Method and system for mild cognitive impairment recognition based on multi-task paradigm fusion of fNIRS data

Abstract

The application discloses a mild cognitive impairment recognition method and system based on multi-task paradigm fusion fNIRS data, and the method comprises the following steps: obtaining fNIRS data of a subject based on resting state tasks and n-back working memory tasks; extracting fNIRS related features based on the obtained fNIRS data; performing feature selection on the extracted fNIRS related features based on a feature selection algorithm to form a feature set; and inputting the formed feature set into a trained deep learning model to output a mild cognitive impairment recognition result. By obtaining fNIRS data under resting state and n-back working memory cognitive tasks, calculating multi-dimensional fNIRS related features, and performing feature selection based on a Gradient Boosting algorithm, the application finally combines a deep learning model to accurately recognize MCI. The method can not only effectively extract high-dimensional features, but also realize automatic MCI diagnosis through model training and reasoning, and has high accuracy and strong potential for popularization and application.

Description

TECHNICAL FIELD

[0001] The application relates to a mild cognitive impairment recognition method and system based on multi-task paradigm fusion of fNIRS data. BACKGROUND

[0002] Mild cognitive impairment (MCI) is a common cognitive function decline state in the elderly population, which is usually manifested as slight impairment in memory, attention, language ability, etc., but not enough to reach the level of dementia. MCI is often regarded as a precursor to neurodegenerative diseases such as Alzheimer's disease, therefore, early and accurate diagnosis of MCI and intervention can effectively delay the progression of the disease and improve the quality of life of patients. At present, the diagnosis of MCI mainly relies on clinical evaluation, including medical history collection, cognitive function evaluation, neuroimaging examination and laboratory examination, etc. Common cognitive evaluation tools include Montreal Cognitive Assessment Scale (MoCA), Mini-Mental State Examination (MMSE), etc. These scales evaluate multiple dimensions such as memory, language ability, attention and executive function to assist doctors in making preliminary diagnosis. However, although these tools are widely used in clinical practice, they still have some problems and limitations.

[0003] On the one hand, traditional cognitive evaluation scales rely on the experience of doctors and are easily affected by factors such as the psychological state of patients, educational background, cultural differences, etc., which may lead to misdiagnosis or missed diagnosis. Especially in the early stage of MCI, the cognitive function of patients declines slightly, and traditional evaluation methods may not be able to timely detect mild cognitive impairment; on the other hand, although existing neuroimaging techniques (such as magnetic resonance imaging MRI, positron emission tomography PET, etc.) have played an important role in the diagnosis of neurodegenerative diseases, these techniques not only have high cost and complex operation, but also have high requirements for equipment, which limits their application in a wide range of population. Especially in the diagnosis of early MCI, imaging examination often fails to reveal minor brain changes, and the diagnostic sensitivity is low.

[0004] In recent years, near-infrared functional imaging technology (fNIRS) has been widely used in cognitive neuroscience research as a non-invasive brain function imaging technology. fNIRS can reflect neural activity by measuring blood oxygen changes in the cerebral cortex region, has good spatial and temporal resolution, and the equipment is relatively light and easy to operate, suitable for large-scale screening.

[0005] The application of fNIRS in cognitive function testing mainly includes two types of task modes: cognitive tasks and resting state. In the study of mild cognitive impairment, the working memory n-back working memory task can be used to assess working memory and attention. When patients perform tasks, fNIRS records the dynamic changes in cerebral blood flow and analyzes the activation strength of different brain regions during cognitive tasks. Resting state tasks can be used to assess the strength of network connections between different brain regions. Brain region activation during the n-back working memory task and the strength of network connections between brain regions during the resting state have certain diagnostic value for identifying cognitive impairments such as MCI. However, fNIRS data in any single task mode usually has only limited discriminatory power, and in-depth fusion analysis and processing of the data are required to extract features with diagnostic value.

[0006] Currently, fNIRS-based MCI identification technology still faces the following challenges and shortcomings:

[0007] 1. Feature selection and high-dimensional data processing issues: fNIRS data typically contains a large amount of time-series signals and blood oxygenation changes across multiple brain regions. These data are often high-dimensional and contain redundant information. Traditional methods struggle to extract effective features, resulting in low recognition accuracy.

