A near real-time consciousness detection system for improving CRS-R accuracy and for pDoC intelligent rehabilitation

Abstract

This invention provides a near-real-time consciousness detection system for improving CRS-R accuracy and for pDoC intelligent rehabilitation, belonging to the field of medical equipment and detection technology. It includes: an integrated wearable hardware device integrating a BIS module, an open / closed eye scene camera module, a VR headset, and a VR eye tracker, all components packaged in the same wearable shell and worn on the patient's head via a detachable headband; a software evaluation module including a trigger unit, a two-level VR eye-tracking evaluation unit, and a data processing unit, which performs two-level evaluation to quantify consciousness levels, determine a personalized arousal-consciousness fluctuation curve, automatically mark intervention time windows for rehabilitation, send optimal intervention timing signals to intelligent rehabilitation devices, and map eye movement trajectory features to CRS-R visual sub-item scores through a machine learning model. This effectively solves the problems of reliance on manual evaluation, difficulty in high-frequency detection, and high misdiagnosis rates, while providing accurate intervention timing signals for intelligent rehabilitation devices, improving the targeting and effectiveness of rehabilitation treatment.

Description

Technical Field

[0001] This invention relates to the field of medical equipment and testing technology, and in particular to a near real-time consciousness detection system for improving the accuracy of CRS-R and for use in pDoC intelligent rehabilitation. Background Technology

[0002] Chronic disorder of consciousness (pDoC) refers to a pathological state of loss of consciousness exceeding 28 days caused by various brain injuries, such as traumatic brain injury, stroke, and hypoxic-ischemic encephalopathy. Patients with chronic disorder of consciousness (pDoC), often simply referred to as pDoC patients, are currently classified into two main categories based on their level of consciousness: vegetative state (VS) and minimally conscious state (MCS). Accurate assessment of the level of consciousness in pDoC patients is crucial for developing personalized rehabilitation plans. Currently, three main types of techniques are used: behavioral scales, neuroimaging techniques, and neurophysiological techniques. Each has its advantages and limitations, and clinical guidelines recommend using a multimodal approach. All major guidelines recommend the CRS-R (Continuous Responsibility Scale-Report) as a primary screening and follow-up tool.

[0003] Theoretically, behavioral scales, fMRI, EEG, and fNIRS can all relatively quickly reveal a patient's state of consciousness to a certain extent, possessing the potential for real-time assessment. However, some techniques require pre-assessment data collection and processing before generating results. For example, EEG microstate detection systems require at least ten minutes of data collection for analysis, but the pre-collection setup and other steps are very time-consuming, and the resulting classification of consciousness is also quite coarse. Furthermore, the equipment is expensive, preventing large-scale deployment for individual patients.

[0004] Therefore, due to limitations in clinical practice, only behavioral scales are currently widely used for real-time assessment of consciousness. However, existing behavioral scale assessments rely heavily on patients' motor function and doctors' frequent and detailed examinations. Since pDoC patients have limited motor function and doctors lack the energy and resources to conduct frequent and detailed examinations on every patient, it is easy to miss the optimal moment of patients' consciousness. The combination of these factors leads to a misdiagnosis rate of up to 40% in assessing the consciousness of pDoC patients, resulting in some patients not being allocated appropriate medical resources due to misjudgment.

[0005] Therefore, the present invention provides a near real-time consciousness detection system for improving CRS-R accuracy and for pDoC intelligent rehabilitation. Summary of the Invention

[0006] This invention provides a near real-time consciousness detection system for improving CRS-R accuracy and for pDoC intelligent rehabilitation, in order to solve the aforementioned technical problems.

[0007] This invention provides a near real-time consciousness detection system for improving CRS-R accuracy and for pDoC intelligent rehabilitation, comprising: The integrated wearable hardware device integrates a BIS module, an open and closed eye scene camera module, a VR headset and a VR eye tracker. All components are packaged in the same wearable shell and worn on the patient's head via a detachable headband. The software evaluation module includes a trigger unit, a two-level VR eye-tracking evaluation unit and a data processing unit. The triggering unit is used to activate the two-level VR eye-tracking assessment unit when the BIS module detects that the patient's arousal value is higher than a preset threshold and the eye-opening and closing scene camera module confirms that the patient's eyes are open based on a visual algorithm. The two-level VR eye-tracking evaluation unit is used when the first level performs visual tracking judgment. At this time, the VR headset presents a moving target, and the eye tracker collects the eye movement trajectory and compares it with the CRS-R visual tracking standard. Once the first-level assessment is passed, the patient automatically proceeds to the second-level assessment. At this point, a multiple-choice question with images and text is presented using a VR headset. The patient answers by focusing on specific points, and the level of consciousness is quantified based on the correctness of the answer and the reaction time. The data processing unit is used to continuously record BIS values, eye-opening events, and two-level assessment results, generate a personalized arousal-consciousness fluctuation curve with time as the axis, automatically mark the time window for intervention and rehabilitation, send the optimal intervention timing signal to the intelligent rehabilitation device, and map eye movement trajectory features into CRS-R visual sub-item scores through a machine learning model and send them to both ends. The dual ends include a doctor-side APP and a patient-side APP equipped with applications, which are used to communicate and interact with the integrated wearable hardware device.

[0008] Preferred options also include: An LED status indicator is installed on the shell of the integrated wearable hardware device. When the first level of evaluation is passed, it receives the first indication signal transmitted by the two-level VR eye-tracking evaluation unit and changes from the first display color to the second display color. The LED status indicator is used to receive the second indication signal transmitted by the two-level VR eye-tracking evaluation unit after the second-level evaluation is passed, and change from the second display color to the third display color, and maintain the third display color before the next evaluation; The total evaluation time for the first-level evaluation and the second-level evaluation shall last at least N minutes.

[0009] Preferably, the software evaluation module further includes: The fatigue protection unit is used to activate the VR display screen when the two-level VR eye-tracking evaluation unit is started, and after the total evaluation time of the first and second level evaluations lasts for N minutes, send a screen-off instruction to the VR display screen and enter the screen-off state. The VR display screen is a component of the VR headset. The pressure ulcer protection unit is used to control the movable automatic traction bracket to pull the VR eye tracker up when it receives a signal that the VR display screen has entered the screen-off state under the screen-off instruction, and to control the movable automatic traction bracket to automatically return the VR eye tracker to its original position when it receives a signal that it needs to be re-evaluated.

[0010] Preferably, the doctor's app is used to scan the QR code of the integrated wearable hardware device and automatically associate it with the patient's bed number, hospital number, name, and cause of illness. The doctor-side APP is also used to send instructions to the integrated wearable hardware device to force the BIS module and the two-level VR eye-tracking assessment unit to start or stop. The doctor-side APP is also used to display the VS status with a first display color, the MCS- status with a second display color, and the MCS+ / eMCS status with a third display color on the first doctor display interface. The colors are synchronized in real time with the most recent two-level VR eye-tracking assessment results. The second doctor display interface simultaneously displays a personalized arousal-consciousness fluctuation curve, a prompt for interventional rehabilitation time window, and the CRS-R visual sub-item score.

[0011] Preferably, the patient-side APP is used to present BIS values, eye-opening events, two-level assessment results, and changes in consciousness level in a timeline manner on the first patient display interface, and to display a segment highlighted on the personalized arousal-consciousness fluctuation curve based on the interventional rehabilitation time window on the second patient display interface, so that the patient can click on the highlighted segment on the third patient display interface to jump to the brain-computer interface rehabilitation device control interface. The patient-side app is also used to obtain the generation password authorized by the associated doctor and to receive additional assessments from family members within a limited time.

[0012] Preferably, the mechanical and circuit structure of the integrated wearable hardware device is as follows: the BIS electrode is arranged inside the wearable shell corresponding to the patient's forehead, the camera of the VR eye tracker and the scene camera are embedded in the front end of the VR headset, and the VR display screen is installed inside the wearable shell corresponding to the patient's eye position.

[0013] Preferred options also include: The control module is used to control the VR display screen to present at least one dynamic test element to conduct a preliminary test on the patient before presenting the text and image multiple-choice questions based on the VR headset, obtain quantitative auxiliary factors, and optimize the quantification process of answer correctness and reaction time to obtain the quantified consciousness level.

