A system for objective evaluation of dynamic characteristics of pupil light reflex

Abstract

An objective evaluation system for dynamic characteristics of pupil light reflex, comprising: acquiring a non-doubt type area and an interference area of a color pupil image; constructing a pupil diameter-time change curve of a pupil sub-block coverage area in the non-doubt type area, acquiring key judgment parameters of the pupil diameter-time change curve of the patient under different scene information conditions, and additionally generating a key judgment parameter safety threshold interval under an anesthesia depth detection mode and a key judgment parameter individualized derivative parameter under a neurology follow-up mode; constructing a clinical evaluation reference table, evaluating the dynamic parameter set according to the clinical evaluation reference table, and acquiring a reflex grade corresponding to the dynamic parameter set; generating a standardized diagnosis and treatment suggestion and a clinical early warning signal of the patient according to the reflex grade, and generating a personalized evaluation conclusion of the patient's neurological function according to the key judgment parameter individualized derivative parameter, thereby providing a fine reference basis for early screening and severity determination of the disease.

Description

Technical Field

[0001] This invention relates to the field of pupillary light reflection characteristic assessment technology, specifically an objective assessment system for the dynamic characteristics of pupillary light reflection. Background Technology

[0002] The pupillary light reflex is a core physiological reflex regulated by the central nervous system. Its dynamic characteristics, such as the amplitude, rate, latency, and recovery time of pupillary contraction / dilation, are key indicators for assessing a patient's state of consciousness, the degree of central nervous system damage, the depth of anesthesia, and the prognosis of the condition. In clinical diagnosis and treatment, the objectivity and accuracy of pupillary light reflex assessment directly affect the formulation and adjustment of treatment plans.

[0003] In existing pupillary light reflex assessment technologies, the software module, as the core processing unit, suffers from numerous technical defects, becoming a key bottleneck restricting assessment accuracy and clinical applicability: clinical assessments use a single parameter reference range, failing to consider the differences in physiological and pathological characteristics among different age groups and disease populations, and often rely on a single parameter to determine results, making misjudgments and omissions easy, resulting in insufficient adaptability of assessment results to actual clinical diagnosis and treatment needs; a few software programs with dynamic analysis functions can only calculate 2-3 basic parameters such as contraction amplitude and recovery time, lacking quantification of core dynamic characteristics such as latency, rate, and stability, and the parameter calculation logic is crude, failing to reflect subtle changes in pupillary reflex, making it difficult to meet the needs of precise clinical assessment. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide an objective evaluation system for the dynamic characteristics of pupil light reflection, comprising a cloud platform, a detection terminal, a segmentation threshold determination module, a data processing module, a reflection evaluation module, and an evaluation and early warning module.

[0005] The segmentation threshold determination module converts the color pupil image sequence to grayscale frame by frame, performs personalized denoising on each grayscale image, divides the denoised grayscale image into several rectangular sub-blocks, obtains the segmentation threshold of each rectangular sub-block, and performs regional feature recognition on each rectangular sub-block based on the segmentation threshold to obtain the non-indiscriminate type region and the interference region.

[0006] The data processing module is used to construct the pupil diameter-time change curve of the pupil sub-block coverage area in the unambiguous type region, obtain the key judgment parameters of the pupil diameter-time change curve of patients under different scenario information conditions, and additionally generate the safety threshold range of key judgment parameters under the anesthesia depth detection mode and the personalized derivative parameters of key judgment parameters under the neurological follow-up mode.

[0007] The reflex assessment module is used to construct a clinical assessment checklist, evaluate the dynamic parameter set based on the clinical assessment checklist, and obtain the reflex level corresponding to the dynamic parameter set.

[0008] The assessment and early warning module is used to generate standardized diagnosis and treatment suggestions and clinical early warning signals for patients based on reflex levels, and to generate personalized assessment conclusions of patients' neurological functions based on personalized derived parameters derived from key judgment parameters.