[0008] 2. Limitations of Traditional Machine Learning Methods: While existing recognition methods based on traditional machine learning algorithms have improved MCI recognition capabilities to a certain extent, these methods typically rely on manually selected features, perform poorly with high-dimensional data, and are unable to effectively capture complex nonlinear relationships.

[0009] 3. Lack of Multi-Level Model Integration: Many existing fNIRS data analysis methods lack the deep integration of multiple features and multi-level data, and cannot fully explore the complex relationships potentially present in the data. Therefore, although they can achieve a certain level of accuracy in certain tasks, their generalizability across different individuals is poor.

[0010] 4. Lack of automated and intelligent reasoning methods: Traditional MCI identification methods often rely on subjective judgment by experts and cumbersome analysis steps, and lack an intelligent model that can automatically perform fast and accurate reasoning on new samples. Summary of the Invention

[0011] To address the shortcomings of existing MCI diagnostic methods, the present invention provides a method and system for identifying mild cognitive impairment (MCI) based on fNIRS data in a multi-task fusion paradigm. This method acquires fNIRS data from resting-state and n-back working memory tasks, calculates multidimensional fNIRS-related features, and uses a gradient boosting algorithm for feature selection. Finally, it integrates a deep learning model for accurate MCI identification. This method not only effectively extracts high-dimensional features but also enables automated MCI diagnosis through model training and inference, demonstrating high accuracy and strong potential for widespread application.

[0012] According to one aspect of the present invention, a method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion is provided, comprising:

[0013] Obtain fNIRS data from the subjects based on the resting-state task and the n-back working memory task;

[0014] Extract fNIRS-related features based on the acquired fNIRS data;

[0015] Based on the feature selection algorithm, the extracted fNIRS-related features are selected to form a feature set;

[0016] The formed feature set is input into the trained deep learning model to output the mild cognitive impairment recognition results.

[0017] As a further technical solution, fNIRS-related features are extracted based on the acquired fNIRS data, including:

[0018] The fNIRS feature values ​​under the resting-state task and n-back working memory task were extracted respectively to form multi-dimensional fNIRS features.

[0019] As a further technical solution, the fNIRS feature values ​​extracted under the resting-state task include: functional connectivity, betweenness centrality, degree centrality, node clustering coefficient, node efficiency, node local efficiency, node shortest path length, global efficiency, local efficiency, clustering coefficient and small-world property.

[0020] As a further technical solution, the fNIRS feature values ​​extracted under the n-back working memory task include: beta value, mean, integral, K-activation and peak value.

[0021] As a further technical solution, feature selection is performed on the extracted fNIRS-related features, including: using a gradient boosting algorithm to perform feature selection on the extracted fNIRS-related features.

[0022] As a further technical solution, the training of the deep learning model includes:

[0023] Construct fNIRS data samples under resting-state task and n-back working memory task;

[0024] Extract fNIRS-related features of fNIRS data samples;

[0025] Perform feature selection on the extracted fNIRS-related features to form a feature set;

[0026] Label the features in the feature set to form a training data set;

[0027] The formed training data set is used to train the fully connected neural network and output the trained deep learning model.

[0028] As a further technical solution, after obtaining the fNIRS data of the subjects based on the resting-state task and the n-back working memory task, it also includes:

[0029] The acquired fNIRS data were preprocessed, including conversion of raw light intensity to optical density, artifact detection, filtering, and conversion of optical density to relative concentration.

[0030] According to one aspect of the present invention, a system for identifying mild cognitive impairment based on fNIRS data in a multi-task paradigm fusion is provided, comprising:

[0031] Data input module, used to obtain fNIRS data of subjects based on resting-state task and n-back working memory task;

[0032] A feature extraction module, used to extract fNIRS-related features based on the acquired fNIRS data;

[0033] A feature selection module is used to select the extracted fNIRS-related features based on a feature selection algorithm to form a feature set;

[0034] The feature recognition module is used to input the formed feature set into the trained deep learning model and output the mild cognitive impairment recognition results.

[0035] According to one aspect of the present invention, a device for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion is provided, comprising a processor and a memory; the memory stores at least one instruction, the at least one instruction being executed by the processor to implement the steps of the method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion.

[0036] According to one aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion are implemented.