[0014] Preferably, the control module includes: The control unit is used to control the VR display screen to present at least one dynamic test element before presenting text and image multiple-choice questions based on the VR headset, and to preset a target attribute combination mode for each dynamic test element. The visual attribute dimensions of each dynamic test element include at least color, size, and brightness. The acquisition unit is used to synchronously acquire the patient's raw eye movement data based on the eye tracker during the test; The real-time processing unit is used to process the raw eye movement data in real time and parse it into a sequence of eye movement microstates in fine time windows. The eye movement microstates include: stable fixation, target-oriented saccades, retrospection, exploratory fixation, and targetless drift. The pattern generation unit is used to generate a theoretically expected eye-tracking pattern based on the features of the corresponding visual attribute dimension in the preset target attribute combination pattern. A line construction unit is used to time-align the eye-tracking microstate sequence with the theoretically expected eye-tracking pattern, determine the fluctuation difference of the comparative states at the same alignment time point, and form a fluctuation trend line. The parameter value determination unit is used to perform time alignment between the raw eye-tracking data and the standard eye-tracking data and divide them into N continuous analysis windows, and determine the spatial dwell ratio Zb, velocity adaptation index Zs, and trajectory morphology similarity Zs of each continuous analysis window to obtain consistency. The range determination unit is used to determine the overlap ratio range based on all consistency factors. ,in, These are the lower and upper limits of the overlap ratio range, respectively; The window division unit is used to start from the first initial window and sequentially set the starting point of the next window based on a random percentage of the overlap ratio range until the time alignment segment division is completed. The overlapping window and the continuous analysis window are the same size but may have different time starting points, and the initial window is the first window in the overlap division result. The comparison analysis unit is used to time-align the first division result of continuous division with the second division result of overlapping division, and obtain the set of overlapping windows that are in comparison with each continuous analysis window, wherein the set of overlapping windows involves at least one overlapping window. A division unit is used to divide the fluctuation trend line according to a continuous analysis window to obtain the fluctuation trend item of each continuous analysis window; The confidence determination unit is used to obtain the response confidence of the corresponding visual attribute dimension based on the overlapping window set and the consistency standard deviation of the corresponding continuous analysis window, and in combination with the consistency and fluctuation trend term of the corresponding continuous analysis window, and to construct the response confidence vector of the corresponding visual attribute dimension. The factor determination unit is used to calculate the stability index within each vector and the cross-dimensional matching index between each vector based on the response confidence vectors of all visual attribute dimensions. The results are input into the pre-trained intent inference model, and the output is the overall intent confidence that represents the patient's integrated response intent to the target attribute combination pattern as a quantification auxiliary factor to optimize the quantification process.

[0015] Compared with the prior art, the beneficial effects of this application are as follows: By integrating core functional components into an all-in-one wearable hardware device, and combining the dual-condition triggering, two-level progressive assessment, and data processing logic of the software assessment module, near real-time, all-weather monitoring of the consciousness state of pDoC patients is achieved. The generated personalized arousal-consciousness fluctuation curve and CRS-R visual sub-item score provide objective data support for clinical diagnosis. The communication linkage between the two-end applications ensures that doctors and families can monitor the patient's status in real time, effectively solving the problems of reliance on manual assessment, difficulty in high-frequency detection, and high misdiagnosis rate in existing technologies. At the same time, it provides precise intervention timing signals for intelligent rehabilitation devices, improving the pertinence and effectiveness of rehabilitation treatment.

[0016] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0017] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a structural diagram of a near-real-time consciousness detection system for improving CRS-R accuracy and for pDoC intelligent rehabilitation, as described in an embodiment of the present invention. Figure 2 This is a structural diagram of the traction hanger in an embodiment of the present invention. Detailed Implementation

[0019] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0020] This invention provides a near real-time consciousness detection system for improving CRS-R accuracy and for pDoC intelligent rehabilitation, such as... Figure 1 As shown, it includes: The integrated wearable hardware device integrates a BIS module, an open and closed eye scene camera module, a VR headset and a VR eye tracker. All components are packaged in the same wearable shell and worn on the patient's head via a detachable headband. The software evaluation module includes a trigger unit, a two-level VR eye-tracking evaluation unit and a data processing unit. The triggering unit is used to activate the two-level VR eye-tracking assessment unit when the BIS module detects that the patient's arousal value is higher than a preset threshold and the eye-opening and closing scene camera module confirms that the patient's eyes are open based on a visual algorithm. The two-level VR eye-tracking evaluation unit is used when the first level performs visual tracking judgment. At this time, the VR headset presents a moving target, and the eye tracker collects the eye movement trajectory and compares it with the CRS-R visual tracking standard. Once the first-level assessment is passed, the patient automatically proceeds to the second-level assessment. At this point, a multiple-choice question with images and text is presented using a VR headset. The patient answers by focusing on specific points, and the level of consciousness is quantified based on the correctness of the answer and the reaction time. The data processing unit is used to continuously record BIS values, eye-opening events, and two-level assessment results, generate a personalized arousal-consciousness fluctuation curve with time as the axis, automatically mark the time window for intervention and rehabilitation, send the optimal intervention timing signal to the intelligent rehabilitation device, and map eye movement trajectory features into CRS-R visual sub-item scores through a machine learning model and send them to both ends. The dual ends include a doctor-side APP and a patient-side APP equipped with applications, which are used to communicate and interact with the integrated wearable hardware device.

[0021] Preferred options also include: An LED status indicator is installed on the shell of the integrated wearable hardware device. When the first level of evaluation is passed, it receives the first indication signal transmitted by the two-level VR eye-tracking evaluation unit and changes from the first display color to the second display color. The LED status indicator is used to receive the second indication signal transmitted by the two-level VR eye-tracking evaluation unit after the second-level evaluation is passed, and change from the second display color to the third display color, and maintain the third display color before the next evaluation; The total evaluation time for the first-level evaluation and the second-level evaluation shall last at least N minutes.

[0022] Preferably, the software evaluation module further includes: The fatigue protection unit is used to activate the VR display screen when the two-level VR eye-tracking evaluation unit is started, and after the total evaluation time of the first and second level evaluations lasts for N minutes, send a screen-off instruction to the VR display screen and enter the screen-off state. The VR display screen is a component of the VR headset. The pressure ulcer protection unit is used to control the movable automatic traction bracket to pull the VR eye tracker up when it receives a signal that the VR display screen has entered the screen-off state under the screen-off instruction, and to control the movable automatic traction bracket to automatically return the VR eye tracker to its original position when it receives a signal that it needs to be re-evaluated.

[0023] Preferably, the doctor's app is used to scan the QR code of the integrated wearable hardware device and automatically associate it with the patient's bed number, hospital number, name, and cause of illness. The doctor-side APP is also used to send instructions to the integrated wearable hardware device to force the BIS module and the two-level VR eye-tracking assessment unit to start or stop. The doctor-side APP is also used to display the VS status with a first display color, the MCS- status with a second display color, and the MCS+ / eMCS status with a third display color on the first doctor display interface. The colors are synchronized in real time with the most recent two-level VR eye-tracking assessment results. The second doctor display interface simultaneously displays a personalized arousal-consciousness fluctuation curve, a prompt for interventional rehabilitation time window, and the CRS-R visual sub-item score.

[0024] Preferably, the patient-side APP is used to present BIS values, eye-opening events, two-level assessment results, and changes in consciousness level in a timeline manner on the first patient display interface, and to display a segment highlighted on the personalized arousal-consciousness fluctuation curve based on the interventional rehabilitation time window on the second patient display interface, so that the patient can click on the highlighted segment on the third patient display interface to jump to the brain-computer interface rehabilitation device control interface. The patient-side app is also used to obtain the generation password authorized by the associated doctor and to receive additional assessments from family members within a limited time.

[0025] Preferably, the mechanical and circuit structure of the integrated wearable hardware device is as follows: the BIS electrode is arranged inside the wearable shell corresponding to the patient's forehead, the camera of the VR eye tracker and the scene camera are embedded in the front end of the VR headset, and the VR display screen is installed inside the wearable shell corresponding to the patient's eye position.

[0026] In this embodiment, the integrated wearable hardware device refers to a portable device that integrates multiple functional components into the same shell and is fixed to the patient's head by wearing. The surface is rounded to avoid sharp edges from scratching the patient. It is worn on the patient's head by a detachable headband (the headband is made of breathable nylon, 3cm wide, and adjustable in length from 45 to 60cm). The inside of the headband is equipped with an anti-slip silicone strip to ensure that the device does not shift when the patient moves slightly.

[0027] The BIS module, or Bispectral Index module, is a hardware module that collects electroencephalogram (EEG) signals from the patient's forehead and analyzes the bispectral index to quantitatively assess the patient's level of arousal. The BIS value ranges from 0 to 100, with higher values ​​indicating higher levels of arousal. The module contains four silver chloride electrode pads that come into contact with the patient's forehead skin through medical conductive gel. The sampling frequency is 128 Hz, and the BIS value is output in real time.

[0028] The open / closed eye scene camera module refers to a camera component that integrates scene shooting and eye state recognition functions. It is used to collect images of the patient's eye area and determine the opening / closing state of the patient's eyes through visual algorithms.

[0029] VR headsets, or virtual reality headsets, are devices that provide patients with a closed visual environment and present visual content, used to deliver visual stimuli needed for assessment.

[0030] VR eye trackers are devices integrated into VR headsets to collect real-time data on a patient's eye movement trajectory, capturing parameters such as the patient's eye rotation angle, gaze point position, and saccade speed.

[0031] The software evaluation module refers to the software system running on the control motherboard and dual-end applications of the integrated wearable hardware device.

[0032] In this embodiment, the judgment logic of the triggering unit adopts the logic of satisfying two conditions at the same time. By reading the output data of the BIS module and the judgment result of the eye-opening and closing scene camera module in real time, when both conditions are satisfied and the duration is ≥3 seconds, a start signal is sent to the two-level VR eye-tracking evaluation unit to avoid false triggering caused by instantaneous signal fluctuations.

[0033] The preset threshold refers to a pre-set BIS value standard used to determine whether a patient's level of arousal meets the assessment requirements. This threshold is determined based on clinical data statistics and expert consensus. For example, if the preset threshold is set to 60, it means that when the BIS value detected by the BIS module is ≥60, the patient's level of arousal is considered to meet the assessment requirements. If the patient has a special pathological condition (such as the influence of sedation drugs), the doctor can adjust the preset threshold to any value between 55 and 65 through the doctor's app, with an adjustment step of 1.

