Space neglect rehabilitation training system and equipment based on artificial intelligence

Abstract

The invention discloses a space neglect rehabilitation training system based on artificial intelligence. The system comprises a space neglect evaluation module used for evaluating the space neglect condition of a patient through a space neglect evaluation program; the rehabilitation program setting module is used for setting a rehabilitation program of the mobile wearable equipment by utilizing a rehabilitation program setting model according to the related data of the patient; the wearable equipment execution training command module is used for executing a corresponding rehabilitation task by the mobile wearable equipment according to the rehabilitation program set in the rehabilitation program setting module and recording the completion condition of the rehabilitation task; and the terminal collection feedback module is used for reading the training data of the patient, evaluating and judging the training compliance of the patient by utilizing the training data of the patient and constructing a training effect analysis model, and generating an adjustment strategy of a rehabilitation training scheme. Through the system provided by the invention, the training compliance and effect of the patient can be comprehensively and objectively evaluated, and the quality and efficiency of rehabilitation treatment are improved.

Description

Technical Field

[0001] The present invention generally relates to the field of artificial intelligence technology. More specifically, the present invention relates to an artificial intelligence-based spatial neglect rehabilitation training system and equipment. Background Art

[0002] Neglect, also known as Unilateral Spatial Neglect (USN), is a cognitive dysfunction that occurs after a stroke (especially a stroke lesion in the right temporoparietal junction) with neglect as the main manifestation, and is unable to correctly report stimulation on the opposite side of the brain lesion. As the name suggests, patients can only be aware of objects in half of the space, but there is nothing wrong with their senses: they may "not see" objects in the left half of the visual field, but their vision is normal. For example, patients with left-sided neglect will not be able to pay attention to objects on the left, dress the right side of the body, eat food on the right side of the plate, read words on the right side of the page, or even the right half of a word. Neglect not only affects the patient's post-stroke rehabilitation, but also affects the patient's quality of life and life safety. Therefore, early rehabilitation intervention for neglect is particularly important.

[0003] Existing rehabilitation training applied to USN is mostly instrument-assisted training under the guidance of traditional medical staff, such as wooden peg boards. This training method is time-consuming and labor-intensive, with a high proportion of medical staff manpower costs, and is restricted by training equipment, so it is impossible to complete rehabilitation training in real time and portably. Some scholars use the principle of prism refraction to carry out prism adaptation training (invention patent: a prism adaptation device and training method for promoting visual spatial recovery of patients with lateral neglect CN109984925A). After using the prism adaptation device, the patient's visual deviation is induced. This method belongs to guided rehabilitation training that combines the device with the patient's finger-touch target. It can adaptively allow the patient to adjust to the neglected side of the visual field, thereby exploring the lateral neglected space and achieving the effect of guided rehabilitation training. However, the disadvantage of this type of training is that the stimulation method is single and the patient's training compliance cannot be monitored. With the gradual advancement of science and technology, some scholars have developed a brain-computer interface system for rehabilitation training of neglect (invention patent: Unilateral spatial neglect rehabilitation training method and device, equipment, storage medium CN118356580 A), which stimulates the patient's spatial neglect side through photoelectric visual stimulation of the computer screen and electrical stimulation of the cervical nerves. However, this design requires a computer host and an electroencephalogram recorder, and an electrode cap needs to be worn. The rehabilitation training venue is fixed, which is not convenient enough, limiting the development of home rehabilitation training and is not conducive to the implementation of early and full-course rehabilitation training for patients.

[0004] In view of this, there is an urgent need to provide an artificial intelligence-based spatial neglect rehabilitation training system so as to complete rehabilitation training in real time and portably. Summary of the invention

[0005] In order to at least solve one or more of the technical problems mentioned above, the present invention proposes an artificial intelligence-based spatial neglect rehabilitation training system and equipment in multiple aspects.

[0006] In a first aspect, the present invention provides an artificial intelligence-based spatial neglect rehabilitation training system, comprising: a spatial neglect assessment module: used to assess the patient's spatial neglect through a spatial neglect assessment program; the spatial neglect assessment program comprises: a line segment elimination test unit, a spontaneous clock drawing test unit, a line segment division test unit and a copying flower drawing test unit.

[0007] Rehabilitation program setting module: used to set the rehabilitation program of the mobile wearable device according to the patient-related data and using the rehabilitation program setting model, including a sensory input training unit, a music stimulation unit, a trunk rotation training unit, a healthy eye occlusion training unit and / or a cross-facilitation training unit.

[0008] Wearable device execution training command module: used for the mobile wearable device to execute corresponding rehabilitation tasks according to the rehabilitation program set in the rehabilitation program setting module, and record the completion status of the rehabilitation tasks.

[0009] Terminal collection feedback module: used to read the patient's training data, and use the patient's training data to build a training effect analysis model to evaluate and determine the patient's training compliance and generate an adjustment strategy for the rehabilitation training program.

[0010] In some embodiments, the spatial neglect assessment program may further include a spatial positioning test unit, a three-dimensional spatial perception test unit, and a visual search task unit.

[0011] In some embodiments, the mobile wearable device performs corresponding rehabilitation tasks and records the completion status of the rehabilitation tasks, including: a precise motion monitoring and recording unit: used to integrate multiple high-precision sensors in the mobile wearable device, and use built-in microprocessors and algorithms to perform real-time analysis on the data obtained by the sensors.

[0012] Multimodal feedback mechanism unit: used to provide visual feedback to patients using an electronic display, optimizing tactile feedback using vibration, and providing auditory feedback through a speaker.

[0013] Adaptive training adjustment unit: used to collect and analyze patient data during the training process, and dynamically adjust the training plan based on real-time evaluation data.

[0014] In some embodiments, the construction of the training effect analysis model includes: extracting meaningful training features from the collected diverse data; wherein the diverse data includes patient training data, physiological data, and training environment data. The training features are normalized and standardized to eliminate differences between different data types and magnitudes. The processed training feature data are divided into a training set and a test set. The training effect analysis model is trained with the training set using a deep learning algorithm, and the prediction error of the model is minimized by adjusting the parameters of the training effect analysis model. The trained training effect analysis model is evaluated using a test set; and the training effect analysis model is optimized based on the evaluation results.

[0015] In a second aspect, the present invention provides a spatial neglect rehabilitation training device, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to implement an artificial intelligence-based spatial neglect rehabilitation training system as described in any one of the first aspects when executing the computer program.

