An intelligent system for assessing and training cognitive functions of hemispatial spatial hierarchical activation and multimodal unilateral spatial neglect

Abstract

The application provides a kind of affected side spatial stratification activation and multi-modal unilateral spatial neglect (USN) cognitive function evaluation and training intelligent system, aiming at solving the pain points such as low efficiency of evaluation, training homogeneity, effect difficult to transfer in USN patient rehabilitation, and building a whole-process intelligent rehabilitation system of "accurate evaluation-personalized training-dynamic optimization-effect transfer". Through the combination of rapid quantitative evaluation system, behavior data and electroencephalogram signal, the objective grading of USN severity is realized, which gets rid of artificial dependence and greatly improves the evaluation efficiency and accuracy; through the whole-process automatic operation and machine learning self-adaptive optimization, the operation cost of clinical personnel is reduced, the rehabilitation effect is traceable and evaluable, and an accurate and personalized intelligent rehabilitation system is built, which provides a whole-process solution for USN patients and has important significance for promoting the intelligent development of post-stroke neurological dysfunction rehabilitation.

Description

Technical Field

[0001] This application relates to the interdisciplinary fields of rehabilitation medicine, neuroengineering, virtual reality (VR) technology, machine learning and sensor technology, and specifically to an intelligent system for assessing and training cognitive functions of unilateral spatial hierarchical activation and multimodal unilateral spatial neglect. Background Technology

[0002] With the increasing incidence of stroke year by year, unexplained neuralgia (USN), a common neurological dysfunction after stroke, has become one of the key issues affecting patients' rehabilitation prognosis. USN patients mainly exhibit a loss of perception, attention, and motor execution abilities on the affected side. Specifically, this manifests as ignoring the sleeve on the affected side when dressing, not noticing food on the affected side of the plate when eating, and bumping into obstacles on the affected side when walking. This severely impairs patients' ability to perform daily living activities, reduces their quality of life, and places a heavy caregiving burden on families and society. Therefore, accurate assessment and effective rehabilitation training for USN have become a research hotspot and urgent need in the field of rehabilitation medicine.

[0003] In recent years, with the rapid development of information technologies such as virtual reality (VR), sensor technology, and machine learning, the medical rehabilitation field has ushered in an opportunity for digital transformation. Virtual simulation rehabilitation systems are gradually being applied to the rehabilitation training of USN patients. Existing technologies, by constructing virtual environments, provide patients with relatively safe and repeatable training scenarios, to some extent compensating for the shortcomings of traditional rehabilitation training, such as limited scenarios and strong subjectivity. Simultaneously, the integration of technologies such as electroencephalogram (EEG) acquisition, eye tracking, and limb motion capture helps to quantify three core indicators: affected-side spatial reaction time, target recognition rate, and bilateral spatial allocation bias. This provides more objective data support for the functional assessment of USN patients, promoting the development of rehabilitation assessment from qualitative to quantitative methods.

[0004] However, current USN rehabilitation technology still faces many challenges: First, the assessment process is cumbersome and complex, relying heavily on manual operation by rehabilitation physicians. Assessment indicators lack standardized criteria, quantitative accuracy is insufficient, and it's difficult to quickly adapt to the patient's real-time rehabilitation status. Second, training programs are highly homogenized, failing to adequately consider individual patient differences and unable to dynamically adjust training difficulty and content based on rehabilitation progress. This leads some patients to give up due to overly difficult training or to fail to achieve ideal rehabilitation results due to overly easy training. Third, training scenarios are disconnected from real daily life, often involving abstract task designs, making it difficult to effectively transfer rehabilitation effects to real life, resulting in a dilemma of "effective training, ineffective life." Fourth, the feedback mechanism is simplistic, primarily relying on visual feedback and lacking synergistic feedback from physiological stimulation and tactile perception. This makes it difficult to effectively strengthen the patient's perceptual association with the affected side of the space, leading to low training compliance. Fifth, the lack of comprehensive remote management functions fails to meet the needs of patients undergoing home rehabilitation, making it difficult to ensure the continuity of rehabilitation training and affecting overall rehabilitation outcomes.

