A post-stroke depression combined eye movement assessment and rehabilitation training system

Abstract

The application relates to the technical field of neurological rehabilitation system, and discloses a post-stroke depression combined eye movement evaluation and rehabilitation training system, which comprises an eye movement data acquisition module, a post-stroke depression evaluation module, a rehabilitation training execution module, a cooperative control module and a data storage and interaction module. The application deeply combines eye movement tracking technology and post-stroke depression evaluation, breaks through the limitation that traditional scale evaluation is strong in subjectivity and is easily influenced by patient cooperation degree, collects multi-scene eye movement data through a non-invasive device, realizes objective quantification and accurate grading of depression state through pretreatment, feature fusion and model calculation, can early identify latent depression, greatly improves evaluation efficiency and reliability, meanwhile, the eye movement evaluation is non-invasive in the whole process and is simple to operate, is suitable for the characteristics that the limbs of a stroke patient are limited in movement and the cognition of the stroke patient is decreased, and reduces the evaluation threshold.

Description

Technical Field

[0001] This invention relates to the field of neurological rehabilitation system technology, specifically a system for post-stroke depression combined with eye movement assessment and rehabilitation training. Background Technology

[0002] Post-stroke depression is one of the most common neuropsychiatric complications of stroke, seriously affecting patients' rehabilitation process and quality of life. Its onset is closely related to neurotransmitter imbalance caused by brain tissue damage and psychosocial factors. Patients often exhibit depressed mood, loss of interest, and lowered self-esteem, and may also resist rehabilitation training, exacerbating limb dysfunction, creating a vicious cycle, and even increasing the risk of stroke recurrence and death. Therefore, early and accurate assessment and effective rehabilitation intervention are crucial for improving patient prognosis.

[0003] According to the search, the authorized announcement number is CN120600227B, which discloses an active stroke rehabilitation auxiliary training system. This system analyzes training execution efficiency scores, patient active movement characteristics, and equipment-assisted performance characteristics in real time, and dynamically adjusts the assistance intensity and task difficulty based on the relative contribution relationship between the two. This achieves a clear distinction between the patient's autonomous effort and the boundaries of equipment assistance during rehabilitation, effectively improving the dynamic adaptability and accuracy of training.

[0004] However, the above-mentioned technical solutions and existing technologies still have certain limitations. They do not consider the impact of post-stroke depression, a core neuropsychiatric complication, on the rehabilitation process, do not set up an assessment module for post-stroke depression, and do not establish a collaborative linkage mechanism between depression assessment and rehabilitation training. They cannot capture changes in the patient's depressive state through objective assessment methods, nor can they dynamically adjust the training program according to the degree of depression, making it difficult to meet the comprehensive rehabilitation needs of stroke patients with depression. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a post-stroke depression combined with eye-tracking assessment and rehabilitation training system. This system deeply integrates eye-tracking technology with post-stroke depression assessment, overcoming the limitations of traditional scale assessments that are highly subjective and easily affected by patient compliance. It achieves synergistic linkage and dynamic adaptation between depression assessment and rehabilitation training, thus solving the aforementioned technical problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a post-stroke depression combined with eye movement assessment and rehabilitation training system, comprising: an eye movement data acquisition module, a post-stroke depression assessment module, a rehabilitation training execution module, a collaborative control module, and a data storage and interaction module; The eye-tracking data acquisition module is used to collect raw eye-tracking data such as the patient's fixation point coordinates and saccade speed under a preset eye-tracking assessment scenario. It uses a non-invasive eye-tracking device with a sampling frequency of ≥120Hz and has motion artifact removal function. The post-stroke depression assessment module receives raw eye movement data collected by the eye movement data acquisition module and performs preprocessing, feature extraction, and fusion to construct an eye movement depression feature vector. Through a built-in trained assessment model, it outputs a quantitative depression score and state grading results. The feature vector construction and score calculation are implemented through the following formula: in: This is the feature vector of eye movement depression. For eye movement feature components, , The weight coefficients of the eigenvectors are obtained through principal component analysis. The mean of the time-domain features, The result is the frequency domain feature transformation. It is a non-linear characteristic index; in: Quantifying depression scores, The feature weight matrix, For bias terms, This is the calibration coefficient, and its value range is [1.2, 1.5]. This is a correction term, and its value range is [0.01, 0.05]. Let be the dimension of the eye-tracking depression feature vector. For summation index variables; The rehabilitation training execution module is used to receive personalized training instructions issued by the collaborative control module, execute limb, cognitive, and emotion regulation rehabilitation training, and integrates motion capture, force feedback and multi-sensory stimulation units. The collaborative control module receives the depression quantitative score and state grading results output by the post-stroke depression assessment module, retrieves the patient's baseline and historical training data, generates and issues personalized training instructions, and receives training data fed back by the rehabilitation training execution module in real time to form a closed-loop control, dynamically adjusting training parameters. The training parameter adaptation coefficient and the final training intensity parameter are calculated in conjunction using the following formula: in: The value range is [0.4, 0.6]. , All are weighting coefficients. Normalized values ​​for baseline limb function data. Normalized values ​​for baseline cognitive function data. For historical training times, For the first Training effect score, This is an adjustment coefficient, and its value range is [-0.8, -0.3]. This represents the change in depression quantification scores. For the final training intensity parameters; The data storage and interaction module is communicatively connected to the eye-tracking data acquisition module, the post-stroke depression assessment module, the rehabilitation training execution module, and the collaborative control module, respectively. It stores various types of system data, supports dual local and cloud storage, and has data encryption, query, export, and remote interaction functions.