[0016] Through the artificial intelligence-based spatial neglect rehabilitation training system provided above, the embodiment of the present invention can accurately determine the patient's spatial neglect situation and severity through a variety of digital and intelligent evaluation procedures and newly added evaluation methods. Further, in some embodiments, by executing the training command module through the wearable device, the accuracy of the patient's training movements can be ensured, the patient's training experience and participation can be improved, and the rehabilitation process can be optimized according to the patient's real-time training situation, so that the training is always adapted to the patient's rehabilitation state, avoiding excessive or insufficient training. Further, in some embodiments, through the terminal collection feedback module, through diversified data collection and secure storage, the construction of an accurate training effect analysis model, and the generation of personalized and dynamic rehabilitation training program adjustment strategies, the patient's training compliance and effect can be comprehensively and objectively evaluated, and detailed and accurate analysis reports and adjustment suggestions can be provided to medical staff, so as to achieve real-time optimization of the rehabilitation training program, further improve the quality and efficiency of rehabilitation treatment, and ultimately promote the improvement and rehabilitation of patients' spatial neglect symptoms. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0018] Figure 1 An exemplary prototype structure diagram of a mobile wearable device for spatial neglect rehabilitation training according to an embodiment of the present invention is shown;

[0019] Figure 2 An exemplary structural block diagram of a spatial neglect rehabilitation training system based on artificial intelligence according to some embodiments of the present invention is shown;

[0020] Figure 3 An exemplary structural block diagram of a spatial neglect rehabilitation training device according to an embodiment of the present invention is shown.

[0021] Figure 1 Middle: 1. Mobile wearable device body; 2. Electronic display screen; 3. Cold air outlet; 4. Skin-friendly elastic fixing belt; 5. Skin-friendly elastic fixing belt adhesive area; 6. Speaker outlet; 7. Lithium-ion battery area; 8. Temperature sensing patch. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0023] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0024] Figure 1 An exemplary prototype structure diagram of a mobile wearable device for spatial neglect rehabilitation training according to an embodiment of the present invention is shown.

[0025] like Figure 1 As shown, the structure of the mobile wearable device includes: a mobile wearable device body 1, which can be used as the core frame of the entire device. Electronic display screen 2: can be used to display training instructions, progress and other information. Cold air outlet 3: used to provide cold stimulation to the patient. Skin-friendly elastic fixing belt 4: used to fix the device on the patient's wrist. Skin-friendly elastic fixing belt adhesive area 5: auxiliary fixation device. Speaker outlet 6: used to play music, instructions and other sounds. Lithium-ion battery area 7: provides power for the device. Temperature sensing patch 8: used to provide thermal stimulation to the patient.

[0026] The above mobile wearable device can be used as an embodiment carrier of the spatial neglect training system of the present invention. When a patient wears the above mobile wearable device, the following advantages can be achieved:

[0027] A. Skin-friendly elastic bracelet, the color of the bracelet can be set to red, which serves as a reminder for medical staff to carry out daily medical care activities on the neglected side of the patient's space as much as possible.

[0028] B. Mobile wearable devices are small and portable, not restricted by time and space, and patients can perform spatial neglect rehabilitation training at any time.

[0029] C. The temperature-sensing patch fits the skin and can quickly provide thermal stimulation to the patient's affected upper limb according to instructions.

[0030] D. This device performs rehabilitation based on assessment, and the terminal collects training data and feeds it back to medical staff, ultimately forming a closed-loop management.

[0031] Combine the following Figure 1 ,pass Figure 2 The specific process of spatial neglect rehabilitation training for patients using the spatial neglect rehabilitation training system with the assistance of mobile wearable devices is described in detail.

[0032] Figure 2 An exemplary structural block diagram of a spatial neglect rehabilitation training system based on artificial intelligence according to some embodiments of the present invention is shown.

[0033] like Figure 2 As shown, an artificial intelligence-based spatial neglect rehabilitation training system provided by an embodiment of the present invention includes: a spatial neglect assessment module 101: used to assess the patient's spatial neglect through a spatial neglect assessment program; the spatial neglect assessment program includes: a line segment elimination test unit, a spontaneous clock drawing test unit, a line segment division test unit and a copying flower drawing test unit.

[0034] In one embodiment, the line segment elimination test unit configuration includes: presenting a line segment elimination task on an electronic display screen of a mobile wearable device, and automatically adjusting the length, density and arrangement of the line segments according to the patient's recovery status to evaluate the patient's spatial neglect.

[0035] By presenting the line segment erasure task on the electronic display screen of the mobile wearable device, the touch sensing technology is used to accurately record the patient's touch trajectory and erasure action. This can avoid the human errors that may occur in traditional paper erasure tests, such as paper movement and unclear markings.

[0036] Timing software can be used to automatically time and record the time it takes the patient to complete the task, and to quantify the score based on the accuracy and speed of the elimination, such as the number of accurately eliminated lines, the number of incorrectly eliminated lines, and the total time it takes to complete the task.

[0037] As the patient's training progresses, the task difficulty can be gradually increased by automatically adjusting the length, density and arrangement of the line segments according to the patient's recovery. For example, from simple horizontal and vertical line segments to diagonal and staggered line segments, the patient's spatial neglect improvement can be more accurately assessed.

[0038] In one embodiment, the spontaneous clock drawing test unit configuration includes intelligent drawing analysis, which analyzes the characteristics of the clock face drawn by the patient on the electronic display screen using image recognition technology. For more serious patients, auxiliary prompts are provided through the display screen to judge the degree of spatial neglect based on the clock face drawn.

[0039] When the patient draws a clock on an electronic display screen, the device software can use image recognition technology to analyze the characteristics of the clock face drawn, such as the order of numbers, the position and length of the hands, and the integrity of the clock face.

[0040] By comparing with the standard clock face template, the software can give a detailed deviation analysis report, including parameters such as the offset angle and distance of each number and pointer, providing medical staff with a more objective evaluation basis.

[0041] In addition, for patients with more serious conditions, some auxiliary prompts can be provided on the display screen, such as a faint circular outline, partial digital positions, etc., and the prompt content can be gradually reduced to observe the patient's recovery.

[0042] In one embodiment, the line segmentation test unit configuration includes: generating a personalized line segmentation task by analyzing the patient's condition assessment results; and implementing the interactive operation of the line segmentation task on a mobile wearable device by utilizing animation demonstration.

[0043] Implement interactive operations for line segmentation tasks on mobile wearable devices. For example, patients can divide line segments by touching, sliding, and other operations, and the device will provide real-time feedback on the operation results. At the same time, the animation demonstration function can be introduced to show the standard line segmentation method before the patient starts the operation, helping the patient to better understand the task requirements.

