A device and system for identifying depression based on near-infrared data

Abstract

This application provides an apparatus and system for identifying depression based on near-infrared data. The apparatus includes a processor configured to acquire near-infrared data from multiple acquisition channels; convert the near-infrared data into Gram difference field images and Gram sum field images using Gram angular field transform, and obtain a first feature image group based on the Gram difference field images and Gram sum field images; obtain functional connectivity parameters between acquisition channels at different time periods based on the near-infrared data, and obtain a second feature image group using the functional connectivity parameters for a preset time period; sort the first and second feature image groups to obtain a fused third feature image group; input the third feature image group into a classification model to obtain a classification result for the subject, the classification result indicating the subject's tendency to suffer from depression. This improves the accuracy of identifying whether a subject has a predisposition to depression.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a device and system for identifying depression based on near-infrared data. Background Technology

[0002] Depression, a chronic and relapsing mental illness, is a significant risk factor for suicide. Its clinical symptoms primarily include depressed mood, cognitive impairment, feelings of worthlessness, and social phobia. Currently, the clinical diagnosis of depression mainly relies on physician clinical interviews and assessments. However, due to the significant heterogeneity of the disease, individual differences in symptom expression, and the high dependence of the patient's subjective report and the physician's clinical experience on the diagnostic process, the diagnosis is prone to confusion, making objective and accurate early identification difficult.

[0003] Functional near-infrared spectroscopy (FNIRS) combined with verbal fluency task (VFT) testing provides psychiatrists with important insights into patient conditions and disease interventions. FNIRS-based classification and identification techniques have become a research hotspot in the auxiliary diagnosis of depression. However, existing classification methods, based on single-type features, struggle to comprehensively describe the complete characteristics of the data, resulting in insufficient classification generalization performance and low accuracy in identifying depression, failing to meet the practical needs of clinical auxiliary diagnosis. Summary of the Invention

[0004] This application addresses the aforementioned technical problems in the existing technology. The purpose of this application is to provide a device and system for identifying depression based on near-infrared data. By fusing temporal and spatial feature images, it can simultaneously capture dynamic changes in brain function and abnormalities in brain region coordination, thereby enriching the feature information used for classification. This not only improves classification generalization performance but also enhances the accuracy and reliability of classifying and identifying predisposition to depression.

[0005] According to a first aspect of this application, a device for identifying depression based on near-infrared data is provided. The device includes a processor configured to: acquire near-infrared data from multiple acquisition channels of a target brain region when a subject performs a cognitive task; convert the normalized near-infrared data into Gram difference field images and Gram sum field images using Gram angle field transform, and obtain a first feature image group based on the Gram difference field images and Gram sum field images; obtain functional connectivity parameters between acquisition channels at different time periods based on the near-infrared data, and obtain a second feature image group using the functional connectivity parameters for a preset time period; sort the first feature image group and the second feature image group to obtain a fused third feature image group; input the third feature image group into a classification model to obtain a classification result for the subject, the classification result indicating the subject's tendency to be a patient with depression.

[0006] According to a second aspect of this application, a device for identifying mental illness based on near-infrared data is provided. The device includes a processor configured to: acquire near-infrared data from multiple acquisition channels of a target brain region when a subject performs a cognitive task; convert the normalized near-infrared data into Gram angle difference field images and Gram angle sum field images using Gram angle field transform, and obtain a first feature image group based on the Gram angle difference field images and Gram angle sum field images; obtain functional connectivity parameters between acquisition channels at different time periods based on the near-infrared data, and obtain a second feature image group using the functional connectivity parameters for a preset time period; sort the first feature image group and the second feature image group to obtain a fused third feature image group; input the third feature image group into a classification model to obtain a classification result for the subject, wherein the classification result indicates that the subject has a predisposition to mental illness.

[0007] According to a third aspect of this application, a system for identifying depression based on near-infrared data is provided. The system includes: a headgear worn on the head of a subject, the headgear having multiple probes for transmitting and / or receiving near-infrared signals to acquire near-infrared data from multiple acquisition channels; and a processor configured to: acquire near-infrared data from multiple acquisition channels of a target brain region when the subject performs a cognitive task; convert the normalized near-infrared data into Gram angle difference field images and Gram angle sum field images using Gram angle field transform, and obtain a first feature image group based on the Gram angle difference field images and Gram angle sum field images; obtain functional connectivity parameters between acquisition channels at different time periods based on the near-infrared data, and obtain a second feature image group using the functional connectivity parameters for a preset time period; sort the first feature image group and the second feature image group to obtain a fused third feature image group; input the third feature image group into a classification model to obtain a classification result for the subject, the classification result indicating the subject's tendency to be a patient with depression.

