CN115349857B

CN115349857B - Dynamic rehabilitation evaluation method and system based on fNIRS brain functional map

Info

Publication number: CN115349857B
Application number: CN202210843314.2A
Authority: CN
Inventors: 李增勇; 霍聪聪; 徐功铖; 谢晖; 张静莎; 张腾宇; 吕泽平
Original assignee: National Research Center for Rehabilitation Technical Aids
Current assignee: National Research Center for Rehabilitation Technical Aids
Priority date: 2022-07-18
Filing date: 2022-07-18
Publication date: 2023-06-20
Anticipated expiration: 2042-07-18
Also published as: CN115349857A

Abstract

The invention provides a dynamic rehabilitation evaluation method and a system based on an fNIRS brain functional map, wherein the method comprises the following steps: selecting a brain function detection paradigm matching the exercise capacity of a user from a plurality of brain function detection paradigms stored in advance based on the exercise capacity of the user; acquiring brain blood oxygen signals which are induced and generated by a user under the matched brain function detection paradigm and acquired by the near infrared brain function acquisition device in real time; based on the acquired cerebral blood oxygen signals, calculating near-infrared multi-mode brain function feature parameters, wherein the multi-mode is an activation mode and a connection mode, and the parameters form a multi-mode brain function feature combination vector; based on the obtained multi-mode brain function feature combination vector, performing brain function feature matching classification by an integrated multi-classification method to obtain a matching classification result; and providing a visual assessment report through a brain function state assessment module based on the acquired matching classification result.

Description

Dynamic rehabilitation evaluation method and system based on fNIRS brain functional map

Technical Field

The invention relates to the technical field of brain function rehabilitation evaluation, in particular to a dynamic rehabilitation evaluation method and system based on an fNIRS brain function map.

Background

Recovery from post-stroke motor dysfunction is often due to functional remodeling or reorganization of the cerebral cortex. The real-time monitoring of brain function status is of great importance for stroke rehabilitation. Because the injury type, degree and position of the apoplexy patient have great individual variability, different modes of nerve remodeling and migration phenomena are presented in the rehabilitation process. However, due to the lack of dynamic quantitative evaluation and real-time feedback means, the stroke rehabilitation technology still has the key problems of unclear action mechanism, inaccurate action targets, ambiguous evaluation indexes and the like, and the clinical rehabilitation effect is not uniform. For the root cause of impaired cortex function after cerebral apoplexy, how to individually select a clinical rehabilitation intervention scheme and how to perform layered treatment on patients can promote the patients to obtain greater rehabilitation benefit is still a clinical core problem.

Developments in neuroimaging technology provide new approaches to study and guide neural plasticity. Common brain function detection technologies such as functional nuclear magnetic resonance (fMRI) and electroencephalogram (EEG) are difficult to acquire dynamic brain function information under exercise or rehabilitation intervention due to poor anti-electromagnetic interference and anti-exercise interference capabilities. As an optical non-invasive brain function imaging technology, the functional near infrared spectroscopy (fNIRS) provides noninvasive visual brain nerve activity information by detecting cerebral blood oxygen metabolism, has the advantages of simplicity in operation, strong anti-interference performance, good electromagnetic compatibility and the like, can realize rapid real-time detection of brain functions of stroke patients in multiple scenes, and has important clinical application value in the field of nerve rehabilitation. However, the current fNIRS imaging system still has the following key problems in stroke rehabilitation: the condition of brain function remodeling is complex by limiting brain function auxiliary evaluation due to the lack of standard detection paradigms and imaging marks, and the characteristic quantity obtained by a single brain function analysis method cannot fully represent the complex brain function network characteristics.

Therefore, there is a need for a dynamic assessment system that can more accurately detect patient brain function rehabilitation.

Disclosure of Invention

In view of the foregoing, embodiments of the present invention provide a method and system for dynamic rehabilitation assessment based on fnigs brain functional maps to obviate or ameliorate one or more of the disadvantages of the prior art.

One aspect of the present invention provides a dynamic rehabilitation assessment method based on fnrs brain function maps, the method comprising the steps of:

selecting a brain function detection paradigm matching the exercise capacity of a user from a plurality of brain function detection paradigms stored in advance based on the exercise capacity of the user;

acquiring brain blood oxygen signals which are induced and generated by a user under the matched brain function detection paradigm and acquired by the near infrared brain function acquisition device in real time;

based on the acquired cerebral blood oxygen signals, calculating near-infrared multi-mode brain function feature parameters, wherein the multi-mode is an activation mode and a connection mode, and the parameters form a multi-mode brain function feature combination vector;

based on the obtained multi-mode brain function feature combination vector, performing brain function feature matching classification by an integrated multi-classification method to obtain a matching classification result, wherein the matching classification result is good in rehabilitation effect of the exercise function, general in effect and poor in effect;

And providing a visual assessment report through a brain function state assessment module based on the acquired matching classification result.

In some embodiments of the invention, the pre-stored plurality of brain function detection paradigms includes a resting state and a typical fnrs brain function detection paradigm of motor-oriented execution neural loop remodeling and sensory-motor neural loop remodeling; the detection paradigm for remodelling the motor-oriented execution nerve loop is specifically based on repeated extension of a target-oriented hemiplegic upper limb; the sensory-motor nerve loop remodeling-oriented detection paradigm is specifically a central-peripheral combined magneto-electric stimulation technology, and activates a motor-related brain region and induces proprioception to pass into the central cortex.

In some embodiments of the invention, the method comprises combining network information of an activation mode and a connection mode by adopting a brain function multimode characteristic fusion method, wherein the activation mode is used for describing activation distribution conditions of various areas of the brain, and the connection mode is used for representing information interaction process under typical tasks of the brain.

In some embodiments of the invention, the method further comprises time-frequency converting the cerebral blood oxygen signal using a wavelet transform time-frequency analysis method based on the ability of the wavelet transform time-frequency analysis method to decouple signal components and provide local phase information.

In some embodiments of the invention, the near infrared multi-mode brain function feature parameters include parameters such as hemisphere activation sideslip, hemisphere connection sideslip, hemisphere autonomous coefficient, undirected weighted topology parameters, directed weighted topology parameters, and the like.

In some embodiments of the invention, the method further comprises classifying movement dysfunction including mild, moderate and severe dysfunction using cluster analysis; clustering the movement function evaluation scale set into mild, moderate and severe movement dysfunction subsets by using a K-means clustering algorithm; the athletic performance assessment scale set includes an athletic performance assessment (Fugl-Meyer, FM) scale, a grip index, and an upper limb movement study ARAT scale.