[0009] Furthermore, the cloud communicates with the detection terminals within a preset range. The detection terminals are used to upload the collected color pupil image sequences and the patient's population type to the detection terminals. The cloud is equipped with a segmentation threshold determination module, a data processing module, a reflex assessment module, an assessment and early warning module, and a database. The database contains several blockchain nodes, which are interconnected to form a blockchain network. Each blockchain node is linked to a data upload end, which is used to store the dynamic parameter set and reflex level of the patient generated by the reflex assessment module to the blockchain node and mark the upload time.

[0010] Furthermore, the process of obtaining the unambiguous type region and the interference region includes:

[0011] Set the local region size, divide the current frame grayscale image into several rectangular sub-blocks according to the local region size, and obtain the preliminary type of each rectangular sub-block based on the grayscale mean and grayscale extreme value difference of each rectangular sub-block;

[0012] Obtain the 3D core features of each rectangular sub-block, obtain grayscale feature benchmarks of different preliminary types, determine whether the 3D core features of each rectangular sub-block match the grayscale feature benchmarks of each preliminary type of rectangular sub-block, if they match, determine the peak type of the rectangular sub-block based on the 3D core features of the rectangular sub-block, and obtain the segmentation threshold of the rectangular sub-block based on the peak type and preliminary type of the rectangular sub-block.

[0013] Based on the preliminary type and segmentation threshold of each rectangular sub-block, the preliminary type segmentation region of each rectangular sub-block is obtained, the edge pixels in the preliminary type segmentation region are obtained, and the gradient direction variance of the preliminary type segmentation region is obtained based on the gray-level gradient direction of the edge pixels. It is determined whether the gradient direction variance of each preliminary type segmentation region matches the gradient feature benchmark of the region with the same preliminary type. If they match, the preliminary type segmentation region is marked as a region without a distinct type. If they do not match, the preliminary type segmentation region is marked as an interference region.

[0014] Furthermore, the process of constructing pupil diameter-time variation curves for the pupil sub-block coverage area within the undoubtedly type region, and obtaining key determination parameters for patients' pupil diameter-time variation curves under different scenario-based information conditions, includes:

[0015] The center of the pupil sub-block covered area in the current frame grayscale image is located, and the center coordinates and diameter of the pupil sub-block covered area are obtained;

[0016] The diameter of the pupil sub-block coverage area of ​​each grayscale image frame is extracted in real time and marked with the corresponding timestamp, generating a pupil diameter-time variation curve.

[0017] Obtain patient contextual information, including emergency rapid screening mode, routine testing mode, anesthesia depth testing mode, and neurological follow-up mode;

[0018] If the patient is in the routine testing mode, the static features of the pupil diameter-time change curve are extracted. Based on the static features and the pupil diameter-time change curve, the dynamic parameter set of the pupil diameter-time change curve is obtained. The static features and the dynamic parameter set are used as key judgment parameters.

[0019] If the patient is in the emergency rapid screening mode, the static features of the pupil diameter-time change curve are masked, and the maximum contraction rate and complete recovery time are extracted from the dynamic parameter set as key judgment parameters.

[0020] Furthermore, the process of additionally generating the safety threshold range for key judgment parameters under the anesthesia depth detection mode includes:

[0021] If the patient is in anesthesia depth detection mode, a preset anesthesia depth-pupil parameter safety threshold mapping library is used. This library includes safety thresholds for key judgment parameters corresponding to different anesthesia depths and population types. The dynamic parameter set of pupil diameter and pupil diameter-time change curve is used as key judgment parameters. The patient's current anesthesia depth and population type are input into the anesthesia depth-pupil parameter safety threshold mapping library, which outputs the safety threshold range of key judgment parameters corresponding to the population type and current anesthesia depth. The key judgment parameters are compared with their corresponding safety threshold ranges. If any key judgment parameter exceeds its corresponding safety threshold range, an audible and visual warning is immediately triggered. At the same time, any key judgment parameter is marked as an abnormal parameter, and the abnormal parameter is compared with its safety threshold range to generate the degree of deviation. The abnormal parameter and its degree of deviation are then fed back to the detection terminal.