[0005] While some existing technologies attempt to incorporate VR scenarios or simple feedback mechanisms, a complete closed-loop system encompassing "assessment-training-optimization-transfer" has not yet been established, failing to fundamentally address core issues such as assessment accuracy, personalized training, effect transferability, multidimensional feedback, and scenario adaptability. Therefore, developing an intelligent rehabilitation system capable of rapid quantitative assessment, personalized tiered training, multimodal synchronous feedback, real-life scenario-based transfer, adaptive incentives, and remote collaborative management is crucial for improving USN rehabilitation outcomes and promoting the intelligent development of rehabilitation medicine, and represents an inevitable trend in current technological advancements. Summary of the Invention

[0006] To address the aforementioned technical issues, this invention provides an intelligent system for lateral spatial hierarchical activation and multimodal USN assessment and training. It integrates five core technologies: lateral spatial hierarchical activation, multimodal synchronous feedback, assessment-training closed loop, life-scenario-based transfer, and adaptive remote management. This system constructs a comprehensive intelligent rehabilitation system encompassing "precise assessment, personalized training, dynamic optimization, and effect transfer," achieving quantitative, personalized, efficient, and scenario-based USN rehabilitation, significantly improving patient rehabilitation outcomes and quality of life.

[0007] The specific solution of this application is as follows: A VR method for rehabilitation of unilateral spatial neglect (USN) includes the following steps:

[0008] S1: After the system is started, it automatically enters the rapid quantitative assessment process to detect the patient's spatial reaction time on the affected side, target recognition rate and bilateral spatial allocation deviation. At the same time, the acquisition device captures the patient's spatial attention-related EEG components. Combining these EEG signals with behavioral data, the severity of USN is graded.

[0009] S2: Based on the evaluation results of S1, personalized training parameters adapted to the patient's current ability are automatically generated through machine learning algorithms, including the initial stage of the affected side spatial activation, stimulus intensity, task difficulty level, training scenario type, and feedback parameter configuration.

[0010] S3: Conduct layered activation training in the order of “affected side edge stimulation → affected side center stimulation → bilateral spatial integration stimulation”. First, set low-saturation dynamic markers at the edge of the affected side visual field in the VR scene. After the patient stabilizes their gaze, move the virtual task target to the center of the affected side visual field. Then, design a linkage task across both visual fields. Monitor eye movement and head rotation data in real time during training and dynamically adjust the spatial crowding threshold.

[0011] S4: Multimodal feedback is triggered synchronously during training. When the patient's limb enters the target area on the affected side and initiates the task action, low-intensity functional electrical stimulation within the safe threshold of 1-5mA is applied to the corresponding muscle on the affected side. The VR device presents visual feedback, and the VR glove tactile module simulates the feeling of touching an object, forming a sensory closed loop.

[0012] S5: After every 3 training sessions, a new round of rapid quantitative assessment is automatically triggered. The rehabilitation progress is analyzed by combining training data and EEG signals, and the training plan is dynamically adjusted to achieve a closed loop of "assessment-training-reassessment-optimization".

[0013] Preferably, in step S1, an EEG acquisition device is used to capture the patient's spatial attention-related EEG components, including P300 amplitude and alpha wave inhibition level.

[0014] Preferably, the specific manifestation of "affected side edge stimulation → affected side center stimulation → bilateral spatial integration stimulation" in S3 is as follows:

[0015] Phase 1: Ill-side peripheral stimulation phase: In the VR scene, low-saturation, dynamically flashing virtual markers are set at the edge of the patient's visual field on the affected side; this phase aims to awaken basic spatial perception on the affected side with minimal cognitive load.

[0016] Phase 2: Ill-sided central stimulation phase: Once the system detects that the patient can stably fixate on the edge marker through eye movement and head movement data, the virtual task target that requires interaction is moved to the center of the visual field on the affected side; in this phase, the patient is forced to actively focus their attention on the affected side and perform the purposeful action;

[0017] Phase 3: Bilateral Spatial Integration Phase. This phase involves setting up collaborative tasks that require movement across the body's midline. It trains attention allocation, planning, and motor coordination in bilateral space, directly simulating complex daily life activities.