In some embodiments of the invention, the method further comprises selecting brain function response features based on a combination of group level statistical tests and function-related regression fusion analysis;

the group-level statistical test is used as a feature screening and filtering method to extract brain function feature indexes with statistical significance;

the function correlation regression fusion analysis is to fuse clinical function evaluation information on the basis of preliminary screening of features by group-level statistical test, establish a segmented regression model, select features with statistical significance, identify an optimal feature subset related to dysfunction in a specific state, and select multi-mode fusion features.

In some embodiments of the invention, the method further comprises employing an integrated multi-classification Support Vector Machine (SVM) model to address the imbalance problem for each set of data based on the imbalance of the clinical data; the integrated multi-classification support vector machine SVM model adopts radial basis function RBF to carry out high-dimensional mapping on near-infrared brain functional characteristics, and adopts a grid optimizing method to search the best punishment factors and nuclear radius parameters.

In another aspect, the present invention provides a dynamic rehabilitation assessment system based on fnrs brain function atlas, the system comprising: the motion detection paradigm selection module is used for selecting a brain function detection paradigm matched with the motion capability of a user from a plurality of brain function detection paradigms stored in advance according to the motion capability of the user;

the near infrared brain function acquisition module is used for acquiring multichannel brain blood oxygen signals induced by a user under the matched brain function detection range in real time;

the brain function response characteristic analysis module is used for preprocessing and calculating near-infrared multi-mode brain function characteristic parameters through the fNIRS data acquisition and multi-mode characteristic calculation module according to the acquired brain blood oxygen signals to obtain brain function multi-mode response characteristic vectors;

The fNIRS brain function map module is used for constructing a mapping model of fNIRS brain function characteristics and exercise rehabilitation according to an integrated multi-classification method, and inputting the obtained brain function multi-mode response characteristic vector into the module for matching classification;

the brain function state evaluation report module is used for outputting the matching result obtained by the fNIRS brain function map module and providing a visual evaluation report.

In some embodiments of the invention, the fnrs brain function map module comprises: the system comprises a motion function evaluation module, an fNIRS data acquisition and multi-mode feature calculation module, a multi-task-multi-mode fusion feature selection module and an fNIRS brain function and motion rehabilitation mapping module;

the exercise function evaluation module is used for storing the limb exercise function evaluation results of users with different time nodes, including comprehensive exercise function disorder evaluation and comprehensive exercise function rehabilitation evaluation results;

the fNIRS data acquisition and multi-mode feature calculation module is used for acquiring fNIRS data under a specific detection paradigm by utilizing the near infrared brain function imaging device and carrying out preprocessing and multi-mode brain function feature calculation;

the multi-task-multi-mode fusion feature selection module is used for extracting multi-task-multi-mode fNIRS brain function fusion features related to the exercise capacity of a user based on two-step feature dimension reduction formed by group-level statistics and function-related regression fusion analysis;

The fNIRS brain function and motion rehabilitation mapping module is used for screening out fNIRS brain function fusion characteristics and rehabilitation effect classification labels from the multi-task-multi-mode fusion characteristic selection module, constructing an integrated multi-classification Support Vector Machine (SVM) model, and selecting parameters with highest classification accuracy as the fNIRS brain function and motion rehabilitation mapping classification model through training and parameter optimization to construct an fNIRS brain function map.

In some embodiments of the invention, the system further comprises a processor and a memory, the memory having stored therein computer instructions for executing the computer instructions stored in the memory, the system implementing the steps of the method as described above when the computer instructions are executed by the processor.

In another aspect of the invention, there is also provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method as described above.

According to the dynamic rehabilitation evaluation method and system based on the fNIRS brain functional map, the neural functional states of the people with dyskinesia after brain injury are dynamically evaluated by combining a near infrared brain functional imaging technology with a specific movement detection paradigm, and an important imaging basis is provided for rehabilitation scheme formulation and optimization; the multi-mode feature fusion method can be adopted to combine the activation and connection network information, and various brain function analysis methods are adopted to describe the brain function map more abundantly; the integrated multi-classification support vector machine model can be used for solving the limitation of clinical data unbalance and solving the mapping prediction model bias phenomenon caused by the data unbalance. The brain function specific response and remodeling characteristics under various typical task ranges can be fused, a dynamic mapping model of near infrared brain function indexes and motor function rehabilitation is constructed, namely, an fNIRS brain function map for motor rehabilitation is realized, and the brain function state evaluation of patients with different rehabilitation levels is realized; the method can utilize the technical advantages of the near infrared brain function imaging technology in resisting motion and electromagnetic interference, design a typical monitoring paradigm facing a motor nerve loop, fully acquire brain function characteristics related to the motor function, and enrich brain stroke nerve function assessment except for a resting state; the method can extract the fNIRS multi-mode brain function response characteristics under the resting state and typical detection task paradigm, establish a mapping model of the fNIRS multi-mode brain function response characteristics and the motor function rehabilitation, and establish an fNIRS brain function map oriented to motor dysfunction so as to realize dynamic evaluation of brain function states of brain injury patients.

Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description and drawings.

It will be appreciated by those skilled in the art that the objects and advantages that can be achieved with the present invention are not limited to the above-described specific ones, and that the above and other objects that can be achieved with the present invention will be more clearly understood from the following detailed description.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate and together with the description serve to explain the invention. In the drawings:

fig. 1 is a schematic flow chart of a dynamic rehabilitation evaluation method based on an fnrs brain functional map according to an embodiment of the invention.

FIG. 2 is a block diagram of a dynamic rehabilitation evaluation system based on an fNIRS brain functional map according to an embodiment of the present invention.

FIG. 3 is a block diagram of a brain function response feature analysis module according to an embodiment of the invention.

FIG. 4 is a block diagram of an fNIRS brain function map module according to an embodiment of the invention.

FIG. 5 is a block diagram of a multi-mode fusion feature selection module according to one embodiment of the invention.

FIG. 6 is a diagram of a model framework for integrating multiple classification support vector machines in accordance with an embodiment of the present invention.

FIG. 7 is a flow chart illustrating the operation of a dynamic rehabilitation evaluation system based on the fNIRS brain function map according to an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the following embodiments and the accompanying drawings, in order to make the objects, technical solutions and advantages of the present invention more apparent. The exemplary embodiments of the present invention and the descriptions thereof are used herein to explain the present invention, but are not intended to limit the invention.

It should be noted here that, in order to avoid obscuring the present invention due to unnecessary details, only structures and/or processing steps closely related to the solution according to the present invention are shown in the drawings, while other details not greatly related to the present invention are omitted.

It should be emphasized that the term "comprises/comprising" when used herein is taken to specify the presence of stated features, elements, steps or components, but does not preclude the presence or addition of one or more other features, elements, steps or components.

It is also noted herein that the term "coupled" may refer to not only a direct connection, but also an indirect connection in which an intermediate is present, unless otherwise specified.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar components, or the same or similar steps.

