Rehabilitation assisting method and device suitable for cognitive disorder treatment and medium

Abstract

The invention discloses a rehabilitation assisting method and device suitable for cognitive disorder treatment and a medium, and relates to medical rehabilitation. A tablet personal computer, a millimeter wave radar and the like are included, artificial intelligence technologies such as signal processing and deep learning are utilized, and rehabilitation training of patients with mild cognitive impairment caused by brain injury and cerebral apoplexy is assisted. Basic information, medical history and current cognitive conditions of the patient are recorded and collected through the tablet computer, and a rehabilitation file is generated; based on the cognitive disorder type and severity of the patient, a rehabilitation training plan is intelligently made; a patient performs rehabilitation training through a gamified training interface, and the physiological state in the training process is monitored in real time in a non-contact mode based on millimeter wave radar; a deep learning model is used for analyzing training data of the patient, and rehabilitation progress is evaluated; training and evaluation data are summarized, and a subsequent rehabilitation scheme is optimized. Full-process automatic management from patient information collection, training plan making, training implementation, rehabilitation evaluation to data analysis report generation can be rapidly and accurately achieved, and the cognitive rehabilitation efficiency and effect are improved.

Description

Technical Field

[0001] The present invention relates to the field of medical rehabilitation technology, and in particular to a rehabilitation assistance method, device and medium based on a deep neural network suitable for the treatment of cognitive impairment in patients with mild cognitive impairment caused by brain injury or stroke. The use of artificial intelligence technologies such as signal processing and deep learning can reduce the workload of medical staff and improve the efficiency and effect of cognitive rehabilitation. Background Art

[0002] Brain injury and stroke are common neurological diseases in clinical practice, which often lead to different degrees of cognitive impairment in patients, such as distraction, memory loss, decreased calculation ability, slow thinking, impaired executive function, and visual tracking difficulties. These cognitive impairments not only seriously affect the daily quality of life of patients, but also bring huge economic and psychological burdens to their families and society.

[0003] Cognition is the intelligent processing process of the body to recognize and acquire knowledge, involving a series of voluntary, psychological and social behaviors such as learning, memory, language, thinking, spirit, emotion, etc. Cognitive impairment refers to abnormalities in the brain's high-level intelligent processing processes related to the above-mentioned learning, memory, thinking and judgment, which causes severe learning and memory disorders, accompanied by aphasia, apraxia, agnosia, or apraxia. Mild cognitive impairment can be treated through purposeful and targeted training, but this type of treatment often requires doctors to conduct long-term rehabilitation training and follow-up of the patient's condition, which increases the burden on medical care and patients.

[0004] Traditional cognitive rehabilitation methods mainly rely on manual intervention, such as manual guidance from physical therapists, speech therapists, and occupational therapists. However, this method has many limitations. Manual intervention requires a lot of time and human resources, and it is difficult to ensure that every patient can receive continuous and personalized rehabilitation services. Traditional methods lack a scientific evaluation system and cannot accurately judge the patient's rehabilitation progress and treatment effect. In addition, due to the different degrees and types of cognitive impairment in patients, traditional methods often fail to meet the personalized needs of patients, resulting in poor rehabilitation effects.

[0005] With the rapid development of artificial intelligence technology, especially the application of deep neural networks in the medical field, new possibilities have been provided for cognitive rehabilitation. Deep neural networks have powerful data processing and pattern recognition capabilities, and can automatically extract useful information from large amounts of data and use it to guide rehabilitation training and effect evaluation. In addition, deep neural networks can also dynamically adjust training content and difficulty based on real-time feedback and physiological data from patients, thus achieving personalized rehabilitation services. Summary of the invention

[0006] The purpose of the present invention is to provide a rehabilitation auxiliary method, device and medium suitable for the treatment of cognitive impairment. It integrates advanced signal processing technology, deep learning algorithm and user-friendly interactive interface to achieve accurate classification of patients, personalized rehabilitation training, dynamic difficulty adjustment and rehabilitation effect evaluation.

[0007] The present invention provides a cognitive rehabilitation auxiliary method based on a deep neural network, comprising the following steps:

[0008] 1) User information input and classification: a) The user inputs information data through the touch screen or external input device, and the system generates a rehabilitation file after standardizing the data and saves it in the back-end storage module; b) Based on the decision tree algorithm, the patient data is analyzed, and the best matching cognitive rehabilitation training path is selected from the preset program library through multi-dimensional features such as evaluation scale scores, medical information, personal characteristics and physical condition;

[0009] 2) Personalized rehabilitation training: a) Rehabilitation training provides six modules including attention training, memory training, calculation ability training, thinking ability training, executive ability training and eye tracking training; according to the patient's classification results, game tasks of different difficulty levels and numbers are intelligently extracted from each module to form a personalized training plan; b) During the training process, the system dynamically adjusts the difficulty and content of the game through interactive scoring calculation methods and reinforcement learning frameworks, combined with physiological data such as heart rate and breathing collected in real time by millimeter-wave radar, to ensure that the training is both challenging and not too difficult, thereby stimulating the enthusiasm of patients and promoting the rehabilitation process;

[0010] 3) Evaluation and visualization of rehabilitation effects: a) After the training, the system uses data mining techniques such as clustering algorithms and regression analysis to deeply analyze the patient's training data and identify potential rehabilitation patterns and trends; b) Visual tools are used to display the rehabilitation progress, helping doctors and nursing staff to intuitively understand the patient's rehabilitation status and provide a basis for formulating and adjusting subsequent rehabilitation plans.