[0018] Preferably, during the S3 training process, the system monitors the patient's eye movement trajectory and head rotation data in real time, and dynamically adjusts the spatial crowding threshold (such as target spacing and frequency of occurrence) to ensure that the training difficulty is always within the adaptation range of "slightly higher than the current ability".

[0019] Preferably, the specific steps of S4 are as follows:

[0020] First, the patient's limb movement trajectory is tracked in real time by a limb motion capture camera and VR glove sensors. When the patient's limb enters the target area on the affected side and initiates a task action, multimodal feedback is immediately triggered. Then, low-intensity functional electrical stimulation with an intensity within the safe threshold of 1-5mA is applied to the corresponding muscles of the affected limb. The stimulation duration is synchronized with the duration of the VR task action to strengthen the muscle-nerve connection. Finally, the VR display device synchronously presents visual feedback of "highlighting the affected limb + flashing the target + filling the progress bar". The VR glove simulates the pressure sensation of touching an object through a tactile module.

[0021] This application also provides an intelligent system for assessing and training cognitive functions of unilateral spatial hierarchical activation and multimodal unilateral spatial neglect, which is applied to the methods described above.

[0022] Preferably, the system consists of a hardware module and a software module system. The hardware module includes a VR display device, a sensing and monitoring module, a multimodal feedback module, a data processing module, and a remote management terminal.

[0023] Preferably, the VR display device includes a VR headset and VR gloves; the sensing and monitoring module includes an eye tracker, a head motion sensor integrated into the VR headset, an EEG brainwave acquisition device, and a limb motion capture camera; the multimodal feedback module includes a low-intensity functional electrical stimulation device and a tactile feedback module; and the data processing module includes a high-speed data converter and a dedicated server.

[0024] Preferably, the head motion sensor integrated into the VR headset includes a gyroscope and an accelerometer.

[0025] Preferably, the software module includes an evaluation algorithm module, a hierarchical activation training module, a multimodal feedback control module, a scene rendering module, an adaptive excitation module, and a remote data interaction module.

[0026] Compared with the prior art, this application includes at least the beneficial effects on pancreatic cancer:

[0027] This invention integrates five core technologies: It employs a three-stage hierarchical activation logic of "affected-side limbic stimulation → affected-side central stimulation → bilateral spatial integration stimulation" to dynamically adapt to the patient's attention recovery progress; it constructs a multimodal synchronization mechanism of "electrical stimulation + visual feedback + tactile feedback" to strengthen spatial perception and motor association on the affected side; it designs a rapid quantitative assessment system that combines EEG signals to objectively grade the severity of USN, forming a closed-loop process of "assessment-training-reassessment-optimization"; it recreates everyday life scenarios such as dressing and eating to facilitate the transfer of rehabilitation effects to real life; and it incorporates machine learning to adaptively adjust difficulty and gamified incentives to support remote home rehabilitation. The system collects patient behavior and EEG data through a sensor monitoring module, analyzes the data through a data processing module to generate personalized training plans, and uses VR devices to construct immersive scenes, synchronously triggering multimodal feedback. The plan is automatically iterated and optimized after every three training sessions. This system enables the quantification, personalization, and scenario-based rehabilitation of USN (Ulcerative Collapse), significantly improving assessment efficiency, training compliance, and the ability to transfer rehabilitation effects, while reducing clinical operation costs. It provides a safe and efficient end-to-end rehabilitation solution for USN patients after stroke, and is of great significance for promoting the intelligent development of rehabilitation for neurological dysfunction. Attached Figure Description

[0028] Figure 1 This is a system architecture diagram of an intelligent system for assessing and training cognitive functions of unilateral spatial hierarchical activation and multimodal unilateral spatial neglect, as described in one embodiment of this application.