In order to solve the problems that the prior art lacks standard detection paradigm and imaging marks for cerebral apoplexy rehabilitation, the cerebral function remodeling condition is complex, the cerebral function analysis method is single and the like, the invention provides a dynamic rehabilitation evaluation method and system based on an fNIRS cerebral function map, which dynamically evaluates the nerve function state of a crowd with dyskinesia after cerebral injury by combining a near infrared cerebral function imaging technology with a specific motion detection paradigm, and provides an important imaging basis for the establishment and optimization of a rehabilitation scheme; the multi-mode feature fusion method can be adopted to combine the activation and connection network information, and various brain function analysis methods are adopted to describe the brain function map more abundantly; the integrated multi-classification support vector machine model can be used for solving the limitation of clinical data unbalance and solving the mapping prediction model bias phenomenon caused by the data unbalance. The brain function specific response and remodeling characteristics under various typical task ranges can be fused, and a dynamic mapping model for near infrared brain function indexes and motor function rehabilitation is constructed; the method can utilize the technical advantages of the near infrared brain function imaging technology in resisting motion and electromagnetic interference, design a typical monitoring paradigm facing a motor nerve loop, fully acquire brain function characteristics related to the motor function, and enrich brain stroke nerve function assessment except for a resting state; the method can extract the fNIRS multi-mode brain function response characteristics under the resting state and typical detection task paradigm, establish a mapping model of the fNIRS multi-mode brain function response characteristics and the motor function rehabilitation, and establish an fNIRS brain function map oriented to motor dysfunction so as to realize dynamic evaluation of brain function states of brain injury patients. The brain function rehabilitation dynamic assessment is more real-time, comprehensive, continuous, objective, accurate and targeted, so that the brain function state assessment of a cerebral apoplexy patient can be carried out, the clinical function assessment and the treatment response detection are assisted, and a treatment target is provided for personalized intervention.

Fig. 1 is a schematic flow chart of a dynamic rehabilitation evaluation method based on fnrs brain functional atlas according to an embodiment of the present invention, and as shown in fig. 1, the method of the embodiment includes the following steps:

step S110, selecting a brain function detection paradigm matched with the exercise capacity of a user from a plurality of brain function detection paradigms stored in advance based on the exercise capacity of the user.

In this step, the pre-stored plurality of brain function detection paradigms includes a resting state and a typical fnrs brain function detection paradigm of motor-directed nerve loop remodeling and sensory-motor nerve loop remodeling; the detection paradigm for remodelling the motor-oriented execution nerve loop is specifically based on repeated extension of a target-oriented hemiplegic upper limb; the sensory-motor nerve loop remodeling-oriented detection paradigm is specifically a central-peripheral combined magneto-electric stimulation technology, and activates a motor-related brain region and induces proprioception to pass into the central cortex.

In this embodiment, for example, a typical fnrs brain function detection paradigm for motor-oriented loop remodeling and sensory-motor neural loop remodeling is designed, specifically, a task paradigm for inducing stroke users to exercise and perform cortical reorganization based on a target-oriented hemiplegic upper limb repetitive stretching motor paradigm; based on the central combined peripheral magneto-electric stimulation technology, the motor-related brain region is activated and the proprioception is induced to be transmitted into the central cortex task mode. The exercise executing task paradigm requires a certain exercise capacity of participants, and the participants can participate in the paradigm under the condition of assisting or initiative; the sensory-motor task paradigm is completed by combining central magnetic stimulation with peripheral electrical stimulation, and does not need the participators to perform specific activities; therefore, aiming at the users without any exercise ability, the magnetoelectric tabu possibility is eliminated, and the resting state and the sensory-exercise task paradigm are selected preferentially; aiming at a user with exercise capability, eliminating the possibility of magnetoelectric tabu, and preferentially selecting a resting state and an exercise execution task paradigm; if the magnetoelectric tabu is eliminated, the resting state, the exercise execution task model and the sensory-exercise task model are preferably selected.

Step S120, acquiring brain blood oxygen signals induced by a user under the matched brain function detection paradigm by the near infrared brain function acquisition device in real time.

In this step, based on a brain function detection paradigm matched with the exercise capacity of the user, a near-red brain function acquisition device is adopted to acquire brain blood oxygen signals induced by the user under the matched brain function detection paradigm in real time, wherein the brain blood oxygen signals comprise an oxyhemoglobin concentration delta [ oxy-Hb ] signal, a deoxyhemoglobin concentration delta [ oxygen-Hb ] signal and a total oxyhemoglobin concentration delta [ total-Hb ] signal.

In an embodiment of the present invention, for example, a multichannel near infrared brain function imaging system is used to acquire cerebral blood oxygen signals of a stroke patient, and the near infrared brain function imaging system includes 17 transmitters and 16 receivers, forming 52 fnrs acquisition channels. According to international 10-20 system arrangement, respectively covering left and right hemispherical forehead leaves, motor areas, sensory and motor and occipital cortex; the distance between the light source and the detector is 3cm, and the sampling frequency is 10Hz. Each user completes fNIRS data acquisition according to the matched brain function detection paradigm, the focused attention frequency band based on the near infrared brain blood oxygen signals is 0.009-0.01 Hz, the lowest frequency is 0.009Hz, namely the period of the lowest frequency is 112s, the lowest measurement duration is 560s, the fluctuation 40s when the state is not entered before the data acquisition is started is considered, namely the total acquisition duration is ensured to be 600s in order to ensure the effectiveness of near infrared brain function indexes in at least 5 low frequency cycles in one measurement, and therefore, the acquisition time of each group of tasks is not less than 10 minutes.

Step S130, calculating near-infrared multi-mode brain function feature parameters based on the acquired brain blood oxygen signals, wherein the multi-mode is an activation mode and a connection mode, and the parameters form a multi-mode brain function feature combination vector.

In this step, based on the acquired brain blood oxygen signals, the brain blood oxygen signals are input into a brain function response feature analysis module, and the acquired data are preprocessed and near infrared multi-mode brain function parameters are calculated by an fnigs data acquisition and multi-mode feature calculation module in the brain function response feature analysis module, as shown in fig. 3.