[0011] In step 1), user information entry and classification includes the following sub-steps:

[0012] (1) The front-end interactive interface collects the user's personal information, medical information, cognitive status and other raw data through a touch screen or external input device, and stores it as structured fields;

[0013] (2) Clean and standardize the collected data to ensure data consistency and integrity. Based on the pre-built five-layer decision tree algorithm model, a multi-level conditional judgment structure is used to achieve intelligent matching of user data with the preset rehabilitation program path;

[0014] (3) The generated personalized rehabilitation plan is combined with the user's personal profile and stored in the database, which includes three main parts: user profile, rehabilitation plan and training record, for easy subsequent reference and tracking.

[0015] In step 2), personalized rehabilitation training includes the following sub-steps:

[0016] (1) Six training modules were defined, each of which contained multiple self-written game tasks designed to train patients’ specific cognitive abilities; the specific contents were as follows:

[0017] Attention training module: including target capture, flickering object recognition, target tracking, and reaction time challenge;

[0018] Memory training module: including pattern matching, number sequence memory, vocabulary association, and memory grid;

[0019] Computational ability training module: including digital filling, fast calculation, chain calculation, and life application;

[0020] Thinking ability training module: including logical reasoning, graphic reasoning, memory maze, and strategy sandbox;

[0021] Execution capability training module: including task sequencing, route planning, command and dispatch, and response challenges;

[0022] Eye tracking training module: including target tracking, visual search, target switching, and dynamic recognition;

[0023] (2) According to the preset difficulty level and quantity requirements of the user's corresponding rehabilitation plan, game tasks matching the user's current plan are extracted from different game modules to form a rehabilitation game queue, which is presented to the user in order;

[0024] (3) The system calls the backend game engine module to render the game scene in real time, and displays it through the front-end user interface, supporting interactive methods such as touch screen operation and external device input;

[0025] (4) During the game, the user's training data is recorded from three aspects and stored as training records:

[0026] Game operation: The system records the patient's key operations, including operation type (sliding track, click location), timestamp, and operation result;

[0027] Game score: The final comprehensive score is calculated based on the total number of training items participating in the comprehensive scoring, the standardized score of the items (ranging from 0 to 100), and the weight;

[0028] Millimeter wave radar: monitors and records the user's heart rate and breathing data sequence;

[0029] (5) Based on user training data, the long short-term memory network (LSTM) and reinforcement learning framework (Deep Q-Network) are used to optimize the game feedback mechanism and adjust the game difficulty to maximize patient engagement and training effect.

[0030] In step 3), rehabilitation effect evaluation and visualization includes the following sub-steps:

[0031] (1) Quantify, compare, and summarize the end-user training data; including the following processing methods:

[0032] Mean: Quantify the average level of each indicator of the user to obtain the user's current recovery status;

[0033] Variance: quantify the degree of dispersion of each user's indicators and obtain the stability of the user's recovery status;

[0034] Clustering algorithm (K-means): cluster the patient's game data and divide the patients into different rehabilitation stages or cognitive level groups;

[0035] Logistic regression: explore the relationship between the patient's recovery process and various indicators, and predict the patient's future recovery trend;

[0036] (2) The obtained data is visualized on the front end so that doctors can better understand the user's cognitive level and problems and make targeted training, adjustments and treatments.

[0037] The present invention provides a cognitive rehabilitation auxiliary device based on a deep neural network, comprising:

[0038] A tablet computer: equipped with a high-performance processor and large-capacity memory, loaded with a cognitive rehabilitation assistance system, and is the core interactive platform of the entire device; it supports complex deep learning algorithms and data processing tasks;

[0039] A millimeter-wave radar component: used to monitor the user's physiological indicators in real time, especially heart rate and breathing rate; these physiological signals will be input into the deep neural network model as an important basis for adjusting the content and difficulty of cognitive rehabilitation;

[0040] An eye-tracking component that tracks the user’s gaze direction and eye movement patterns to assess attention and visual cognition;

[0041] The cognitive rehabilitation auxiliary system includes the following modules:

[0042] A data receiving module, used to receive the user's heart rate and breathing data monitored by the millimeter wave radar device; the module includes a signal preprocessing unit, which is used to filter out noise and interference, and perform preliminary analysis on the data and transmit it to the back-end module;

[0043] A front-end module for uploading basic data such as the patient's basic information, medical history, and current cognitive status, and displaying the game interface for rehabilitation training, using a responsive design to interact and provide feedback based on user input;

[0044] A back-end module, consisting of a storage module, a function selection module, and a game engine module; the storage module is responsible for receiving and saving the data to be processed, the function selection module is used to call the corresponding data processing module according to the user's situation, and the game engine module is responsible for generating and managing the virtual scenes and interactive content of the rehabilitation training game;

[0045] a data processing module, including a rehabilitation training submodule, a game interaction submodule, a user monitoring submodule and a data analysis submodule;

[0046] The rehabilitation training submodule generates a rehabilitation file based on the patient's basic information, medical history data, and current cognitive status;

[0047] The game interaction submodule is responsible for calculating the interactive scores in the game and adjusting the game difficulty and content in real time based on the patient's operation records and reaction speed;

[0048] The user monitoring submodule analyzes changes in the user's heart rate, breathing and other vital signs, and dynamically adjusts the patient's rehabilitation plan;

[0049] The data analysis submodule conducts in-depth analysis of the patient's historical data, optimizes subsequent rehabilitation plans, and helps doctors and caregivers better understand the patient's rehabilitation process in a visual way.

[0050] Furthermore, the front-end module is also used to receive the processing results sent by the back-end module and display them on the interface.