[0022] Furthermore, the process of generating personalized derived parameters for key diagnostic parameters in the neurological follow-up model includes:

[0023] If the patient is in neurological follow-up mode, the patient's full historical data is extracted from the blockchain node. The full historical data includes the patient's population type, individual key judgment parameter baseline data, previous and previous follow-up test data, individual baseline threshold range, and the weight of neurological function indicators corresponding to the disease type. The static features and dynamic parameter set of the current pupil diameter-time change curve are used as key judgment parameters. The absolute difference and relative change rate between the current key judgment parameters and the individual key judgment parameter baseline data and the previous follow-up test data are obtained. Based on the current key judgment parameters and previous follow-up test data, a linear regression algorithm is used to obtain the change trend and change rate of each key judgment parameter. At the same time, the static features and dynamic parameter sets of the pupil diameter-time change curves of the patient's left and right pupils are compared to obtain the bilateral parameter symmetry. The bilateral parameter symmetry is compared with the previous follow-up test data to obtain the asymmetry change coefficient.

[0024] The current key judgment parameters are compared with the individual baseline threshold range, and key judgment parameters that exceed the individual baseline threshold range are marked as abnormal parameters.

[0025] Furthermore, the process of constructing a clinical assessment checklist includes:

[0026] Using different population types and contextual information as cluster centers, the static features and dynamic parameter sets of several different population types and contextual information in the database are clustered to obtain the static features and dynamic parameter sets of different population types under different contextual information conditions. The distribution type test is performed on each group of dynamic parameters in the static features and dynamic parameter sets of different population types to obtain the distribution type of the static features and each group of dynamic parameters of different population types under different contextual information conditions. Based on the distribution type, the general reference range of the static features and each group of dynamic parameters of different population types under different contextual information conditions is obtained. The clinical classification of the static features and each group of dynamic parameters is obtained. Based on the clinical classification and general reference range of the static features and each group of dynamic parameters, the critical interval of the reflex level of the static features and each group of dynamic parameters is obtained.

[0027] A clinical assessment comparison table was constructed based on the static characteristics of different population types under different contextual information conditions and the critical intervals of the reflex levels of each group of dynamic parameters.

[0028] Furthermore, the process of evaluating key judgment parameters for patients in emergency rapid screening mode, routine testing mode, or anesthesia depth testing mode based on a clinical assessment checklist, and obtaining the reflex levels corresponding to the key judgment parameters, includes:

[0029] The key parameters of the pupil diameter-time change curve and the patient's population type are used as evaluation indicators. The indicator weights of the evaluation indicators are set, and the clinical assessment control table is used as the fuzzy evaluation rule table. The membership matrix of the evaluation indicators for different reflex levels is obtained through fuzzy comprehensive evaluation. The reflex level corresponding to the evaluation indicator is obtained according to the membership matrix and the indicator weights.

[0030] Furthermore, the process of generating standardized diagnostic and treatment recommendations and clinical warning signals for patients based on reflex levels, and generating personalized neurological function assessment conclusions for patients based on personalized derived parameters derived from key judgment parameters, includes:

[0031] A pre-defined inline clinical knowledge base is provided, which includes standardized treatment suggestions and clinical warning signals corresponding to different reflex levels and population types. If the patient is in the emergency rapid screening mode, routine testing mode, or anesthesia depth testing mode, the patient's population type and reflex level are input into the inline clinical knowledge base, and the standardized treatment suggestions and clinical warning signals for the patient are output.

[0032] If the patient is in neurological follow-up mode, and the preset neurological function grading and quantitative judgment rules are used, then the patient's personalized neurological function assessment conclusion will be generated based on the neurological function grading and quantitative judgment rules, the absolute difference, relative rate of change, trend of change, rate of change, asymmetric change coefficient, and whether there are abnormal parameters.