In the embodiment of the invention, the activation distribution situation of each area of the brain can be described based on the activation mode, and the connection mode among different brain areas can represent the information interaction process under the typical task of the brain, so that the embodiment combines the activation and connection network information by adopting the brain function multi-mode feature fusion method, and comprehensively constructs the brain function map. Establishing indexes such as brain activation, brain function connection, brain effect connection and the like based on a wavelet transformation time-frequency analysis method, and constructing a near-infrared brain function analysis model; on the basis, indexes such as hemisphere activation sideslip, hemisphere connection sideslip, hemisphere autonomous coefficient, undirected weighted topological parameter, directed weighted topological parameter and the like are further calculated, and near infrared brain function network characteristics are comprehensively analyzed. The signal needs to be preprocessed before the near infrared multi-mode brain function parameter is calculated, and the specific method is as follows:

(1) Preliminary visual inspection and signal-to-noise ratio inspection: and checking the signal-to-noise ratio of the signal through light intensity data acquired by the near infrared light with two wavelengths, and removing a noise channel. To estimate the signal-to-noise ratio of the data channel, the signal-to-noise ratio of the channel data is estimated by calculating the relative coefficients of variation (CV,%) for the 760nm and 850nm raw signals, the relative coefficients of variation being expressed as follows:

CV＝σ/μ×100％

where μ is the signal data mean and σ is the standard deviation of the signal. And calculating the CV value of each channel in the whole task paradigm acquisition process, and eliminating the measuring channels with CV values of any wavelength signals larger than 15%.

(2) Band-pass filtering: the original light intensity signal is subjected to band-pass filtering (0.005 Hz-2 Hz) by using a zero-phase sixth-order Butterworth filter so as to remove high-frequency and ultra-low-frequency interference components.

(3) Concentration conversion: according to the modified Beer-Lambert law, i.e., the basic law of light absorption, the detected light intensity changes to brain blood oxygen concentration signals, including an oxyhemoglobin concentration delta [ oxy-Hb ] signal, a deoxyhemoglobin concentration delta [ oxy-Hb ] signal, and a total oxyhemoglobin concentration delta [ total-Hb ] signal.

(4) Motion artifact removal: and eliminating obvious abnormal points in the signals by adopting a moving average method, and then removing motion artifacts by adopting a method based on a moving standard deviation and cubic spline interpolation.

(5) Removing physiological interference: scalp signals acquired by the short-distance data acquisition channels are utilized, regression analysis is carried out on blood oxygen signals of each channel to remove scalp interference noise, and principal component analysis (independent component analysis, ICA) is combined to reduce physiological interference in fNIRS measurement, including heartbeat, respiratory signals, blood pressure signals and the like. All ICA decomposed signal components were passed through a visual inspection one by one to determine components that may be correlated with noise and artifacts. The component of interest is visually identified based on the criterion that the brain function signal should have a significant low frequency (in the range of 0.01-0.08 Hz) oscillation spectrum, indicating that the acquired signal contains brain functional hemodynamic response information.

(6) Wavelet transform time-frequency analysis: the time-frequency analysis method based on wavelet transformation has the capability of decoupling signal components and providing local phase information, is suitable for analysis of biological oscillation signals, and in the embodiment of the invention, for example, complex Morlet wavelet is selected as a wavelet basis function to perform time-frequency conversion on cerebral blood oxygen signals, and the time point t is at a specific frequency f _n The complex wavelet coefficients obtained by wavelet transformation are defined as:

the resulting wavelet coefficients appear as complex numbers in the time-frequency plane, and the absolute amplitude W of each frequency and time can be defined _i (f,t _n ) And instantaneous phase information

Wherein i is represented as a near red brain blood oxygen signal of a certain channel, a _i (f,t _n ) Is the real part of complex number, b _i (f,t _n ) Is the imaginary part of the complex number; amplitude and phase information may be used to represent the brain activation and the relationship between the different oscillation signals, respectively. On the basis, indexes such as hemisphere activation sideslip, hemisphere connection sideslip, hemisphere autonomous coefficient, undirected weighted topological parameter, directed weighted topological parameter and the like are further calculated, near infrared brain function network characteristics are comprehensively analyzed, and brain function multimode response characteristic vectors are obtained, wherein the specific analysis method is as follows:

1) Brain activation response and hemispheric lateral deviation

Based on wavelet transformation, the amplitude information of the cerebral blood oxygen signal at a certain frequency and time point is defined as W _i (f,t _n )：

Wavelet Amplitude (WA) is defined as the average result of a time-frequency signal in the time domain, and can be expressed as a power index reflecting the oscillation amplitudes of different frequency components of the original signal and can be used for representing the activity intensity of the cortex; the WA index of each fnrs channel is calculated over the frequency band of interest to represent the hemodynamic oscillation amplitude of the covered brain region.

In addition, in order to observe the balance among hemispheres of the activation condition of a stroke user, the activation hemisphere lateral deviation index is further calculated based on WA results LI _WA ):

Wherein WA _ipsi WA values for the affected side hemisphere fNIRS channel; WA (Wireless LAN area) _contra WA index for the tonic hemisphere fNIRS channel; LI (LI) _WA The magnitude of the value is between-1 and-1, -1 indicates full healthy side hemisphere activation, and 1 indicates full diseased side hemisphere activation.

2) Brain function connection analysis

The dynamic phase information can be used to study the relationship between the different oscillations. In an embodiment of the invention, a wavelet phase coherence index (WPCO) is used to describe the brain Function Connection (FC), which quantifies the tendency of the phase difference between two signals to remain constant over a particular frequency. WPCO can be defined as:

wherein the method comprises the steps of

Representing the instantaneous phase difference of the two fnrs channel signals. In the embodiment of the invention, whether the calculated WPCO value is effective or not is checked by calculating the WPCO value between every two fNIRS channels and adopting an amplitude self-adaptive Fourier transform substitution signal, and after the substitution signal is checked, a functional connection matrix corresponding to the number of the fNIRS channels is generated.

In addition, to observe the inter-hemisphere balance of the stroke user's functional connected network, the connected hemisphere bias index (LI _WPCO )：

N _ipsi-i Representing the number of functional connections to the affected hemisphere channel i; t (T) _ipsi-i Representing all of the lateral hemispheric passages i The number of possible functional connections; n (N) _contra-j Representing the number of functional connections to the robust side hemisphere channel j; t (T) _contral-j The method is characterized in that the method comprises the steps of representing all possible functional connection quantity of a healthy side hemisphere channel j, wherein the channel i and the channel j are distributed in hemispherical symmetry; LI (LI) _WPCO The magnitude of the value is between-1 and 1, -1 represents a complete healthy side hemisphere connection, and 1 represents a complete diseased side hemisphere connection.

In addition, to observe the intra-hemisphere and inter-hemisphere connection differences of the stroke user function connection network, the connection hemisphere autonomy coefficient (AI) is further calculated based on WPCO results _WPCO )：

N _i The number of channels with functional connection between the hemispheres at the same side and the channel i is represented; n (N) _c The number of channels in the contralateral hemisphere that are functionally connected to channel i; t (T) _i And T is _c The total number of channels of the same side hemisphere and the opposite side hemisphere are respectively represented; the larger the AI value is, the higher the connection strength between the channel and the hemispheres is, the AI value of the comparison double-sided hemispheres can be used for reflecting the autonomous condition of the hemispheres of the brain, the imbalance of the functional connection between the hemispheres is reflected on the basis of the comparison of the functional connection between the hemispheres, and the lateral deviation of the cortex is indirectly reflected by comparing the AI sizes of the left hemispheres and the right hemispheres.

