DIKWP-driven individualized brain-map interaction feedback mechanism

Abstract

The invention discloses a DIKWP-driven individualized brain-map interactive feedback system, which is used for neural rehabilitation and brain-computer interface training. The system obtains brain activity data of a patient through electroencephalogram acquisition equipment, constructs a personal brain-semantic map in combination with cognitive evaluation, and establishes a mapping relation between semantic units and brain region responses. In the training process, the DIKWP semantic analysis module performs multi-layer analysis on indexes such as reaction time, accuracy and intention achievement, generates multi-mode instant feedback such as visual sense, auditory sense or tactile sense, and performs directional reinforcement on a weak semantic domain. The system has a dynamic target optimization capability, and the difficulty can be automatically adjusted according to training performance; when attention distraction, emotion abnormity or semantic deviation is detected, a safety intervention mechanism is automatically triggered, and training effectiveness and safety are guaranteed. According to the method, closed-loop individualized rehabilitation interaction is realized, the adaptation degree and efficiency of brain-computer interface training are remarkably improved, and the method is suitable for various rehabilitation scenes such as languages, movement and cognition and has a good industrial application prospect.

Description

Technical Field

[0001] This invention relates to the fields of neurorehabilitation and artificial intelligence, specifically a personalized brain-map interactive feedback mechanism for brain-computer interface-assisted rehabilitation training. The system connects the patient's brain state to a semantic map and adjusts the feedback strategy in real time based on semantic analysis results driven by the DIKWP model, thereby achieving semantic adaptation and dynamic goal optimization during the rehabilitation process. This mechanism belongs to the field of intelligent rehabilitation training systems. Background Technology

[0002] Brain-computer interfaces (BCIs) have shown great potential in rehabilitation medicine, such as for post-stroke motor function training and speech disorder training. Traditional BCI systems typically employ a universal, fixed training paradigm. However, due to significant individual differences in sensory, cognitive, and brain function among patients, a universal BCI often fails to meet the specific needs of individuals. Researchers have found that brain signal characteristics can differ significantly among different users performing the same brain-computer task, leading to a decline in the performance of universal decoding models across individuals and requiring cumbersome personalized calibration. This has spurred the development of BCI technology towards personalization, improving interaction efficiency by customizing systems for each user. For example, some researchers have proposed selecting the most suitable brain signal characteristics and channels based on user features to customize personalized BCI models.

[0003] Rehabilitation training also requires dynamic adjustments to goals and strategies. Currently, many rehabilitation systems suffer from fixed content and a lack of adaptability, making it difficult to adjust training goals in a timely manner based on patient progress. This can lead to some patients receiving tasks that are too simple to provide sufficient stimulation, while others face tasks that are too difficult, causing frustration and hindering the maximization of rehabilitation outcomes. Furthermore, real-time feedback is crucial in rehabilitation: a two-way closed-loop BCI rehabilitation system allows patients to actively regulate brain function through neurofeedback. For example, in motor imagery BCI training, visual or auditory feedback allows patients to understand changes in their brain signals, thereby learning to adjust their imagery strategies, significantly improving rehabilitation effectiveness. Visual feedback methods, in particular, can present users' brain activity characteristics in real time, enabling users to self-assess and adjust their imagery methods during interaction, achieving "what you think is what you see," significantly improving BCI performance. However, existing feedback often focuses on low-level signals or task success rates, lacking feedback mechanisms for high-level semantic understanding and intent. For example, in language rehabilitation, it's not simply about training correct pronunciation; it's also about training patients' grasp of semantics and the realization of their communicative intentions, requiring a more advanced feedback system.

[0004] Therefore, there is an urgent need for a novel rehabilitation training feedback mechanism that can integrate individualized and semantic-level adjustments: taking into account each patient's unique brain functional atlas, and dynamically adjusting the training content and difficulty based on the patient's current semantic understanding and achievement of rehabilitation goals. This invention proposes a solution to this need. Summary of the Invention

[0005] This invention proposes a DIKWP-driven personalized brain-map interaction feedback mechanism for use in brain-computer interface rehabilitation systems, enabling semantic adaptation and dynamic goal optimization during the training process. The core idea is to map the patient's individual brain function patterns onto a semantic map, forming a "brain-map" interaction; and to analyze the patient's semantic performance at each stage of training using the DIKWP (Data-Information-Knowledge-Wisdom-Purpose) model, thereby personalizing the adjustment of feedback strategies and rehabilitation goals.