[0029] Figure 2 This is a flowchart of the affected side spatial layer activation and multimodal USN evaluation and training in one embodiment of this application. Detailed Implementation

[0030] Please see Figure 1 This application provides an intelligent system for assessing and training cognitive functions of unilateral spatial hierarchical activation and multimodal unilateral spatial neglect, which consists of hardware modules and software modules.

[0031] In one embodiment, the hardware module includes a VR display device, a sensing and monitoring module, a multimodal feedback module, a data processing module, and a remote management terminal. The VR display device includes a VR headset and VR gloves. The sensing and monitoring module includes an eye tracker, a head motion sensor integrated into the VR headset, an EEG brainwave acquisition device, and a limb motion capture camera. The head motion sensor integrated into the VR headset includes a gyroscope and an accelerometer. The multimodal feedback module includes a low-intensity functional electrical stimulation device and a tactile feedback module. The data processing module includes a high-speed data converter and a dedicated server.

[0032] The software modules include an evaluation algorithm module, a hierarchical activation training module, a multimodal feedback control module, a scene rendering module, an adaptive excitation module, and a remote data interaction module.

[0033] The sensor monitoring module collects real-time data on patients' eye movements, head movements, limb trajectories, and electroencephalogram (EEG) signals. This data is then transmitted to the software module for analysis via the data processing module. The software module generates personalized assessment reports and training plans based on the analysis results. An immersive training scenario is constructed using a VR display device, and the multimodal feedback module is simultaneously controlled to output stimulation signals. The remote management terminal enables real-time viewing of rehabilitation data and remote adjustment of training plans, supporting two-way interaction between clinicians and patients.

[0034] In addition, such as Figure 2 As shown, this application also provides a USN-specific evaluation and training method, which includes the following steps:

[0035] S1: USN Exclusive Rapid Quantitative Assessment

[0036] After system startup, it automatically enters the "rapid assessment process" without manual intervention, achieving quantitative detection of three core indicators: Affected-sided spatial reaction time: recording the patient's response speed to targets at different locations in the affected-sided visual field, with 1.2 seconds set as a reference threshold; Target recognition rate: statistically analyzing the proportion of correct recognition of near, intermediate, and far targets within the affected-sided visual field; Bilateral spatial allocation bias: capturing the patient's gaze trajectory using an eye tracker and calculating the proportion of gaze duration in both visual fields. Simultaneously, the EEG acquisition device captures EEG components related to spatial attention (such as P300 amplitude and alpha wave inhibition level). The software module combines EEG signals and behavioral data to analyze the neural activation level of spatial attention on the affected side and complete the USN severity grading (mild, moderate, severe).

[0037] S2: Personalized Training Program Generation

[0038] Based on the evaluation results of step 1, personalized training parameters adapted to the patient's current abilities are automatically generated using a supervised learning-based classification and regression algorithm. These parameters include: the initial stage of spatial activation on the affected side, stimulus intensity, task difficulty level, training scenario type, and feedback parameter configuration. For example, for patients with severe neglect (affected side reaction time > 2 seconds), the initial training parameters are set as "affected side limb stimulation stage + low-intensity electrical stimulation + basic daily life scenario + low-difficulty task"; for patients with moderate neglect (affected side recognition rate < 50%), the parameters are set as "affected side central stimulation stage + moderate-intensity feedback + complex daily life scenario + moderate-difficulty task".

[0039] S3: Spatial Layer Activation Training on the Affected Side

[0040] The system employs a three-stage progressive activation logic: “affected side peripheral stimulation → affected side central stimulation → bilateral spatial integration stimulation”. All stages are embedded with highly lifelike virtual scenarios and a task gradient from easy to difficult is set to ensure that the training effect can be systematically transferred to daily life.

[0041] The "low-saturation dynamic labeling," "virtual task objectives," and "bilateral linkage tasks" used in these three stages are not designed in isolation, but rather form an integrated training chain. They correspond to the rehabilitation objectives of "sensory arousal," "active attention enhancement," and "functional integration and transfer," respectively. They are dynamically linked according to the patient's abilities, forming a complete rehabilitation path from easy to difficult, from unilateral to bilateral, and from basic perception to complex operations.