3) Brain effect connection analysis

In the embodiment of the invention, for example, a coupling phase oscillation model between every two signals is constructed in the interested frequency band by utilizing the dynamic phase information after wavelet transformation:

w _i (t) is a parameter of natural frequency, ζ _i (t) Gaussian white noise, two oscillator phases φ _i And phi _j Is a function q of (2) _i Representing a coupling equation; describing the model using dynamic Bayesian inference computationAnd thus obtain a parameter describing the coupling relationship between the two oscillating signals, i.e. the Effect Connection (EC). On the basis, an amplitude self-adaptive Fourier transform method is adopted to check whether the calculated effect connection value is valid or not; after the substitution signal is checked, the corresponding effect connection matrix of the fNIRS channel number is generated.

Based on this, the above connection-side bias index LI can be used _EC Hemispherical autonomous coefficient AI _EC Is calculated based on the definition of the two indices of the effect-based connected network.

4) Brain function network index calculation

In an embodiment of the present invention, brain regions of interest are partitioned according to fNIRS channel coverage locations, and based on functional connection and effect connection results, an undirected weighted network and directed weighted network feature parameters (G) are calculated, respectively, the undirected weighted network feature parameters including global characteristic topology parameters (small world network, global efficiency, rich club and hierarchy), the directed weighted network feature parameters including node characteristic topology parameters (cluster coefficients, shortest path, node efficiency, local efficiency of nodes, click centrality and intermediary centrality).

In summary, the fnigs multi-mode brain function feature combination vector V for each state is calculated:

V＝[WA，LI _WA ，FC，LI _FC ，AI _FC ，EC，LI _EC ，AI _EC ，G]

step S140, based on the obtained multi-mode brain function feature combination vector, performing brain function feature matching classification by an integrated multi-classification method to obtain a matching classification result, wherein the matching classification result is good in rehabilitation effect, general in effect and poor in effect of the exercise function.

In the step, the fNIRS multi-mode brain function feature combination vector V input calculated in the step S130 is constructed into an fNIRS brain function map module related to exercise rehabilitation by utilizing an integrated multi-classification method, the matching classification of brain function features is carried out, and the neural function state of a user is judged according to the matching classification result; the fNIRS brain function map module comprises a motion function evaluation module, an fNIRS data acquisition and multi-mode feature calculation module, a multi-task-multi-mode fusion feature selection module and an fNIRS brain function and motion rehabilitation mapping module; as shown in fig. 4, the fnrs brain function atlas module uses the exercise function assessment result to group the user's exercise dysfunctions and calculate a comprehensive exercise assessment score; grouping the exercise function rehabilitation effects of the user and calculating the comprehensive exercise rehabilitation score according to the exercise function evaluation results of the front and back two times in the rehabilitation process; based on a near-infrared brain function imaging device, acquiring brain blood oxygen signals of a stroke user under various typical task norms and performing brain function multi-mode feature analysis, and realizing near-infrared brain function feature and dysfunction index fusion feature selection by using group level statistics and functional regression; the multi-task-multi-mode fNIRS brain function fusion index is input into an integrated multi-classification classifier to realize classification mapping of rehabilitation effects of cerebral apoplexy users, wherein the classification mapping comprises good rehabilitation effects of motor functions, general effects and poor effects.

In an embodiment of the present invention, the exercise function evaluation module is configured to record basic information of a user, where the basic information includes: age, sex, type of injury (ischemic/hemorrhagic), lesion location. Evaluating motor functions of a brain injury patient by using an evaluation scale and an index, wherein the evaluation scale comprises a grip index, a Fugl-Meyer (FM) scale and an ARAT (automatic) scale for researching upper limb actions, the grip index represents the relative grip strength of the patient's hand relative to the healthy hand, and the average grip strength (affected side)/the average grip strength (healthy side) is calculated to be multiplied by 100; the FM scale is used to assess the ability of a patient to move a single joint under synergy, including assessment of motor control and strength; the ARAT gauge is used for evaluating complex movements of upper limbs in daily life, including grasping, holding, pinching and gross movement under flexion and extension of shoulder and elbow. The module stores the evaluation results of the limb movement functions of users at different time nodes, and can calculate the evaluation result of the comprehensive movement dysfunction and the rehabilitation evaluation result of the comprehensive movement function.

Aiming at comprehensive exercise function rehabilitation evaluation, a K-means clustering (K-means) algorithm is adopted, an exercise function evaluation scale set (comprising an FM scale, a grip index and an ARAT scale) is divided into three types, namely, mild, moderate and severe exercise dysfunction subsets, and a principal component analysis (Principal component analysis, PCA) method is further adopted as a precursor process of a subsequent feature fusion module in each dysfunction subclass to construct a comprehensive exercise evaluation index M based on clinical exercise function evaluation results (comprising the FM scale, the grip index and the ARAT scale). For each subclass, PCA using FM scale, ARAT scale and grip index as input variables generates a single factor solution, and the number of main components is determined according to the principle that the cumulative variance contribution rate is not lower than 90%. In the embodiment of the present invention, the factor value of the first principal component, which is analyzed by the FM scale, ARAT, and grip index as the principal components of the input variables, is defined as the integrated motor function score M.

For comprehensive exercise function rehabilitation evaluation, firstly, calculating the relative difference value of each evaluation index (including FM scale, grip index and ARAT scale) at different time points, namely

d(x,y)＝(x-y)/(1+y)

Where d represents the evaluation index difference, x represents the previous time node evaluation score, and y represents the next time node evaluation score. After z-standardization processing is carried out on each assessment index difference value, a K-means clustering (K-means) algorithm is adopted, and a z-standardized motion function assessment difference value set (comprising an FM scale difference value, a grip index difference value and an ARAT scale difference value) is divided into three types, namely, good recovery, general recovery and poor recovery subsets which are used as the preamble process of a subsequent classification mapping task module. In addition, each evaluation index difference value after z-normalization processing is input into PCA to construct a comprehensive exercise rehabilitation index R. And determining the number of the main components according to the principle that the cumulative variance contribution rate is not lower than 90%. The principal component scores are weighted and summed according to the proportion of the variance contribution rate of each principal component to the accumulated variance contribution rate of the extracted principal component:

PCA with FM scale, ARAT scale and grip index as input variables gives a comprehensive motor rehabilitation score index R, where negative values do not represent poor motor ability, but rather relatively less motor improvement than the whole group.