[0044] According to the patient's preliminary assessment results, the device software automatically generates personalized segmentation tasks. For example, for patients with spatial neglect biased to one side, the difficulty of segmentation on that side is increased, such as increasing the number of segments or shortening the length of the segments.

[0045] In one embodiment, the flower copying test unit is configured in a dynamic copying mode, which presents a dynamic flower pattern to be copied on an electronic display screen and uses pressure sensing technology to record the patient's pen pressure and speed during the copying process to analyze their hand-eye coordination and spatial perception abilities; and uses an image comparison algorithm to analyze the difference between the copied pattern and the original pattern.

[0046] The flower pattern to be copied is presented on an electronic display screen. The flower pattern can be dynamic, such as the opening and closing of petals, the rotation of the flower, etc., to increase the difficulty and interest of copying.

[0047] Using pressure sensing technology, the pressure and speed of the patient's pen strokes during the copying process are recorded to analyze their hand-eye coordination and spatial perception abilities.

[0048] The device assists patients in copying by providing copying auxiliary tools, such as zooming in and out functions, and local detail prompts. After copying is completed, the device software uses image comparison algorithms to analyze in detail the differences between the copied pattern and the original pattern, including the shape of the petals, the size of the flower, and the proportions of each part.

[0049] In one embodiment, the spatial neglect assessment program may further include a spatial positioning test unit, a three-dimensional space perception test unit, and a visual search task unit.

[0050] The spatial positioning test unit randomly displays target points at different locations on the electronic display screen of the mobile wearable device, and the patient needs to click the target point with his finger as quickly as possible. The device records the accuracy of the click (whether the target point is clicked) and the reaction time. The target point can have different sizes and colors, and the difficulty is gradually adjusted according to the patient's performance. For example, the target point gradually becomes smaller, and the contrast between the color and the background gradually decreases.

[0051] The spatial positioning test unit can directly evaluate the patient's perception and response ability to spatial position, and reflect the patient's spatial neglect in spatial positioning.

[0052] The three-dimensional space perception test unit uses the electronic display screen of the device to present simple three-dimensional graphics, such as cubes and spheres, and the patient needs to judge the shape, position and direction of the graphics. The test can be performed by asking the patient to rotate the graphics and judge the overlapping relationship of the graphics. It can also be combined with virtual reality (VR) or augmented reality (AR) technology to allow patients to operate in a more realistic three-dimensional space environment. For example, the patient wears simple VR glasses to find and operate specific three-dimensional objects in a virtual scene.

[0053] Through the three-dimensional space perception test unit test, the patient's perception ability in three-dimensional space can be evaluated, because spatial neglect may not only exist in the two-dimensional plane, but may also be reflected in three-dimensional space. This test can more comprehensively evaluate the patient's spatial neglect.

[0054] The visual search task unit presents a scene containing multiple different elements on an electronic display screen, such as graphics of different shapes and colors mixed together, in which a specific target element is hidden. The patient needs to find the target element within a specified time. This task unit can observe the patient's visual search ability and spatial neglect by changing the complexity of the scene, such as increasing the number of elements and reducing the size of the target element.

[0055] Visual search tasks can reflect patients' ability to focus on specific spatial areas in complex visual environments, and help detect spatial neglect problems that may occur in patients in daily scenes.

[0056] Furthermore, in the rehabilitation program setting module 102: it is used to set the rehabilitation program of the mobile wearable device according to the patient-related data and using the rehabilitation program setting model, including a sensory input training unit, a music stimulation unit, a trunk rotation training unit, a healthy eye occlusion training unit and / or a cross-facilitation training unit.

[0057] In one embodiment, the sensory input training unit may include vibration stimulation and temperature stimulation, wherein the vibration stimulation may adopt a multi-mode vibration mode.

[0058] The above vibration can be divided into two modes: timed vibration and external stimulation vibration. Timed vibration means that the portable mobile wearable device can complete the timed and fixed frequency vibration, thereby providing external sensory input; external stimulation vibration uses infrared sensors. When there is an object moving within 2 meters, the wearable device automatically turns on the vibration prompt to remind the patient to pay attention to the movement of the object on the ignored side, thereby completing the stimulation of the visual center and peripheral level at the same time. In addition, the vibration reminder can play a reminder and protection role in outdoor environments.

[0059] In addition to timed vibration and external stimulation vibration, patterned vibration is added. For example, different natural vibration rhythms can be simulated, such as heartbeat rhythm (regular alternation of strong and weak), wave rhythm (regular fluctuation intensity), etc. The switching of different vibration modes is achieved through built-in algorithms and vibration motor control.

[0060] In vibration stimulation training, there is also a vibration intensity adaptive function. By adding acceleration sensors and pressure sensors to the device, the sensor monitors the patient's limb movement amplitude and pressure changes in daily activities, and automatically adjusts the vibration intensity based on these data. For example, when the patient is in motion, the vibration intensity is appropriately increased to ensure effective sensory input; when the patient is at rest, the vibration intensity is reduced to avoid discomfort.

[0061] Multi-mode vibration can provide patients with diverse sensory stimulation, avoid adaptive fatigue caused by a single vibration mode, and improve the effectiveness of rehabilitation training. The adaptive vibration intensity can be dynamically adjusted according to the actual situation of the patient, which not only ensures effective stimulation under different activity states, but also improves the patient's wearing comfort and enhances the patient's compliance with rehabilitation training.

[0062] In the temperature stimulation training in the sensory input training unit, a thermoelectric cooling sheet can be used to give the patient's upper limb on the visual neglect side cold stimulation through the cold air outlet of the wearable device. The cold air temperature can be set according to the patient's acceptance range. There is a temperature-sensing patch at the junction of the back of the portable mobile wearable device and the skin. The material is a thermal stimulation graphene heating sheet, which can set the heating temperature within the skin's safe range, thereby providing thermal stimulation to the upper limb on the visual neglect side; alternating hot and cold stimulation completes the sensory input training of the upper limb on the visual neglect side.

[0063] In addition, gradient temperature stimulation can also be included: based on the original cold and hot stimulation, a gradient change in temperature can be achieved. For example, during cold stimulation, the temperature gradually decreases from room temperature to a set low temperature, or during hot stimulation, the temperature gradually increases from a low temperature to a set high temperature. The linear change in temperature is achieved by precisely controlling the power of the thermoelectric cooling sheet and the temperature sensing patch.