[0008] According to the fourth aspect of this application, a system for identifying mental illness based on near-infrared data is provided. The system includes: a headgear for wearing on a subject's head, the headgear having multiple probes for transmitting and / or receiving near-infrared signals to acquire near-infrared data from multiple acquisition channels; and a processor configured to: acquire near-infrared data from multiple acquisition channels of a target brain region when the subject performs a cognitive task; convert the normalized near-infrared data into Gram angle difference field images and Gram angle sum field images using Gram angle field transform, and store the Gram angle difference field images and Gram angle sum field images to obtain a first feature image group; obtain functional connectivity parameters between acquisition channels at different time periods based on the near-infrared data, and obtain a second feature image group using the functional connectivity parameters for a preset time period; sort the first feature image group and the second feature image group to obtain a fused third feature image group; input the third feature image group into a classification model to obtain a classification result for the subject, the classification result indicating that the subject has a tendency to suffer from mental illness.

[0009] Compared with the prior art, the beneficial effects of the embodiments of this application are as follows: The device for identifying depression based on near-infrared data provided in this application utilizes Gram angle field transform to convert near-infrared data into Gram angle difference field images and Gram angle sum field images. A first feature image set is obtained based on these images, preserving the temporal correlation of brain functional signals and information on brain activation intensity. A second feature image set, based on spatial features constructed from dynamic functional connectivity, captures the dynamic trends of inter-brain region coordination relationships. The first and second feature image sets are sorted to obtain a fused third feature image set, which simultaneously contains information on the temporal correlation of brain functional signals, information on brain activation intensity, and information on the dynamic trends of inter-brain region coordination relationships. Therefore, when training a classification model using the obtained third feature image set of training samples, it is beneficial for the classification model to simultaneously learn the dynamic changes in brain activation intensity and the characteristics of inter-brain region coordination relationships, significantly improving the classification model's generalization ability.

[0010] When identifying a subject's predisposition to depression, inputting the obtained third-feature image set of the subject into a trained classification model can improve the accuracy of identifying whether the subject has a predisposition to depression. Validation results using an external test set show that the device provided in this application achieves an accuracy rate of 71.3% in identifying depression, significantly improving the accuracy and robustness of clinical auxiliary diagnosis and providing reliable support for early screening and clinical auxiliary diagnosis of depression.

[0011] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above description and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0012] In drawings that are not necessarily drawn to scale, the same reference numerals may describe similar parts in different views. Similar reference numerals with different letter suffixes may indicate different examples of similar components. The drawings generally illustrate various embodiments by way of example rather than limitation, and are used together with the specification and claims to illustrate the disclosed embodiments. Such embodiments are illustrative and exemplary, and are not intended to be exhaustive or exclusive embodiments of the method, apparatus, system, or non-transitory computer-readable medium having instructions for implementing the method.

[0013] Figure 1 A schematic diagram of a device for identifying depression based on near-infrared data according to an embodiment of this application is shown.

[0014] Figure 2A flowchart illustrating processor execution in an apparatus for identifying depression based on near-infrared data according to an embodiment of this application is shown.

[0015] Figure 3 A schematic diagram is shown showing the fusion of a first feature image group and a second feature image group to obtain a third feature image group according to an embodiment of this application.

[0016] Figure 4 A schematic diagram of the classification process of inputting a third feature image group into a classification model according to an embodiment of this application is shown.

[0017] Figure 5 A schematic block diagram of the network architecture of a shuffling network according to an embodiment of this application is shown.

[0018] Figure 6 A schematic diagram of a system for identifying depression based on near-infrared data according to an embodiment of this application is shown. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions of this application, the application will be described in detail below with reference to the accompanying drawings and specific embodiments. The embodiments of this application will be further described in detail below with reference to the accompanying drawings and specific examples, but these are not intended to limit the scope of this application.

[0020] The terms "first," "second," and similar words used in this application do not indicate any order, quantity, or importance, but are merely used for distinction. The terms "including" or "comprising," etc., used in this application mean that the element preceding the word encompasses the elements listed after the word, and do not exclude the possibility of encompassing other elements. In this application, the arrows shown in the figures for each step are merely examples of the execution order, not limitations. The technical solution of this application is not limited to the execution order described in the embodiments. The steps in the execution order can be combined, broken down, or rearranged, as long as the logical relationship of the executed content is not affected.

[0021] All terms used in this application (including technical or scientific terms) have the same meaning as understood by one of ordinary skill in the art to which this application pertains, unless otherwise specifically defined. It should also be understood that terms defined in general dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant art, and not as idealized or highly formalized, unless expressly defined herein. Technologies and equipment known to one of ordinary skill in the art may not be discussed in detail, but where appropriate, such technologies and equipment should be considered part of the specification.

[0022] Figure 1 A schematic diagram of a device for identifying depression based on near-infrared data according to an embodiment of this application is shown, such as... Figure 1 As shown, the device 100 includes a processor 101.

[0023] The processor 101 is configured to execute Figure 2 The steps S201-S205 are shown.

[0024] In this embodiment, the arrows shown in the figure for each step are merely examples of the execution order and not limitations. The technical solution of this application is not limited to the execution order described in the embodiment. The steps in the execution order can be combined, decomposed, or their order can be changed, as long as the logical relationship of the execution content is not affected.

[0025] In step S201, near-infrared data from multiple acquisition channels of the target brain region are acquired when the subject performs a cognitive task.