[0051] Furthermore, the rehabilitation training submodule uses a decision tree to classify patients and assigns them to different rehabilitation training paths according to their basic information and medical history.

[0052] Furthermore, the game interaction submodule is responsible for calculating the interaction score in the game, using a comprehensive score function and a reinforcement learning framework to optimize the game feedback mechanism, and adjusting the game difficulty and content based on the patient's operation record and reaction speed to improve the patient's participation and training effect.

[0053] Furthermore, the cognitive rehabilitation auxiliary equipment includes six modules: attention training, memory training, calculation ability training, thinking ability training, execution ability training, and eye tracking training, which carry out different types of rehabilitation training respectively and have an evaluation mechanism for rehabilitation assessment.

[0054] Furthermore, the user monitoring submodule uses a long short-term memory network (LSTM) to analyze data changes and dynamically adjust the patient's rehabilitation plan; this sub-algorithm captures long-term dependencies and is suitable for processing long-term vital sign data.

[0055] Furthermore, the data analysis submodule uses clustering algorithms, regression analysis and other methods to explore potential patterns and trends in the data and optimize subsequent rehabilitation plans; this submodule also provides data visualization tools to help doctors and caregivers better understand the patient's rehabilitation process.

[0056] The present invention provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the cognitive rehabilitation assisting method as described above is implemented.

[0057] When the computer program is loaded into a computer and executed by the computer, it enables the computer to perform the following steps:

[0058] (A) Receive basic patient information, medical history and current cognitive status input by the user and generate a personalized rehabilitation file;

[0059] (B) Use decision trees to classify patients and assign them the most suitable rehabilitation training path;

[0060] (C) Based on the classification results, game tasks are extracted from multiple training modules to form a personalized training plan;

[0061] (D) Dynamically adjust the game difficulty and content during training;

[0062] (E) After training, the patient’s training data is analyzed in depth and the rehabilitation progress is displayed through visualization tools.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] 1. The present invention provides a cognitive rehabilitation auxiliary device based on a deep neural network. The device uses artificial intelligence technologies such as signal processing and deep learning to train patients with mild cognitive impairment caused by brain injury and stroke for rehabilitation and evaluate its effect; it can effectively reduce the workload of medical staff and improve the efficiency and effect of cognitive rehabilitation.

[0065] 2. The device system of the present invention adopts a front-end and back-end separation architecture, with Vue used on the front end, Spring Boot used on the back end, and Flask used on the algorithm end. Users interact and give feedback through the front-end input, the back-end stores data and calls algorithms for processing, and the processing results are returned to the front end, which has the characteristics of high efficiency, flexibility, scalability, and ease of use.

[0066] 3. The present invention uses a decision tree to classify patient data and assigns them to different rehabilitation training paths according to their basic information and medical history. The classification method of the present invention has the advantages of high accuracy, low latency, high degree of automation, and strong adaptability.

[0067] 4. The present invention uses a reinforcement learning algorithm to optimize the game feedback mechanism based on the user's operation records and reaction speed. Different evaluation methods are set for different game modes, and the game difficulty and content are adjusted according to user feedback. The game interaction method of the present invention has the advantages of high personalization, good effect, and improved user experience.

[0068] 5. The present invention analyzes the changes in the user's vital sign data based on the long short-term memory network and adjusts the rehabilitation plan. This method has the advantages of high precision, high robustness, and low computational complexity.

[0069] 6. The present invention analyzes and visualizes the data obtained from rehabilitation training so that doctors can better understand the patient's rehabilitation status and problems and make targeted training, adjustments and rehabilitation. The data analysis and visualization display method of the present invention has the advantages of being intuitive, clear and easy to understand. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is the overall architecture diagram of the present invention;

[0071] Figure 2 It is a flow chart of the overall module in the present invention;

[0072] Figure 3 is a flow chart of the rehabilitation training submodule in the present invention;

[0073] Figure 4 is a schematic diagram of the game evaluation mechanism in the present invention;

[0074] Figure 5 is a schematic diagram of the game interaction process in the present invention;

[0075] Figure 6 It is a schematic diagram of the data analysis and visualization method of the present invention. DETAILED DESCRIPTION

[0076] In order to make the technical solution of the present invention easy to understand and achieve the purpose.

[0077] The present invention provides a cognitive rehabilitation auxiliary device based on deep neural network. The device uses artificial intelligence technologies such as signal processing and deep learning to train patients with mild cognitive impairment caused by brain injury and stroke for rehabilitation and evaluate its effect. The system adopts a front-end and back-end separated architecture. The front-end mainly uses the Vue framework to provide an intuitive and easy-to-use interactive experience. The back-end mainly uses the Spring Boot framework, which is responsible for data processing, algorithm calling and storage management. The algorithm end uses the Flask framework to deploy deep learning models to achieve efficient data analysis and processing. Users interact and give feedback through front-end input, and the back-end stores data and calls algorithms for processing, and the processing results are returned to the front-end.

[0078] A cognitive rehabilitation assistive device based on a deep neural network, the hardware device includes the following components:

[0079] A tablet computer equipped with a high-performance processor and large-capacity memory, loaded with a cognitive rehabilitation assistance system, is the core interactive platform of the entire device, supporting complex deep learning algorithms and data processing tasks;

[0080] A millimeter wave radar component is used to monitor the user's physiological indicators in real time, especially heart rate and breathing rate. These physiological signals will be input into the deep neural network model as an important basis for adjusting the content and difficulty of cognitive rehabilitation.