In the embodiment of the invention, with reference to fig. 4, the fnrs data acquisition and multi-mode feature calculation module is configured to acquire fnrs data under a specific detection paradigm by using the near infrared brain function imaging device and perform preprocessing and multi-mode brain function feature calculation. Specific detection paradigms of the fnrs data acquisition and multi-mode feature computation module include a typical fnrs brain function detection paradigm of resting state and motor-directed loop remodeling and sensory-motor neural loop remodeling; the detection paradigm for reconstructing the function of the motor-oriented nerve loop is based on repeated extension of the target-oriented hemiplegic upper limb; the detection paradigm facing the sensory-motor nerve loop function reconstruction is based on a central-peripheral combined magnetoelectric stimulation technology to activate a motor-related brain region and induce a proprioceptive task.

In the embodiment of the invention, the multi-task-multi-mode fusion feature selection module is used for realizing multi-task-multi-mode fNIRS brain function fusion feature extraction related to the exercise function of a user based on two-step feature dimension reduction formed by group level statistics and function related regression. To fully describe the brain function state, the system designs various typical task paradigms and starts from an active mode and a connected mode, calculates fNIRS multi-mode brain function feature combination vectors V under the various typical task paradigms, and the brain function feature index formed by the combination vectors V is large in dimension and number, in which case if all brain function features are used as features in a classifier, excessive fitting problems can be caused, a large amount of redundant information exists, classification performance is not ideal, and a brain function network related to specific dysfunction is usually concentrated on a small part of all possible indexes. In order to reduce redundancy of index features, feature selection is realized by combining group-level statistical inspection and functional correlation regression fusion analysis, and the resolving power of feature subsets is ensured. The specific steps are shown in fig. 5:

Based on the near-infrared multi-mode brain function feature vector, a group-level statistical test is used for preliminarily selecting patient task-specific near-infrared brain function response features with statistical significance, the group-level statistical test is aimed at resting fNIRS brain function response features, and the system utilizes the significance difference between the patient and a healthy control group to perform feature selection; aiming at the fNIRS brain function response characteristics under the exercise execution task and the sensory-exercise task, the system utilizes the significant difference between the task state and the resting state of the patient to perform characteristic selection. The multi-mode near infrared brain functional characteristics under the multi-task are primarily screened through the inter-group significance test. The group-level statistical test in the system does not use the label information of the sample to select the characteristics, so that the bias of the classification result is reduced.

Based on the preliminary screening of the features by group-level statistical test, clinical function evaluation information is fused, a regression model is established, features with statistical significance are selected, an optimal feature subset related to dysfunction in a specific state is identified, and multi-task-multi-mode fusion feature selection is realized. The motion function classification label is provided for the module through a K-means algorithm based on the motion function evaluation module, and the comprehensive motion evaluation score in each motion dysfunction subclass is calculated based on a PCA algorithm. With reference to fig. 5, a multi-segment regression model is respectively established according to task-specific near-infrared brain function response characteristics and motion function comprehensive evaluation scores, comprehensive motion function scores of all motion function disorder subclasses are selected as segment nodes, the comprehensive motion scores are used as dependent variables, the near-infrared brain function characteristics are used as independent variables, the mapping relation analysis between fNIRS brain function indexes and motion function disorders of a patient suffering from stroke in a resting state is realized, and resting-state fNIRS brain function maps of the patient suffering from mild, moderate and severe motion function disorders of the stroke are constructed; analyzing the mapping relation between the fNIRS brain function response index of the stroke patient induced by the exercise execution task and the exercise dysfunction, and constructing fNIRS brain function maps related to the task execution of the stroke patient with mild, moderate and severe exercise dysfunction; and (3) analyzing the mapping relation between the fNIRS brain function response index of the cerebral apoplexy patient induced by the sensory-motor task and the motor dysfunction, and constructing fNIRS brain function maps related to the sensory-motor task of the cerebral apoplexy patient with mild, moderate and severe motor dysfunction. In the regression process, the least square sum of the difference between the given function value and the model predicted value is taken as a loss function, and the least square method and the gradient descent method are used for calculating the final fitting parameters. The function of the module is to carry out regression fusion analysis on specific response indexes of brain functions under specific tasks and movement dysfunction assessment, comprehensively screen specific response characteristics of the multi-task multi-mode near infrared brain functions of the brain stroke patients obviously related to the movement functions, and construct fNIRS brain function maps of the brain stroke patients with different movement dysfunction degrees under the multi-task state. In addition, the module further realizes dimension reduction of feature data, filtering and extraction of fusion feature vectors of fNIRS brain function indexes and clinical exercise evaluation, and provides near infrared brain function features obviously related to exercise functions for the near infrared indexes and stroke rehabilitation mapping module.

In the embodiment of the invention, the fNIRS brain function and exercise rehabilitation mapping module is used for inputting the multi-task-multi-mode fNIRS brain function fusion index and the comprehensive exercise rehabilitation index into the integrated classifier to realize the classification mapping of the fNIRS brain function and the rehabilitation effect. The rehabilitation effect evaluation module provides a sample label for the module according to the multi-task fNIRS brain function index and clinical evaluation fusion characteristics screened by the multi-task-multi-mode fusion characteristic selection module, wherein the sample label comprises three types of good recovery, general recovery and poor recovery; because clinical data is often unbalanced, and the recovery situation proportion of cerebral apoplexy patients is also unbalanced, an integrated multi-classification Support Vector Machine (SVM) model is designed, as shown in FIG. 6. The integrated model is used for solving the unbalanced problem of each group of data, and mainly consists of N multi-classification SVM models, wherein the value of N is determined according to the situation; each SVM model will input a balanced sub-training set containing the same number of well-recovered, general-recovered, poorly recovered patients. The generation principle of the sub training set is that the sample quantity consistent with the number of the minority class samples is randomly selected from the majority class samples according to the number of the minority class samples; each sub-training set formed in this way is different, but a few classes of samples are fully utilized, and each sub-training set is guaranteed to be balanced. Construction on each multi-classification SVM model

A classifier, k=3 in the embodiment of the present invention, thus constructsThe number of k is merely an example of 3 two-classification SVM models, and the present invention is not limited thereto. The model adopts Radial Basis Function (RBF) to carry out high-dimensional mapping on near-infrared brain functional characteristics, adopts a grid optimizing method to find the optimal punishment factor C and nuclear radius sigma parameters, and comprises the following specific processes:

assuming that the effective range distribution of the punishment factor C and the kernel radius sigma is a and b and the search step length is k, the effective range distribution is calculated in a coordinate system (C _i ,σ _i ) Middle C _i ＝C*i/k,σ _i =σ.i/k, according to the cross-validation calculation procedure, for different C _i ,σ _i And performing traversal combination, calculating the classification accuracy of the sample, and selecting classifier parameters with the maximum cross-validation classification accuracy after traversal is completed to establish a classifier model. And determining the classification result of each multi-classification SVM according to the voting method through the weight. The integrated SVM model combines a plurality of SVMs by adopting a bagging majority voting method to form a strong learner.