[0007] Compared with the prior art, the present invention provides a system for post-stroke depression combined with eye movement assessment and rehabilitation training, which has the following beneficial effects: 1. This invention deeply integrates eye-tracking technology with post-stroke depression assessment, overcoming the limitations of traditional scale assessments, which are highly subjective and easily affected by patient cooperation. It collects eye-tracking data from multiple scenarios using non-invasive devices, and through preprocessing, feature fusion, and model calculation, achieves objective quantification and accurate grading of depressive states. This allows for early identification of hidden depression, significantly improving assessment efficiency and reliability. Simultaneously, the eye-tracking assessment is entirely non-invasive and easy to operate, adapting to the characteristics of stroke patients with limited limb movement and cognitive decline, lowering the assessment threshold, and providing objective evidence for rapid clinical screening of post-stroke depression. This effectively addresses the pain points of traditional assessment methods, such as lag and large biases, facilitating early intervention.

[0008] 2. This invention achieves synergistic linkage and dynamic adaptation between depression assessment and rehabilitation training; the system can automatically calculate training intensity based on depression level, physical and cognitive baseline, and historical training effects, and adjust the training plan in a personalized manner. It also matches corresponding emotional intervention parameters through multi-sensory stimulation units to avoid training resistance or inefficient rehabilitation caused by depression. Compared with traditional rehabilitation systems, this solution takes into account both physical function and emotional rehabilitation, breaks the single rehabilitation model, forms a comprehensive intervention mechanism, and significantly improves rehabilitation compliance of stroke patients with depression. Attached Figure Description

[0009] Figure 1 This is a logic block diagram showing the relationships between the modules of this invention; Figure 2 This is a schematic diagram of the workflow of the system of the present invention. Detailed Implementation

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0011] Example: Please see Figures 1-2 This invention provides a post-stroke depression combined with eye movement assessment and rehabilitation training system, comprising: an eye movement data acquisition module, a post-stroke depression assessment module, a rehabilitation training execution module, a collaborative control module, and a data storage and interaction module. The modules are bidirectionally connected via wired or wireless communication to achieve real-time data transmission, command issuance, and status feedback, forming a closed-loop rehabilitation control system of assessment, analysis, decision-making, execution feedback, and reassessment. The specific implementation methods of each module are as follows: I. Eye-tracking data acquisition module In a specific embodiment of the present invention, the eye-tracking data acquisition module employs a non-invasive eye-tracking device with a sampling frequency of 120Hz, such as the Tobii Pro Spectrum or EyeLink 1000 Plus. This device integrates a high-precision infrared light source and a binocular camera, forming an adaptive adjustment unit that enables pupil center positioning and corneal reflection tracking without contact with the patient's eyes. The device incorporates a motion artifact removal algorithm, which, through Kalman filtering and adaptive threshold setting, effectively identifies and removes abnormal data points caused by momentary head movements, blinking, or eyelid occlusion, ensuring the continuity and reliability of the raw eye-tracking data.