[0081] An eye-tracking component that tracks the user's gaze direction and eye movement patterns to assess attention and visual cognition.

[0082] The cognitive rehabilitation auxiliary system includes the following modules:

[0083] A data receiving module is used to receive the user's heart rate and breathing data monitored by the millimeter wave radar device. The module includes a signal preprocessing unit for filtering out noise and interference, and performing preliminary analysis on the data before transmitting it to the back-end module.

[0084] A front-end module for uploading basic data such as the patient's basic information, medical history, and current cognitive status, and displaying the game interface for rehabilitation training, using a responsive design to interact and provide feedback based on user input;

[0085] A backend module, consisting of a storage module, a function selection module, and a game engine module. The storage module is responsible for receiving and saving the data to be processed, the function selection module is used to call the corresponding data processing module according to the user's situation, and the game engine module is responsible for generating and managing the virtual scenes and interactive content of the rehabilitation training game.

[0086] a data processing module, including a rehabilitation training submodule, a game interaction submodule, a user monitoring submodule and a data analysis submodule;

[0087] The rehabilitation training submodule generates a rehabilitation file based on the patient's basic information, medical history data, and current cognitive status;

[0088] The game interaction submodule is responsible for calculating the interactive scores in the game and adjusting the game difficulty and content in real time based on the patient's operation records and reaction speed;

[0089] The user monitoring submodule analyzes changes in the user's heart rate, breathing and other vital signs, and dynamically adjusts the patient's rehabilitation plan;

[0090] The data analysis submodule conducts in-depth analysis of the patient's historical data, optimizes subsequent rehabilitation plans, and helps doctors and nurses better understand the patient's rehabilitation process in a visual way;

[0091] The front-end module is also used to receive the processing results sent by the back-end module and display them on the interactive interface.

[0092] The rehabilitation training submodule uses a decision tree neural network to classify patients and assign them to different rehabilitation training paths based on their basic information and medical history.

[0093] The game interaction submodule is responsible for calculating the interactive score in the game, using a comprehensive score function and a reinforcement learning framework to optimize the game feedback mechanism, and adjusting the game difficulty and content based on the patient's operation records and reaction speed to improve the patient's participation and training effect.

[0094] The cognitive rehabilitation auxiliary equipment includes six modules: attention training, memory training, calculation ability training, thinking ability training, executive ability training, and eye tracking training, which carry out different types of rehabilitation training. The final comprehensive score is calculated based on the total number of training items participating in the comprehensive score, the standardized score of the training items (calculated based on the game interaction behavior), and the weight of the training items (determined by the preset rehabilitation training path), and the standardized range is between 0 and 100.

[0095] The user monitoring submodule uses a long short-term memory network (LSTM) to analyze data changes and dynamically adjust the patient's rehabilitation plan. This sub-algorithm captures long-term dependencies and is suitable for processing long-term vital sign data.

[0096] The data analysis submodule uses clustering algorithms, regression analysis and other methods to mine potential patterns and trends in the data and optimize subsequent rehabilitation plans. This submodule also provides data visualization tools to help doctors and nursing staff better understand the patient's rehabilitation process.

[0097] The following will further describe the use process of the present invention:

[0098] like Figure 1 As shown in the figure, users can create personal accounts through the front-end module and upload basic data such as the patient's basic information, medical history, and current cognitive status. Personal information includes but is not limited to gender, year of birth, BMI, blood pressure, occupation, years of education, and handedness; medical information covers the course of disease, cause, past medical history, medication in the past two weeks, smoking, and drinking. Cognitive status is composed of scores of CRS-R (Rehabilitation Assessment Scale), GCS (Glasgow Coma Score), DRS (Disability Assessment Scale), and RLA (Functional Independence Assessment Scale). User input data can also be obtained by interacting with the game interface of the front-end rehabilitation training. The front-end module sends the user data to the back-end module. After receiving the user data, the back-end module calls the corresponding data processing module for processing.

[0099] Figure 2 The schematic diagram of the flow of the overall module of the system is shown. In this embodiment, the data processing module called by the back-end module includes a rehabilitation training submodule, a game interaction submodule, a user monitoring submodule and a data analysis submodule.

[0100] like Figure 3 As shown, after receiving the basic data such as the patient's basic information, medical history and current cognitive status, the rehabilitation training submodule preprocesses the data, cleans up missing values ​​and outliers, and uses One-Hot Encoding to convert the categorical data into numerical form. The decision tree algorithm is used to classify patients and assign them to different rehabilitation training paths based on their basic information and medical history. The decision tree D is a hierarchical decision model consisting of nodes and directed edges, which can be formally defined as follows:

[0101]

[0102] in, Represents the node set of the decision tree, including the non-leaf node set N and the leaf node set L, that is, V = N ∪L, E represents the directed edge set between nodes, and R represents the decision rule set.

[0103] The decision tree model receives standardized user profile data and selects branches at each judgment layer node based on the preset clinical indicator threshold, including:

[0104] Root node layer: preliminary classification based on standardized assessment scale scores to determine the user's baseline level of cognitive function;

[0105] The second judgment layer: segmentation based on the user's basic medical information to determine the safety boundaries and taboos of training;

[0106] The third judgment layer: stratify users according to their personal characteristics and adjust the presentation method and difficulty baseline of training content;

[0107] The fourth judgment layer: determines the rhythm and intensity of training based on the user's physical condition indicators;

[0108] Fifth judgment layer (leaf node layer): comprehensively consider all the above parameters and select the most suitable cognitive rehabilitation training program template.