[0006] The mechanism of this invention contains three key features:

[0007] 1) Personalized Brain Mapping: Based on each patient's EEG / brain imaging data, a unique semantic cognitive atlas is built, linking the patient's brain state with specific semantic units. This is equivalent to creating a "map" of the patient's brain responses to different tasks and semantic concepts. Using this map, the system can identify which semantic categories the patient is weak in or strong in, thus enabling targeted training.

[0008] 2) Semantic-Driven Feedback Modulation: Unlike traditional feedback based solely on success rate or brainwave amplitude, this invention uses the DIKWP model to analyze the patient's semantic performance during training, achieving deep feedback. For example, in language training, if a patient is detected to be slow to respond to a certain type of vocabulary (knowledge layer K), the system increases the frequency of practice for that type of vocabulary; in motor imagery training, if a patient repeatedly fails at a specific situational goal (goal layer P), the training strategy is changed or the goal is broken down. This semantic adaptation ensures that the training content and methods match the patient's current comprehension ability and rehabilitation stage.

[0009] 3) Dynamic Goal Optimization: Rehabilitation goals are no longer static but dynamically adjusted based on the patient's real-time progress. The system sets an initial rehabilitation goal (such as achieving a specific score on a language test) and continuously assesses the patient's performance during training. When it detects that the patient is close to the current goal and the training effect is approaching a plateau, the goal is automatically increased (e.g., increasing the required vocabulary difficulty or communication fluency); conversely, if the patient is unable to achieve the current goal for a long period, it is broken down into several sub-goals to be completed gradually. The entire process is similar to a game mechanism that automatically adjusts the difficulty, always keeping the patient at a challenging level that is "within reach but requires effort," in order to maximize neuroplasticity.

[0010] In summary, the interactive feedback mechanism provided by this invention fully embodies the principles of semantic sovereignty and human-centeredness: each patient's training semantic path is dominated by their own characteristics, and the system only provides guidance and support within a semantic framework, rather than forcibly imposing a uniform template. This mechanism will greatly improve the effectiveness and personalization of rehabilitation training, and has significant novelty and promising prospects for industrial application.

[0011] The technical solution of this invention includes the following specific steps and module designs:

[0012] 1. Initialization of the patient's semantic brain atlas: Brain signal data (such as EEG, fMRI, etc.) in the patient's resting and task states are collected and combined with cognitive assessment results to establish an individualized brain-semantic atlas. The specific method involves having the patient participate in a series of baseline tasks related to rehabilitation content, such as recognizing object images, listening to words and imagining their meaning, and attempting motor imagery, while simultaneously recording their brain activity patterns. Using pattern recognition algorithms, different stimuli / tasks are associated with the patient's brain signal characteristics and mapped to semantic labels. For example, if a patient shows significant occipital lobe evoked potentials when seeing animal images and a significant decrease in motor cortex beta waves when listening to action verbs, then the atlas records the association between the node "animal" and visual cortex activation, and the association between "action" and motor-related rhythms. The atlas is stored in a graph database or multidimensional matrix format, with nodes representing semantic concepts and edges representing the response intensity of corresponding brain regions or frequency bands. This brain atlas is completely personalized, reflecting the patient's unique semantic cognitive structure. After initialization, the system obtains the patient's basic ability distribution across different semantic tasks.

[0013] 2. DIKWP Semantic Analysis Module: During training, the system incorporates Professor Duan Yucong's DIKWP model to perform multi-layer semantic analysis on the real-time training data. Taking language rehabilitation as an example, when a patient attempts to select words using BCI, the system not only records correct / incorrect selections (data layer D), but also analyzes their reaction time and EEG characteristic changes (information layer I), combines this with the patient's brain map to determine their level of understanding of the word's meaning (knowledge layer K), and infers whether their communication intention has been achieved (wisdom layer W and intention layer P). This five-layer analysis comprehensively depicts the patient's current training state. For example, if a patient selects incorrectly multiple times but their EEG shows N400 waves related to the correct word meaning, it indicates that they understand the word meaning but made a mistake in the selection operation; this contradiction suggests a problem at the execution level rather than the comprehension level. Conversely, if the EEG does not show the expected semantically relevant potentials, it suggests a need to strengthen training in the semantic understanding of the word. The DIKWP model enables the system to "know not only what" but also "why" about the patient: not only seeing the surface performance but also understanding the underlying reasons and intentions, thus providing a basis for feedback adjustments.