Based on the above processing of the fNIRS data set and the clinical evaluation set of the brain injury patient in the clinically collected multitask state, a data set composed of brain function indexes and function rehabilitation labels is formed, the data set is used as the input of an integrated learner, training and parameter optimizing are carried out, and the parameter with the highest classification accuracy is selected as an fNIRS brain function and exercise rehabilitation mapping model, namely, a near infrared brain function map is constructed for brain function dynamic evaluation and rehabilitation effect prediction.

In the embodiment of the invention, the fNIRS brain function map is a data set formed by forming brain function indexes and function rehabilitation labels after the collected fNIRS data set and clinical evaluation set under a specific detection paradigm of a brain injury user are used for processing, the data set is used as input of an integrated learner for training and parameter optimizing, and the parameter with the highest classification accuracy is selected as a fNIRS brain function and exercise rehabilitation mapping classification model to construct a fNIRS brain function map knowledge base oriented to exercise rehabilitation.

Step S150, a visualized evaluation report is provided by the brain function state evaluation module based on the acquired matching classification result.

In the step, based on the fNIRS brain function map knowledge base for exercise rehabilitation constructed in the step S140, brain function response characteristics of a cerebral apoplexy patient under a specific detection paradigm are input into the fNIRS brain function map module for classification matching, the brain function state of the patient can be judged according to the classification matching result, a brain function state evaluation result is output, a visual evaluation report is provided, and the brain function response characteristics are fed back to the patient and medical staff.

Correspondingly, the invention also provides a system, as shown in fig. 2, comprising: the system comprises a motion detection paradigm selection module, a near infrared brain function acquisition module, a brain function response characteristic analysis module, an fNIRS brain function atlas module and a brain function state evaluation report module; the motion detection paradigm selection module is used for selecting a specific brain function detection paradigm according to the motion function of a user; the near-infrared brain function acquisition module is used for acquiring multichannel brain blood oxygen signals of a user under a specific detection paradigm in real time; the brain function response characteristic analysis module is used for preprocessing and calculating near-infrared multi-mode brain function characteristic parameters according to the acquired brain blood oxygen signals under the specific detection paradigm, and constructing brain function response characteristic vectors induced by the specific detection paradigm; the fNIRS brain functional map module is used for constructing a mapping model of fNIRS brain functional characteristics and exercise rehabilitation according to an integrated multi-classification method, namely an fNIRS brain functional map knowledge base related to exercise rehabilitation; inputting the brain function response characteristics induced by the specific detection paradigm of the user obtained by the brain function response characteristic analysis module into the module for matching classification, and judging the neural function state of the user; the brain function state evaluation report module is used for outputting the matching result obtained by the fNIRS brain function map module, providing a visual evaluation report and feeding back to a user and medical staff.

In an embodiment of the present invention, the system further includes a computer device, where the computer device includes a processor and a memory, where the memory stores computer instructions, and the processor is configured to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, as shown in fig. 7, perform the following steps:

step S1, according to the movement function condition of a brain injury patient, eliminating the possibility of magnetoelectric tabu, and selecting a detection paradigm in a movement detection paradigm selection module.

Step S2: and acquiring cerebral blood oxygen signals of a user under a specific detection paradigm by a near infrared brain function acquisition device, wherein each state is not less than 10 minutes.

Step S3: and inputting the acquired cerebral blood oxygen signals under the specific detection paradigm into a brain function response characteristic analysis module, and calculating near-infrared multi-mode brain function parameters to obtain brain function multi-mode response characteristics induced by the specific detection paradigm.

Step S4: training the constructed integrated multi-classifier by taking the multi-task-multi-mode fNIRS brain functional characteristics and the clinical rehabilitation effect classification labels as data sets, and forming an optimal detection model after saving training, parameter adjustment and testing, namely forming a near infrared brain functional map for rehabilitation of cerebral apoplexy exercise functions; and inputting brain function response characteristics induced by the specific detection paradigm into an fNIRS brain function atlas knowledge base to carry out matching classification, and judging the neural function state of the user.

Step S5: and (3) matching and outputting the fNIRS brain functional spectrum to obtain a result for visual reporting, so as to realize brain functional state evaluation of a cerebral apoplexy patient, and feeding back the result to a user and medical staff to assist clinical functional evaluation and treatment response detection, and providing a treatment target for personalized intervention.

The invention provides a dynamic rehabilitation evaluation method and a system based on an fNIRS brain functional map, which dynamically evaluate the nerve functional state of a crowd with dyskinesia after brain injury by combining a near infrared brain functional imaging technology with a specific movement detection paradigm, and provide important imaging basis for rehabilitation scheme formulation and optimization; the multi-mode feature fusion method can be adopted to combine the activation and connection network information, and various brain function analysis methods are adopted to describe the brain function map more abundantly; the integrated multi-classification support vector machine model can be used for solving the limitation of clinical data unbalance and solving the mapping prediction model bias phenomenon caused by the data unbalance. The brain function specific response and remodeling characteristics under various typical task ranges can be fused, and a dynamic mapping model for near infrared brain function indexes and motor function rehabilitation is constructed; the method can utilize the technical advantages of the near infrared brain function imaging technology in resisting motion and electromagnetic interference, design a typical monitoring paradigm facing a motor nerve loop, fully acquire brain function characteristics related to the motor function, and enrich brain stroke nerve function assessment except for a resting state; the method can extract the fNIRS multi-mode brain function response characteristics under the resting state and typical detection task paradigm, establish a mapping model of the fNIRS multi-mode brain function response characteristics and the motor function rehabilitation, and establish an fNIRS brain function map oriented to motor dysfunction so as to realize dynamic evaluation of brain function states of brain injury patients. The brain function rehabilitation dynamic assessment is more real-time, comprehensive, continuous, objective, accurate and targeted, so that the brain function state assessment of a cerebral apoplexy patient can be carried out, the clinical function assessment and the treatment response detection are assisted, and a treatment target is provided for personalized intervention. The brain function state evaluation method is simple to operate, and can be used for evaluating the brain function states of patients with different rehabilitation levels.

The embodiments of the present invention also provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the edge computing server deployment method described above. The computer readable storage medium may be a tangible storage medium such as Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disk, a removable memory disk, a CD-ROM, or any other form of storage medium known in the art.

Those of ordinary skill in the art will appreciate that the various illustrative components, systems, and methods described in connection with the embodiments disclosed herein can be implemented as hardware, software, or a combination of both. The particular implementation is hardware or software dependent on the specific application of the solution and the design constraints. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, a plug-in, a function card, or the like. When implemented in software, the elements of the invention are the programs or code segments used to perform the required tasks. The program or code segments may be stored in a machine readable medium or transmitted over transmission media or communication links by a data signal carried in a carrier wave.