[0109] According to the classification results, the user's rehabilitation training plan is assigned. The rehabilitation training plan includes six modules of training: attention training, memory training, calculation ability training, thinking ability training, executive ability training, and eye tracking training. Each module contains 4 game tasks.

[0110] The definition and task objectives of each training module can refer to the following design, as long as the relevant functions can be achieved:

[0111] (1) Attention training: Improve the patient's ability to concentrate and the duration of attention, including target capture, flickering object recognition, target tracking, and reaction time challenges. The evaluation mechanism includes accuracy, reaction time, retention, success rate and other factors.

[0112] Target Capture: Multiple geometric shapes of different colors move randomly on the screen. Only specific targets need to be clicked by the user, and other interfering targets need to be ignored. Within a limited time, the user must click or select a target object of a specific color or shape on the screen. As the difficulty increases, the speed of the geometric shapes will gradually increase. The purpose of this game training is to improve the user's selective attention and help them quickly identify and select target objects in complex environments.

[0113] Flashing object recognition: Before starting the game, a specific picture will be randomly selected from the picture library and displayed to the user for 2 seconds. After starting the game, multiple different pictures will appear on the screen in turn at random, and the picture sequence includes the specific picture. Within a limited time, the user clicks the button at the bottom of the screen when the specific picture appears randomly for the first time. As the difficulty increases, the number of specific pictures that need to be recognized will increase. The purpose of this game training is to improve the user's concentration and help them maintain their focus on a specific task for a long time.

[0114] Target tracking: The game will specify a specific moving target, which will move along the maze route and leave at a certain exit. The user needs to continue to focus on and track the object on the screen, while ignoring the surrounding interference targets and pointing out the final direction of the object's departure. As the difficulty increases, the number of moving targets that need to be paid attention to will increase. This training is designed to train users' ability to maintain attention in a dynamic environment.

[0115] Reaction time challenge: The game will play a rhythmic music. At the rhythmic drum beat, the music will randomly pause. The user needs to press the button at the bottom of the screen as soon as the music pauses. As the difficulty increases, the rhythm of the drum beat pause will speed up. The training purpose of this game is to enhance the user's attention to sound signals and coordination ability.

[0116] (2) Memory training: Enhance the patient's short-term and long-term memory abilities, including games such as graphic matching, number sequence memory, vocabulary association, and memory grid. The evaluation mechanism of this type of game includes factors such as accuracy, reaction time, error type, and difficulty adaptability.

[0117] Graphics matching: After starting the game, a group of graphics (such as geometric figures, common objects, etc.) will be displayed on the screen for a few seconds before hiding. Then another group of graphics with a larger number will be presented, requiring the user to quickly identify which graphics are consistent with the previously displayed ones. As the game progresses, the number of graphics and the difficulty can be gradually increased. The purpose of this game training is to enhance the user's visual short-term memory ability.

[0118] Number sequence memory: After starting the game, flashing numbers will appear on the screen. Each number will pause for 1 second and disappear after the flashing. The user needs to select all the numbers in order after all the numbers are displayed. As the difficulty changes, the number of numbers to be remembered and the flashing time also change. The purpose of this game training is to improve the user's short-term memory and working memory ability.

[0119] Vocabulary Association: Vocabulary Display: At the beginning of the game, a word will be randomly displayed on the screen, such as "apple". The user needs to type or voice input the word that comes to mind within the specified time, such as "fruit", "red", "sweet", etc. As the difficulty changes, the number of correct words that need to be submitted to pass the level and the input time will also change. The game is designed to enhance the user's semantic memory and vocabulary association ability.

[0120] Memory Grid Game: In a grid-like grid, some patterns will be randomly displayed. After a short display, the patient is asked to recall the shape of the pattern and mark the location where it appeared before. As the game progresses, the number of squares will be increased, interference items will be added, etc., aiming to enhance the patient's visual-spatial memory ability.

[0121] (3) Calculation ability training: Improve the patient's basic arithmetic and logical reasoning abilities, including games such as number filling, rapid calculation, chain calculation, and life application. The evaluation mechanism of this type of game includes factors such as accuracy, reaction time, error type, and fatigue analysis.

[0122] Number Fill-in-the-Blank: In this game, users will see a series of equations with some numbers replaced by blanks. For example, "5 + __ = 10" is displayed. Users need to fill in the correct numbers according to the equation. After each question is displayed, users have 15 seconds to answer. As the difficulty changes, multi-step calculations or complex content such as fractions and decimals will be introduced, aiming to improve users' basic arithmetic skills and number sense.

[0123] Quick calculation: Users will see a series of simple math problems (such as addition, subtraction, multiplication and division) on the screen. Each problem is displayed for 2 seconds, and users need to quickly select the correct answer before the time is up. The system will record the user's accuracy and reaction time. As the difficulty changes, multi-step calculations or complex content such as fractions and decimals will be introduced to improve the user's calculation speed and accuracy.

[0124] Chain calculation: The game provides a series of interrelated equations, and the user needs to solve the next equation immediately after completing it. For example, "3 + 5 = 8, then 8 × 2 = __". The user must maintain the consistency of the calculation in each step, and the time limit will decrease as the difficulty increases. Chain calculation games are designed to improve users' comprehensive calculation ability.

[0125] Life Application: In the game, users will encounter math problems related to daily life, such as calculating the total cost of shopping, the change amount, or budget allocation. The game will give actual scenarios, such as "If you buy 3 apples at a price of 5 yuan each, and pay 20 yuan, how much change will you get?" You need to make calculations based on the situation. As the difficulty increases, the complexity of the application will increase, aiming to enhance users' ability to apply mathematics to real life.