[0064] Gradient temperature stimulation can simulate a more natural temperature change process, produce richer stimulation to the patient's skin receptors, help further activate neural pathways, and improve the effect of sensory input training.

[0065] It also includes local temperature stimulation: using microfluidic technology to set up microchannels on the surface of the device that contacts the skin, and by controlling the flow of hot or cold fluids in different channels, temperature stimulation of specific local areas of the upper limbs (such as fingers, wrists, forearms, etc.) is achieved.

[0066] Through local temperature stimulation, precise training can be performed on different parts of the upper limbs. For patients with more obvious local sensory impairments, rehabilitation treatment can be carried out more specifically to improve rehabilitation efficiency.

[0067] In one embodiment, the sensory input training unit further includes a music stimulation unit. The portable mobile wearable device worn on the visual neglect side upper limb can play music at a fixed time according to the patient's music preferences to increase the auditory input on the visual neglect side.

[0068] Specifically, personalized music generation can be used: based on the patient's music preferences, a music generation algorithm is used to generate customized music that matches the patient's preferred style. For example, for patients who like classical music, new music clips are generated based on the style of their preferred composers; for patients who like pop music, new songs are created based on the timbre and rhythm characteristics of their favorite singers. Music creation is performed by analyzing the patient's existing music preference data and combining music theory with artificial intelligence algorithms.

[0069] Personalized music generation can greatly improve patients' interest and acceptance of music stimulation and enhance the effectiveness of music stimulation, because patients are more likely to have positive emotional responses to music of their favorite style, thereby promoting recovery.

[0070] You can also synchronize music with other stimuli: synchronize music playback with other stimuli (such as vibration, temperature stimulation). For example, at the climax of the music, the vibration intensity is increased or the temperature changes rapidly to form a multi-sensory synergistic stimulation. This is achieved by co-programming the music playback and other stimulation modules through the central control unit of the device.

[0071] Synchronizing music with other stimuli can produce a multi-sensory integration effect, strengthen the brain's perception of the association between different stimuli, help activate more neural pathways, and improve the overall effect of rehabilitation training.

[0072] In one embodiment, the sensory input training unit also includes a trunk rotation training unit, which is configured to use virtual reality technology to create a virtual training scene for the patient. The portable mobile wearable device is equipped with a displacement sensor. After the patient's trunk rotates to the neglected side of the space, the rotation angle and amplitude are set. When the target is achieved, the device prompts that the action is completed, and this is repeated to set the training time and training frequency.

[0073] Specifically, in training, real-time posture correction can be used: multiple high-precision posture sensors (such as gyroscopes, accelerometers, etc.) are added to the device to monitor the patient's trunk rotation posture in real time. When the patient's rotation posture does not meet the set requirements, the device will correct the patient's movements in time through voice prompts or vibration prompts. For example, if the patient's rotation angle is too large or too small, the device will immediately issue a prompt to guide the patient to adjust the movement.

[0074] Real-time posture correction can ensure that each trunk rotation movement of the patient meets the requirements of rehabilitation training, avoid repetition of incorrect movements, and improve the accuracy and effectiveness of training.

[0075] Virtual reality-assisted training can also be used: virtual reality technology is combined to create a virtual training scene for patients. When patients are doing trunk rotation training, they seem to be in a specific environment (such as a virtual forest, indoor scene, etc.), and they interact with objects in the virtual environment by rotating their bodies. For example, patients need to rotate their bodies to touch flowers or objects in the virtual scene, which enhances the fun and immersion of the training.

[0076] Virtual reality-assisted training can increase the fun and immersion of training, making patients more willing to actively participate in training. At the same time, the interactive elements in the virtual environment can further stimulate the patient's brain and body coordination and promote recovery.

[0077] In one embodiment, the sensory input training unit further includes a healthy eye masking training unit. By covering the right eye of the left neglecter, the patient's attention level to the left object is improved. The portable mobile wearable device automatically enters a countdown, and when the set time is reached, it vibrates to remind the healthy eye masking training task to be completed.

[0078] Specifically, it can be enhanced by configuring visual feedback: while covering the healthy eye, visual feedback is provided to the patient on the electronic display screen of the mobile wearable device. For example, when the patient's right eye is covered for training, the real-time scene in the patient's left visual field (captured by the camera) is displayed on the display screen, and important objects in the scene are highlighted to guide the patient to focus on the objects on the left.

[0079] Enhanced visual feedback can help patients understand their left visual field more intuitively, and through highlighting prompts and other methods, it can more effectively guide patients to focus on objects on the ignored side, thereby improving the effect of healthy eye occlusion training.

[0080] You can also configure attention monitoring and adjustment: use eye tracking technology to monitor the patient's attention direction during training. If the patient's attention is found to be deviating from the left visual field, the device will guide the patient to refocus on the left side by changing the visual prompts on the display (such as changing the highlight color, flashing frequency, etc.) or issuing sound prompts.

[0081] Through attention monitoring and adjustment, we can ensure that the patient's attention is focused on the training goals in real time, avoid the patient's unconscious deviation from the training requirements during the training process, and ensure the quality and effectiveness of the training.

[0082] In one embodiment, the sensory input training unit further comprises a cross-facilitation training unit. The cross-facilitation training unit is configured to grade the difficulty of the cross-facilitation training movements according to the patient's rehabilitation progress, and combine multi-sensory feedback to strengthen the patient's perception and memory of the correct movements. The multi-sensory feedback includes pressure feedback, sound and vibration feedback.

[0083] The portable mobile wearable device is equipped with a pressure receptor. The patient needs to use the upper limb on the non-spatial neglect side to cross the midline of the body, and use the fingers to press hard at a specific position on the mobile wearable device to reach the target pressure value, completing a pressing action, and repeating the action tactile stimulation. (The cross-promotion training program starts: the voice broadcast starts, please press hard 1, reach the target value, complete; 2 reach the target value, complete; repeat).

[0084] Specifically, it is equipped with action difficulty grading and adaptive functions, which can grade the difficulty of cross-promotion training according to the patient's rehabilitation progress. For example, from simple one-handed pressing movements, gradually increase to double-handed coordinated pressing, pressing areas of different shapes and sizes, etc. The device automatically determines the patient's training level through pressure receptor data feedback and built-in algorithms, and adaptively adjusts the difficulty of the next training.

[0085] Through movement difficulty grading and adaptation, training can be made more in line with the patient's rehabilitation process, avoiding slow rehabilitation progress due to too low training difficulty, or frustration in patients due to too high training difficulty, thereby improving training efficiency and patient enthusiasm.