[0034] In this embodiment, the first and second level assessments adopt a progressive logic. If the first level assessment fails, the second level assessment will not be initiated. Assessment data is recorded in real time during the assessment process. If the assessment is interrupted (such as when the patient's eyes are closed), the collected data is saved and will continue to be executed when the assessment is triggered again.

[0035] Visual tracking assessment is an evaluation process to determine whether a patient has the ability to track objects. For example, the CRS-R visual tracking standard is as follows: the patient's eyes can follow the movement of a moving target with a movement angle ≥15° and a continuous following time ≥3 seconds. The moving target presented by the VR headset is a white circular light spot (1cm in diameter), the movement trajectory is a horizontal reciprocating motion, the movement speed is 0.5cm / s, and the movement range covers 80% of the horizontal area of ​​the VR display screen. The eye tracker collects the movement trajectory of the patient's eyes following the light spot, and the software algorithm calculates the movement angle and duration. If the CRS-R standard is met, the first level of assessment is considered passed.

[0036] The CRS-R visual tracking standard refers to the standard used in CRS-R to assess a patient's visual tracking ability, and it is a clinically recognized reference for consciousness assessment.

[0037] Image-text multiple-choice questions refer to multiple-choice questions presented on a VR headset during the Level 2 assessment. These questions combine images and text and are used to test the patient's selective attention and cognitive response abilities. For example, the question stem for an image-text multiple-choice question is accompanied by a voice prompt: "Please look at an item you use daily." The options consist of two images and text: on the left is an image of a cup (3cm x 3cm) with the text "cup" (in bold, 24pt font); on the right is an image of a stone (3cm x 3cm) with the text "stone" (in bold, 24pt font).

[0038] Fixation-based answering refers to a method where patients select the correct option from the text and images to complete the question. The eye tracker determines the patient's choice by collecting the fixation point position and fixation duration. For example, if a fixation duration of ≥2 seconds is set as a valid selection, the eye tracker collects the patient's fixation point coordinates in real time. If the patient fixates on the left "cup" option for ≥2 seconds, the patient is considered to have selected that option; if the patient fixates on the right "stone" option for ≥2 seconds, the patient is considered to have selected that option; if the fixation duration for both options is less than 2 seconds or the fixation point is frequently switched, the answer is considered invalid, and the question is re-presented (maximum of 2 re-presentations).

[0039] The Personalized Awakening-Consciousness Fluctuation Curve is a curve generated by plotting time on the horizontal axis and the patient's BIS value, eye-opening state, and level of consciousness on the vertical axis, reflecting changes in the patient's awakening and state of consciousness over a period of time. The horizontal axis represents time (in hours), and the vertical axis is divided into three dimensions: BIS value (0-100), eye-opening state (0=eyes closed, 1=eyes open), and level of consciousness (0=VS, 1=MCS-, 2=MCS+ / eMCS). The curve uses different colors and line types to distinguish the three dimensions (BIS value is represented by a solid blue line, eye-opening state by a dashed red line, and level of consciousness by a green bar chart). The generated curve can be zoomed and panned, and data is retained for 7 days.

[0040] The intervention-friendly rehabilitation time window refers to the time period automatically identified based on the personalized arousal-consciousness fluctuation curve when the patient's consciousness is in a good state and suitable for rehabilitation intervention. For example, when the patient's consciousness level reaches MCS- or above, and the duration of this state is ≥15 minutes, the time period is automatically marked as the intervention-friendly rehabilitation time window. The starting point of the time window is the moment when the consciousness level reaches MCS-, and the ending point is the moment when the consciousness level drops to VS or the duration ends. The marking method is to highlight the time period with a green background on the fluctuation curve.

[0041] A machine learning model refers to an algorithmic model used to map eye movement trajectory features to CRS-R visual sub-item scores, which have predictive capabilities after being trained on training data. For example, a machine learning model can be built using the random forest algorithm. The training data includes eye movement trajectory data from 100 pDoC patients and corresponding manual scores for CRS-R visual sub-items. The model input is eye movement trajectory features (fixation duration, saccade amplitude, accuracy, etc.), and the output is the CRS-R visual sub-item score (0-4 points). The model's prediction accuracy is ≥85%. After training, the model is stored in the control motherboard's storage unit and supports online fine-tuning.

[0042] Eye movement characteristics refer to parameters collected by an eye tracker that reflect a patient's visual cognitive abilities, including fixation duration, saccade amplitude, saccade speed, fixation point stability, and answer accuracy. For example, fixation duration refers to the duration (in seconds) a patient focuses on a target; for instance, the duration of focusing on the "cup" option is 3.5 seconds. Saccades refer to the angle (in degrees) the eye moves from one fixation point to another; for example, the saccade amplitude from the left option to the right option is 25 degrees. Answer accuracy refers to the ratio of the number of correctly answered questions to the total number of questions in the Level 2 assessment (e.g., answering 2 out of 3 questions correctly results in an accuracy rate of 66.7%).

[0043] The CRS-R visual sub-item score refers to the score on the CRS-R scale for assessing a patient's visual function, ranging from 0 to 4 points. A higher score indicates a stronger visual cognitive ability. For example, the CRS-R visual sub-item score standards are 0 points (no visual response), 1 point (responds to light), 2 points (able to track objects visually), 3 points (able to recognize objects), and 4 points (able to complete visual commands). This invention uses a machine learning model to map eye movement trajectory features to this score. For example, if a patient passes the first level of visual tracking assessment, the mapped score is 2 points; if they pass the second level of cognitive response assessment, the mapped score is 3 points.

[0044] Communication linkage refers to the data transmission and command interaction between the dual-end application and the integrated wearable hardware device, including the hardware device sending evaluation data to the dual-end and the dual-end sending control commands to the hardware device. For example, the hardware device sends real-time data (BIS value, current awareness level, LED light status) to the dual-end every 30 seconds. When the dual-end sends a control command (such as forcibly starting the evaluation), the hardware device responds and executes the command within 5 seconds, and the execution result is fed back to the dual-end APP.

[0045] In this embodiment, the LED status indicator uses 3-color LED beads, installed on the front of the device casing near the patient's forehead for easy observation by doctors and family members. The first display color is set to red, with an emission wavelength of 620-630nm, symbolizing a lower level of consciousness for the patient. The second display color is set to yellow, with an emission wavelength of 580-590nm, symbolizing that the patient has basic visual tracking ability and a slightly improved level of consciousness. The third display color is set to green, with an emission wavelength of 520-530nm, symbolizing that the patient has a higher level of cognitive response ability and a higher level of consciousness. It should be noted that the three-color display design enables intuitive visualization of the patient's level of consciousness, allowing doctors and family members to quickly understand the patient's status without having to check the app, thus improving the convenience of information acquisition.

[0046] In this embodiment, N is set to 10 minutes, that is, the cumulative time of the two-level assessment is ≥10 minutes. If the first-level assessment takes 4 minutes, the second-level assessment must take at least 6 minutes. If the assessment is interrupted due to changes in the patient's condition (such as closing the eyes), the interruption time is not included in the total assessment time. The time will continue to be accumulated after the assessment resumes until 10 minutes are reached.

[0047] In this embodiment, the anti-fatigue protection unit and the two-level VR eye-tracking assessment unit are activated synchronously. The accumulated time is evaluated in real time through a timer. When the accumulated time reaches N minutes (e.g., 10 minutes), a screen-off indication signal is sent to the VR display screen. The VR display screen turns off the backlight within 1 second and enters the screen-off state. During the screen-off period, the BIS module and the eye-opening and closing scene camera module continue to work to monitor the patient's status.

[0048] A movable automatic traction hanger is a mechanical structure that connects to a VR eye tracker and automatically moves the VR eye tracker up and down. Its core components include a micro motor, traction rope, and slide rail. The micro motor is a DC geared motor, the traction rope is a high-strength nylon rope, and the slide rail is a linear slide rail made of ABS material. The VR eye tracker connects to the slide rail via a slider. The traction hanger has a traction stroke of 5-10mm, ensuring a 3-5mm separation between the device and the patient's facial skin. This avoids pressure sores and does not affect the accuracy of repositioning in subsequent assessments. Figure 2 As shown.

[0049] The VR eye tracker is pulled upwards by the pressure ulcer protection unit, which controls the movable automatic traction frame to start. The micro motor drives the traction rope to retract, pulling the VR eye tracker upwards along the slide rail. The upward pulling speed is 1mm / s, and the duration is 5-10 seconds (adjusted according to the traction stroke). During the upward pulling process, the position of the VR eye tracker is detected in real time by the position sensor. When the preset upward pulling height is reached, the motor stops working, and the VR eye tracker remains in the upward pulling position.

[0050] In this embodiment, after the doctor scans the QR code, an information input box pops up on the APP interface. The doctor can manually enter the patient's bed number (e.g., bed 5 in ward 3), hospitalization number (e.g., 20240512003), name (e.g., Zhang San), and cause of illness (e.g., traumatic brain injury). Alternatively, the doctor can directly query and import patient information by entering the hospital's HIS system account and password. After the information is entered, the doctor clicks "bind," and the APP associates and stores the patient information with the device serial number. Once associated, the information cannot be modified, ensuring the accuracy of the information.

[0051] In this embodiment, the first doctor's display interface adopts a list layout, with each row displaying the information of a patient, including bed number, name, online status of the device (green dot for online, gray dot for offline), and color indicator of the current state of consciousness (red / yellow / green square). The interface supports filtering by bed number, name, and state of consciousness, and can display up to 50 patient information at the same time, with a refresh rate of 1 time / 30 seconds.