[0026] The near-infrared data can be light intensity change data directly acquired by the acquisition module, or light density change data obtained by converting light intensity change data, or relative change data of hemoglobin concentration obtained by further utilizing the modified Beer-Lambert law. Specifically, the relative change data of hemoglobin concentration can include at least one of the relative change of oxygenated hemoglobin (HbO) concentration, the relative change of deoxygenated hemoglobin (HbR) concentration, and the relative change of total hemoglobin (HbT) concentration.

[0027] Specifically, the processor 101 can acquire light intensity change data for subsequent processing, or it can directly acquire light density change data or relative change data of hemoglobin concentration obtained by other devices for subsequent processing. This application does not make specific limitations on this, as long as the processor can perform subsequent processing based on the acquired near-infrared data to identify the tendency of the examinee to be a patient with depression.

[0028] For example, near-infrared brain functional imaging equipment can be used to acquire light intensity change data. The near-infrared brain functional imaging equipment has at least a headgear, which is worn on the head of the subject to acquire light intensity change data of the target brain region. For example, the headgear may have multiple probes for transmitting and / or receiving near-infrared light to acquire light intensity change data of multiple corresponding acquisition channels. Alternatively, the headgear may have multiple mounting positions for detachable assembly of individual probes; in use, the probes can be mounted onto the headgear through these mounting positions. Each of the multiple probes can be configured as a transmitting probe or a receiving probe, and each pair of probes can form an acquisition channel. The arrangement of the transmitting and receiving probes on the headgear can be determined based on the distribution location of the target brain region of the subject.

[0029] For example, the processor 101 can acquire light intensity change data of the target brain region of the subject in real time, collected by the probe. Alternatively, the light intensity change data of the target brain region collected by the probe can be saved to a designated storage space, which the processor 101 can access via wired or wireless means to obtain the light intensity change data of the target brain region of the subject. Alternatively, after converting the acquired light intensity change data into relative change data of hemoglobin concentration, the processor 101 can directly acquire the relative change data of hemoglobin concentration, perform normalization processing, and obtain Gram angle difference field image and Gram angle sum field image. It is understood that before normalizing the near-infrared data, the processor can also perform other conventional processing on the near-infrared data, such as filtering and baseline drift removal, and the processing methods and steps can be selected according to actual needs.

[0030] Optionally, the target brain region may be a brain region related to the identification of depression and / or of interest to the user. In some preferred embodiments, the target brain region may be the frontal lobe and / or temporal lobe. Experimental verification has shown that near-infrared data based on the frontal lobe and / or temporal lobe has higher accuracy in identifying depression.

[0031] In some embodiments, the cognitive task may be, for example, a Stroop task, an n-back task, a spatial working memory task, or a language fluency task, etc., which are merely examples and do not exclude other cognitive tasks related to the assessment of depression.

[0032] Preferably, the cognitive task is a language fluency task. Experimental verification has shown that, compared to other cognitive tasks, near-infrared data from the frontal and / or temporal lobes of the subject performing the language fluency task yields higher accuracy in identifying depression. For ease of understanding, in all embodiments shown in this application, subsequent processing and analysis are based on the near-infrared data obtained from the subject performing the language fluency task.

[0033] In step S202, the near-infrared data after normalization is converted into Gram difference field image and Gram sum field image using Gram angle field transform, and a first feature image group is obtained based on the Gram difference field image and Gram sum field image.

[0034] Specifically, before normalizing the near-infrared data, preprocessing can be performed on the near-infrared data. Prior to preprocessing, data quality can be checked. For example, the coefficient of variation can be calculated to check data quality. If the number of channels with a coefficient of variation greater than 10% accounts for more than 10% of the total number of channels, it indicates that the acquired near-infrared data has too much noise and is unreliable; this data set can be directly excluded to avoid interference with subsequent analysis.

[0035] For example, the preprocessing may include motion artifact correction, physiological noise removal, or baseline correction, or it may include data on the relative change in hemoglobin concentration obtained using a modified Beer-Lambert law. This application does not limit this, and the processing methods and steps can be selected according to actual needs.

[0036] In some embodiments of this application, before normalizing the near-infrared data, the time series of the near-infrared data can be divided into multiple consecutive time periods, and a representative value of the near-infrared data in each time period can be obtained. The representative values ​​corresponding to each time period are then recombined according to the chronological order of the corresponding time periods to obtain a compressed time series. In this way, the length of the time series of dynamic functional connectivity can be matched and calculated while preserving the trend of brain activation intensity changes using the compressed time series.

[0037] Specifically, near-infrared data can be compressed using the Piecewise Aggregate Approximation (PAA) algorithm. Of course, those skilled in the art can also choose other compression methods or not compress near-infrared data according to actual needs, and this application does not make specific limitations in this regard.

[0038] Specifically, the modified Beer-Lambert law can be used to obtain time series data for each acquisition channel, which can be used to represent the distribution of the relative change in hemoglobin concentration at different acquisition times.