[0086] The cross-facilitation training unit configuration can also include a combination of multi-sensory feedback: when the patient is doing cross-facilitation training, in addition to the original pressure feedback, sound and vibration feedback are combined. For example, when the patient reaches the target pressure value, the device sends a confirming sound prompt and a short vibration feedback, and displays the completion progress and score on the electronic display. Through the combination of multi-sensory feedback, the patient's perception and memory of the correct action are strengthened.

[0087] The combination of multi-sensory feedback can enhance patients' perception and memory of training movements, and stimulate the brain through multiple sensory channels at the same time, which helps to better promote the recovery of neurological function and improve hand-eye coordination.

[0088] Furthermore, the wearable device executes a training command module 103: for executing the corresponding rehabilitation task according to the rehabilitation program set in the rehabilitation program setting module, and recording the completion status of the rehabilitation task.

[0089] In one embodiment, the mobile wearable device performs corresponding rehabilitation tasks and records completion status of the rehabilitation tasks, including:

[0090] Precision motion monitoring and recording unit: It is used to integrate multiple high-precision sensors into mobile wearable devices and use built-in microprocessors and algorithms to perform real-time analysis on the data obtained by the sensors.

[0091] First, data is collected through high-precision sensor integration. Specifically, a variety of high-precision sensors can be integrated into mobile wearable devices, such as high-precision pressure sensors, angle sensors, acceleration sensors, etc.

[0092] For example, when performing cross-promotion training, the pressure sensor can not only detect whether the target pressure value is reached, but also accurately measure the pressure change curve, including the speed of pressure rise, the time the pressure is maintained, etc.; in trunk rotation training, the angle sensor and acceleration sensor can accurately record the angle, angular velocity, angular acceleration and other parameters of the trunk rotation.

[0093] Then, real-time motion capture and analysis is performed. By using the built-in microprocessor and algorithm, the data obtained by the sensor is analyzed in real time.

[0094] For example, when patients perform rehabilitation training movements, the device can determine in real time whether the movements meet the set specifications. For healthy eye occlusion training, eye tracking technology (which can be integrated into the device with a micro camera and corresponding algorithms) is used to accurately monitor the occlusion of the healthy eye to prevent patients from cheating or irregular operations; in music stimulation training, it monitors whether the patient makes corresponding reaction movements when the music is playing (such as swinging limbs to the rhythm of the music, etc.).

[0095] Multimodal feedback mechanism unit: used to provide visual feedback to patients using an electronic display, optimizing tactile feedback using vibration, and providing auditory feedback through a speaker.

[0096] When the mobile wearable device performs the corresponding rehabilitation task, the completion status is monitored and fed back through the configured multimodal feedback mechanism. Specifically, it can include visual feedback enhancement. By making full use of the functions of the electronic display screen, richer visual feedback can be provided to the patient.

[0097] For example, in sensory input training, when temperature stimulation is performed, the display screen can display the current temperature value, temperature change curve, and skin sensation diagram corresponding to different temperature stimulations in real time; in the experiment of the spatial neglect assessment module, after the patient completes the operation, an analysis diagram of the operation results will be immediately presented on the display screen, such as displaying the un-crossed-out line area in the line segment elimination test, and displaying the deviation between the drawn clock face and the standard clock face in the clock drawing test.

[0098] In addition, it is also equipped with tactile feedback optimization, which further optimizes the tactile feedback in addition to the original vibration function.

[0099] For example, when the patient completes or fails to complete a training task, different information is conveyed through vibrations of different frequencies, intensities, and patterns. In cross-promotion training, when the patient does not reach the target pressure value, the device sends a low-frequency, short-duration vibration prompt; when the target pressure value is reached, a high-frequency, long-duration vibration feedback is sent, and at the same time, a gradual vibration sense of pressure release is used to simulate a sense of accomplishment of completing a task.

[0100] In addition, it is also equipped with diversified auditory feedback. By expanding the functions of the speaker, diversified auditory feedback is provided.

[0101] For example, during rehabilitation training, in addition to playing music, different sound prompts can be issued according to the patient's movements. In trunk rotation training, when the patient's rotation angle approaches the set target, a prompt tone with a gradually increasing frequency is issued; when the target angle is reached, a confirmation sound effect is issued. During training, encouraging voice prompts can also be played according to the patient's training progress and performance, such as "You are doing great, keep it up!" or "If you adjust your movements slightly, you can do better."

[0102] Adaptive training adjustment unit: used to collect and analyze patient data during the training process, and dynamically adjust the training plan based on real-time evaluation data.

[0103] Specifically, the training plan can be dynamically adjusted through real-time training data evaluation.

[0104] The equipment continuously collects and analyzes the patient's data during the training process, including the completion of movements, reaction time, training duration, etc. For example, in the spatial neglect assessment module, the improvement of the patient's spatial neglect is evaluated based on the patient's performance data in different tests; in the various trainings in the rehabilitation program setting module, the accuracy and efficiency of the patient's completion of various training tasks within a certain period of time (such as one week) are counted.

[0105] In the dynamic adjustment of the training plan, the training plan is dynamically adjusted according to the real-time evaluation data. If the patient shows good adaptability and progress in a certain training item (such as vibration sensory input training), the device automatically increases the difficulty of the training, such as increasing the vibration frequency, changing the vibration mode, etc.; on the contrary, if the patient progresses slowly or shows resistance in a certain training (such as healthy eye masking training), the device can appropriately reduce the difficulty of the training, such as shortening the masking time, adding visual prompts, etc., to maintain the patient's training enthusiasm and effectiveness.

[0106] Through the above-mentioned embodiment of the wearable device execution training command module provided by the present invention, the patient's training movements can be accurately monitored through high-precision sensors and real-time motion capture and analysis, ensuring that the patient performs rehabilitation training in accordance with the correct specifications, and avoiding the rehabilitation effect being affected by irregular movements. For example, in trunk rotation training, accurate angle and posture monitoring can ensure that the patient's rotation movement meets the rehabilitation requirements, so that the training can accurately stimulate the patient's related nerves and muscles, accelerating the rehabilitation process.

[0107] The multimodal feedback mechanism can provide patients with rich feedback information from multiple aspects such as vision, touch and hearing, allowing patients to understand their training status and effects more intuitively.

[0108] Diverse feedback methods can increase the interest of training. For example, in music stimulation training, combining visual and tactile feedback allows patients to enjoy music while being better engaged in training through stimulation of other senses, thereby improving patients' acceptance and participation in rehabilitation training.