[0014] 3. Interactive Feedback Engine: Based on DIKWP analysis results, the system generates personalized feedback to patients in real time. Feedback can take the form of visual, auditory, or other sensory channels. For example, visual feedback can utilize visual neurofeedback: presenting the difference between the patient's current brain activity and the target state in an intuitive graphic. In motor training, the system can display the brainwave power distribution during left and right hand imagery on the screen. If the difference in activation between the left and right hemispheres is not significant (indicating insufficient discrimination), the system will mark a prompt on the graph and guide the patient to adjust their imagery strategy. In language training, visual feedback can use semantic network diagrams: highlighting vocabulary nodes that the patient has mastered and dimming those that they haven't mastered, allowing the patient to clearly see which semantic areas they need to strengthen their practice in. For auditory feedback, the system generates voice prompts. For example, when it detects that a patient repeatedly makes mistakes on a certain exercise and shows frustration, the voice assistant will provide timely encouragement such as, "Don't be discouraged, let's try another method." Tactile feedback (such as vibration cues) can also be used to instantly reward correct actions or remind attention. Through multimodal interactive feedback, patients receive timely information guidance and psychological support throughout the training process, thus participating more actively in rehabilitation training.

[0015] 4. Dynamic Goal Adjustment Module: This module continuously monitors the patient's training progress and compares it with the predetermined rehabilitation goals. On one hand, if the patient's performance significantly exceeds the current stage requirements, indicating accelerated rehabilitation progress, the module will automatically upgrade the training goals. For example, if the original plan was to master 50 words this week, and the patient completes this task ahead of schedule with EEG indicators showing proficiency in processing these words (e.g., a shortened P300 latency), the system will add more challenging vocabulary or situational dialogue practice to ensure the training intensity keeps pace with the patient's improved abilities. On the other hand, if the patient struggles to achieve a certain goal and shows signs of fatigue, the module will break down that goal into several sub-goals and adjust the training process. For example, if the goal of "speaking 5 complete sentences consecutively" is too difficult, the system can break it down into "speaking 1 sentence" to gradually accumulate; or it can use more engaging practice formats to reignite motivation. Dynamic goal adjustment also leverages reinforcement learning algorithms for continuous optimization: the system uses different adjustment measures and their effects as experience to continuously update strategies, making future decisions more intelligent. Through this module, the rehabilitation plan is no longer a rigid task list for patients, but a dynamically evolving personalized plan—just like a level in a game that automatically adjusts its difficulty according to the player's skill level, always maintaining an appropriate level of challenge and sense of accomplishment.

[0016] 5. Semantic Safety and Bug Management: This invention also incorporates semantic safety and awareness bug theory to improve the reliability of the feedback mechanism. When the system analyzes the patient's semantic state or adjusts the goals, if it detects anomalies (e.g., the patient's brain activity is completely mismatched with the semantic task, suspecting a "cognitive bug" such as inattention) or detects that certain training content triggers negative emotions in the patient (e.g., a surge in stress levels), the system will pause the current training and trigger a safety mechanism. One measure is to switch to relaxing content, such as playing soothing music or guiding meditation, to help the patient's emotions stabilize. Simultaneously, the background system analyzes the cause of the anomaly, whether it's due to excessive task difficulty or the patient's poor condition, and adjusts subsequent plans accordingly. This mechanism ensures that rehabilitation training always remains within the patient's acceptable physiological and psychological range, preventing adverse effects from excessive challenges. It embodies the proactive correction idea in awareness bug theory: the training system acts like a coach with "metacognition," self-correcting the strategy when it realizes a problem, to protect the patient's rehabilitation motivation and safety.

[0017] Through the above technical solutions, this invention creatively combines personalized patient brain mapping with semantic feedback mechanisms, achieving personalized rehabilitation training interaction for each individual. A series of studies and practical experiences have shown that personalized BCI programs and neurofeedback can significantly improve training effectiveness; this invention theoretically and practically meets this need, and further introduces dynamic optimization at the semantic level, making it a pioneering solution both domestically and internationally. Attached Figure Description

[0018] Figure 1 A schematic diagram of the overall architecture of the individualized brain-map interaction feedback mechanism.

[0019] Figure 2 This is a diagram illustrating the visualization of neurofeedback.