[0126] (4) Thinking ability training: Promote patients' creative thinking and problem-solving abilities, including games such as logical reasoning, graphic reasoning, memory mazes, and strategy sandboxes. The evaluation mechanism of this type of game includes factors such as accuracy, reaction time, and error type.

[0127] Logical reasoning: Users will face a series of situation-based logical questions, which may involve reasoning, classification, relationship building, etc. For example, "In a room, there are three people: Alice, Bob, and Charlie. Alice is on Bob's left and Charlie is on Bob's right. Who is in the middle?" Users need to deduce the answer based on the given information. As the difficulty increases, time limits will be introduced, requiring users to complete more questions in a shorter time. It aims to improve users' analytical thinking and reasoning abilities.

[0128] Graphic reasoning: In the game, users will see a series of graphic sequences or patterns, and users need to predict the next graphic based on the rules. For example, given a set of graphic sequences (such as: circle, triangle, circle, triangle), users need to choose whether the next graphic is a circle or a triangle. As the difficulty increases, the complexity of the change rules increases, such as comprehensive changes in color, size, and shape. It aims to improve users' spatial imagination and pattern recognition ability.

[0129] Memory Maze Game: Users will explore a maze, each node has a different sign, and users need to remember the signs they have passed. When users reach the end, the system will ask users to recall and choose the correct path sign. As the difficulty increases, the complexity of the maze will be gradually increased, and more branches and paths will be added. It aims to improve users' spatial cognition and navigation ability in complex environments.

[0130] Strategy sandbox game: Users will develop strategies to deal with different challenges in a given virtual scenario. For example, develop development plans, allocate resources and solve problems with limited resources. As the difficulty increases, multiple variables and complex situations are gradually introduced to increase the complexity of decision-making. It aims to enhance users' creative thinking and decision-making abilities.

[0131] (5) Executive ability training: Enhance the patient's ability to plan, organize and execute tasks, including games such as task sequencing, path planning, command and dispatch, and reaction challenges. The evaluation mechanism of this type of game includes factors such as completion rate, accuracy, and completion time.

[0132] Task sorting: Users will see a set of events in a disrupted order and need to sort these tasks correctly based on logical relationships or chronological order. For example, users may need to sort the tasks of "getting up", "eating breakfast", and "going to work" in the correct order. As the difficulty increases, interference information is introduced, such as adding irrelevant tasks or time limits. It aims to improve users' logical thinking ability and overall task management ability, and enhance the efficiency of performing various activities in daily life.

[0133] Path planning: Users need to plan an optimal path on a grid-shaped map, avoiding obstacles and minimizing the number of steps. As the difficulty increases, multiple restrictions are provided, such as "must pass through a specific node" or "arrive within a time limit." It aims to enhance users' logical thinking and strategic planning abilities, helping them better deal with complex situations in real life.

[0134] Command and dispatch: Users will be responsible for coordinating the execution of multiple tasks, and need to reasonably allocate resources and arrange tasks according to the priority and time limit of the tasks. For example, users may need to dispatch multiple vehicles to complete different delivery tasks in the city. As the difficulty increases, the number and complexity of tasks will increase, requiring users to coordinate multiple tasks at the same time. It aims to improve users' execution, organizational and adaptability.

[0135] Reaction Challenge: After the game starts, different colored texts (such as red, blue, green, and yellow) will appear randomly on the screen from time to time, but the color of the text is different from the description itself. Each pattern is displayed for 1-2 seconds, and the user needs to choose the correct color of the text within the time limit. As the difficulty increases, interference information is introduced, such as displaying other wrong color texts at the same time when the correct stimulus appears. The main purpose of the training is to enhance the user's execution ability and immediate reaction ability.

[0136] (6) Eye tracking training: Improve the patient's eye tracking and visual processing abilities, including games such as target tracking, visual search, target switching, and dynamic recognition. The evaluation mechanism of this type of game includes factors such as accuracy and stability.

[0137] Target tracking: After starting the game, users need to track the movement of the target on the screen through eye movement. The target will move on a random trajectory, and users need to keep their eyes on the target as much as possible. As the game progresses, the target's movement speed will gradually increase, making tracking more difficult. It is designed to help users improve their ability to quickly and accurately track targets in a dynamic environment, thereby improving their visual processing ability in daily life.

[0138] Visual Search: After starting the game, a cluttered image will be displayed on the screen, and the user needs to quickly find a specific target in the image, such as an object of a specific shape, color, or pattern. As the difficulty increases, the number of targets to be found gradually increases. The goal of the visual search game is to train the flexibility and quick response of eye movements, and enhance the user's ability to quickly identify and process information in real life.

[0139] Target switching: Several different graphics will be displayed on the screen at the same time, and the user needs to quickly adjust their gaze according to the current instructions (such as color or shape) and switch from one target to another. As the difficulty increases, the switching frequency increases, so that the user needs to adapt to the target changes more quickly. It aims to improve the user's eye movement flexibility and switching ability.

[0140] Dynamic recognition: At the beginning of the game, the user will see a target (such as an animal, object, etc.), and then the target will randomly swap positions with other similar targets and move. The user needs to identify the position of the target as accurately as possible during the change process. As the difficulty increases, the frequency of target changes increases, forcing the user to make quick judgments and identifications. The purpose is to enhance the user's visual recognition ability and dynamic information processing ability.