[0052] The second doctor's display interface is divided into three areas. The upper part displays the patient's basic information (bed number, hospital number, name, cause of illness, and serial number of the bound device). The middle part displays the personalized awakening-consciousness fluctuation curve (supports viewing data for 1 day, 3 days, and 7 days). The lower part displays the interventional rehabilitation time window prompts (displaying the predicted time window for the next 24 hours in a list format) and the CRS-R visual sub-item score (displaying the current score and the trend of score changes in the last 3 times). The interface supports data export (export format is PDF / Excel) for easy medical record archiving.

[0053] The first patient's display interface uses time (in hours) as the horizontal axis and BIS values ​​(0-100) on the left and consciousness levels (0=VS, 1=MCS-, 2=MCS+ / eMCS) on the right. BIS values ​​are represented by a blue curve, and consciousness levels are represented by bars of different colors (red=VS, yellow=MCS-, green=MCS+ / eMCS). Eye-opening events are marked with red dots on the time axis. The interface supports swiping to view historical data, and clicking on any time point allows you to view detailed assessment parameters for that moment (such as BIS value, fixation duration, etc.).

[0054] The second patient display interface shows the fluctuation curve in full screen. The interventional rehabilitation time window is highlighted with a green semi-transparent background. The start and end times of the highlighted area are marked below the curve. The estimated duration of the time window is also displayed (e.g., "09:30-10:00, lasting 30 minutes"). The interface supports zooming to facilitate family members to view details.

[0055] The third patient display interface shows the name of the bound smart rehabilitation device (such as a brain-computer interface rehabilitation training device), the device's online status, and the current working mode. It also has three buttons: Start Rehabilitation, Pause Rehabilitation, and End Rehabilitation. After the family member clicks to start rehabilitation, the APP sends a start command to the smart rehabilitation device. After receiving the command, the device starts rehabilitation training according to the preset program. During the training process, the training progress is fed back to the APP in real time.

[0056] The linked doctor authorization password refers to a one-time password generated by the doctor through the doctor's app to authorize family members to conduct additional assessments. This password is time-sensitive and unique. For example, the password is a 6-digit number combination (000000-999999), randomly generated by the doctor's app. The doctor can set the password's validity period (e.g., 1 hour / 2 hours / 4 hours). After generation, it is sent to the family member via SMS or in-app message. The family member can then initiate additional assessments within the validity period (up to 3 times) after entering the password in the patient's app.

[0057] Additional assessments by family members within a specified timeframe refer to extra consciousness assessments initiated by family members through the patient's app within a time limit after they enter a valid password. The assessment process is consistent with the automatic assessment, avoiding unnecessary frequent assessments while meeting the needs of family members to proactively understand the patient's condition, further enhancing the system's usability and user experience.

[0058] The beneficial effects of the above technical solution are as follows: By integrating core functional components into an integrated wearable hardware device, and combining the dual-condition triggering, two-level progressive assessment, and data processing logic of the software assessment module, near real-time, all-weather detection of the consciousness state of pDoC patients is achieved. The generated personalized arousal-consciousness fluctuation curve and CRS-R visual sub-item score provide objective data support for clinical diagnosis. The communication linkage between the two-end applications ensures that doctors and family members can grasp the patient's status in real time, effectively solving the problems of reliance on manual assessment, difficulty in high-frequency detection, and high misdiagnosis rate in existing technologies. At the same time, it provides precise intervention timing signals for intelligent rehabilitation devices, improving the pertinence and effectiveness of rehabilitation treatment.

[0059] This invention provides a near real-time consciousness detection system for improving CRS-R accuracy and for pDoC intelligent rehabilitation, and also includes: The control module is used to control the VR display screen to present at least one dynamic test element to conduct a preliminary test on the patient before presenting the text and image multiple-choice questions based on the VR headset, obtain quantitative auxiliary factors, and optimize the quantification process of answer correctness and reaction time to obtain the quantified consciousness level.

[0060] Preferably, the control module includes: The control unit is used to control the VR display screen to present at least one dynamic test element before presenting text and image multiple-choice questions based on the VR headset, and to preset a target attribute combination mode for each dynamic test element. The visual attribute dimensions of each dynamic test element include at least color, size, and brightness. The acquisition unit is used to synchronously acquire the patient's raw eye movement data based on the eye tracker during the test; The real-time processing unit is used to process the raw eye movement data in real time and parse it into a sequence of eye movement microstates in fine time windows. The eye movement microstates include: stable fixation, target-oriented saccades, retrospection, exploratory fixation, and targetless drift. The pattern generation unit is used to generate a theoretically expected eye-tracking pattern based on the features of the corresponding visual attribute dimension in the preset target attribute combination pattern. A line construction unit is used to time-align the eye-tracking microstate sequence with the theoretically expected eye-tracking pattern, determine the fluctuation difference of the comparative states at the same alignment time point, and form a fluctuation trend line. The parameter value determination unit is used to perform time alignment between the raw eye-tracking data and the standard eye-tracking data and divide them into N continuous analysis windows, and determine the spatial dwell ratio Zb, velocity adaptation index Zs, and trajectory morphology similarity Zs of each continuous analysis window to obtain consistency. The range determination unit is used to determine the overlap ratio range based on all consistency factors. ,in, These are the lower and upper limits of the overlap ratio range, respectively; The window division unit is used to start from the first initial window and sequentially set the starting point of the next window based on a random percentage of the overlap ratio range until the time alignment segment division is completed. The overlapping window and the continuous analysis window are the same size but may have different time starting points, and the initial window is the first window in the overlap division result. The comparison analysis unit is used to time-align the first division result of continuous division with the second division result of overlapping division, and obtain the set of overlapping windows that are in comparison with each continuous analysis window, wherein the set of overlapping windows involves at least one overlapping window. A division unit is used to divide the fluctuation trend line according to a continuous analysis window to obtain the fluctuation trend item of each continuous analysis window; The confidence determination unit is used to obtain the response confidence of the corresponding visual attribute dimension based on the overlapping window set and the consistency standard deviation of the corresponding continuous analysis window, and in combination with the consistency and fluctuation trend term of the corresponding continuous analysis window, and to construct the response confidence vector of the corresponding visual attribute dimension. The factor determination unit is used to calculate the stability index within each vector and the cross-dimensional matching index between each vector based on the response confidence vectors of all visual attribute dimensions. The results are input into the pre-trained intent inference model, and the output is the overall intent confidence that represents the patient's integrated response intent to the target attribute combination pattern as a quantification auxiliary factor to optimize the quantification process.

[0061] In this embodiment, the formula for calculating consistency is as follows: ; ; in, For the consistency of the i-th consecutive analysis window; , , The weights of Zb, Zs, and Zs are respectively. As a synergistic enhancement factor; is the standard deviation of the absolute values ​​of the Euclidean distance differences between the actual trajectory points and the expected trajectory points based on all gaze points in the i-th continuous analysis window; It is the minimum value; is the average of the absolute values ​​of the Euclidean distance differences between the actual trajectory points and the expected trajectory points based on all gaze points in the i-th continuous analysis window; In this embodiment, the formula for calculating the overlap ratio range is as follows: ; ; in, The average value of the trend characteristic components that satisfy a normal distribution with a probability of 0.8 across all continuous analysis windows; Let be the trend characteristic component of the i-th continuous analysis window, and , , These represent the consistency of the (i+1)th and (i-1)th consecutive analysis windows, where i = 2, 3, ..., M-1; This refers to the steepness parameter of the Sigmoid function; The threshold for trend significance is M; M is the number of continuous analysis windows. This is the average of the consistency across all consecutive analysis windows; For all Adjacent units sorted from largest to smallest The maximum value among the differences; for , The maximum of the two; In this embodiment, among the clinical indicators Priority > Priority > The priority, at this time, , , The weights are 0.4, 0.3, and 0.3 respectively. The reasonable range of values ​​is 0.1 to 0.3.

[0062] Based on the consistency data validation of 100 pDoC patients, when and It can accurately identify the trend changes in consciousness while avoiding noise interference, and adapt to the calculation requirements of the formula for trend feature components.

[0063] In this embodiment, the original quantitative rules for the second-level assessment are as follows: 2 points for each correct answer, 1 point for a reaction time ≤ 5 seconds, and a maximum score of 5 points. If the patient answers 2 questions correctly with an average reaction time of 4 seconds, the original quantitative score is 2×2+1=5 points. With a quantitative auxiliary factor of 1.1, the optimized quantitative score is 5×1.1=5.5 points (rounded to one decimal place), corresponding to an awareness level of MCS+ / eMCS. If the quantitative auxiliary factor is 0.9, the optimized quantitative score is 5×0.9=4.5 points, and the corresponding awareness level is still MCS+ / eMCS, ensuring that the quantitative results reflect both the patient's actual ability and the basic visual response status.

[0064] In this embodiment, the target attribute combination mode refers to the combination and change of visual attribute dimensions (color, size, brightness) set for each dynamic test element. Different modes correspond to different change patterns. For example, mode 1 is color gradient + size scaling (color gradients once every 5 seconds, size scales once every 3 seconds), mode 2 is color gradient + brightness adjustment (color gradients once every 5 seconds, brightness adjusts once every 4 seconds), and mode 3 is size scaling + brightness adjustment (size scales once every 3 seconds, brightness adjusts once every 4 seconds). Doctors can select the test mode through the doctor's app, with mode 1 being the default.