[0037] Compared with existing technologies, this invention has significant advantages in the early identification of MCI by combining fNIRS data, feature selection, machine learning, and deep learning techniques:

[0038] 1. Improve diagnostic accuracy:

[0039] Through efficient feature extraction and selection of fNIRS data, especially the multi-dimensional features obtained by combining n-back and resting-state tasks, the present invention can capture subtle changes that are difficult to identify with traditional methods, significantly improving the accuracy of early diagnosis of MCI.

[0040] 2. Automated diagnosis:

[0041] The present invention uses a deep learning model for automated reasoning, which can diagnose new samples efficiently and without subjective bias, reduce manual intervention, lower the misdiagnosis rate, and significantly shorten the time required for diagnosis.

[0042] 3. Reduce costs and increase accessibility:

[0043] Compared to traditional imaging tests (such as MRI and PET), fNIRS equipment is relatively inexpensive and easy to use. This invention enables data collection using more widely available equipment, reducing the cost of MCI screening and increasing its potential for application in primary care hospitals and health checkup centers.

[0044] 4. Strong adaptability:

[0045] The technical solution of the present invention can be adapted to different patient groups, especially by assessing cognitive function through a multi-task combination (VFT and n-back), which enhances the sensitivity and generalization ability of the method and can adapt to the early diagnosis of different types of cognitive impairment. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A flowchart of a method for identifying mild cognitive impairment based on fNIRS data in a multi-task paradigm fusion according to an embodiment of the present invention.

[0047] Figure 2 A schematic diagram of feature extraction results provided by an embodiment of the present invention.

[0048] Figure 3 A schematic diagram of the model training process provided by an embodiment of the present invention.

[0049] Figure 4 Schematic diagram of a mild cognitive impairment identification system based on fNIRS data under a multi-task paradigm fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] With the rapid development of machine learning and deep learning technologies, artificial intelligence-based methods for MCI identification have gradually become a research hotspot. In recent years, algorithms such as gradient boosting, deep neural networks (DNNs), and convolutional neural networks (CNNs) have been widely used in analyzing various types of medical data, demonstrating outstanding performance in processing large-scale datasets, feature selection, and pattern recognition.

[0051] Machine learning, particularly ensemble learning and deep learning methods, offers significant advantages for fNIRS data. Ensemble learning algorithms, such as Gradient Boosting, effectively improve model accuracy and stability through multiple rounds of learning and error correction. Deep learning models can automatically learn deeper feature representations in high-dimensional, nonlinear data, further enhancing the model's generalization and reasoning capabilities.

[0052] Based on this, the present invention proposes a novel method for identifying MCI. This method obtains fNIRS results during resting-state and n-back working memory tasks, calculates multidimensional fNIRS-related features, and uses a GradientBoosting algorithm for feature selection. Finally, it integrates a deep learning model for accurate MCI identification. This method not only effectively extracts high-dimensional features but also enables automated MCI diagnosis through model training and inference, demonstrating high accuracy and strong potential for widespread application.

[0053] The following will clearly and completely describe the technical solutions of various embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0054] Existing MCI identification methods mainly rely on clinical assessment and neuroimaging examinations. Although these methods can provide help to some extent, they still face many problems in practical applications: (1) Traditional cognitive assessment tools are highly subjective and easily affected by factors such as the patient's psychological state and educational background; (2) Although imaging examinations such as MRI and PET can provide information on brain structural changes, they have low sensitivity for early identification of MCI, and the equipment is expensive and complex to operate, making them unable to be widely used in large-scale screening; (3) Existing machine learning methods face difficulties in selecting high-dimensional features when processing fNIRS data, and lack effective feature extraction and model training methods, resulting in low diagnostic accuracy.

[0055] To solve the above technical problems, the present invention provides a method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion. Figure 1 As shown, first, fNIRS data of the subjects based on the resting-state task and the n-back working memory task are obtained; then, fNIRS-related features are extracted based on the acquired fNIRS data; then, based on the feature selection algorithm, feature selection is performed on the extracted fNIRS-related features to form a feature set; then, the formed feature set is input into the trained deep learning model to output the mild cognitive impairment recognition result.

[0056] Based on the aforementioned technical solution, this embodiment of the present invention combines two task modes (resting state and n-back working memory tasks) to assess cognitive function from the two dimensions of resting brain networks and cognitive task activation, and extracts feature values. In addition, feature selection algorithms (such as gradient boosting) are used for feature selection to avoid feature redundancy.