[0012] This eye-tracking device also features adaptive adjustment. The built-in distance sensor has a measurement range of 0.3-1.5m and an accuracy of ±1cm, enabling real-time monitoring of the relative distance between the patient's eyes and the device. The facial recognition unit employs a deep convolutional neural network-based facial landmark detection algorithm, which can locate key feature points such as the center of the patient's eyes, corners of the eyes, and brow ridges in real time, and automatically adjust the acquisition angle and parameters based on the patient's head position and eye condition. The acquisition angle adjustment step is 1°, and the sampling frequency can adaptively switch between 120-150Hz. When the patient's head moves slightly, the system can complete the acquisition angle adjustment within 50ms; when the patient experiences eyelid drooping due to fatigue, the system can automatically increase the intensity of the infrared light source to enhance pupil contrast.

[0013] The aforementioned adaptive function can effectively adapt to the acquisition needs of patients in different positions, ensuring that high-quality eye movement data can still be stably acquired even when the patient's head moves slightly due to limb dysfunction or fatigue, with a acquisition success rate of over 95%.

[0014] During eye-tracking assessment, the system presents four assessment scenarios in a preset order, with a 15-second rest period between each scenario to avoid visual fatigue in the patient. 1. Emotional Face Gazing Scenario: The system sequentially displays images of three standard emotional faces—happy, sad, and neutral—on a 23.8-inch high refresh rate display screen, with four images of each face type, totaling 12 images. Each image is displayed for 4 seconds, with a 1.5-second image switching interval. The image presentation order is randomized to avoid sequence effects. The system records the patient's eye movement characteristics in real time at a sampling rate of 250Hz, including the initial gaze position, gaze duration percentage, and pupil diameter change rate for different emotional faces. This data is used to assess the patient's attentional bias towards emotional stimuli and their emotional processing ability.

[0015] 2. Visual Search Scenario: The screen simultaneously displays 8 visual targets and 12 interfering targets. The targets are circular patterns, while the interfering targets are square or irregular patterns. The visual difference between the targets and the interfering targets is controlled to be around 25%. The system records parameters such as the patient's saccade path length, number of fixations, target recognition reaction time, and saccade accuracy to evaluate visual search efficiency and saccade control ability.

[0016] 3. Sustained gaze scenario: A fixed red visual target with a diameter of 6mm is displayed in the center of the screen, and the patient is required to gaze at it continuously for 12 seconds. The system records parameters such as fixation point drift amplitude, gaze stability index, and micronystagmus frequency to assess fixation maintenance ability.

[0017] 4. Dynamic Emotion Stimulus Tracking Scenario: A 20-second dynamic emotion video clip at 30 frames per second is played. The video content includes a gradual emotional transition from calm to pleasure and from calm to sadness, using standardized stimulus materials from a professional emotion material library. The system records parameters such as the patient's gaze trajectory, pupillary dynamic response curve, and gaze tracking ability under dynamic stimuli at a sampling rate of 240Hz, to assess the dynamic eye-movement characteristics of emotional processing.

[0018] II. Post-stroke depression assessment module The post-stroke depression assessment module receives raw eye-tracking data transmitted from the eye-tracking data acquisition module and performs preprocessing. The preprocessing process is as follows: 1. Data Denoising: Gaussian filtering is used with a kernel size of 3×3 and a cutoff frequency of 12Hz to remove high-frequency noise and artifacts caused by blinking. For brief data loss due to blinking, the system uses a weighted average of adjacent valid data for smoothing to ensure the continuity of the data sequence.

[0019] 2. Missing Value Imputation: Cubic spline interpolation is used with an interpolation step size of 0.15 seconds. This interpolation method can better preserve the non-linear characteristics of the eye-tracking trajectory, ensuring that the deviation between the imputed data and the original data does not exceed 5%. When consecutive missing data exceeds 50ms, the system marks the data segment as invalid and performs special processing in subsequent feature extraction to avoid the accumulation of interpolation errors.

[0020] 3. Baseline Calibration: Eye movement data collected over 45 seconds in a calm state is used as the baseline. This baseline data is collected after the patient enters the assessment room but before the formal assessment begins. During data collection, the patient fixates on a fixed target point in the center of the screen, and the system records the eye movement parameters in a stable state. The raw eye movement data is then baseline-calibrated using a difference correction method, where the eye movement parameters at each time point are subtracted from the baseline average to eliminate intra-individual differences caused by fatigue or emotional fluctuations.