[0065] Visual attribute dimensions refer to the visual characteristic parameters of dynamic test elements, including color, size, and brightness. Changes in these dimensions are used to stimulate patients to produce visual responses. For example, the color dimension uses the RGB color space, with a range of red (255,0,0) → orange (255,165,0) → yellow (255,255,0) → red (255,0,0); the size dimension uses diameter as an indicator, with a range of 2cm → 3cm → 2cm; the brightness dimension uses relative brightness as an indicator, with a range of 50% → 100% → 50%. All changes in these dimensions are smooth transitions to avoid abrupt stimuli.

[0066] In this embodiment, raw eye-tracking data refers to the unprocessed raw signal data directly collected by the VR eye tracker, which includes the raw parameters of the patient's eye movements. For example, each record of raw eye-tracking data includes a timestamp, horizontal fixation angle (e.g., 15.2°), vertical fixation angle (e.g., 8.3°), and pupil diameter (e.g., 4.5mm). The data is output in the form of a data stream, without any missing or duplicate data.

[0067] A fine-grained time window refers to the time interval used to divide eye-tracking data. Each window is relatively short to ensure accurate capture of changes in the microscopic state of eye movements. The duration is set according to the sampling frequency of the eye tracker. For example, if the duration of the fine-grained time window is set to 100ms, then each 100ms window is a time window, and each window corresponds to 12 frames of eye-tracking data (sampling frequency of 120Hz). By analyzing the data within each window, the dominant microscopic state of eye movements within that window is determined, ensuring the accuracy of state interpretation.

[0068] Stable gaze refers to the eye remaining still, with a gaze point fluctuation range of ≤1° and a duration of ≥100ms; target-oriented saccade refers to the eye moving rapidly toward the direction of the dynamic test element, with a movement angle of ≥5° and a duration of ≤200ms; retrograde gaze refers to the eye returning from the current gaze point to the previous gaze point, with the movement trajectory being opposite to the previous saccade trajectory; exploratory gaze refers to the slow movement of the eye within the VR display area without a clear target; targetless drift refers to irregular, minute movements of the eye, with a fluctuation range of >1° and a speed of <10° / s.

[0069] The theoretical expected eye movement pattern refers to a pre-defined ideal sequence of eye movement states based on the combination of target attributes of dynamic test elements and the normal human visual response patterns, used as a comparison benchmark. For example, the theoretical expected eye movement pattern sequence for pattern 1 (color gradient + size scaling) is: stable fixation (0-2000ms) → target-oriented saccade (2000ms, triggered by color change) → stable fixation (2100-4000ms) → target-oriented saccade (4000ms, triggered by size change) → stable fixation (4100-6000ms). The start time and duration of each state are clearly defined, providing a clear benchmark for comparison.

[0070] A fluctuation trend line is a curve that reflects the changing trend of the difference between the actual eye movement microstate sequence and the theoretically expected eye movement pattern. The larger the difference, the more significant the difference. For example, in the fluctuation trend line, if the actual state in a certain time window is consistent with the expected state, the fluctuation difference is 0; if the actual state is target-free drift and the expected state is stable fixation, the fluctuation difference is -1; if the actual state is target-oriented saccade and the expected state is stable fixation, the fluctuation difference is 1. The trend line allows for a direct observation of the deviation between the patient's eye movement response and the ideal state.

[0071] Standard eye-tracking data refers to pre-stored eye-tracking data from healthy individuals or pDoC patients with a clear state of consciousness, serving as a reference benchmark for calculation consistency. For example, standard eye-tracking data is the average of eye-tracking data from 10 healthy volunteers under the same dynamic test element stimulation, including parameters such as horizontal fixation angle, vertical fixation angle, and saccade velocity at each time point. The data has been standardized to ensure comparability with patient eye-tracking data.

[0072] A continuous analysis window refers to a time window used to calculate parameters and ensure consistency. Each window is relatively long to ensure the reliability of parameter calculations, and the duration is set based on the total preliminary test time. For example, if the total preliminary test time is 2 minutes (120 seconds), the number of continuous analysis windows is N=12, each window is 10 seconds long, and there is no overlap between windows, covering the entire preliminary test process. Each window corresponds to 1200 frames of eye-tracking data to ensure a sufficient sample size for parameter calculations.

[0073] The spatial dwell time ratio (Zb) refers to the percentage of time a patient's gaze is within the area containing the dynamic test element within a continuous analysis window. It reflects the patient's level of attention to the target and ranges from 0 to 1. For example, if the area containing the dynamic test element is a circular area with a radius of 2cm centered on the element on a VR display screen, and if the duration of a continuous analysis window is 10 seconds, and the cumulative time the patient's gaze is within this area is 8 seconds, then the spatial dwell time ratio (Zb) = 8 / 10 = 0.8.

[0074] The speed fit index Zs refers to the degree of fit between a patient's saccade speed and the rate of change of dynamic test elements (such as color gradation speed and size scaling speed) within a continuous analysis window. The value ranges from 0 to 1, with values ​​closer to 1 indicating better fit. For example, if the size scaling speed of dynamic test elements is 0.5 cm / s and the patient's average saccade speed is 0.45 cm / s, then the speed fit index Zs = 0.45 / 0.5 = 0.9.

[0075] Trajectory morphology similarity Sg refers to the degree of similarity between a patient's eye movement trajectory and a standard eye movement trajectory within a continuous analysis window. It is calculated using cosine similarity, with a value ranging from 0 to 1, where a value closer to 1 indicates a higher similarity. If the patient's eye movement trajectory and the standard eye movement trajectory are represented as vectors (composed of the gaze angle at each time point), and the cosine similarity between the two vectors is calculated, a similarity of 0.85 indicates a trajectory morphology similarity Sg = 0.85.

[0076] An overlapping window is an analysis window that is the same size as the continuous analysis window but has a different time starting point, resulting in temporal overlap with the continuous analysis window. It is used to increase the data sample size and improve the reliability of the analysis. For example: The continuous analysis windows are as follows: A1: 1-10 seconds; A2: 11-20 seconds; A3: 21-30 seconds (first partition result): In the overlapping window, starting with A1, determine the left boundary line of the second window. The left boundary line is defined as: 10 seconds × a value randomly selected from the overlap ratio range. For example, if the randomly selected value is 0.4, then the left boundary line of the second overlapping window is... =5 seconds. At this time, the first overlapping window is 1-4 seconds, and so on. When determining the left boundary line of the third overlapping window, the analysis is based on the second overlapping window, which is 5-14 seconds. When the last overlapping window is less than 10 seconds, the window with less than 10 seconds can be used for subsequent analysis.

[0077] The second division result refers to the division result of the overlapping window. For example, the first overlapping window: 1-4 seconds, the second overlapping window: 5-12 seconds, the third overlapping window: 13-21 seconds, and the fourth overlapping window: 22-30 seconds.

[0078] The overlapping window set refers to the set of all overlapping windows that overlap with a certain continuous analysis window in time. For example, if A1: 1-10 seconds overlaps with the first overlapping window: 1-4 seconds and the second overlapping window: 5-12 seconds, then the first overlapping window and the second overlapping window are the overlapping window set of A1.

[0079] In this embodiment, the fluctuation trend term refers to the characteristic value of the fluctuation trend line corresponding to each continuous analysis window, reflecting the overall difference trend between the actual eye movement state and the expected eye movement pattern within that window. If the fluctuation trend term is positive, it means that the actual eye movement state within that window is more active than the expected state (e.g., more saccadic movements); if the trend term is negative, it means that the actual eye movement state within that window is smoother than the expected state (e.g., more targetless drift); if the trend term is 0, it means that the actual state is basically consistent with the expected state.

[0080] Consistency standard deviation refers to the standard deviation of consistency among all overlapping windows in a set of overlapping windows corresponding to a continuous analysis window. It reflects the degree of dispersion of the overlapping window data. The smaller the standard deviation, the more stable the data.

[0081] Response confidence is a parameter that measures the reliability of a patient's response to changes in a visual attribute dimension (color, size, brightness). It takes into account consistency, consistency standard deviation, and fluctuation trend term, and its value ranges from 0 to 1. For example, for the color dimension, if the consistency of the continuous analysis window is 0.85, the consistency standard deviation is 0.014, and the fluctuation trend term is -0.1, then the response confidence = 0.85 × (1 - 0.014) + (-0.1) × 0.1 = 0.828, which means that the patient's response to color changes is highly reliable. Here, 0.1 is a preset coefficient.

[0082] The response confidence vector is a vector composed of the response confidence scores for each visual attribute dimension, reflecting the distribution of the patient's response reliability to changes in all visual attribute dimensions. For example, if the visual attribute dimensions include color, size, and brightness, with corresponding response confidence scores of 0.828, 0.795, and 0.812 respectively, then the response confidence vector would be [0.828, 0.795, 0.812]. The vector's dimensions are consistent with the number of visual attribute dimensions.

[0083] The stability index is a parameter that measures the stability of the elements within a single response confidence vector, and is calculated using the standard deviation of the vector elements.

[0084] Cross-dimensional matching metrics are parameters that measure the degree of matching between confidence vectors of responses from different visual attribute dimensions, and are calculated using cosine similarity.