[0042] Phase 1: Affected Side Edge Stimulation Phase. In the VR scene, low-saturation, dynamically flashing virtual markers are placed at the edge of the patient's affected side's visual field. For example, in the "Dressing" scene, the initial (basic) task is simply to perceive the glowing outline of the affected side's sleeve edge; in the "Eating" scene, the initial task is to perceive the semi-transparent outline of the affected side's plate. This phase aims to awaken basic spatial perception on the affected side with minimal cognitive load.

[0043] Phase 2: Ill-Side Central Stimulation Phase. Once the system detects that the patient can stably fixate on the edge markers using eye-tracking and head-movement data, the virtual task target requiring interaction is moved to the center of the affected side's visual field. Task difficulty is increased: In the intermediate task of "dressing," the target becomes a clearly defined "virtual button" on the affected side, requiring the patient to reach out, grasp, and button it; in the "eating" scenario, the target becomes a "virtual spoon" on the affected side, requiring the patient to hold it. This phase compels the patient to actively focus their attention on the affected side and perform the purposeful action.

[0044] Phase 3: Bilateral Spatial Integration Phase. This phase involves collaborative tasks requiring movement across the body's midline. These are advanced tasks; for example, in a "dressing" scenario, the patient is asked to use their affected hand to retrieve clothing from a wardrobe on the left and hand it to their unaffected hand, completing the dressing action together. In a "eating" scenario, the patient is asked to use their affected hand to hold a spoon to retrieve food from a bowl on the left, transport it, and place it on a plate on their unaffected side. This phase trains attention allocation, planning, and motor coordination in bilateral space, directly simulating complex daily life activities.

[0045] During training, the system monitors the patient's eye movement trajectory and head rotation data in real time, and dynamically adjusts the spatial crowding threshold (such as target spacing and frequency of occurrence) to ensure that the training difficulty is always within the appropriate range of "slightly higher than the current ability", avoiding the situation where the difficulty is too high and leads to abandonment of training or the difficulty is too low and there is no rehabilitation effect.

[0046] S4: Enhanced Multimodal Synchronous Feedback

[0047] During the tiered activation training process, a triple synchronous feedback mechanism of "VR task - physiological stimulation - sensory feedback" is constructed. First, the patient's limb movement trajectory is tracked in real time through a limb motion capture camera and VR glove sensors. When the patient's limb enters the target area on the affected side and initiates a task action (such as touching a virtual target on the affected side or grasping an object on the affected side), multimodal feedback is immediately triggered. Then, low-intensity functional electrical stimulation with an intensity within the safe threshold of 1-5mA is applied to the corresponding muscles of the affected limb (such as the biceps brachii, forearm muscles, and quadriceps femoris of the lower limb). The stimulation duration is synchronized with the duration of the VR task action (such as continuous stimulation within 3 seconds of touching the target) to strengthen the muscle-nerve connection. Finally, the VR display device synchronously presents visual feedback of "highlighting the affected limb + flashing the target + filling the progress bar". The VR glove simulates the pressure sensation of touching an object through the tactile module, forming a sensory closed loop of "action-vision-touch", allowing the patient to intuitively perceive the effectiveness of the action on the affected side and break the cognitive bias of "the space on the affected side does not exist".

[0048] S5: Evaluation-Training Closed-Loop Iteration

[0049] After every three training sessions, the system automatically triggers a new round of rapid quantitative assessment. Combining data such as task completion rate, reaction time changes, and EEG signal activation levels from this training session, the system analyzes the patient's rehabilitation progress using machine learning algorithms. If the assessment results show that the patient's spatial attention ability on the affected side has improved (e.g., the reaction time on the affected side has been shortened to 1.5 seconds, and the recognition rate has increased to 60%), the training program is automatically upgraded to the next stage (e.g., transitioning from the limbic stimulation stage to the central stimulation stage), while simultaneously increasing the task difficulty and feedback intensity. If the assessment results show that rehabilitation progress is slow, the current training stage is maintained, and the stimulation parameters and task design are optimized (e.g., increasing the flashing frequency of edge markers and reducing the target spacing) to ensure that the training program always fits the individual rehabilitation rhythm of the patient, achieving a closed-loop process of "assessment-training-reassessment-optimization".