[0020] Figure 3 A schematic diagram for optimizing dynamic objectives. Detailed Implementation

[0021] Example 1: Personalized Language Training for a Stroke Patient with Aphasia. A 58-year-old male stroke patient with aphasia was enrolled in the language rehabilitation program of this invention. The team first constructed a semantic brain map of the patient using fMRI data and language test results. The results showed that the patient's understanding of everyday object nouns was relatively good (normal activation of the Wernicke's area in the brain when hearing related words), but his understanding of abstract concepts and long sentences was poor. Based on this, the system developed individualized training content: initially focusing on practicing short sentences with concrete nouns, and gradually adding abstract vocabulary and complex sentences later. During the training process, the system continuously collected the patient's EEG. When the patient practiced vocabulary, DIKWP analysis showed that his EEG P300 amplitude was higher in categories such as "fruit" (indicating successful semantic recognition), while N400 frequently appeared in "emotional" vocabulary (indicating difficulty in comprehension). Based on this, the system adjusted in real time—increasing the frequency of practice for emotional vocabulary and providing additional explanatory pictures to aid understanding when these words appeared. At each stage, the system dynamically increases the difficulty based on the patient's progress. For example, the initial goal was to be able to name objects; once the patient achieved this, the system upgraded the goal to complete sentences describing the uses of the objects. The entire training was challenging but not overly difficult, and the patient maintained a high level of interest. After six weeks of testing, he showed significant improvement in all sub-items of the AAT Aphasia Test, particularly in complex sentence comprehension, which improved by 40%. This was attributed to the system's targeted training based on his individual weaknesses (abstract semantics) and the continuous, progressive optimization of the training objectives, resulting in a comprehensive improvement in his semantic abilities.

[0022] Example 2: Motor Imagery Rehabilitation for Cerebral Palsy Patients. A 16-year-old cerebral palsy patient with lower limb paralysis hoped to recover some motor function through motor imagery combined with BCI training. We used the mechanism of this invention to assist in his training. First, the patient's EEG was recorded when imagining different movements (such as lifting the leg and walking), and his motor imagery brain map was drawn using the SVM algorithm. The results showed that the patient's μ-wave inhibition when imagining the left leg was significantly weaker than that of the right leg, suggesting that the left motor pathway was more damaged. Therefore, the training program focused on strengthening the imagination of the left lower limb. During training, the patient wore a BCI headband and looked at a screen with a virtual character on it. The system asked him to imagine controlling the virtual character to step with his left leg. If the EEG characteristics reached the threshold, the character successfully took a step; otherwise, it stopped. Initially, the patient failed many times and showed a frustrated expression. The system's DIKWP analysis found that although he understood the task requirements (clear intention level), the activation level of brain signals was insufficient (poor data level), indicating that the imagination level needed to be enhanced. The system then activated visual feedback: a bar graph appeared on the side of the screen dynamically reflecting the activation intensity of his left and right motor imagery, and a red line marked the threshold for achieving the target. Based on this, the patient attempted to adjust their focus, gradually increasing the intensity of their visualization of their left leg, in a strip-like shape. Figure 1Each time the target was successfully met, the virtual human took a step, and the system immediately provided reward feedback (cheers and screen flashes). When the system detected multiple consecutive successes and a decrease in alpha waves on the brain (indicating fatigue), it temporarily lowered the threshold or switched to a simpler task for rest. As training progressed, the system continuously raised the target threshold (dynamically increasing the goal), eventually enabling the patient to consistently achieve twice the initial brain signal output. After several months of training, his clinical lower limb muscle tone and voluntary control improved, and he was able to stand briefly with the brace. This embodiment demonstrates the application value of this invention in motor rehabilitation: by using visualization and diverse feedback to help patients master the skill of "moving with their brains," and by continuously optimizing the training difficulty based on individual progress, it is significantly more effective than a static training program.

[0023] Example 3: Real-time Semantic Monitoring for Attention Rehabilitation. Some patients with cognitive impairment require training in attention and memory. This example uses a patient with mild Alzheimer's disease to demonstrate the role of this mechanism in cognitive training. The training task involves presenting a sentence on a screen with one word intentionally missing. The patient must listen attentively to the sentence and then select the missing word from a rapidly scrolling list of words. The system monitors the patient's response to the target word via EEG. If the patient produces a significant P300 wave when the target word appears and successfully selects the correct word, it indicates that they have maintained focus and remembered the sentence's semantics (training goal achieved); conversely, if the expected ERP does not appear and the wrong word is selected, it indicates insufficient attention or memory lapse. The system adjusts the difficulty based on this semantic monitoring result: when the patient selects the correct word multiple times easily and the P300 wave latency shortens, the system automatically speeds up the scrolling or increases the similarity of distracting words; while when the patient selects the wrong word repeatedly and the EEG shows increased theta waves (possibly indicating inattention), the system slows down the pace and issues a voice reminder to focus. Through this closed-loop adjustment, the patient is always challenged to the edge of their ability without collapsing. After a period of training, the system examined the changes in the patient's electroencephalogram (EEG) compared to the initial state. It was found that the P300 amplitude induced by the word loss detection task was significantly increased compared to before training, and the latency was shortened by approximately 50 milliseconds—indicating improvements in the patient's information processing speed and concentration. Functional magnetic resonance imaging (fMRI) also showed enhanced functional connectivity in language-related brain regions. Ultimately, the patient's score on the clinical attention test scale improved by 25%. Example 3 demonstrates how this invention optimizes attention training by monitoring semantic comprehension levels, resulting in rehabilitation effects reflected in both objective EEG indicators and behavioral assessments.