[0085] The intent inference model refers to a pre-trained machine learning model used to output the overall intent confidence of a patient based on stability indicators and cross-dimensional matching indicators. The support vector machine (SVM) model is used as the intent inference model. The training data includes stability indicators, cross-dimensional matching indicators and corresponding manually assessed intent confidence of 50 pDoC patients. The input of the model is the stability indicators and cross-dimensional matching indicators, and the output is the overall intent confidence, which ranges from 0.8 to 1.2.

[0086] The overall intent confidence score refers to the parameter output by the intent inference model that represents the patient's integrated response intent to the combination of target attributes, i.e., a quantitative auxiliary factor, with a value range of 0.8-1.2. If the patient's stability index is 0.023 and the cross-dimensional matching index is 0.811, and the overall intent confidence score output after inputting into the intent inference model is 1.05, it indicates that the patient has a strong integrated response intent to the dynamic test elements and a stable visual response state.

[0087] The beneficial effects of the above technical solution are as follows: Through the collaborative work of multiple sub-units of the control module, a complete process from dynamic test element presentation, eye movement data acquisition and processing to quantitative auxiliary factor calculation is realized. Each sub-unit has a clear function and rigorous logic, ensuring the accuracy of quantitative auxiliary factors. By introducing techniques such as eye movement micro-state analysis, theoretical expected eye movement pattern comparison, and multi-window comparative analysis, the effective information in eye movement data is fully explored, effectively correcting the limitations of a single assessment indicator. This makes the quantitative results of consciousness level more consistent with the patient's actual state of consciousness, further improving the system's assessment accuracy and reliability, and providing more accurate data support for clinical diagnosis.

[0088] In one embodiment, based on continuously recorded BIS values, eye-opening events, and two-level assessment results, a personalized arousal-consciousness fluctuation curve is generated along a time axis, automatically marking intervention time windows for rehabilitation, specifically including: Online analysis of the BIS data identifies candidate awakening periods that meet the following conditions: the BIS value is consistently higher than the personalized baseline threshold exceeding the first time threshold T1, and the first derivative of its value curve shows a specific pattern of first rising and then stabilizing during this period. The eye-opening events are analyzed, and the start and end times of each sustained eye-opening event are marked. When a candidate arousal period is detected, search within that candidate arousal period for whether there is a valid eye-opening period that overlaps with it in time; If there is overlap, the overlapping time period is defined as a primary candidate time window; Within the primary candidate time window, it is retrieved whether the two-level VR eye-tracking evaluation unit outputs a valid behavioral response result during the time period or the buffer period immediately before and after the time period. If a valid result is found, the primary candidate time window is upgraded to a secondary candidate time window, and the corresponding behavioral response type and integrated intent confidence are associated. Valid behavioral response results include: the first-level visual tracking pass signal or the second-level integrated intent confidence. For each of the secondary candidate time windows, perform verification based on state transition logic: The patterns of BIS data and behavioral response data are analyzed within a preset historical period before the start of the secondary candidate time window, and the smoothness and likelihood probability of the patient's transition from a low-arousal-unresponsive state to a high-arousal-responsive state are calculated. The analysis examines whether there is a sudden drop in BIS values ​​or a disappearance of behavioral responses within a predetermined subsequent period after the end of the time window, in order to verify that the secondary candidate time window represents a complete, definite plateau of available consciousness rather than a momentary fluctuation. If the state transition is smooth and the plateau period is obvious, the verification is successful, and the secondary candidate time window is marked as a verified valid time window. Using time as the horizontal axis, the following three layers of information are plotted simultaneously to form a personalized awakening-consciousness fluctuation curve: First layer: BIS numerical curve, with all candidate awakening periods highlighted; The second layer: behavioral event labeling, which uses discrete symbols to label eye-opening events, first-level assessment pass events, and second-level integration intention confidence; Third layer: Mark all verified valid time windows on the timeline with a highlighted background color; Based on the occurrence patterns of valid time windows verified in historical cycles, a time series prediction model is used to adaptively predict and provide hints for similar time windows that may appear in the future. When the actual data matches the prediction, the verification process of the secondary candidate time windows is accelerated. The latest and proven effective time windows, or the high-probability time windows that are predicted to appear and whose initial data match, are ultimately determined as the time windows for intervention and rehabilitation. On the doctor's and patient's display interfaces of the personalized awakening-consciousness fluctuation curve, the interventional rehabilitation time window is dynamically highlighted and its boundaries are marked. When the interventional rehabilitation time window begins or reaches the optimal intervention point, the data processing unit automatically generates and sends an intervention timing trigger signal containing time window parameters (such as expected duration and current level of consciousness) to the preset intelligent rehabilitation device.

[0089] In this embodiment, a verification chain based on the temporal coupling relationship between multi-source asynchronous signals and state transition logic is proposed. Through a series of progressive and mutually verifying conditional judgments—candidate arousal period → overlap with eye-opening event → association with behavioral response → verification of state transition smoothness—it is ensured that the identified time window is a true synchronous and unified manifestation of physiological arousal, behavioral arousal, and cognitive function, greatly improving the specificity and reliability of the marker.

[0090] The system introduces a two-way verification of state transition smoothness and plateau period integrity. It not only examines the state within the time window but also analyzes in depth how the state arises (pre-initiation history) and how it ends (post-initiation trend). This effectively filters out transient pseudo-awakenings caused by brief external stimuli (such as pain) that cannot sustain rehabilitation training, ensuring that the time window corresponds to the patient's intrinsic and stable stage of consciousness improvement.

[0091] In this embodiment, the personalized baseline threshold is a BIS value benchmark threshold set according to the individual patient's situation, rather than a uniform fixed value, in order to adapt to the physiological differences of different patients. For example, in the first 24 hours after the patient wears the device, the system automatically collects their BIS data and calculates the median BIS value during this period as the personalized baseline threshold. If the patient's median BIS value in the first 24 hours is 55, then the personalized baseline threshold is set to 55. A value higher than this indicates that the patient's level of arousal is higher than their daily baseline level, and the time period for baseline calculation is adjusted by the system backend.

[0092] The first time threshold T1 refers to the shortest duration for which the BIS value is continuously higher than the personalized baseline threshold. It is used to filter out short-term fluctuations and ensure the effectiveness of the candidate awakening period. If T1 is set to 30 seconds, the BIS value must be higher than the personalized baseline threshold for 30 consecutive seconds to be considered as meeting the time condition for the candidate awakening period. If it only lasts for 20 seconds, it will not be included to avoid misjudgment caused by instantaneous awakening. This threshold is adjusted by the doctor through the doctor's app within the range of 20-60 seconds.

[0093] The first derivative of the numerical curve refers to the rate at which the BIS value changes over time, reflecting the trend of the BIS value (rising, falling, or stable). A positive first derivative indicates that the value is rising, zero indicates that it is stable, and a negative first derivative indicates that it is falling. For example, if the BIS value is 55 at time t1 and 57 at time t2 (t2-t1=1 second), then the first derivative for this period is (57-55) / (1)=2; at time t3 (t3-t2=1 second), the value is 58, and the first derivative is 1; from t4 to t10 (a total of 6 seconds), the BIS value remains at 58, and the first derivative is 0. The derivative can be directly calculated through the algorithm code to determine the trend of change.

[0094] The specific pattern (rising first and then stabilizing) refers to the change in the first derivative of the BIS numerical curve, which is initially positive (numerical rise) and then becomes zero after a period of time (numerical stability). This indicates that the patient's level of arousal gradually increases and is maintained at a high level.

[0095] The effective eye-opening period refers to the period during which the eyes remain open for a duration not less than the preset minimum duration in an eye-opening event. It is used to filter out invalid events with brief eye openings, and the preset minimum duration for effective eye opening is 10 seconds.

[0096] The candidate awakening period refers to the period in which the BIS value is continuously higher than the personalized baseline threshold by more than T1 and the first derivative first rises and then stabilizes, which is the basis for subsequent selection of time windows through online analysis of BIS data.

[0097] The primary candidate time window refers to the portion of the candidate awakening period that overlaps with the effective eye-opening period in time. It is a preliminary screening of the period that may be suitable for rehabilitation intervention.

[0098] The buffer period immediately preceding and following this period refers to a pre-defined short period before the start time and after the end time of the primary candidate time window. This period is used to cover any possible response delays in the evaluation unit and ensure that no valid behavioral responses are missed.

[0099] A valid behavioral response result refers to a signal output by the two-level VR eye-tracking assessment unit that demonstrates the patient's conscious response, including a Level 1 visual tracking pass signal and a Level 2 integrated intent confidence score. A valid behavioral response result is defined as one where the assessment unit outputs a visual tracking pass signal within the primary candidate time window, or where the Level 2 integrated intent confidence score is ≥0.8; otherwise, it is considered invalid.

[0100] Secondary candidate time windows refer to the time periods upgraded from primary candidate time windows after valid behavioral response results are retrieved. These time windows are associated with the behavioral response type and the confidence level of integration intent, further narrowing the scope of rehabilitation intervention time periods. For example, if a visual tracking pass signal (behavioral response type MCS-) is retrieved within the primary candidate time window of 08:00:15-08:01:45, then this time period is upgraded to a secondary candidate time window, associated with visual tracking pass and an integration intent confidence level of 0.7. The system automatically completes the upgrade marking and associates the relevant data.