[0050] S6: Adaptive Excitation and Remote Management

[0051] The adaptive incentive module uses machine learning algorithms to analyze patient training data streams in real time, dynamically modeling their ability change curves, and adjusting the difficulty gradient of the next set of tasks accordingly to achieve a "balance between challenge and skill." It incorporates a gamified incentive mechanism, where patients unlock new life scenarios and earn virtual badges after completing tasks on the affected side, improving training adherence. The system supports remote deployment, allowing clinicians to view patient assessment reports, training data, and rehabilitation curves in real time via remote management terminals, and to remotely adjust training plans based on patient conditions. Patients can access the system through home devices for home rehabilitation, with the system automatically recording rehabilitation data and synchronizing it to the clinical end, achieving seamless integration of in-hospital and home rehabilitation.

[0052] Based on the above description, this application achieves objective grading of the severity of USN by combining a rapid quantitative assessment system with behavioral data and EEG signals, eliminating reliance on manual intervention and significantly improving assessment efficiency and accuracy. The deep integration of a three-stage hierarchical activation logic with an assessment-training closed loop allows for dynamic adjustment of training difficulty, stimulation intensity, and task design, precisely adapting to individual patient rehabilitation progress and solving the inefficiency problem of homogeneous traditional rehabilitation training. Based on an immersive, embodied cognition-based design of real-life scenarios, it compels patients to actively focus on the affected side of the space when performing tasks such as dressing and eating, effectively addressing the core pain point of USN patients: "effective training, ineffective daily life." Through visual, tactile, and electrical stimulation... The multimodal synchronous feedback mechanism strengthens the neural connection between spatial attention and motor execution on the affected side, improving patients' perception of the affected side's space and training compliance. The remote management module enables the coordinated advancement of centralized in-hospital rehabilitation and decentralized home rehabilitation, breaking spatial limitations, meeting patients' long-term rehabilitation needs, and expanding the coverage of rehabilitation services. The fully automated operation and machine learning adaptive optimization reduce the operating costs for clinical staff, enabling traceable and assessable rehabilitation effects. Overall, it constructs a precise and personalized intelligent rehabilitation system, providing a full-process solution for USN patients, which is of great significance for promoting the intelligent development of rehabilitation for post-stroke neurological dysfunction.

Claims

1. A VR method for rehabilitation of unilateral spatial neglect (USN), characterized in that: Includes the following steps: S1: After the system is started, it automatically enters the rapid quantitative assessment process to detect the patient's spatial reaction time on the affected side, target recognition rate and bilateral spatial allocation deviation. At the same time, the acquisition device captures the patient's spatial attention-related EEG components. Combining these EEG signals with behavioral data, the severity of USN is graded. S2: Based on the evaluation results of S1, personalized training parameters adapted to the patient's current ability are automatically generated through machine learning algorithms, including the initial stage of the affected side spatial activation, stimulus intensity, task difficulty level, training scenario type, and feedback parameter configuration. S3: Conduct layered activation training in the order of "affected side edge stimulation → affected side center stimulation → bilateral spatial integration stimulation". First, set low-saturation dynamic markers at the edge of the affected side visual field in the VR scene. After the patient stabilizes their gaze, move the virtual task target to the center of the affected side visual field. Then, design a linkage task across both visual fields. Monitor eye movement and head rotation data in real time during training and dynamically adjust the spatial crowding threshold. S4: Multimodal feedback is triggered synchronously during training. When the patient's limb enters the target area on the affected side and initiates the task action, low-intensity functional electrical stimulation within the safe threshold of 1-5mA is applied to the corresponding muscle on the affected side. The VR device presents visual feedback, and the VR glove tactile module simulates the feeling of touching an object, forming a sensory closed loop. S5: After every 3 training sessions, a new round of rapid quantitative assessment is automatically triggered. The rehabilitation progress is analyzed by combining training data and EEG signals, and the training plan is dynamically adjusted to achieve a closed loop of "assessment-training-reassessment-optimization".