[0057] The fNIRS-based MCI identification method provided by the present invention enables early identification of MCI, overcoming the accuracy and universality issues of existing methods. Furthermore, the present invention uses a deep learning model to reason about new samples, enabling automated and rapid MCI diagnosis.

[0058] The method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion provided by the embodiment of the present invention mainly includes the following aspects:

[0059] 1. Acquisition of fNIRS Data:

[0060] The present invention obtains fNIRS data from subjects using a resting-state task and an n-back working memory task. In the resting-state task, subjects are asked to remain calm, with their eyes closed, but not asleep, for several minutes to assess the strength of resting network connectivity between brain regions. In the n-back working memory task, subjects are asked to memorize a series of numbers and determine whether the current number matches the previous one, which is used to assess working memory ability.

[0061] After acquiring fNIRS data, the raw fNIRS signals are also preprocessed as follows:

[0062] (1) Convert the raw light intensity to optical density (OD);

[0063] (2) Motion artifact detection was performed using predefined parameters (tMotion = 0.5 s, STDEVthresh = 20, tMask = 3 s, AMPthresh = 5), followed by spline interpolation correction (p = 0.99);

[0064] (3) bandpass filtering between 0.01 and 0.1 Hz to reduce cardiac and respiratory artifacts;

[0065] (4) Convert the optical density (OD) to relative concentration.

[0066] 2. Calculate fNIRS-related features:

[0067] The acquired fNIRS data includes the time series information of cerebral blood oxygenation changes, which usually has high-dimensional features. In order to effectively utilize this data, the present invention extracts multi-dimensional fNIRS features through signal processing methods. Figure 2 As shown, the integral, mean, and general linear model beta values ​​of the oxygenated hemoglobin concentration during the task period for all channels and brain regions under the n-back working memory task; the wavelet coherence and Pearson correlation values ​​of the changes in oxygenated hemoglobin concentration for all channels and brain regions under the resting-state task mode.

[0068] Specifically, the features of the n-back working memory task include: Beta (beta value), Mean (mean), Integration (integration), K-slope (K-activation), and Peak (peak value).

[0069] The features under the resting-state task include: FC (functional connectivity), aBc (betweenness centrality), aDc (degree centrality), aNCp (node ​​clustering coefficient), aNe (node ​​efficiency), aNLe (node ​​local efficiency), aNLp (node ​​shortest path length), aEg (global efficiency), aEloc (local efficiency), aCp (clustering coefficient), and aSigma (small-world property).

[0070] 3. Select feature set based on Gradient Boosting Classifier:

[0071] After extracting features from fNIRS data, the present invention uses the Gradient Boosting algorithm for feature selection. Gradient Boosting is an ensemble learning algorithm that, through multiple rounds of iterative training, can effectively reduce redundancy and noise in high-dimensional features, improving model accuracy. This algorithm automatically selects the most representative features and performs a weighted combination to generate a feature set suitable for MCI identification.

[0072] 4. Build a deep learning model:

[0073] After feature selection, the selected features are further processed using a deep learning model. Before inputting them into the deep learning model, the features extracted by the two paradigms are preprocessed. This involves normalizing the features in the selected feature set and labeling them as "MCI" to form a training dataset.

[0074] The present invention uses a fully connected neural network (DNN) to build a deep learning model. The DNN model is used to capture the complex nonlinear relationship between features and improve the accuracy of MCI recognition. After training, the model can automatically learn the brain activation patterns under different tasks and accurately classify MCI based on the performance characteristics. The training of the model is as follows Figure 3 shown.

[0075] 5. Reasoning about new samples based on deep learning models:

[0076] Finally, the trained deep learning model can be applied to new samples for inference, automatically identifying MCI. The new sample's fNIRS data is compared with the model's training results, and the model generates a diagnosis. This step requires no human intervention, enabling rapid screening and accurate diagnosis of new patients.

[0077] The implementation basis of each embodiment of the present invention is achieved through programmed processing by a device with processor functionality. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present invention are encapsulated into various modules. Based on this reality, and in addition to the above-mentioned embodiments, an embodiment of the present invention provides a system for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion. This system is used to implement a method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion described in the above-mentioned method embodiment.

