auxiliary assessment model; The ADHD auxiliary assessment model includes a feature extraction module, a feature fusion module and a classification module; the feature extraction module is used to extract spatial features and temporal features from the input dynamic thermal image; the feature fusion module is used to fuse and process the features extracted by the feature extraction module to generate a final feature map; the classification module performs classification based on the final feature map to generate an ADHD detection result; Step 5: Input the dynamic thermal image corresponding to the EEG data of the target detection person into the trained ADHD auxiliary evaluation model to obtain the ADHD auxiliary detection result of the target detection person.
2. The ADHD auxiliary detection method based on image processing according to claim 1, characterized in that: In step 1, the preprocessing includes filtering and re-referencing; the specific process includes: The Butterworth filter is used to filter the EEG signal; Perform re-reference processing on the filtered signal.
3. The ADHD auxiliary detection method based on image processing according to claim 1, characterized in that: In step 2, wavelet transform is used to perform time-frequency analysis on the image to obtain brain wave time-frequency data; the wavelet transform calculation formula is as follows: Among them, W f (a, b) represents the transformed EEG time-frequency data, a represents the scale factor, b represents the displacement factor, and f(t) represents the preprocessed EEG signal. represents the wavelet basis function.
4. The ADHD auxiliary detection method based on image processing according to claim 1, characterized in that: The step 3 specifically includes: According to the spatial position of the electrode points corresponding to the EEG time-frequency data, the EEG topographic map is obtained by mapping it to the plane coordinate system; The EEG time-frequency data corresponding to each electrode point in the EEG topographic map are normalized, and different colors are used to map the EEG data of different frequencies to generate a dynamic EEG thermal image.
5. The ADHD auxiliary detection method based on image processing according to claim 4, characterized in that: According to the spatial position of the electrode points corresponding to the EEG time-frequency data, the EEG topographic map is obtained by mapping it to the plane coordinate system, which specifically includes: Polar coordinate mapping is used to map the spatial position of the electrode points into a plane coordinate system to obtain an EEG topographic map.
6. The ADHD auxiliary detection method based on image processing according to claim 1, characterized in that: The feature extraction module includes a spatial feature extraction module and a temporal feature extraction module for parallel processing; The spatial feature extraction module includes three 3D convolutional layers and a fully connected layer connected in sequence, and each of the 3D convolutional layers is connected to a splicing layer, a BN layer and a maximum pooling layer; the 3D convolutional layer uses three convolution kernels of different sizes for feature extraction, and the sizes of the three convolution kernels are 3×3×3, 5×5×5 and 7×7×7 respectively; the temporal feature extraction module includes multiple LSTM units, a BN layer and a fully connected layer.
7. The ADHD auxiliary detection method based on image processing according to claim 1, characterized in that: The feature fusion module includes a multi-layer perceptron, a flatten layer and a normalization layer.
8. The ADHD auxiliary detection method based on image processing according to claim 1, characterized in that: The classification module introduces the LightBGM algorithm to classify the feature graph output by the feature fusion module.
9. The ADHD auxiliary detection method based on image processing according to claim 1, characterized in that: Step 4 also includes adjusting the model parameters using a cross-loss entropy function when training the ADHD auxiliary assessment model. The cross-loss entropy function is as follows: L(y,p)=-[ylog(p)+(1-y)log(1-p)] Among them, p represents the probability of model output, y represents the sign function, and the ADHD auxiliary assessment model predicts that the accurate output is 1 and the error is 0.
10. An ADHD auxiliary detection system based on image processing, characterized in that: include: The EEG signal acquisition module is used to acquire the EEG signal of the person to be tested and perform preprocessing; The time-frequency data conversion module is used to perform time-frequency analysis on the preprocessed EEG signal to obtain EEG time-frequency data; The thermal map generation module is used to map the spatial position of the electrode points of the test EEG data into a plane coordinate system, normalize the EEG time-frequency data and use different colors to map EEG data of different frequencies to generate a dynamic EEG thermal image; Model building module, used to build and train ADHD auxiliary assessment model; The result output module is used to input the thermal image corresponding to the EEG data of the target detection person into the trained ADHD auxiliary evaluation model to obtain the ADHD auxiliary detection result of the target detection person.