It should be understood that the invention is not limited to the particular arrangements and instrumentality described above and shown in the drawings. For the sake of brevity, a detailed description of known methods is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method processes of the present invention are not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between steps, after appreciating the spirit of the present invention.

In this disclosure, features that are described and/or illustrated with respect to one embodiment may be used in the same way or in a similar way in one or more other embodiments and/or in combination with or instead of the features of the other embodiments.

The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, and various modifications and variations can be made to the embodiments of the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A dynamic rehabilitation assessment system based on fnrs brain function atlas, the system comprising:

a motion detection paradigm selecting module, configured to select a brain function detection paradigm matching with a user's motion ability from a plurality of brain function detection paradigms stored in advance according to the user's motion ability; the pre-stored plurality of brain function detection paradigms includes a resting state and a typical fnrs brain function detection paradigm that faces motor-performed neural loop remodeling and sensory-motor neural loop remodeling; the detection paradigm for remodelling the motor-oriented execution nerve loop is based on repeated extension of the target-oriented hemiplegic upper limb; the sensory-motor nerve loop remodeling-oriented detection range is a central-peripheral combined magnetoelectric stimulation technology, and is used for activating a motor-related brain region and inducing proprioception to be transmitted into a central cortex;

the near-infrared brain function acquisition module is used for acquiring multichannel brain blood oxygen signals induced by a user in the matched brain function detection range in real time;

the brain function response characteristic analysis module is used for preprocessing and calculating near-infrared multi-mode brain function characteristic parameters through an fNIRS data acquisition and multi-mode characteristic calculation module in the fNIRS brain function map module according to the acquired brain blood oxygen signals, and combining multi-mode network information through brain function multi-mode characteristic fusion to obtain brain function multi-mode response characteristic vectors; the multi-mode is an activation mode and a connection mode, wherein the activation mode is used for describing the activation distribution condition of each area of the brain, and the connection mode is used for representing the information interaction process under the typical task of the brain;

The fNIRS brain function map module is used for constructing a mapping model of fNIRS brain function characteristics and exercise rehabilitation according to an integrated multi-classification method, inputting the obtained brain function multi-mode response characteristic vector into the module for matching classification, and obtaining a matching classification result;

the brain function state evaluation report module is used for outputting the matching classification result obtained by the fNIRS brain function map module and providing a visual evaluation report;

the system further comprises:

classifying the movement dysfunction into mild, moderate and severe dysfunction by adopting cluster analysis; clustering the movement function evaluation scale set into mild, moderate and severe movement dysfunction subsets by using a K-means clustering algorithm; the exercise function evaluation scale set comprises an exercise function evaluation scale, a grip index and an upper limb action research ARAT scale;

selecting brain function response characteristics based on combination of group-level statistical tests and function-related regression fusion analysis; the group-level statistical test is used as a feature screening and filtering method to extract brain function feature indexes with statistical significance; the function correlation regression fusion analysis is to fuse clinical function evaluation information on the basis of preliminary screening of features by group-level statistical test, establish a segmented regression model, select features with statistical significance, identify an optimal feature subset related to dysfunction in a specific state, and select multi-task-multi-mode fusion features;

Based on the unbalance of clinical data, adopting an integrated multi-classification Support Vector Machine (SVM) model to deal with the unbalance problem of each group of data; the integrated multi-classification support vector machine SVM model adopts radial basis function RBF to carry out high-dimensional mapping on near-infrared brain functional characteristics, and adopts a grid optimizing method to search the best punishment factors and nuclear radius parameters.

2. The system of claim 1, wherein the fnrs brain function map module comprises: the system comprises a motion function evaluation module, an fNIRS data acquisition and multi-mode feature calculation module, a multi-task-multi-mode fusion feature selection module and an fNIRS brain function and motion rehabilitation mapping module;

the fNIRS brain function and motion rehabilitation mapping module is used for screening out fNIRS brain function fusion characteristics and rehabilitation effect classification labels from the multi-task-multi-mode fusion characteristic selection module, constructing an integrated multi-classification Support Vector Machine (SVM) model, and selecting the parameter with the highest classification accuracy as the fNIRS brain function and motion rehabilitation mapping classification model through training and parameter optimization, so as to construct an fNIRS brain function map.

3. A dynamic rehabilitation assessment system based on fnigs brain function atlas, the system further comprising a processor and a memory, characterized in that the memory has stored therein computer instructions, the processor being configured to execute the computer instructions stored in the memory, the system performing the following steps when the computer instructions are executed by the processor:

selecting a brain function detection paradigm matching the exercise capacity of a user from a plurality of brain function detection paradigms stored in advance based on the exercise capacity of the user; the pre-stored plurality of brain function detection paradigms includes a resting state and a typical fnrs brain function detection paradigm that faces motor-performed neural loop remodeling and sensory-motor neural loop remodeling; the detection paradigm for remodelling the motor-oriented execution nerve loop is based on repeated extension of the target-oriented hemiplegic upper limb; the sensory-motor nerve loop remodeling-oriented detection range is a central-peripheral combined magnetoelectric stimulation technology, and is used for activating a motor-related brain region and inducing proprioception to be transmitted into a central cortex;

Based on the acquired cerebral blood oxygen signals, calculating near-infrared multi-mode brain function feature parameters, and combining multi-mode network information by adopting a brain function multi-mode feature fusion method to form a multi-mode brain function feature combination vector; the multi-mode is an activation mode and a connection mode, wherein the activation mode is used for describing the activation distribution condition of each area of the brain, and the connection mode is used for representing the information interaction process under the typical task of the brain;

providing a visual assessment report through a brain function state assessment module based on the acquired matching classification result;

the system also performs:

4. The system of claim 3, wherein the system further performs the step of preprocessing the acquired cerebral blood oxygen signals before performing the step of calculating the near infrared multi-mode brain function feature parameter when the computer instructions are executed by the processor, comprising: and performing time-frequency conversion on the cerebral blood oxygen signal by adopting a wavelet transformation time-frequency analysis method.

5. The system of claim 3, wherein the near infrared multi-mode brain function feature parameters include hemisphere activation sideslip, hemisphere connection sideslip, hemisphere autonomy coefficients, undirected weighted topology parameters, and directed weighted topology parameters.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that the program is executed by a processor to:

the processor also performs:

7. The storage medium of claim 6, wherein the program, when executed by the processor, further performs the step of preprocessing the acquired cerebral blood oxygen signals before performing the step of calculating the near infrared multi-mode brain function feature parameter, comprising: and performing time-frequency conversion on the cerebral blood oxygen signal by adopting a wavelet transformation time-frequency analysis method.

8. The storage medium of claim 6, wherein the near infrared multi-mode brain function feature parameters include hemisphere activation sideslip, hemisphere connection sideslip, hemisphere autonomy coefficients, undirected weighted topology parameters, and directed weighted topology parameters.