[0024] As can be seen from the above embodiments, the individualized brain-map interactive feedback mechanism of the present invention is applicable to various rehabilitation scenarios such as language, motor, and cognition, and can automatically adapt to the needs of different patients and different stages, significantly improving training efficiency and patient participation. Its licensability lies in the innovative combination of semantic graphs, the DIKWP model, and closed-loop feedback, breaking through the limitations of the traditional BCI training fixed mode; its feasibility for implementation is strong, requiring only existing EEG devices and ordinary computing terminals, and the software algorithm modules also have a mature foundation, possessing a low industrialization threshold and enormous application value.

Claims

1. An individualized brain-atlas interaction feedback method based on a brain-computer interface, characterized by The method comprises the following steps: a) obtaining brain activity data and cognitive assessment results of a specific patient, establishing a brain-semantic map of the patient, and associating different stimulation tasks with their brain electrical characteristics to form a semantic mapping relationship; b) during the patient's rehabilitation training, real-time acquisition of the patient's brain electrical signals, and multi-layer semantic analysis of the signals based on the DIKWP model to evaluate the patient's understanding of the current training content and the intention achievement; c) generating instant feedback according to the analysis results of step b), presenting it to the patient through vision, hearing or other forms to guide the patient to adjust the cognitive or operation strategy; d) continuously monitoring the patient's training performance and comparing it with the preset rehabilitation goal, automatically increasing the training difficulty or upgrading the goal when the patient reaches or approaches the goal, and decomposing the goal to reduce the difficulty when the patient fails to achieve the goal for a long time, forming a closed-loop dynamic goal optimization mechanism; e) repeating steps b) to d) until all rehabilitation training goals are completed. The establishment of the brain-semantic map in step a) includes presenting a series of semantic stimuli related to rehabilitation (including pictures, words, sentences or motor imagination instructions) to induce the patient's brain activity, using machine learning algorithms to map different semantic categories of stimuli with the patient's corresponding brain electrical characteristics, and storing them in the map database, wherein each semantic unit node is associated with the activation mode intensity of the patient's specific brain area or frequency band. The DIKWP model analysis in step b) includes the following aspects of the patient's brain electrical activity: data layer to obtain original performance indicators (such as accuracy, reaction time, etc.), information layer to extract neural signal features (such as evoked potentials, spectral power, etc.), knowledge layer to judge the patient's mastery of the concept involved (such as whether the semantic related potential appears), wisdom layer to infer the patient's comprehensive understanding and application ability, and intention layer to determine whether the patient's training goal is achieved, thereby obtaining a multi-dimensional semantic cognitive state description. The feedback in step c) includes visual neural feedback, which presents the patient's current brain activity characteristics in a graphical way compared with the target state; and semantic prompt feedback, which automatically provides explanations, prompts or demonstrations related to the task when detecting repeated errors in the patient's performance of the task to help the patient understand; at the same time, when detecting abnormal emotions or attention of the patient, the training is paused and soothing or reminding feedback is given.

2. The method of claim 1, wherein: The dynamic goal optimization mechanism in step d) is realized by preset rules and reinforcement learning algorithms, wherein the preset rules include: if the training performance continuously exceeds the threshold, upgrade the goal, and if the training performance continuously falls below the threshold, reduce the difficulty; the reinforcement learning algorithm adjusts the amplitude and frequency of the goal adjustment according to the patient's long-term training data to individualize the patient's progress rhythm.

3. The method of claim 1, wherein: The control calculation module is configured with a brain-semantic map database and a DIKWP semantic analysis engine to execute the above steps a) to e); the feedback output unit includes at least one of a display screen, a speaker and a vibration motor to output multi-modal feedback; the system further includes a safety monitoring sub-module to adjust or suspend the current training when detecting abnormal physiological or semantic state of the patient to ensure the safety and effectiveness of the patient.

4. The method of claim 1, wherein: ​ 5. The method of claim 1, wherein: ​ 6. A brain-computer interface rehabilitation system for implementing the method of any one of claims 1-5, comprising an electroencephalogram acquisition device, a signal processing unit, a feedback output unit and a control computing module, characterized in that: ​

Images