[0039] As a preferred embodiment, the relative change in hemoglobin concentration described in this application is the relative change in oxygenated hemoglobin (HbO) concentration. In other words, the time series data can be the distribution of the relative change in oxygenated hemoglobin concentration at different acquisition times.

[0040] For example, assuming there are 48 acquisition channels, a probe acquisition frequency of 11Hz, and a total acquisition duration of 125 seconds, there are a total of 1375 (i.e., 125*11) acquisition moments, each corresponding to a relative change in hemoglobin concentration. These 1375 data points are divided into 48 consecutive time periods, and the relative change in hemoglobin concentration in each period is averaged to obtain a representative value. Then, the representative values ​​corresponding to these 48 consecutive time periods are recombined in chronological order to obtain a compressed time series. This compressed time series includes representative values ​​corresponding to 48 recombined time points, representing the distribution of the representative value of the relative change in hemoglobin concentration across these 48 recombined time points, enabling the subsequent generation of 48×48 dimensional Gram difference field images and Gram sum field images.

[0041] In some embodiments of this application, the near-infrared data of each acquisition channel is normalized based on the maximum and minimum values ​​obtained from a preset near-infrared data set, which includes near-infrared data from healthy subjects and patients with depression. In other words, this application uses the maximum and minimum values ​​obtained from a preset near-infrared data set that simultaneously includes near-infrared data from healthy subjects and patients with depression as normalization boundaries to normalize the near-infrared data of each acquisition channel for the current subject.

[0042] The maximum and minimum values ​​can be the maximum and minimum values ​​of data directly obtained from a preset near-infrared data set, or they can be the maximum and minimum values ​​obtained based on data in the preset near-infrared data set. For example, the data in the preset near-infrared data set can be sorted from smallest to largest, and the maximum and minimum values ​​can be determined based on a preset proportion of data. For example, the preset proportion can be 5%. The average value of the first 5% (i.e., low values) of the data can be taken as the minimum value, and the average value of the last 5% (i.e., high values) of the data can be taken as the maximum value. Alternatively, the near-infrared data in the preset near-infrared data set can be compressed, and the maximum and minimum values ​​of the compressed data can be used as the normalization boundary.

[0043] The maximum and minimum values ​​can also be preset. Understandably, these preset values ​​should be correlated with the maximum and minimum values ​​obtained from a large sample of data, including healthy individuals and patients with depression, to ensure accuracy. In summary, all data acquisition channels for the current subjects use the same maximum and minimum values ​​for normalization. This allows for a holistic consideration of the subject's brain activation, avoiding the loss of brain activation information caused by normalization at different scales. This improves the accuracy of classification results when identifying the subject's tendency towards depression.

[0044] Specifically, based on the modified Beer-Lambert law, time series data corresponding to each acquisition channel are obtained. The time series can first be divided into multiple consecutive time periods, and representative values ​​of the relative change in hemoglobin concentration within each time period are obtained. These representative values ​​are then recombined according to the chronological order of the corresponding time periods to obtain a compressed time series. This compressed time series represents the distribution of the representative values ​​of the relative change in hemoglobin concentration at different recombined time points. Then, for each acquisition channel, the compressed time series is further analyzed... j Representative value of the relative change in hemoglobin concentration The normalization process is performed using formula (1) to obtain representative values ​​of the relative changes in hemoglobin concentration after normalization, with a distribution range between [-1, 1]. .

[0045] Formula (1).

[0046] in, j It is a positive integer representing the total number of representative values ​​of the relative change in hemoglobin concentration that are less than or equal to the total number of such values. The maximum value is... This is the minimum value.

[0047] The representative values ​​of the relative changes in hemoglobin concentration in the compressed time series corresponding to each acquisition channel are calculated sequentially using formula (1) to obtain the normalized compressed time series.

[0048] The representative value of the relative change in hemoglobin concentration in the normalized compressed time series can be obtained by formula (2). Convert to polar coordinates in the polar coordinate system:

[0049] Formula (2).

[0050] In formula (2), For the first j Representative values ​​of the relative changes in hemoglobin concentration in the compressed time series corresponding to each acquisition channel after normalization. The corresponding polar coordinates in the polar coordinate system (unit: radians). arccos is the inverse cosine function. r j For the first j Recombined time points in a normalized compressed time series t j The corresponding radius coordinates in the polar coordinate system. N A constant factor used to standardize the span of the polar coordinate system.

[0051] After obtaining the polar coordinates of the compressed time series corresponding to each acquisition channel after normalization, the Gram angle difference field image (GADF) and the Gram angle sum field image (GASF) can be obtained by formula (3):

[0052] Formula (3).

[0053] in, and These are a series of angle values ​​obtained by mapping the normalized compressed time series using formula (2). 1 corresponds to the polar coordinate angle of the first data point in the compressed time series. N The number of the corresponding compressed time series N The polar coordinates of the data points. N This represents the total number of data points in the compressed time series.

[0054] A first feature image group can be obtained by arranging GADF images and GASF images alternately or sequentially. For example, the first feature image group includes 48 GADF images and 48 GASF images arranged alternately, and the size of the first feature image group is 48×48×96.