[0021] 4. Normalization: The maximum-minimum normalization method is used to map each eye movement feature data to the [0,1] interval. Through normalization, the influence of the difference in scale between different features on subsequent feature fusion can be eliminated, and the data scale is unified.

[0022] After preprocessing, the post-stroke depression assessment module performs feature extraction and fusion to construct an eye-tracking depression feature vector. Specifically, three types of feature components are extracted from the preprocessed eye-tracking data: 1. Mean temporal characteristics: These include mean fixation duration, mean saccade latency, fixation point density, and mean pupil diameter. These characteristics reflect the patient's basic eye movement control ability and visual attention status.

[0023] 2. Frequency Domain Feature Transformation Results: By performing a Fast Fourier Transform on the gaze trajectory sequence, frequency domain features such as gaze entropy and saccade power spectral density are extracted. Specifically, the gaze point coordinate sequence is used as a time-domain signal, and the power spectral density in the 0.5-10Hz frequency band is calculated after Fast Fourier Transform. This frequency band is closely related to attention maintenance and eye movement rhythm.

[0024] 3. Nonlinear feature index: The pupil dynamics complexity is calculated using the sample entropy algorithm, and the chaotic features of the fixation point sequence are calculated using the Lyapunov index.

[0025] The weight coefficients of each characteristic component are determined using principal component analysis. ,satisfy In practical applications, principal component analysis was performed on eye movement data from 300 patients with post-stroke depression and 100 healthy controls to obtain... =0.42, =0.33, =0.25. Eye movement depression feature vector The calculation formula is: This feature vector integrates time-domain, frequency-domain, and nonlinear features, and its value ranges from [0,1]. The higher the value, the more significant the depression-related eye movement abnormalities.

[0026] Subsequently, the module will use the eye movement depression feature vector Input into the built-in, pre-trained evaluation model. This model was trained based on eye movement data and clinical Hamilton Depression Rating Scale scores from 300 stroke patients with post-stroke depression. Depression Quantification Score The calculation formula is: in, The feature weight matrix, For bias terms, Let be the dimension of the eye-tracking depression feature vector. The calibration coefficient has a value range of [1.2, 1.5]. In this embodiment, it is taken as... =1.35; This is a correction term, with a value range of [0.01, 0.05]. In this embodiment, it is taken as... =0.02.

[0027] Based on ratings The post-stroke depression assessment module classifies depression into mild, moderate, or severe depression, and the classification results are simultaneously fed back to the collaborative control module.

[0028] III. Collaborative Control Module The collaborative control module includes a data processing unit, which integrates a circuit board and a wireless communication module. The collaborative control module receives the quantitative depression score output by the post-stroke depression assessment module. The system retrieves patient baseline data and historical training data stored in the data storage and interaction module, along with the status grading results.

[0029] The collaborative control module, combining the above data, calculates the final training intensity parameter using the following formula. : The following are examples of the parameter settings and calculations in this formula: is the weighting coefficient for the depression score, with a value range of [0.4, 0.6]. In this embodiment, it is taken as... =0.5; These are all weighting coefficients, with values ​​ranging from 0.1 to 0.3. This embodiment uses the default values. =0.25、 =0.2、 =0.15; The normalized value of the baseline data for limb function is obtained by dividing the Fugl-Meyer score by 100. The normalized value of the baseline cognitive function data is obtained by dividing the MMSE score by 30. For historical training times, For the first Training effect score, The adjustment coefficient has a value range of [-0.8, -0.3]. In this embodiment, it is taken as... =-0.5; This represents the change in depression quantification scores.

[0030] Based on the patient's limb function rating and cognitive function score, the collaborative control module can... Make adaptive fine-tuning: when the limb function assessment score is <60, Adjusted upwards to 0.28-0.3; when the cognitive function score is <20, Adjust to 0.25-0.3; when the average historical training performance score is <60 points, Adjusted upwards to 0.25-0.3.

[0031] The final training intensity parameters obtained from the calculation Used to generate personalized training instructions and send them to the rehabilitation training execution module.

[0032] IV. Rehabilitation Training Implementation Module The rehabilitation training execution module integrates a motion capture unit, a force feedback unit, and a multi-sensory stimulation unit. The motion capture unit uses a non-invasive inertial measurement unit sensor, which is worn on key joints of the patient's upper and lower limbs via a fixed structure. The sensor sampling frequency is 100-120Hz, with a measurement accuracy of ±0.5°, and it collects real-time data on the patient's limb movement amplitude, speed, and accuracy at a cycle of 0.5-1 seconds per acquisition. The collected motion data is then fed back to the collaborative control module in real time.