[0101] Behavioral response type refers to the type of consciousness response exhibited by the patient in the two-level assessment, which is associated with the secondary candidate time window and corresponds to different levels of consciousness. For example, behavioral response types include: visual tracking passed (corresponding to MCS- state) and cognitive response effective (corresponding to MCS+ / eMCS state). The system automatically matches and associates types based on the assessment results, providing a basis for subsequent state judgment. The correspondence between types and consciousness levels can be preset in the system.

[0102] The preset historical time period refers to a fixed duration before the start of the secondary candidate time window, used to analyze the patient's state of consciousness before entering that time window. The low arousal-unresponsive state refers to the state in which the patient's BIS value is lower than the personalized baseline threshold and no effective behavioral response occurs within the preset historical time period. The high arousal-responsive state refers to the state in which the patient's BIS value is higher than the personalized baseline threshold and effective behavioral response occurs within the secondary candidate time window. The smoothness of the transition refers to whether the changes in BIS value and behavioral response are continuous and without abrupt changes during the transition from the low arousal-unresponsive state to the high arousal-responsive state. The smoothness value ranges from 0 to 1, with the closer to 1 being smoother. For example, by calculating the variance of the rate of change of BIS value within the historical time period and time window, if the variance is 0.02 (relatively small) and the behavioral response gradually appears from none to one (first opening the eyes, then visual tracking), then the smoothness of the transition is 0.92, indicating a smooth transition. The smoothness is obtained through variance calculation and behavioral response time series analysis.

[0103] Using a logistic regression model, with input features such as the rate of change in BIS and duration of eye opening, the likelihood probability of this state transition was calculated to be 0.88, indicating that the transition was a highly probable and effective change rather than a coincidence. The model can optimize its prediction accuracy by increasing the number of training samples.

[0104] The preset follow-up period refers to a fixed duration after the end of the secondary candidate time window. It is used to verify whether the state is stable after the time window ends, rather than experiencing an instantaneous decline. For example, if the preset follow-up period is 60 seconds, and the secondary candidate time window ends at 08:01:45, then the follow-up period is from 08:01:45 to 08:02:45. The BIS data and behavioral responses within this period are analyzed, and the period length can be set from 30 to 120 seconds.

[0105] A sudden drop in BIS value refers to a decrease in the BIS value within a short period of time (e.g., 10 seconds) that exceeds a preset threshold, indicating a rapid decline in arousal level. For example, if the preset threshold for a sudden drop is 10, and the patient's BIS value drops from 62 to 48 within 10 seconds in the subsequent period, a decrease of 14, exceeding the threshold, this is considered a sudden drop in BIS value. The threshold for a sudden drop can be set between 8 and 15.

[0106] The disappearance of behavioral response refers to the patient no longer exhibiting the previously effective behavioral response within a preset subsequent time period, such as failing the visual tracking assessment or a significant decrease in the confidence level of integrated intent. If, within a subsequent time period, the patient's eye movement trajectory no longer follows the moving light point, the visual tracking assessment fails, and the confidence level of integrated intent drops to 0.3, this is considered the disappearance of behavioral response. The system determines whether the response has disappeared by comparing the assessment results.

[0107] A complete plateau period of usable consciousness refers to a period within a secondary candidate time window where the patient's highly aroused and responsive state remains stable, with a smooth transition before initiation and no sudden drop or disappearance of response after initiation. This is a complete period capable of supporting rehabilitation intervention. For example, in a secondary candidate time window of 08:00:15-08:01:45, if the state transition is smooth before initiation (smoothness 0.92), the state remains stable within the time window, and the BIS remains between 58-60 in the subsequent period after initiation, with continued effective behavioral responses, this constitutes a complete plateau period of usable consciousness. The system verifies and confirms the integrity of the plateau period through multi-dimensional data.

[0108] Transient fluctuations refer to a patient's level of arousal and behavioral response that only appears briefly, cannot be sustained, and then quickly declines, making rehabilitation intervention unsuitable. For example, if a patient's BIS value briefly rises to 60 (lasting 5 seconds), eyes are open for 8 seconds, and visual tracking assessment is not passed, followed by a sharp drop in BIS and disappearance of behavioral response, this is a transient fluctuation. Such periods are not considered valid time windows, and the system filters transient fluctuations based on duration and stability.

[0109] State transition smoothness refers to the gradual change in BIS value and the gradual appearance of behavioral responses as a patient transitions from a low-arousal-unresponsive state to a high-arousal-responsive state, without abrupt changes or jumps, conforming to physiological laws. For example, if the BIS value gradually increases from 53 to 61 (taking 30 seconds), with eye opening occurring first (08:00:05) and then visual tracking (08:00:15), and the change is continuous without sudden increases or decreases, this is considered state transition smoothness. The system determines smoothness by the rate of data change and the timing of behavioral responses.

[0110] A clear plateau period indicates that within the secondary candidate time window, the BIS value fluctuates within a small range, the behavioral response remains stable, the duration meets the requirements, and the characteristics of the plateau period are clearly identifiable.

[0111] The first layer: The BIS numerical curve refers to the bottom layer of the personalized awakening-consciousness fluctuation curve. It is a curve of the BIS value changing over time, drawn with specific colors and line types. It is the basic data presentation.

[0112] Highlighting refers to marking specific time periods (such as candidate arousal periods or effective time windows) in a curve using a prominent visual method to facilitate rapid identification.

[0113] The second layer: Behavioral event markers refer to the middle layer of the curve, which uses different discrete symbols to mark various behavioral events, intuitively showing the time and type of the event.

[0114] Discrete symbols refer to independent graphic symbols used to mark behavioral events. Each symbol corresponds to a specific event type, and they are not easily confused, facilitating quick differentiation. For example, three discrete symbols can be defined: a circle for the "eyes open" event, a triangle for the "pass the first-level assessment" event, and a square for the "second-level confidence" event. The symbol color is strongly correlated with the event type (red - eyes open, yellow - pass the assessment, green - confidence). Technical personnel in the relevant field can quickly identify them, and the symbol system can be clearly defined through system design documents.

[0115] The Level 1 assessment pass event refers to the event in which the patient passes the Level 1 visual tracking assessment. The time point of the event is marked, and it is an important manifestation of the behavioral response. For example, if the patient passes the Level 1 visual tracking assessment at 08:00:15, the system marks that time point on the curve with a yellow triangle symbol and adds the label "Visual Pass" next to the symbol. Detailed data of the event (such as tracking angle and duration) is recorded simultaneously. The event data can be exported and viewed through the data log.

[0116] The third layer: Highlighting the top layer of the curve with a different background color than the bottom layer curves, this layer marks the verified effective time windows and highlights the core rehabilitation intervention period.

[0117] Historical period refers to a complete monitoring period (such as 7 days) in the past, which includes multiple verified effective time windows to analyze the occurrence patterns of time windows.

[0118] The pattern of occurrence of verified effective time windows refers to the stable characteristics of verified effective time windows in terms of time distribution, duration, and interval within historical cycles.

[0119] Time series forecasting models are models built using algorithms based on historical time window data to predict the timing and duration of future time windows. One such model employs a Long Short-Term Memory (LSTM) network, taking into account features such as the occurrence time, duration, and BIS baseline of historical time windows. After training, the model can output predictions of time windows that may occur within the next 24 hours.

[0120] Similar time windows that may appear in the future refer to future periods output by time series prediction models that have similar characteristics to effective time windows in historical cycles (such as the time of occurrence, duration, and level of awareness).

[0121] Adaptive prediction and alerts refer to the ability of the prediction model to dynamically adjust the prediction results based on changes in the patient's real-time data, and to send alert messages to both mobile apps when the prediction time window is about to appear.

[0122] Actual data matching prediction means that the real-time collected BIS data, eye-opening events, etc., are consistent with the results output by the prediction model, achieving the preset matching conditions.

[0123] If the model predicts that an effective time window will appear between 09:10 and 09:12, and real-time data shows that the patient's BIS value has risen above the baseline by 09:09:30 and the patient's eyes are open, the initial data matches the prediction. The prediction probability of this time window is 0.85, which is a high probability time window. The probability threshold can be set to 0.7-0.9.

[0124] The interventional rehabilitation time window refers to the optimal period when the patient's consciousness is good and suitable for intervention by intelligent rehabilitation equipment. It is one of the core outputs of the system. For example, by combining the latest verified effective time window (08:04:10-08:05:40) and the high-probability predicted time window (09:10-09:12), 08:04:10-08:05:40 is finally determined as the current interventional rehabilitation time window, and is simultaneously marked on the curve of the dual-end APP. The selection logic of the time window can be adjusted through priority settings.

[0125] Dynamic highlighting refers to highlighting the interventional rehabilitation time window on the display interface using dynamic visual effects, such as slow flashing or color gradient, to enhance recognizability. Boundary marking refers to clearly marking the start and end times of the interventional rehabilitation time window on the display interface to facilitate accurate understanding of the time range.

[0126] The optimal intervention point refers to the time within the feasible rehabilitation time window when the patient's state of consciousness is most stable and the rehabilitation effect is likely to be best. This is usually the middle of the time window or a period after the patient's state has stabilized. The feasible rehabilitation time window is 08:04:10-08:05:40, lasting 90 seconds. The optimal intervention point is set at 08:04:40, which is the middle point of the time window. At this time, the patient's state is stable, and intervention can achieve a better rehabilitation effect. The optimal intervention point can be set at any point within the 30%-70% range of the time window.