[0055] Returning to the embodiments of this application, in step S203, functional connection parameters between acquisition channels at different time periods are obtained based on the near-infrared data, and a second feature image group is obtained using the functional connection parameters for a preset time period. Exemplarily, the second feature image group can be obtained using functional connection parameters for all time periods, or it can be obtained using functional connection parameters for only a portion of the time periods; this application does not specifically limit this.

[0056] The inventors believe that the trend of brain inter-regional synergy in patients with depression differs significantly from that in healthy individuals. The second feature image set, which constructs spatial features based on dynamic functional connectivity, can capture the temporal trend of brain inter-regional synergy. By combining the first feature image set, which retains the temporal correlation of brain functional signals and information on brain activation intensity, with the second feature image set, which can capture the dynamic trend of brain inter-regional synergy, the predisposition to depression in subjects can be identified, thereby improving the accuracy of identification.

[0057] Specifically, a sliding window can be used to divide the time series of the near-infrared data into multiple consecutive time periods, and the functional connection parameters between the acquisition channels in each time period can be obtained.

[0058] For example, taking the near-infrared data collected from the subject when performing the VFT task as an example, 125 seconds of near-infrared data were obtained, totaling 1375 data points. For example, the sliding window length was set to 10 seconds and the step size was set to 1 second, generating a total of 116 sliding time windows, resulting in 116 periods of near-infrared data.

[0059] In formula (4), r is the Pearson correlation coefficient, n represents the number of data points in each time period, Pi represents the relative change in hemoglobin concentration in acquisition channel P corresponding to the i-th data point, and Qi represents the relative change in hemoglobin concentration in acquisition channel Q corresponding to the i-th data point. This represents the mean of all data points collected in channel P during that time period. This represents the average value of all data points collected in channel Q during that time period.

[0060] Based on the functional connectivity parameters *r* between the acquisition channels within each time period, the functional connectivity matrices for each time period can be obtained. These functional connectivity matrices are then plotted as corresponding second feature images, resulting in a set of second feature images. Preferably, these Pearson correlation coefficients *r* can be converted into z-scores using Fisher's r-to-z transform to improve normality. Using functional connectivity matrices constructed based on z-scores during training is beneficial for improving the stability and accuracy of subsequent classification model predictions.

[0061] In this example, 116 spatial feature images of size 48×48 were constructed based on 116 windows to obtain a second feature image group with a size of 48×48×116. The second feature image group not only contains the brain region coordination state at a single instant, but also the dynamic trend of the brain region coordination state.

[0062] This is merely an example and does not constitute a limitation on any specific solution.

[0063] In step S204, the first feature image group and the second feature image group are sorted to obtain the fused third feature image group.

[0064] Specifically, such as Figure 3 As shown, the first feature image group includes GADF images and GASF images, which are alternately ordered. The first and second feature image groups are arranged in sequence to obtain the fused third feature image group. In this way, richer brain activity information can be obtained compared to single temporal or single spatial feature images.

[0065] In step S205, the third feature image group is input into the classification model to obtain the classification result of the subject, which indicates the subject's tendency to be a patient with depression.

[0066] The third feature image set simultaneously contains information on the temporal correlation of brain functional signals, brain activation intensity, and dynamic trends in brain inter-regional coordination. Therefore, using the obtained third feature image set of the training samples to train a classification model allows the model to simultaneously learn the dynamic changes in brain activation intensity and abnormal features of brain inter-regional coordination, significantly improving the classification model's generalization ability. When identifying a subject's predisposition to depression, inputting the obtained third feature image set of the subject into the trained classification model can improve the accuracy of identifying whether the subject has a predisposition to depression.

[0067] As a preferred embodiment, the classification model is ShuffleNet, a lightweight deep learning architecture that can efficiently perform classification. Experiments have shown that, compared to other classification models, inputting the third feature image group into the ShuffleNetV2 model to identify depression has a higher recognition accuracy.

[0068] Specifically, the shuffling network includes a feature extraction module, multiple shuffling modules, a convolutional module, and a classification module. It should be understood that the network structure of the shuffling network can be added to or removed according to the developer's needs. For example, the ShuffleNetV2 model is used to classify the third feature image group, and the processor 101 executes as follows: Figure 4 Steps S401 to S404 are shown.

[0069] In step S401, the third feature image group is input to the feature extraction module to obtain the image features of the third feature image group.

[0070] The aforementioned feature extraction module is used to perform feature extraction operations on the third feature image group to obtain the first image features of the third feature image group. In some embodiments, the architecture of the aforementioned feature extraction module can be designed with reference to convolutional extraction architecture, autoencoder extraction architecture, etc., and the embodiments of this application are not limited thereto.

[0071] In step S402, the image features or rearranged features obtained by the shuffling module are divided into two groups of data. A convolution operation is performed on one of the two groups of data. The convolutional group of data is then concatenated with the other group of data and rearranged to obtain rearranged features.