[0033] The multi-sensory stimulation unit adaptively adjusts the intensity and type of stimulation based on the depressive state classification results: For patients with mild depression: use moderate intensity stimulation, with visual brightness of 500-800 lux and auditory volume of 50-60 dB, and stimulation types mainly consisting of positive emotional visual materials and soothing music; For patients with moderate depression: use low to medium intensity stimulation, with visual brightness of 300-500 lux and auditory volume of 40-50 dB, and the stimulation type should mainly consist of neutral emotional materials and soft music; For patients with severe depression: use low-intensity stimulation, with visual brightness of 100-300 lux and auditory volume of 30-40 dB, and stimulation types mainly consisting of soft visual light effects and natural white noise.

[0034] V. Data Storage and Interaction Module The data storage and interaction module communicates with the eye-tracking data acquisition module, post-stroke depression assessment module, rehabilitation training execution module, and collaborative control module to store all system data. This module supports both local and cloud storage, and features data encryption, querying, exporting, and remote interaction capabilities, employing a three-tiered access control system.

[0035] VI. The workflow of this system is as follows: 1. Initial Assessment Phase: When a patient uses the system for the first time, medical staff enter the patient's baseline data through the data storage and interaction module, including stroke onset time, limb function rating, cognitive function score, and history of depression. Subsequently, the eye-tracking data acquisition module collects the patient's raw eye-tracking data in scenarios involving emotional face gaze, visual search, sustained gaze, and dynamic emotional stimulus tracking. The post-stroke depression assessment module preprocesses, extracts, and fuses the collected eye-tracking data to construct an eye-tracking depression feature vector, and outputs a quantitative depression score and state grading result through the built-in assessment model. The initial assessment result serves as the patient's baseline data storage.

[0036] 2. Collaborative Decision-Making Phase: The collaborative control module receives the quantitative depression score and state grading results output by the post-stroke depression assessment module, and retrieves the patient's baseline data and historical training data stored in the data storage and interaction module. The module combines multi-dimensional parameters such as depression score, limb function baseline data, cognitive function baseline data, average historical training effects, and changes in depression score, and generates the final training intensity parameter through a built-in linkage calculation model. Based on this parameter, it automatically maps to a preset training program template, generates personalized training instructions, and sends them to the rehabilitation training execution module.

[0037] 3. Rehabilitation Execution Phase: The rehabilitation training execution module receives personalized training instructions from the collaborative control module and executes rehabilitation training in areas such as limb movement, cognition, and emotion regulation. During training, the motion capture unit collects data on the patient's limb movement amplitude, speed, and accuracy in real time at a sampling frequency of 100-120Hz and feeds this data back to the collaborative control module. The force feedback unit dynamically adjusts resistance or assistive force according to the training task, achieving task-oriented training. The multi-sensory stimulation unit adaptively outputs visual, auditory, and tactile stimuli of appropriate intensity and type based on the depression status classification results. The collaborative control module can dynamically adjust training parameters during a single training session based on the real-time feedback motion data, forming a real-time closed-loop control.

[0038] 4. Periodic Reassessment Phase: After each preset number of training sessions, the system automatically triggers an eye-tracking reassessment, repeating the initial assessment process and updating the depression quantification score and state grading results. The collaborative control module, based on the updated depression score and its changes, combined with the latest historical training data, recalculates the training intensity parameters and generates a new training plan. Through periodic assessment and adjustment, the treatment plan is dynamically optimized according to the patient's depressive state and recovery progress.

[0039] 5. Data Recording and Sharing Phase: During system operation, all assessment data, training data, and parameter adjustment records are encrypted and stored through the data storage and interaction module. Medical staff can log in to the system via multiple terminals to view real-time visualization reports such as patient depression score trend graphs, training effect curves, and motor function recovery curves. When a patient's depression score exceeds the severe depression threshold for two consecutive assessments, the system automatically sends an early warning notification to the attending physician, indicating the need for psychological intervention. In remote consultation mode, with patient authorization, data can be shared with collaborating hospitals, facilitating multidisciplinary team collaborative management of the patient's recovery process.