[0127] The intervention timing trigger signal refers to the control signal generated and sent by the data processing unit at the optimal intervention point, which is used to instruct the preset intelligent rehabilitation equipment to start rehabilitation intervention.

[0128] The pre-set intelligent rehabilitation equipment refers to the intelligent devices bound to this system for the rehabilitation training of pDoC patients, such as brain-computer interface rehabilitation devices and visual stimulation rehabilitation devices.

[0129] The time window parameter refers to the key information related to the interventional rehabilitation time window contained in the intervention timing trigger signal, providing a reference for the working parameters of intelligent rehabilitation equipment.

[0130] The expected duration refers to the anticipated duration of the intervention window for rehabilitation, which is predicted by the system based on historical data and real-time status of verified effective time windows.

[0131] The beneficial effects of the above technical solution are as follows: By conducting temporal coupling analysis of BIS data, eye-opening events, and behavioral responses from multiple sources, a complete verification chain is constructed, consisting of candidate arousal period screening, primary time window extraction, secondary time window upgrading, and bidirectional verification of state transitions. This effectively filters out false alarms and missed alarms caused by transient fluctuations, ensuring that the marked interventional rehabilitation time window is a patient's intrinsic, stable period of high arousal and responsiveness. This provides precise intervention timing for intelligent rehabilitation devices, significantly improving the targeting and effectiveness of rehabilitation treatment. Simultaneously, the three-layer visualization design of the personalized arousal-consciousness fluctuation curve, combined with dynamic highlighting and boundary annotation on both mobile and offline apps, allows doctors and family members to intuitively grasp the patterns of changes in the patient's state of consciousness, assisting in optimizing clinical decision-making. This solves the problems of inaccurate matching of rehabilitation timing, unintuitive assessment data, and large errors in human judgment found in existing technologies.

[0132] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A near-real-time consciousness detection system for improving CRS-R accuracy and for use in pDoC intelligent rehabilitation, characterized in that, include: The integrated wearable hardware device integrates a BIS module, an open and closed eye scene camera module, a VR headset and a VR eye tracker. All components are packaged in the same wearable shell and worn on the patient's head via a detachable headband. The software evaluation module includes a trigger unit, a two-level VR eye-tracking evaluation unit and a data processing unit. The triggering unit is used to activate the two-level VR eye-tracking assessment unit when the BIS module detects that the patient's arousal value is higher than a preset threshold and the eye-opening and closing scene camera module confirms that the patient's eyes are open based on a visual algorithm. The two-level VR eye-tracking evaluation unit is used when the first level performs visual tracking judgment. At this time, the VR headset presents a moving target, and the eye tracker collects the eye movement trajectory and compares it with the CRS-R visual tracking standard. Once the first-level assessment is passed, the patient automatically proceeds to the second-level assessment. At this point, a multiple-choice question with images and text is presented using a VR headset. The patient answers by focusing on specific points, and the level of consciousness is quantified based on the correctness of the answer and the reaction time. The data processing unit is used to continuously record BIS values, eye-opening events, and two-level assessment results, generate a personalized arousal-consciousness fluctuation curve with time as the axis, automatically mark the time window for intervention and rehabilitation, send the optimal intervention timing signal to the intelligent rehabilitation device, and map eye movement trajectory features into CRS-R visual sub-item scores through a machine learning model and send them to both ends. The dual ends include a doctor-side APP and a patient-side APP equipped with applications, which are used to communicate and interact with the integrated wearable hardware device.

2. The near real-time consciousness detection system according to claim 1, characterized in that, Also includes: An LED status indicator is installed on the shell of the integrated wearable hardware device. When the first level of evaluation is passed, it receives the first indication signal transmitted by the two-level VR eye-tracking evaluation unit and changes from the first display color to the second display color. The LED status indicator is used to receive the second indication signal transmitted by the two-level VR eye-tracking evaluation unit after the second-level evaluation is passed, and change from the second display color to the third display color, and maintain the third display color before the next evaluation; The total evaluation time for the first-level evaluation and the second-level evaluation shall last at least N minutes.

3. The near real-time consciousness detection system according to claim 2, characterized in that, The software evaluation module also includes: The fatigue protection unit is used to activate the VR display screen when the two-level VR eye-tracking evaluation unit is started, and after the total evaluation time of the first and second level evaluations lasts for N minutes, send a screen-off instruction to the VR display screen and enter the screen-off state. The VR display screen is a component of the VR headset. The pressure ulcer protection unit is used to control the movable automatic traction bracket to pull the VR eye tracker up when it receives a signal that the VR display screen has entered the screen-off state under the screen-off instruction, and to control the movable automatic traction bracket to automatically return the VR eye tracker to its original position when it receives a signal that it needs to be re-evaluated.

4. The near real-time consciousness detection system according to claim 1, characterized in that, The doctor's app is used to scan the QR code of the integrated wearable hardware device and automatically associate it with the patient's bed number, hospital number, name, and cause of illness. The doctor-side APP is also used to send instructions to the integrated wearable hardware device to force the BIS module and the two-level VR eye-tracking assessment unit to start or stop. The doctor-side APP is also used to display the VS status with a first display color, the MCS- status with a second display color, and the MCS+ / eMCS status with a third display color on the first doctor display interface. The colors are synchronized in real time with the most recent two-level VR eye-tracking assessment results. The second doctor display interface simultaneously displays a personalized arousal-consciousness fluctuation curve, a prompt for interventional rehabilitation time window, and the CRS-R visual sub-item score.

5. The near real-time consciousness detection system according to claim 1, characterized in that, The patient-side APP is used to present BIS values, eye-opening events, two-level assessment results, and changes in consciousness level in a timeline manner on the first patient display interface, and to display the highlighted sections on the personalized arousal-consciousness fluctuation curve based on the interventional rehabilitation time window on the second patient display interface, so that the patient can click on the highlighted sections on the third patient display interface to jump to the brain-computer interface rehabilitation device control interface. The patient-side app is also used to obtain the generation password authorized by the associated doctor and to receive additional assessments from family members within a limited time.

6. The near real-time consciousness detection system according to claim 1, characterized in that, The mechanical and circuit structure of the integrated wearable hardware device is as follows: BIS electrodes are arranged inside the wearable shell corresponding to the patient's forehead, the VR eye tracker's camera and scene camera are embedded in the front end of the VR headset, and the VR display screen is installed inside the wearable shell corresponding to the patient's eye position.

7. The near real-time consciousness detection system according to claim 1, characterized in that, Also includes: The control module is used to control the VR display screen to present at least one dynamic test element to conduct a preliminary test on the patient before presenting the text and image multiple-choice questions based on the VR headset, obtain quantitative auxiliary factors, and optimize the quantification process of answer correctness and reaction time to obtain the quantified consciousness level.

8. The near real-time consciousness detection system according to claim 7, characterized in that, The control module includes: The control unit is used to control the VR display screen to present at least one dynamic test element before presenting text and image multiple-choice questions based on the VR headset, and to preset a target attribute combination mode for each dynamic test element. The visual attribute dimensions of each dynamic test element include at least color, size, and brightness. The acquisition unit is used to synchronously acquire the patient's raw eye movement data based on the eye tracker during the test; The real-time processing unit is used to process the raw eye movement data in real time and parse it into a sequence of eye movement microstates in fine time windows. The eye movement microstates include: stable fixation, target-oriented saccades, retrospection, exploratory fixation, and targetless drift. The pattern generation unit is used to generate a theoretically expected eye movement pattern based on the features of the corresponding visual attribute dimension in the preset target attribute combination pattern. A line construction unit is used to time-align the eye-tracking microstate sequence with the theoretically expected eye-tracking pattern, determine the fluctuation difference of the comparative states at the same alignment time point, and form a fluctuation trend line. The parameter value determination unit is used to perform time alignment between the raw eye-tracking data and the standard eye-tracking data and divide them into N continuous analysis windows, and determine the spatial dwell ratio Zb, velocity adaptation index Zs, and trajectory morphology similarity Zs of each continuous analysis window to obtain consistency. The range determination unit is used to determine the overlap ratio range based on all consistency factors. ,in, These are the lower and upper limits of the overlap ratio range, respectively; The window division unit is used to start from the first initial window and sequentially set the starting point of the next window based on a random percentage of the overlap ratio range until the time alignment segment division is completed. The overlapping window and the continuous analysis window are the same size but may have different time starting points, and the initial window is the first window in the overlap division result. The comparison analysis unit is used to time-align the first division result of continuous division with the second division result of overlapping division, and obtain the set of overlapping windows that are in comparison with each continuous analysis window, wherein the set of overlapping windows involves at least one overlapping window. A division unit is used to divide the fluctuation trend line according to a continuous analysis window to obtain the fluctuation trend item of each continuous analysis window; The confidence determination unit is used to obtain the response confidence of the corresponding visual attribute dimension based on the overlapping window set and the consistency standard deviation of the corresponding continuous analysis window, and in combination with the consistency and fluctuation trend term of the corresponding continuous analysis window, and to construct the response confidence vector of the corresponding visual attribute dimension. The factor determination unit is used to calculate the stability index within each vector and the cross-dimensional matching index between each vector based on the response confidence vectors of all visual attribute dimensions. The results are input into the pre-trained intent inference model, and the output is the overall intent confidence that represents the patient's integrated response intent to the target attribute combination pattern as a quantification auxiliary factor to optimize the quantification process.

Images