[0072] The shuffling module can be used to further extract features from image features or rearranged features obtained by other shuffling modules. Multiple shuffling modules can be connected sequentially. For example, if there are three shuffling modules 1 to 3, shuffling module 1 can be connected to shuffling module 2, shuffling module 2 can be connected to shuffling module 3, and shuffling module 3 can be connected to the classification module.

[0073] Each shuffling module can use the same convolution parameters (e.g., kernel size, kernel stride) when performing convolution operations to achieve a lightweight shuffling network. The concatenation operation described above can be performed based on the concat operator in related technologies. The rearrangement operation described above can be used to rearrange the feature points in the concatenated features obtained from the concatenation operation. This rearrangement can be achieved by multiplying the concatenated features by a preset transpose matrix, or by rearranging the concatenated features according to specific rules. Through the processing of the shuffling module, information flow of image features or rearranged features can be realized, which is beneficial to improving the accuracy of classification results.

[0074] In step S403, the rearranged features obtained by the shuffling module connected to the convolution module are convolved to obtain convolutional features.

[0075] The convolution module is used to extract features again from the received rearranged features to improve the representativeness of the rearranged features. The convolution parameters used by the convolution module when performing convolution operations can be determined according to the developer's needs.

[0076] In step S404, the classification module performs a classification operation on the convolutional features to obtain the classification result of the examinee.

[0077] In one example, the classification module can output a confidence score for either a healthy individual or someone with a predisposition to depression, based on the convolutional features, as the classification result. In another example, the label with the highest confidence score, indicating either a healthy individual or someone with a predisposition to depression, can also be used as the classification result.

[0078] See Figure 5 , Figure 5 A schematic block diagram of the network architecture of a shuffling network according to one embodiment of this application is shown. (In conjunction with...) Figure 5The shuffling network may include an input layer, an initial convolutional layer, a pooling layer, multiple network layers stacked through multiple shuffling modules, a first convolutional layer, a global average pooling layer, and a fully connected layer. The input layer receives the test image or training image and inputs it to the initial convolutional layer and pooling layer. In this example, the feature extraction module mentioned above may include the initial convolutional layer and pooling layer. The convolutional module mentioned above may include the first convolutional layer. The classification module mentioned above may include a global average pooling layer and a fully connected layer. Each shuffling module may include a channel segmentation layer, several 1*1 convolutional layers, 3*3 convolutional layers, a channel concatenation layer, and a channel rearrangement layer. The channel segmentation layer is used to divide any one of the image features or rearranged features obtained from other shuffling modules into two groups of data. The 1*1 convolutional layer and 3*3 convolutional layer are used to perform convolution operations on one of the two groups of data. The channel concatenation layer is used to concatenate the convolutionally processed group of data with the other group of data. The aforementioned channel rearrangement layer is used to rearrange the spliced ​​features obtained from the splicing operation to obtain rearranged features.

[0079] Furthermore, the inventors verified the recognition effect on depression by comparing the representation inputs of different near-infrared data with classifier combinations. As shown in Table 1, the different input types include HbO time series without image processing, images with spatiotemporal features (i.e., the third feature image group), Gram angle field images with HbO Int obtained based on the normalization processing method of the embodiments of this application (i.e., the first feature image group), Gram angle field images without HbO Int obtained using the conventional normalization method, and dynamic function connection matrix images (i.e., the second feature image group). Among them, the conventional normalization processing method uses the maximum and minimum values ​​of the near-infrared data of each acquisition channel as the normalization boundary for normalization processing of each acquisition channel. That is, each acquisition channel is normalized based on its own maximum and minimum values, and the normalization scale of each acquisition channel is not uniform.

[0080] Table 1. The effect of different near-infrared data representation inputs on the identification of depression.

[0081] Comparing the results with different input types reveals that the input type affects both the performance and generalization ability of the classification model. External test set validation results show that, overall, images incorporating spatiotemporal features (i.e., the third feature image group) perform better as input, with the ShuNetV2 (ShuffleNetV2) model achieving the highest recognition accuracy of 71.3% ± 2.3%. This indicates that using the third feature image group described in the various embodiments of this application as input to the classification model helps improve the accuracy of identifying predisposition to depression.

[0082] In some embodiments of this application, the sample of patients with depression in the preset near-infrared data set includes patients with mild depression, patients with moderate depression, and / or patients with severe depression.

[0083] Specifically, in some embodiments, the preset near-infrared data set may include near-infrared data of healthy subjects as well as near-infrared data of patients with different degrees of depression, such as near-infrared data of patients with mild, moderate and / or severe depression. In this way, compared with near-infrared data of only healthy subjects and patients with a single degree of depression, it is beneficial to improve the accuracy of identifying whether the examinee has a tendency to depression.

[0084] During training, using samples covering mild, moderate, and severe depression to train the classification model can help the model learn the specific differences in brain function corresponding to different degrees of depression. It can also help the model learn the continuous trend of brain function characteristics from the degree of depression in healthy individuals to those with mild, moderate, and severe depression. This not only helps the classification model identify whether the subject has a tendency to develop depression, but also helps to distinguish the degree of depression.