[0109] By optimizing the rehabilitation process and using the adaptive training adjustment function, the training plan can be dynamically adjusted according to the patient's actual training data, so that the training difficulty always matches the patient's rehabilitation progress.

[0110] The personalized training adjustment provided by the embodiments of the present invention can avoid over-training or under-training. For example, in spatial neglect assessment and rehabilitation training, the training plan can be adjusted in time according to the patient's changing degree of spatial neglect, ensuring that rehabilitation training is carried out within the most effective range, improving rehabilitation efficiency, and shortening the rehabilitation cycle.

[0111] Furthermore, the terminal collects feedback module 104: for reading the patient's training data, and using the patient's training data to construct a training effect analysis model, evaluate and determine the patient's training compliance, and generate an adjustment strategy for the rehabilitation training program.

[0112] First, the terminal collection feedback module collects and stores data. Specifically, diversified collection can be performed. In addition to collecting basic training data recorded by mobile wearable devices, such as training duration, number of training sessions, task completion rate, etc., the patient's physiological data during training should also be collected. For example, wearable heart rate monitoring devices, blood pressure monitoring devices, etc. are used to obtain changes in the patient's heart rate and blood pressure during training, and skin galvanic response sensors are used to obtain the patient's emotional stress state data.

[0113] Collect data from patients in different training scenarios, including indoor and outdoor environments. For example, in outdoor environments, collect data related to the device's infrared sensor trigger frequency (corresponding to external stimulus vibration) and the patient's actual response to evaluate the patient's training effect in real-life scenarios.

[0114] When storing the collected data, a cloud storage platform can be established to encrypt the collected data and store it in the cloud. Distributed storage technology is used to ensure the security and reliability of data and prevent data loss or tampering.

[0115] The data is stored in categories, and classified according to the patient's personal information, data from different training modules, time series, etc., to facilitate subsequent rapid retrieval and analysis.

[0116] Through the above-mentioned diverse data collection and storage, we can fully understand the various performances of patients during rehabilitation training, including training behaviors, physiological reactions, and environmental adaptation. This helps to gain a deeper understanding of the patient's rehabilitation status and avoid the limitations of relying on a single data type for evaluation. For example, combining training movement data and physiological data can more accurately determine whether the patient is fatigued or uncomfortable during a certain training, thereby more accurately evaluating the training effect.

[0117] In one embodiment, constructing a training effect analysis model includes:

[0118] Extract meaningful training features from the collected diverse data; wherein the diverse data includes patient training data, physiological data, and training environment data. Specifically, extract meaningful features from the collected diverse data. For example, extract features such as accuracy, fluency, and symmetry of movements from the patient's training movement data; extract features such as heart rate variability and blood pressure fluctuation amplitude from physiological data; and extract different features of patient training performance in different scenarios from environmental data.

[0119] The training features are normalized and standardized to eliminate differences between different data types and magnitudes so as to better perform model training.

[0120] Then, select a suitable machine learning or deep learning model. For relatively simple linear relationship data, linear regression, logistic regression and other models can be used; for complex nonlinear relationship data, such as the relationship between patient training behavior and physiological and psychological reactions, neural networks (such as multi-layer perceptrons) and support vector machines can be selected.

[0121] The processed feature data is divided into a training set and a test set. For example, a deep learning algorithm is used to train the training effect analysis model using the training set, and the prediction error of the model is minimized by adjusting the parameters of the training effect analysis model (such as the weight and bias of the neural network). During the training process, methods such as cross-validation are used to prevent the model from overfitting.

[0122] Then, the trained training effect analysis model is evaluated using the test set; and the training effect analysis model is optimized according to the evaluation results. When the trained model is evaluated using the test set, the model's accuracy, recall rate, F1-score and other evaluation indicators can be calculated. According to the evaluation results, the model is optimized. For example, if the model has a low accuracy in evaluating patient training compliance, you can try to increase the amount of data, adjust the model structure or use an integrated learning method (such as random forest) to improve the model performance.

[0123] The following example uses a multi-layer perceptron (MLP) to build a training effect analysis model based on a deep learning algorithm to give a specific process for evaluating the effect of spatial neglect rehabilitation training.

[0124] In the model structure design, it is assumed that the MLP model has three layers: input layer, hidden layer and output layer. The number of neurons in the input layer depends on the number of selected features.

[0125] Assume that the following features are selected as input to the model:

[0126] : The average accuracy of patients in completing various training tasks in the past week (range: 0 to 1).

[0127] : Average time (in minutes) that patients completed training in the past week.

[0128] : The patient's most recent spatial neglect assessment test (such as the score of the line segment elimination test, a standardized value ranging from -1 to 1).

[0129] : The average heart rate change rate of the patient during rehabilitation training (the change ratio compared with the resting heart rate, ranging from -1 to 1).

[0130] The number of neurons in the input layer is 4, each neuron corresponds to an input feature, and the input feature vector is expressed as: x = [ x 1 , x 2 , x 3 , x 4 ] T .

[0131] In the hidden layer, 8 neurons can be set and ReLU is used as the activation function. The calculation process of the hidden layer can be expressed as:

[0132] ,

[0133] in, is the weight matrix from the input layer to the hidden layer, with dimension , is the bias vector of the hidden layer, with dimension .

[0134] There are 2 neurons in the output layer, which are used to predict the patient's training compliance (the value range is 0 to 1, indicating the degree of compliance) and the comprehensive score of the rehabilitation training effect (the value range is 0 to 100, indicating the degree of effect).

[0135] The Softmax function was used to normalize the training compliance, and the linear activation function was used to output the rehabilitation training effect score.

[0136] The calculation process of the output layer is:

[0137] y = [ y 1 , y 2 ] T = [ S o f t m a x ( W 2 h + b 2 ) , W 3 h + b 3 ] T .

[0138] in, is the weight matrix from the hidden layer to the training dependency output neuron, with dimension , is the corresponding bias, is the weight matrix from the hidden layer to the rehabilitation training effect score output neuron, with a dimension of , is the corresponding bias.

[0139] Next, in order to train the model, a joint loss function is defined, combining the classification cross entropy loss (for training compliance prediction) and the mean square error loss (for rehabilitation training effect score prediction). The formula may include:

[0140] L = − 1 N ∑ i = 1 N [ y 1 i l o g ( y ^ 1 i ) + ( 1 − y 1 i ) l o g ( 1 − y ^ 1 i ) ] + 1 N ∑ i = 1 N ( y 2 i − y ^ 2 i ) 2 ,

[0141] in, is the number of training samples, and They are The true training compliance labels of samples and the training compliance probabilities predicted by the model, and They are The actual rehabilitation training effect scores of the samples and the rehabilitation training effect scores predicted by the model.