[0040] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A system for combined eye-tracking assessment and rehabilitation training for post-stroke depression, characterized in that, include: The eye-tracking data acquisition module is used to collect raw eye-tracking data such as the patient's gaze coordinates and saccade speed in a preset eye-tracking assessment scenario. It uses a non-invasive eye-tracking device with a sampling frequency of ≥120Hz and has motion artifact removal capabilities. The post-stroke depression assessment module receives raw eye movement data collected by the eye movement data acquisition module, performs preprocessing, feature extraction and fusion, constructs an eye movement depression feature vector, and outputs a quantitative depression score and state classification result through a built-in trained assessment model. Its feature vector construction and score calculation are achieved through the following formula: in: This is the feature vector of eye movement depression. For eye movement feature components, , The weight coefficients of the eigenvectors are obtained through principal component analysis. The mean of the time-domain features, The result is the frequency domain feature transformation. It is a non-linear characteristic index; in: Quantifying depression scores, The feature weight matrix, For bias terms, This is the calibration coefficient, and its value range is [1.2, 1.5]. This is a correction term, and its value range is [0.01, 0.05]. Let be the dimension of the eye-tracking depression feature vector. For summation index variables; The rehabilitation training execution module receives personalized training instructions from the collaborative control module and executes rehabilitation training in the areas of limb, cognition, and emotion regulation. It integrates motion capture, force feedback, and multi-sensory stimulation units. The collaborative control module receives the depression quantitative score and state grading results output by the post-stroke depression assessment module, retrieves the patient's baseline and historical training data, generates and issues personalized training instructions, and receives training data feedback from the rehabilitation training execution module in real time to form a closed-loop control, dynamically adjusting training parameters. The training parameter adaptation coefficient and the final training intensity parameter are calculated in conjunction using the following formula: in: The value range is [0.4, 0.6]. , All are weighting coefficients. Normalized values ​​for baseline limb function data. Normalized values ​​for baseline cognitive function data. For historical training times, For the first Training effect score, This is an adjustment coefficient, and its value range is [-0.8, -0.3]. This represents the change in depression quantification scores. For the final training intensity parameters; The data storage and interaction module is communicatively connected to the eye-tracking data acquisition module, the post-stroke depression assessment module, the rehabilitation training execution module, and the collaborative control module, respectively. It stores various types of system data, supports dual local and cloud storage, and has data encryption, query, export, and remote interaction functions.

2. The post-stroke depression combined with eye movement assessment and rehabilitation training system according to claim 1, characterized in that: The eye-tracking data acquisition module's eye-tracking assessment scenarios specifically include: emotional face fixation scenario, visual search scenario, continuous gaze scenario, and dynamic emotional stimulus tracking scenario. The emotional face fixation scenario includes standard emotional face images of three categories: happy, sad, and neutral. Each type of face image is displayed for 3-5 seconds, switching every 1-2 seconds. The visual search scenario sets 5-8 visual targets and 10-15 interference targets, with the visual difference between the targets and interference targets controlled at 20%-30%. The continuous gaze scenario sets a single fixed visual target point with a diameter of 5-8mm, requiring a gaze duration of 10-15 seconds. The dynamic emotional stimulus tracking scenario uses dynamic emotional video clips, with a video duration of 15-20 seconds and a frame rate of 24-30 frames per second, to specifically capture the differences in the patient's eye-tracking responses to different emotional stimuli.

3. The post-stroke depression combined with eye movement assessment and rehabilitation training system according to claim 1, characterized in that: The preprocessing procedure of the post-stroke depression assessment module specifically includes: Data denoising was performed by removing high-frequency interference signals from the original eye-tracking data. The filtering parameters were set to kernel size 3×3 and cutoff frequency 10-15Hz. Missing values ​​are imputed using interpolation with an interpolation step size of 0.1-0.2 seconds to ensure that the imputed data deviates from the original data by no more than 5%. Baseline calibration is performed by using eye movement data from 30-60 seconds of the patient's resting state as the baseline to correct the baseline of the raw eye movement data. Normalization is performed by mapping the data to the [0,1] interval to eliminate interference factors in the original eye-tracking data, unify the data scale, and provide a basis for feature extraction and fusion.