[0085] In some embodiments of this application, a device for identifying mental illness based on near-infrared data is provided. The device includes a processor configured to: acquire near-infrared data from multiple acquisition channels of a target brain region when a subject performs a cognitive task; convert the normalized near-infrared data into Gram angle difference field images and Gram angle sum field images using Gram angle field transform, and obtain a first feature image group based on the Gram angle difference field images and Gram angle sum field images; obtain functional connectivity parameters between acquisition channels at different time periods based on the near-infrared data, and obtain a second feature image group using the functional connectivity parameters for a preset time period; sort the first feature image group and the second feature image group to obtain a fused third feature image group; input the third feature image group into a classification model to obtain a classification result for the subject, the classification result indicating the subject's predisposition to mental illness.

[0086] Thus, the fused third feature image set simultaneously contains information on the temporal correlation of brain functional signals, brain activation intensity, and dynamic trends in brain inter-regional coordination. As input to the classification model, this can improve the accuracy of identifying the subject's predisposition to mental illness. Furthermore, training the classification model for identifying the subject's predisposition to mental illness using the obtained training sample's third feature image set allows the model to simultaneously learn the dynamic changes in brain activation intensity and abnormal features in brain inter-regional coordination, significantly improving the classification model's generalization ability.

[0087] It should be noted that the processor can execute the various steps performed by the processor in the near-infrared data-based depression identification device of the above embodiments, which will not be repeated here. When identifying the mental illness tendency of the examinee, the device can also adopt the normalization processing method described above. The corresponding preset near-infrared data set is set according to the type of mental illness to be identified, for example, it includes samples from both healthy subjects and patients with the disease.

[0088] The embodiments in this application are only illustrated using depression as an example. However, the inventive concept of this application can not only be used to identify depression, but also to other mental illnesses besides depression, such as schizophrenia, anxiety disorder, depression combined with anxiety disorder, etc. These will not be listed one by one here. Those skilled in the art can make adaptive adjustments to the corresponding steps or content according to the application scenario.

[0089] In some embodiments of this application, a system for identifying depression based on near-infrared data is provided, such as... Figure 6 As shown, the system 600 includes a headgear 601 for wearing on a subject's head and a processor 602. The headgear 601 is equipped with multiple probes for transmitting and / or receiving near-infrared signals to acquire near-infrared data from multiple acquisition channels. The processor 602 is configured to acquire near-infrared data from multiple acquisition channels of a target brain region when the subject performs a cognitive task; convert the normalized near-infrared data into Gram angle difference field images and Gram angle sum field images using Gram angle field transform; obtain a first feature image group based on the Gram angle difference field images and Gram angle sum field images; obtain functional connectivity parameters between acquisition channels at different time periods based on the near-infrared data; obtain a second feature image group using the functional connectivity parameters for a preset time period; sort the first and second feature image groups to obtain a fused third feature image group; input the third feature image group into a classification model to obtain a classification result for the subject, the classification result indicating the subject's tendency to be a patient with depression.

[0090] Thus, when identifying the depressive tendency of subjects, inputting the third feature image group, which simultaneously contains the time-series correlation of brain functional signals, information on brain activation intensity, and dynamic trend information on the synergistic relationship between brain regions, into the trained classification model can improve the accuracy of identifying whether subjects have a tendency to suffer from depression.

[0091] It should be noted that the steps executed by the processor in the device for identifying depression based on near-infrared data in the above embodiments will not be repeated here. In some embodiments of this application, a system for identifying mental illness based on near-infrared data is provided. The system includes: a headgear for wearing on a subject's head, the headgear having multiple probes for transmitting and / or receiving near-infrared signals to acquire near-infrared data from multiple acquisition channels; and a processor configured to: acquire near-infrared data from multiple acquisition channels of a target brain region when the subject performs a cognitive task; convert the normalized near-infrared data into Gram angle difference field images and Gram angle sum field images using Gram angle field transform, and obtain a first feature image group based on the Gram angle difference field images and Gram angle sum field images; obtain functional connectivity parameters between acquisition channels at different time periods based on the near-infrared data, and obtain a second feature image group using the functional connectivity parameters for a preset time period; sort the first and second feature image groups to obtain a fused third feature image group; and input the third feature image group into a classification model to obtain a classification result for the subject, the classification result indicating that the subject has a predisposition to mental illness.

[0092] It should be noted that the processor executes the various steps of the near-infrared data-based depression identification device in the above embodiments, which will not be repeated here. This system can be used to identify depression, as well as other mental illnesses besides depression, such as schizophrenia, anxiety disorders, and depression combined with anxiety disorders, which will not be listed here.