[0142] For parameter learning in the model, the stochastic gradient descent (SGD) algorithm can be used to update the model parameters. In each iteration, a mini-batch of samples is randomly selected from the training data set, the gradient of the loss function to the model parameters is calculated, and the parameters are updated. The formula is:

[0143] ,

[0144] in, Including all weight matrices and the bias vector , is the learning rate, Indicates the number of iterations. It is the gradient of the loss function with respect to the model parameters. The gradient represents the rate of change of the loss function in the parameter space, and its direction points to the direction where the loss function grows fastest.

[0145] The parameters in the above example are further described in detail below in combination with spatial neglect rehabilitation training.

[0146] In the Input Feature Parameters: Feature (Average accuracy of training tasks): reflects the patient's accuracy in various rehabilitation training tasks. The higher the accuracy, the better the patient's mastery of the training, and may be positively correlated with the training effect and compliance. (Average training duration): To some extent, it reflects the patient's commitment to the training, but a duration that is too long or too short may require further analysis in combination with other factors. For example, a duration that is too long may indicate that the patient is having difficulty, while a duration that is too short may indicate poor compliance or that the training is too simple. (Spatial neglect assessment score): directly reflects the improvement of the patient's spatial neglect symptoms. The higher the score, the better the rehabilitation effect may be. It is highly correlated with the rehabilitation training effect score and has important reference value for adjusting the training program. Features (Heart rate variability): It reflects the patient's stress state and physical adaptation during training from a physiological perspective. Excessive heart rate changes may indicate that the training intensity is too high or the patient is not adapted, affecting training compliance and results.

[0147] For the model weight parameters ( ) and the bias parameter ( ): These parameters are learned by the model during training. The weight parameters determine the influence of different input features on the output and the connection strength between neurons in the hidden layer and the output layer. The bias parameters provide an additional learnable constant offset for neurons to help the model better fit the data. For example, in the weight matrix In the elements Represents the input layer neurons to the hidden layer The connection weights of neurons in the model are as follows: a large positive value may indicate that the input feature has a greater promoting effect on the activation of neurons in the hidden layer, and conversely, a large negative value may indicate an inhibitory effect. By continuously adjusting these parameters, the model can learn the complex nonlinear relationship between input features and outputs (training compliance and rehabilitation training effect scores), thereby achieving accurate evaluation and prediction of the patient's training status and providing strong support for the adjustment of rehabilitation training programs.

[0148] In actual application scenarios, a large amount of patient training data is needed to train and optimize the model in order to obtain an accurate and reliable training effect analysis model and provide a more accurate decision-making basis for spatial neglect rehabilitation training. At the same time, the model structure, input features and hyperparameters (such as learning rate, number of hidden layer neurons, etc.) can be further adjusted according to actual conditions to improve the performance and adaptability of the model.

[0149] Furthermore, in the generation of rehabilitation training program adjustment strategies, the strategies can be adjusted by formulating personalized adjustment rules. Specifically, personalized rehabilitation training program adjustment rules are formulated based on the results of the training effect analysis model. For example, if the model analysis finds that the patient's training effect in a certain training module (such as vibration sensory input training) reaches a certain threshold and the improvement trend slows down, a rule can be formulated to reduce the training intensity or frequency of the module and increase the training volume of other related modules (such as cross-promotion training).

[0150] Different adjustment strategies can also be developed by considering individual differences of patients, such as age, gender, underlying diseases, etc. For example, for elderly patients, when adjusting the training program, more attention should be paid to the gradualness of the training intensity to avoid excessive fatigue.

[0151] During the rehabilitation training process, it is necessary to dynamically adjust the training strategy and provide feedback on the training effect. Through the dynamic adjustment mechanism of the rehabilitation training program, the analysis model is updated according to the patient's data after each training, and the adjustment strategy is generated in real time. For example, when a patient shows resistance to a certain training method during a certain stage of training (reflected by physiological data such as skin electrical response), the training program is adjusted in time, the training method is changed, or the training parameters are adjusted.

[0152] Provide timely feedback of the adjusted training program to medical staff and patients. Detailed analysis reports and adjustment suggestions are pushed to medical staff through mobile applications or emails, while easy-to-understand adjustment prompts are displayed on the patient's mobile wearable device to guide the patient in subsequent training.

[0153] The training effect analysis model constructed above can objectively evaluate the patient's training compliance. The model is trained and learned based on a large amount of data, and can automatically identify the patient's training behavior patterns, reducing the subjectivity and errors of human evaluation. For example, the model can accurately determine whether the patient is training in accordance with the prescribed rehabilitation plan by analyzing data such as the patient's training duration, training frequency, and consistency of training movements, and provide medical staff with reliable compliance assessment results.

[0154] In addition, the personalized adjustment rules and dynamic adjustment mechanism generated based on the training effect analysis model can make the rehabilitation training program better adapt to the individual needs and rehabilitation progress of patients, avoiding the one-size-fits-all phenomenon that may occur in the adjustment of traditional rehabilitation programs and improving the pertinence and effectiveness of rehabilitation training.

[0155] For example, for patients at different stages of rehabilitation and with different individual characteristics, the training intensity, content and methods can be adjusted in a timely manner according to the results of model analysis, so that patients can always receive the most suitable training plan throughout the rehabilitation process, accelerate the rehabilitation process and improve the quality of rehabilitation.

[0156] In summary, the spatial neglect rehabilitation training system provided by the present invention can achieve the following beneficial effects:

[0157] 1) Accurate assessment and personalized rehabilitation: Through the expanded spatial neglect assessment module, a variety of digital and intelligent assessment procedures and newly added assessment methods, such as digital optimization of line segment elimination test and three-dimensional space perception test, can more accurately determine the patient's spatial neglect and severity. On this basis, the rehabilitation program setting module can formulate personalized rehabilitation training programs for patients according to the assessment results, including diversified sensory input training (such as multi-modal vibration, gradient temperature stimulation, etc.), music stimulation (personalized music generation and synchronization with other stimuli), targeted trunk rotation training (real-time posture correction and virtual reality assistance), healthy eye occlusion training (visual feedback enhancement and attention monitoring) and cross-facilitation training (movement difficulty grading and adaptation, multi-sensory feedback combination), to meet the specific needs of different patients and improve the pertinence and effectiveness of rehabilitation training.