[0078] See also Figure 4The system includes: a data input module for acquiring fNIRS data of subjects obtained based on a resting-state task and an n-back working memory task; a feature extraction module for extracting fNIRS-related features based on the acquired fNIRS data; a feature selection module for performing feature selection on the extracted fNIRS-related features based on a feature selection algorithm to form a feature set; and a feature recognition module for inputting the formed feature set into a trained deep learning model to output a mild cognitive impairment recognition result.

[0079] The embodiment of the present invention provides a mild cognitive impairment identification system based on fNIRS data under multi-task paradigm fusion, which addresses the shortcomings of existing MCI diagnosis methods. Figure 4 Several modules in the system acquire fNIRS data during resting-state and n-back working memory tasks, calculate multi-dimensional fNIRS-related features, perform feature selection using the GradientBoosting algorithm, and ultimately combine it with a deep learning model to accurately identify MCI. This method not only effectively extracts high-dimensional features but also enables automated MCI diagnosis through model training and inference, demonstrating high accuracy and strong potential for widespread application.

[0080] It should be noted that the system embodiments provided by the present invention are not only used to implement the methods in the above-mentioned method embodiments, but also used to implement the methods in other method embodiments provided by the present invention. The only difference is the setting of corresponding functional modules. The principles thereof are basically the same as those of the above-mentioned system embodiments provided by the present invention. As long as those skilled in the art refer to the specific technical solutions in other method embodiments on the basis of the above-mentioned system embodiments, obtain corresponding technical means and technical solutions composed of these technical means by combining technical features, and on the premise of ensuring the practicality of the technical solutions, improve the modules in the above-mentioned system embodiments to obtain corresponding system-type embodiments, which are used to implement the methods in other method-type embodiments. For example:

[0081] Based on the content of the above system embodiment, as a preferred embodiment, the embodiment of the present invention provides a mild cognitive impairment identification system based on fNIRS data under a multi-task paradigm fusion, wherein the feature extraction module is further configured to execute the following instructions:

[0082] The fNIRS feature values ​​under the resting-state task and n-back working memory task were extracted respectively to form multi-dimensional fNIRS features.

[0083] Based on the contents of the above system embodiments, as a preferred embodiment, an embodiment of the present invention provides a mild cognitive impairment identification system based on fNIRS data under a multi-task paradigm fusion, wherein the fNIRS feature values ​​under the resting-state task extracted by the feature extraction module include: functional connectivity, betweenness centrality, degree centrality, node clustering coefficient, node efficiency, node local efficiency, node shortest path length, global efficiency, local efficiency, clustering coefficient, and small-world property.

[0084] Based on the contents of the above system embodiment, as a preferred embodiment, the embodiments of the present invention provide a mild cognitive impairment identification system based on fNIRS data under a multi-task paradigm fusion. The fNIRS feature values ​​extracted by the feature extraction module under the n-back working memory task include: beta value, mean, integral, K-activation, and peak value.

[0085] Based on the content of the above system embodiment, as a preferred embodiment, the embodiment of the present invention provides a mild cognitive impairment identification system based on fNIRS data under a multi-task paradigm fusion, wherein the feature selection module is further used to execute the following instructions:

[0086] The gradient boosting algorithm was used to select the extracted fNIRS-related features.

[0087] Based on the content of the above system embodiment, as a preferred embodiment, the embodiment of the present invention provides a mild cognitive impairment identification system based on fNIRS data under a multi-task paradigm fusion, wherein the feature recognition module is further configured to execute the following instructions:

[0088] Construct fNIRS data samples under resting-state task and n-back working memory task;

[0089] Extract fNIRS-related features of fNIRS data samples;

[0090] Perform feature selection on the extracted fNIRS-related features to form a feature set;

[0091] Label the features in the feature set to form a training data set;

[0092] The formed training data set is used to train the fully connected neural network and output the trained deep learning model.

[0093] Based on the content of the above system embodiment, as a preferred embodiment, the embodiment of the present invention provides a mild cognitive impairment identification system based on fNIRS data under multi-task paradigm fusion, further comprising:

[0094] The preprocessing module is used to preprocess the acquired fNIRS data, including conversion of raw light intensity to optical density, artifact detection, filtering, and conversion of optical density to relative concentration.

[0095] Based on the same inventive concept as the above embodiment, an embodiment of the present invention further provides a device for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion, including a processor and a memory; the memory stores at least one instruction, and the at least one instruction is used to be executed by the processor to implement the steps of the method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion.