[0093] It should be understood that in the embodiments of this application, the processor may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0094] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0095] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0096] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0097] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of the invention. However, this approach should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with fewer features than all of those in a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0098] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules in the apparatus for assessing schizophrenic tendencies based on near-infrared data according to embodiments of the present invention. The present invention can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0099] The above description is merely a specific embodiment of the present invention or an explanation of that embodiment. The scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A device for identifying depression based on near-infrared data, characterized in that, The device includes a processor, the processor being configured to: Acquire near-infrared data from multiple acquisition channels of the target brain region when the subject performs a cognitive task; The near-infrared data after normalization is converted into Gram difference field image and Gram sum field image using Gram angle field transform, and a first feature image group is obtained based on the Gram difference field image and Gram sum field image. Based on the near-infrared data, functional connection parameters between acquisition channels at different time periods are obtained, and a second feature image group is obtained using the functional connection parameters for a preset time period. The first and second feature image groups are sorted to obtain the fused third feature image group; The third feature image group is input into the classification model to obtain the classification result of the subject, which indicates the subject's tendency to be a patient with depression.

2. The apparatus according to claim 1, characterized in that, The normalization process includes: The near-infrared data of each acquisition channel is normalized based on the maximum and minimum values ​​obtained from a preset near-infrared data set, which includes near-infrared data from healthy subjects and patients with depression.

3. The apparatus according to claim 2, characterized in that, The sample of patients with depression in the preset near-infrared data set includes patients with mild depression, moderate depression, and / or severe depression.

4. The apparatus according to claim 1, characterized in that, Based on the near-infrared data, functional connection parameters between acquisition channels at different time periods are obtained, specifically including: The time series of near-infrared data is divided into multiple consecutive time periods using a sliding window, and the functional connection parameters between the acquisition channels in each time period are obtained.

5. The apparatus according to claim 1, characterized in that, The classification model is a shuffling network, which includes a feature extraction module, multiple shuffling modules, a convolution module, and a classification module. The third feature image group is input into the classification model to obtain the classification result of the subject, including the following operations: The third feature image group is input into the feature extraction module to obtain the image features of the third feature image group; The shuffling module divides the image features or rearranged features obtained by other shuffling modules into two groups of data. A convolution operation is performed on one of the two groups of data. The convolutional group of data is then concatenated with the other group of data before a rearrangement operation is performed to obtain the rearranged features. The convolution module is used to perform a convolution operation on the rearranged features obtained by the shuffling module connected to the convolution module to obtain convolutional features. The classification module performs a classification operation on the convolutional features to obtain the classification result of the examinee.

6. The apparatus according to claim 1 or 2, characterized in that, Before normalizing the near-infrared data, the processor is further configured to: The time series of near-infrared data is split into multiple consecutive time periods, and the representative value of the near-infrared data in each time period is obtained. The representative values ​​of each time period are then recombined according to the time order of the corresponding time periods to obtain a compressed time series.

7. The apparatus according to claim 1, characterized in that, The cognitive task is a language fluency task, and the target brain regions are the frontal lobe and / or temporal lobe.

8. A device for identifying mental illnesses based on near-infrared data, characterized in that, The device includes a processor, the processor being configured to: Acquire near-infrared data from multiple acquisition channels of the target brain region when the subject performs a cognitive task; The near-infrared data after normalization is converted into Gram difference field image and Gram sum field image using Gram angle field transform, and a first feature image group is obtained based on the Gram difference field image and Gram sum field image. Based on the near-infrared data, functional connection parameters between acquisition channels at different time periods are obtained, and a second feature image group is obtained using the functional connection parameters for a preset time period. The first and second feature image groups are sorted to obtain the fused third feature image group; The third feature image group is input into the classification model to obtain the classification result of the subject, which indicates the subject's tendency to suffer from mental illness.

9. A system for identifying depression based on near-infrared data, characterized in that, The system includes: A headgear for wearing on a subject's head, the headgear being equipped with multiple probes for transmitting and / or receiving near-infrared signals to acquire near-infrared data from multiple acquisition channels; and The processor is configured as follows: Acquire near-infrared data from multiple acquisition channels of the target brain region when the subject performs a cognitive task; The near-infrared data after normalization is converted into Gram difference field image and Gram sum field image using Gram angle field transform, and a first feature image group is obtained based on the Gram difference field image and Gram sum field image. Based on the near-infrared data, functional connection parameters between acquisition channels at different time periods are obtained, and a second feature image group is obtained using the functional connection parameters for a preset time period. The first and second feature image groups are sorted to obtain the fused third feature image group; The third feature image group is input into the classification model to obtain the classification result of the subject, which indicates the subject's tendency to be a patient with depression.

10. A system for identifying mental illnesses based on near-infrared data, characterized in that, The system includes: It includes a headgear for wearing on a subject's head, the headgear being equipped with multiple probes for transmitting and / or receiving near-infrared signals to acquire near-infrared data from multiple acquisition channels; and The processor is configured as follows: Acquire near-infrared data from multiple acquisition channels of the target brain region when the subject performs a cognitive task; The near-infrared data after normalization is converted into Gram difference field image and Gram sum field image using Gram angle field transform, and a first feature image group is obtained based on the Gram difference field image and Gram sum field image. Based on the near-infrared data, functional connection parameters between acquisition channels at different time periods are obtained, and a second feature image group is obtained using the functional connection parameters for a preset time period. The first and second feature image groups are sorted to obtain the fused third feature image group; The third feature image group is input into the classification model to obtain the classification result of the subject, which indicates the subject's tendency to suffer from mental illness.

Images