[0158] 2) Enhanced training execution and feedback: Extended technical implementation of wearable device execution training command modules, such as precise motion monitoring and recording (high-precision sensor integration and real-time motion capture analysis), multimodal feedback mechanism (optimization and diversification of visual, tactile, and auditory feedback), and adaptive training adjustment (real-time training data evaluation and dynamic adjustment of training plans), to ensure the accuracy of patients' training movements, improve patients' training experience and participation, and optimize the rehabilitation process according to the patient's real-time training situation, so that the training always adapts to the patient's rehabilitation status and avoids excessive or insufficient training.

[0159] 3) Optimize training effect evaluation and program adjustment: The terminal collection and feedback module can comprehensively and objectively evaluate patient training compliance and effectiveness through diversified data collection and secure storage, build accurate training effect analysis models (feature engineering, model selection and training, evaluation and optimization), and generate personalized and dynamic rehabilitation training program adjustment strategies (personalized adjustment rule formulation and dynamic adjustment and feedback). It can also provide medical staff with detailed and accurate analysis reports and adjustment suggestions, realize real-time optimization of rehabilitation training programs, further improve the quality and efficiency of rehabilitation treatment, and ultimately promote the improvement and rehabilitation of patients' spatial neglect symptoms, help patients better restore their ability to take care of themselves and their quality of life, and reduce the burden on families and society. In addition, during the rehabilitation process, patients can train in a relatively comfortable and positive environment, which is conducive to the overall recovery of the body and mind.

[0160] Figure 3 An exemplary structural block diagram of a spatial neglect rehabilitation training device according to an embodiment of the present invention is shown.

[0161] like Figure 3 As shown, in the device, it includes: a memory 301 and a processor 302, the memory is used to store a computer program, and the processor is used to implement the following when executing the computer program: Figure 1 The artificial intelligence-based spatial neglect rehabilitation training system described in any one of the embodiments of the system shown.

[0162] Although multiple embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art may conceive of many changes, modifications, and alternatives without departing from the thought and spirit of the present invention. It should be understood that in the process of practicing the present invention, various alternatives to the embodiments of the present invention described herein may be adopted. The appended claims are intended to define the scope of protection of the present invention, and therefore cover equivalents or alternatives within the scope of these claims.

Claims

1. A spatial neglect rehabilitation training system based on artificial intelligence, characterized in that: include: Spatial Neglect Assessment Module: used to assess the patient's spatial neglect through the spatial neglect assessment procedure; The spatial neglect assessment procedure includes: a line segment elimination test unit, a spontaneous clock drawing test unit, a line segment division test unit and a flower copying test unit; Rehabilitation program setting module: used to set the rehabilitation program of the mobile wearable device according to the patient-related data and using the rehabilitation program setting model, including a sensory input training unit, a music stimulation unit, a trunk rotation training unit, a healthy eye occlusion training unit and / or a cross-facilitation training unit; Wearable device execution training command module: used for the mobile wearable device to execute corresponding rehabilitation tasks according to the rehabilitation program set in the rehabilitation program setting module, and record the completion status of the rehabilitation tasks; Terminal collection feedback module: used to read the patient's training data, and use the patient's training data to build a training effect analysis model to evaluate and determine the patient's training compliance and generate an adjustment strategy for the rehabilitation training program.

2. The system according to claim 1, characterized in that The line segment elimination test unit configuration includes: The line segment elimination task is presented on the electronic display screen of the mobile wearable device, and the length, density and arrangement of the line segments are automatically adjusted according to the patient's recovery status to assess the patient's spatial neglect.

3. The system according to claim 1, characterized in that The spontaneous clock drawing test unit configuration includes intelligent drawing analysis, in which the patient draws a clock on an electronic display screen and uses image recognition technology to analyze the features of the clock face drawn; for heavier patients, auxiliary prompts are provided through the display screen to judge the degree of spatial neglect based on the drawn clock face.

4. The system according to claim 1, characterized in that The line segmentation test unit configuration includes: generating a personalized line segmentation task by analyzing the patient evaluation results; and implementing the interactive operation of the line segmentation task on the mobile wearable device by utilizing animation demonstration.

5. The system according to claim 1, characterized in that The flower copying test unit is configured in a dynamic copying mode, which presents a dynamic flower pattern to be copied on an electronic display screen and uses pressure sensing technology to record the patient's pen pressure and speed during the copying process to analyze their hand-eye coordination and spatial perception abilities; and uses an image comparison algorithm to analyze the difference between the copied pattern and the original pattern.

6. The system according to claim 1, characterized in that The spatial neglect assessment program may also include a spatial positioning test unit, a three-dimensional space perception test unit, and a visual search task unit.

7. The system according to claim 1, characterized in that The trunk rotation training unit configuration includes using virtual reality technology to create a virtual training scene for the patient; the cross-promotion training unit is configured to grade the difficulty of the cross-promotion training movements according to the patient's rehabilitation progress, and combine multi-sensory feedback to strengthen the patient's perception and memory of the correct movements. The multi-sensory feedback includes pressure feedback, sound and vibration feedback.

8. The system according to claim 1, characterized in that The mobile wearable device performs corresponding rehabilitation tasks and records completion of the rehabilitation tasks, including: Precision motion monitoring and recording unit: It is used to integrate multiple high-precision sensors into mobile wearable devices and use built-in microprocessors and algorithms to analyze the data obtained by the sensors in real time; Multimodal feedback mechanism unit: used to provide visual feedback to patients using an electronic display, optimizing tactile feedback using vibration, and providing auditory feedback through a speaker; Adaptive training adjustment unit: used to collect and analyze patient data during the training process, and dynamically adjust the training plan based on real-time evaluation data.

9. The system according to claim 1, characterized in that The constructing of the training effect analysis model comprises: Extracting meaningful training features from the collected diverse data; wherein the diverse data includes patient training data, physiological data, and training environment data; Normalizing and standardizing the training features to eliminate differences between different data types and magnitudes; Divide the processed training feature data into a training set and a test set; Using a deep learning algorithm, the training effect analysis model is trained through a training set, and the prediction error of the model is minimized by adjusting the parameters of the training effect analysis model; The trained training effect analysis model is evaluated using a test set; and the training effect analysis model is optimized according to the evaluation result.

10. A spatial neglect rehabilitation training device, characterized in that: It comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to implement the artificial intelligence-based spatial neglect rehabilitation training system as described in any one of claims 1 to 9 when executing the computer program.

Images