[0096] In an embodiment of the present invention, the memory may be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), or a volatile memory (volatile memory), such as a random-access memory (RAM). The memory is any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory in an embodiment of the present invention may also be a circuit or any other device that can implement a storage function, for storing program instructions and / or data.

[0097] In the embodiments of the present invention, the processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, and may implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention may be directly implemented and executed by a hardware processor, or by a combination of hardware and software modules within the processor.

[0098] Based on the same inventive concept as the above embodiment, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of implementing the method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion are as follows:

[0099] Obtain fNIRS data from the subjects based on the resting-state task and the n-back working memory task;

[0100] Extract fNIRS-related features based on the acquired fNIRS data;

[0101] Based on the feature selection algorithm, the extracted fNIRS-related features are selected to form a feature set;

[0102] The formed feature set is input into the trained deep learning model to output the mild cognitive impairment recognition results.

[0103] In summary, the present invention obtains fNIRS results during resting-state and n-back working memory tasks, calculates multidimensional fNIRS-related features, performs feature selection based on a gradient boosting algorithm, and ultimately integrates a deep learning model for accurate identification of MCI. This method not only effectively extracts high-dimensional features but also enables automated MCI diagnosis through model training and inference, demonstrating high accuracy and strong potential for widespread application.

[0104] 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 replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion, characterized by: include: Obtain fNIRS data from the subjects based on the resting-state task and the n-back working memory task; Extraction of fNIRS-related features based on the acquired fNIRS data, including: extracting fNIRS feature values ​​for the resting-state task and the n-back working memory task to form multidimensional fNIRS features; extracted fNIRS feature values ​​for the resting-state task, including functional connectivity, betweenness centrality, degree centrality, node clustering coefficient, node efficiency, node local efficiency, node shortest path length, global efficiency, local efficiency, clustering coefficient, and small-world property; extracted fNIRS feature values ​​for the n-back working memory task, including beta value, mean, integral, K-activation, and peak value; Based on the feature selection algorithm, the extracted fNIRS-related features are selected to form a feature set; The formed feature set is input into the trained deep learning model to output the mild cognitive impairment recognition results.

2. The method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion according to claim 1, characterized in that: Feature selection is performed on the extracted fNIRS-related features, including: using a gradient boosting algorithm to perform feature selection on the extracted fNIRS-related features.

3. The method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion according to claim 1, characterized in that: The training of the deep learning model includes: Construct fNIRS data samples under resting-state task and n-back working memory task; Extract fNIRS-related features of fNIRS data samples; Perform feature selection on the extracted fNIRS-related features to form a feature set; Label the features in the feature set to form a training data set; The formed training data set is used to train the fully connected neural network and output the trained deep learning model.

4. The method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion according to claim 1, characterized in that: After acquiring the fNIRS data of the subjects based on the resting-state task and the n-back working memory task, it also includes: The acquired fNIRS data were preprocessed, including conversion of raw light intensity to optical density, artifact detection, filtering, and conversion of optical density to relative concentration.

5. A mild cognitive impairment identification system based on fNIRS data under a multi-task paradigm fusion, characterized by: include: Data input module, used to obtain fNIRS data of subjects based on resting-state task and n-back working memory task; A feature extraction module is used to extract fNIRS-related features based on the acquired fNIRS data, including extracting fNIRS feature values ​​under the resting-state task and the n-back working memory task to form multidimensional fNIRS features; the extracted fNIRS feature values ​​under the resting-state task include functional connectivity, betweenness centrality, degree centrality, node clustering coefficient, node efficiency, node local efficiency, node shortest path length, global efficiency, local efficiency, clustering coefficient, and small-world property; the extracted fNIRS feature values ​​under the n-back working memory task include beta value, mean, integral, K-activation, and peak value; A feature selection module is used to select the extracted fNIRS-related features based on a feature selection algorithm to form a feature set; The feature recognition module is used to input the formed feature set into the trained deep learning model and output the mild cognitive impairment recognition results.

6. A device for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion, characterized in that: The method comprises a processor and a memory; the memory stores at least one instruction, and the at least one instruction is used to be executed by the processor to implement the steps of the method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for identifying mild cognitive impairment based on fNIRS data under a multi-task paradigm fusion according to any one of claims 1 to 4 are implemented.

Images