[0141] like Figure 4 As shown in the figure, the classification results quantify the demand intensity of each module into different levels (such as high, medium, and low), and the system assigns different numbers and levels of training tasks based on user classification. The comprehensive formula for game scores is as follows:

[0142]

[0143] Among them, n is the total number of training items involved in the comprehensive scoring is the standardized score of the ith training item (ranging from 0 to 100). The weight of each training item changes according to the intensity of the demand for each module. It is the sum of all item weights, ensuring that the final composite score ranges between 0 and 100.

[0144] The overall game interaction process is as follows Figure 5 As shown in the figure, the user interacts with the game through the front end, and the system records every operation of the patient, including the operation type, timestamp, operation result, etc., and passes it to the game interaction submodule. The game interaction submodule is responsible for calculating the interaction score in the game.

[0145] Based on the patient's engagement and progress, a reinforcement learning framework (Deep Q-Network) is used to optimize the game feedback mechanism. The model can learn how to adjust the game difficulty in different situations to maximize patient engagement and training results. The reinforcement learning framework (Deep Q-Network) is a reinforcement learning algorithm that combines deep learning and Q-learning, and is mainly used to solve high-dimensional state space problems. Q-learning is a model-free reinforcement learning algorithm that finds the optimal strategy by learning the value (Q value) of the state-action pair. The Q-learning update formula is:

[0146]

[0147] in, and are the current state and the current action respectively. is the next state, is the current reward. is the learning rate, is the discount factor, which indicates the importance of future rewards.

[0148] DQN uses neural networks to approximate the Q function, solving the Q table explosion problem encountered by traditional Q-learning in high-dimensional state space.

[0149] During the user's game, the data receiving module receives the user's heart rate and breathing data monitored by the millimeter wave radar device, filters out noise and interference from the data, and transmits it to the user monitoring submodule. The user monitoring submodule uses a long short-term memory network (LSTM) to analyze data changes and dynamically adjust the patient's rehabilitation plan. This sub-algorithm captures long-term dependencies and is suitable for processing long-term vital sign data.

[0150] Consider factors such as reaction speed, operation accuracy and game progress. The following basic formula can be used:

[0151]

[0152] in, Represents the factors based on heart rate data, obtained by the LSTM network analysis of the user monitoring submodule. Represents the factors based on breathing data, also analyzed by LSTM network. It is the comprehensive score of the user in the current round of the game. , , It is the weight of each factor, and the specific value is adjusted according to actual conditions and needs.

[0153] Based on the calculated interaction score, a dynamic adjustment strategy is adopted. If the patient's score is high, the number of high-difficulty games is increased in the training task to reduce the demand intensity. Conversely, the number of low-difficulty games is increased to increase the demand intensity.

[0154] After the planned rehabilitation training game is completed, the data analysis submodule evaluates the obtained data and gives the patient's cognitive level and rehabilitation progress. Figure 6 As shown, the data analysis submodule quantifies, compares, and summarizes the patient's game scores, and can use the following statistical methods:

[0155] 1. Use the mean to calculate the average level of the patient's accuracy, reaction time and other indicators. represents the number of samples, Represents the value of the i-th sample.

[0156]

[0157] 2. Use variance to calculate the degree of dispersion of the patient's accuracy, reaction time and other indicators, reflecting the stability and consistency of the patient's recovery status. represents the number of samples, Represents the value of the i-th sample.

[0158]

[0159] The data analysis submodule uses clustering algorithms, regression analysis and other methods to explore potential patterns and trends in the data and optimize subsequent rehabilitation plans. The clustering algorithm (K-means) is used to cluster the patient's game data and divide the patients into different rehabilitation stages or cognitive level groups. Logistic regression is used to explore the relationship between the patient's rehabilitation process and various indicators to predict the patient's future rehabilitation trend.

[0160] Cluster analysis is an unsupervised learning method used to group data (i.e. cluster) so that data points in the same group are more similar in some sense, while data points between different groups are more different. K-means clustering is one of the most commonly used clustering algorithms. Its goal is to divide data points into K clusters so that the similarities between data points within a cluster are maximized and the differences between clusters are minimized. The K value (the number of clusters) can be determined by the Elbow Method. The goal of K-means clustering is to minimize the sum of the distances between the data points in each cluster and the center of the cluster. The objective function (also called the cost function) can be expressed as follows:

[0161]

[0162] Where: K is the number of clusters, is the kth cluster, is the ith data point, is the centroid (center point) of the kth cluster. It is a data point To cluster center The Euclidean distance of .

[0163] Regression analysis is a supervised learning method used to explore and model the relationship between variables, especially the impact of independent variables (inputs) on dependent variables (outputs). Logistic regression is often used for binary or multi-classification problems, such as predicting whether a patient has reached a certain rehabilitation standard. The formula is:

[0164]

[0165] in, is the probability of an event occurring.

[0166] According to the results of clustering algorithm and regression analysis, a more targeted rehabilitation training plan can be designed. For example, the training difficulty can be increased for patients who recover faster, and the difficulty can be appropriately reduced or the training content can be adjusted for patients who recover slower.

[0167] This submodule also performs data visualization, presenting the data in the form of graphs, tables, curves, etc., to help doctors and nursing staff better understand the patient's recovery process. The data module returns the processing results to the back-end module. The back-end module sends the processing results to the front-end module. After receiving the processing results, the front-end module displays them on the interface and provides various functional options, such as saving, sharing, printing, etc.

[0168] The present invention can quickly and accurately realize the full process automated management from patient information collection, personalized training plan formulation, training implementation, rehabilitation assessment to data analysis report generation, reduce the workload of medical staff, and improve the efficiency and effect of cognitive rehabilitation.