4. The post-stroke depression combined with eye movement assessment and rehabilitation training system according to claim 1, characterized in that: The status grading results of the post-stroke depression assessment module are specifically divided into three levels: mild depression, moderate depression, and severe depression. The quantitative depression score is as follows: This is mild depression, which corresponds to patients experiencing mild low mood and does not affect daily rehabilitation training. The patient is diagnosed with moderate depression, characterized by significant low mood and loss of interest. The intensity of rehabilitation training needs to be adjusted appropriately. The patient is classified as having severe depression, which corresponds to severe low mood and resistance to training. High-intensity training should be suspended and emotional intervention should be prioritized. The classification results are simultaneously fed back to the collaborative control module.

5. The post-stroke depression combined with eye movement assessment and rehabilitation training system according to claim 1, characterized in that: The motion capture unit of the rehabilitation training execution module uses a non-invasive inertial measurement unit sensor with a sampling frequency of 100-120Hz and a measurement accuracy of ±0.5°. The sensor is worn on the key joints of the patient's upper and lower limbs to collect data on the patient's limb movement amplitude (range 0-180°), movement speed (0-50cm / s), and movement accuracy (threshold ≥80%) in real time. The collection cycle is 0.5-1 seconds / time, and the collected motion data is fed back to the collaborative control module in real time for training effect evaluation and parameter adjustment.

6. The post-stroke depression combined with eye movement assessment and rehabilitation training system according to claim 1, characterized in that: The patient baseline data retrieved by the collaborative control module specifically includes: the patient's stroke onset time (accurate to the day), limb function rating (using the Fugl-Meyer rating scale, with a score range of 0-100), cognitive function rating (using the Mini-Mental State Examination (MMSE), with a score range of 0-30), and relevant data on past history of depression (including whether there is a history of depression, the number of previous depressive episodes, and treatment). The baseline data is pre-entered into the system by medical staff. The collaborative control module combines the above baseline data with the depression quantitative score to assist in generating personalized rehabilitation training instructions.

7. The post-stroke depression combined with eye movement assessment and rehabilitation training system according to claim 1, characterized in that: The collaborative control module All are fixed values ​​between 0.1 and 0.3, where The default value is 0.

25. The default value is 0.

2. The default value is 0.15, and it can be adaptively fine-tuned based on the patient's limb function rating and cognitive function score. When the limb function rating is <60 points, it is possible to Adjusted upwards to 0.28-0.3, when the cognitive function score is <20. It can be adjusted upwards to 0.25-0.3 when the average historical training effect score is <60. It can be adjusted upwards to 0.25-0.

3.

8. The post-stroke depression combined with eye movement assessment and rehabilitation training system according to claim 1, characterized in that: The data storage and interaction module encrypts and protects patient privacy data (including eye-tracking data, depression assessment data, and basic personal information). The encryption permission management function adopts a three-level permission setting. The first-level permission allows access to all data and system settings, the second-level permission allows access to all assessment and training data of the patients under their care, and the third-level permission allows access only to patient training data.

9. A post-stroke depression combined with eye movement assessment and rehabilitation training system according to claim 1, characterized in that: The eye-tracking device of the eye-tracking data acquisition module also has an adaptive adjustment function. The device has a built-in distance sensor (measurement range 0.3-1.5m, measurement accuracy ±1cm) and a facial recognition unit. It can automatically adjust the acquisition angle and parameters according to the patient's head position (horizontal adjustment range ±30°, vertical adjustment range ±20°) and eye status (pupil diameter change). The acquisition angle adjustment step is 1°, and the sampling frequency can be adaptively switched between 120-150Hz to meet the acquisition needs of patients in different positions (sitting, semi-recumbent, supine).

10. A post-stroke depression combined with eye movement assessment and rehabilitation training system according to claim 1, characterized in that: The multi-sensory stimulation unit of the rehabilitation training execution module can adaptively adjust the stimulation intensity and type based on the depression status grading results output by the post-stroke depression assessment module. For patients with mild depression, moderate-intensity stimulation (visual brightness 500-800 lux, auditory volume 50-60 dB) is used, with the stimulation type mainly consisting of positive emotional visual materials and soothing music; For patients with moderate depression, low to moderate intensity stimulation (visual brightness 300-500 lux, auditory volume 40-50 dB) is used, with the stimulation type mainly consisting of neutral emotional material and soft music. Patients with severe depression are given low-intensity stimulation (visual brightness 100-300 lux, auditory volume 30-40 dB), with stimulation types mainly consisting of soft visual light effects and natural white noise. The stimulation intensity for patients with severe depression is lower than that for patients with mild and moderate depression.

Images