2. The VR method for unilateral spatial neglect (USN) rehabilitation according to claim 1, characterized in that: In S1, an EEG acquisition device is used to capture the patient's spatial attention-related EEG components, including P300 amplitude and alpha wave inhibition level.

3. The VR method for unilateral spatial neglect (USN) rehabilitation according to claim 2, characterized in that: The specific manifestations of "affected-side marginal stimulation → affected-side central stimulation → bilateral spatial integration stimulation" in S3 are as follows: Phase 1: Ill-side peripheral stimulation phase: In the VR scene, low-saturation, dynamically flashing virtual markers are set at the edge of the patient's visual field on the affected side; this phase aims to awaken basic spatial perception on the affected side with minimal cognitive load. Phase 2: Ill-sided central stimulation phase: Once the system detects that the patient can stably fixate on the edge markers through eye movement and head motion data, the virtual task target that requires interaction is moved to the center of the affected visual field; During this stage, patients are forced to focus their attention on the affected side and perform a purposeful action; Phase 3: Bilateral Spatial Integration Phase: Set up linked tasks that require cooperation across the body midline; this phase trains attention allocation, planning and movement coordination in bilateral space, directly simulating complex daily life activities.

4. The VR method for unilateral spatial neglect (USN) rehabilitation according to claim 3, characterized in that: During the S3 training process, the system monitors the patient's eye movement trajectory and head rotation data in real time, and dynamically adjusts the spatial crowding threshold (such as target spacing and frequency of occurrence) to ensure that the training difficulty is always within the adaptation range of "slightly higher than the current ability".

5. A VR method for unilateral spatial neglect (USN) rehabilitation according to claim 4, characterized in that: The specific steps of S4 are as follows: First, the patient's limb movement trajectory is tracked in real time by a limb motion capture camera and VR glove sensors. When the patient's limb enters the target area on the affected side and initiates a task action, multimodal feedback is immediately triggered. Then, low-intensity functional electrical stimulation with an intensity within the safe threshold of 1-5mA is applied to the corresponding muscles of the affected limb. The stimulation duration is synchronized with the duration of the VR task actions to strengthen the muscle-nerve connection. Finally, the VR display device synchronously presents visual feedback of "highlighting the affected limb, flashing the target, and filling the progress bar," while the VR gloves simulate the pressure sensation of touching an object through a tactile module.

6. A smart system for assessing and training cognitive functions of unilateral spatial hierarchical activation and multimodal unilateral spatial neglect, characterized in that: Applied to the methods shown in claims 1-5.

7. The intelligent system for assessing and training unilateral spatial hierarchical activation and multimodal unilateral spatial neglect cognitive function according to claim 6, characterized in that: The system consists of hardware modules and software modules. The hardware modules include a VR display device, a sensing and monitoring module, a multimodal feedback module, a data processing module, and a remote management terminal.

8. The intelligent system for assessing and training unilateral spatial hierarchical activation and multimodal unilateral spatial neglect cognitive function according to claim 7, characterized in that: The VR display device includes a VR headset and VR gloves. The sensing and monitoring module includes an eye tracker, a head motion sensor integrated into the VR headset, an EEG brainwave acquisition device, and a limb motion capture camera. The multimodal feedback module includes a low-intensity functional electrical stimulation device and a tactile feedback module. The data processing module includes a high-speed data converter and a dedicated server.

9. The intelligent system for assessing and training unilateral spatial hierarchical activation and multimodal unilateral spatial neglect cognitive function according to claim 8, characterized in that: The head motion sensors integrated into the VR headset include a gyroscope and an accelerometer.

10. The intelligent system for assessing and training unilateral spatial hierarchical activation and multimodal unilateral spatial neglect cognitive function according to claim 7, characterized in that: The software modules include an evaluation algorithm module, a hierarchical activation training module, a multimodal feedback control module, a scene rendering module, an adaptive excitation module, and a remote data interaction module.

Images