[0169] The above embodiments are only preferred embodiments of the present invention and cannot be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A rehabilitation-assisted method for the treatment of cognitive impairment, characterized in that The following steps are involved: 1) User information input and classification: a) The user inputs the patient's personal information, medical information and cognitive status assessment scale score through the touch screen or external input device. The system generates a rehabilitation file after standardizing the data and saves it in the back-end storage module; b) The patient data is analyzed based on a five-layer decision tree algorithm. Through the multi-dimensional characteristics of the assessment scale score, medical information, personal characteristics and physical condition, the best matching cognitive rehabilitation training path is selected from the preset program library; 2) Personalized rehabilitation training: a) Rehabilitation training provides six modules including attention training, memory training, calculation ability training, thinking ability training, executive ability training and eye tracking training; According to the patient's classification results, game tasks of different difficulty levels and numbers are intelligently extracted from each module to form a personalized training plan; b) During the training process, the system uses an interactive scoring method and reinforcement learning framework, combined with heart rate and respiratory physiological data collected in real time by millimeter-wave radar, to dynamically adjust the difficulty and content of the game to ensure that the training is challenging but not too difficult, thereby stimulating the patient's enthusiasm and promoting the rehabilitation process; 3) Evaluation and visualization of rehabilitation effects: a) After the training, the system uses clustering algorithms and regression analysis to deeply analyze the patient's training data and identify potential rehabilitation patterns and trends; b) Visual tools are used to display the progress of rehabilitation, helping doctors and nursing staff to intuitively understand the patient's rehabilitation status and provide a basis for formulating and adjusting subsequent rehabilitation plans.

2. A cognitive rehabilitation assistive device based on deep neural network, characterized in that include: A tablet computer: equipped with a high-performance processor and large-capacity memory, loaded with a cognitive rehabilitation assistance system, and is the core interactive platform of the entire device; Support complex deep learning algorithms and data processing tasks; A millimeter wave radar component: used to monitor the user's physiological indicators in real time, especially heart rate and breathing rate; these physiological signals will be input into the deep neural network model as an important basis for adjusting the content and difficulty of cognitive rehabilitation; An eye-tracking component that tracks the user’s gaze direction and eye movement patterns to assess attention and visual cognition; The cognitive rehabilitation auxiliary system includes the following modules: A data receiving module, used to receive the user's heart rate and breathing data monitored by the millimeter wave radar device; the module includes a signal preprocessing unit, which is used to filter out noise and interference, and perform preliminary analysis on the data and transmit it to the back-end module; A front-end module for uploading the patient's basic information, medical history, and current cognitive status data, and displaying the game interface for rehabilitation training, using a responsive design to interact and provide feedback based on user input; A back-end module, consisting of a storage module, a function selection module, and a game engine module; the storage module is responsible for receiving and saving the data to be processed, the function selection module is used to call the corresponding data processing module according to the user's situation, and the game engine module is responsible for generating and managing the virtual scenes and interactive content of the rehabilitation training game; a data processing module, including a rehabilitation training submodule, a game interaction submodule, a user monitoring submodule and a data analysis submodule; The rehabilitation training submodule generates a rehabilitation file based on the patient's basic information, medical history data, and current cognitive status; The game interaction submodule is responsible for calculating the interactive scores in the game and adjusting the game difficulty and content in real time based on the patient's operation records and reaction speed; The user monitoring submodule analyzes changes in the user's heart rate and respiratory signs information data, and dynamically adjusts the patient's rehabilitation plan; The data analysis submodule conducts in-depth analysis of the patient's historical data, optimizes subsequent rehabilitation plans, and helps doctors and caregivers better understand the patient's rehabilitation process in a visual way.

3. A cognitive rehabilitation assistive device based on a deep neural network as claimed in claim 2, characterized in that The front-end module is also used to receive the processing results sent by the back-end module and display them on the interface.

4. A cognitive rehabilitation assistive device based on a deep neural network as claimed in claim 2, characterized in that The rehabilitation training submodule uses a decision tree to classify patients and assigns them to different rehabilitation training paths based on their basic information and medical history.

5. A cognitive rehabilitation assistive device based on a deep neural network as claimed in claim 2, characterized in that The game interaction submodule is responsible for calculating the interactive score in the game, using a comprehensive score function and a reinforcement learning framework to optimize the game feedback mechanism, and adjusting the game difficulty and content based on the patient's operation records and reaction speed to improve the patient's participation and training effect.

6. A cognitive rehabilitation assistive device based on a deep neural network as claimed in claim 2, characterized in that The cognitive rehabilitation auxiliary equipment includes an attention training module, a memory training module, a calculation ability training module, a thinking ability training module, an execution ability training module, and an eye tracking training module. Each module is used to perform different types of rehabilitation training and has an evaluation mechanism for rehabilitation assessment.

7. A cognitive rehabilitation assistive device based on a deep neural network as claimed in claim 2, characterized in that The user monitoring submodule uses a long short-term memory network to analyze data changes and dynamically adjust the patient's rehabilitation plan; this sub-algorithm captures long-term dependencies and is suitable for processing long-term vital sign data.

8. A cognitive rehabilitation assistive device based on a deep neural network as claimed in claim 2, characterized in that The data analysis submodule uses clustering algorithms and regression analysis to explore potential patterns and trends in the data and optimize subsequent rehabilitation plans; this submodule also provides data visualization tools to help doctors and caregivers better understand the patient's rehabilitation process.

9. A storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed by a processor, the method according to claim 1 is implemented.

Images