[0033] Compared with the prior art, the beneficial effects of the present invention are:

[0034] 1. This invention addresses four core scenarios in clinical practice: rapid emergency screening, routine testing, anesthesia depth monitoring, and neurological follow-up. It designs personalized key parameter extraction rules and analysis logic, achieving a deep match between testing modes and clinical needs: the rapid emergency screening mode filters out non-core parameters, extracting only the maximum contraction rate and full recovery time, ensuring shorter analysis time and adapting to rapid triage requirements; the anesthesia depth monitoring mode constructs a dedicated safety threshold range, enabling real-time audio-visual warnings for parameters exceeding thresholds, assisting in adjusting anesthetic drug dosages; the neurological follow-up mode generates personalized derivative parameters for key judgment parameters, enabling quantitative analysis and trend determination of parameter changes; and the routine testing mode extracts all static and dynamic parameters to meet basic clinical assessment needs. This scenario-based design breaks through the existing system's "one-size-fits-all" parameter extraction mode, avoiding interference from non-core parameters in clinical decision-making, while achieving a balance between testing efficiency and assessment accuracy in different scenarios. This allows the system to cover the testing needs of multiple departments, including emergency medicine, anesthesiology, neurology, and ophthalmology, significantly expanding its clinical application scope.

[0035] 2. Based on the physiological and pathological characteristics of different population groups, dynamic parameter reference ranges for each population are constructed through cluster analysis and distribution type tests. This breaks through the limitations of traditional single reference standards that can be applied to all populations, and fully considers the differences in pupillary reflex characteristics among children, the elderly, and patients with different diseases. Simultaneously, parameters are categorized into superior, inferior, and stable parameters according to their clinical significance, and critical intervals for reflex level are customized for different types of parameters, constructing a scientific and hierarchical clinical assessment checklist. Furthermore, fuzzy comprehensive evaluation is introduced, and index weights are set in conjunction with clinical expert experience to achieve comprehensive judgment of multiple parameters. This avoids misjudgments caused by abnormalities in a single parameter, making the assessment results more aligned with actual clinical diagnostic and treatment needs. It can not only accurately determine the normality and abnormality of pupillary reflexes but also classify the degree of abnormality, providing a refined reference for early disease screening and severity assessment. Attached Figure Description

[0036] Figure 1 This is a flowchart of an objective evaluation system for the dynamic characteristics of pupil light reflection, according to an embodiment of this application. Detailed Implementation

[0037] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0038] like Figure 1 As shown, an objective evaluation system for the dynamic characteristics of pupil light reflection includes a cloud platform, a detection terminal, a segmentation threshold determination module, a data processing module, a reflection evaluation module, and an evaluation and early warning module.

[0039] The segmentation threshold determination module converts the color pupil image sequence to grayscale frame by frame, performs personalized denoising on each grayscale image, divides the denoised grayscale image into several rectangular sub-blocks, obtains the segmentation threshold of each rectangular sub-block, and performs regional feature recognition on each rectangular sub-block based on the segmentation threshold to obtain the non-indiscriminate type region and the interference region.

[0040] The data processing module is used to construct the pupil diameter-time change curve of the pupil sub-block coverage area in the unambiguous type region, obtain the key judgment parameters of the pupil diameter-time change curve of patients under different scenario information conditions, and additionally generate the safety threshold range of key judgment parameters under the anesthesia depth detection mode and the personalized derivative parameters of key judgment parameters under the neurological follow-up mode.

[0041] The reflex assessment module is used to construct a clinical assessment checklist, evaluate the dynamic parameter set based on the clinical assessment checklist, and obtain the reflex level corresponding to the dynamic parameter set.

[0042] The assessment and early warning module is used to generate standardized diagnosis and treatment suggestions and clinical early warning signals for patients based on reflex levels, and to generate personalized assessment conclusions of patients' neurological functions based on personalized derived parameters derived from key judgment parameters.

[0043] It should be further explained that, in the specific implementation process, the cloud communicates with the detection terminals within a preset range. The detection terminals are used to upload the collected color pupil image sequences and the patient's population type to the detection terminals. The cloud is equipped with a segmentation threshold determination module, a data processing module, a reflex assessment module, an assessment and early warning module, and a database. The database is pre-built based on 10,000+ clinical samples, covering parameters corresponding to population types such as healthy adults, healthy children (3-12 years old), healthy elderly (≥60 years old), and patients of different ages with diabetes, cerebrovascular diseases, and glaucoma. This system collaborates with multiple medical institutions, obtaining data from the hospital's HIS system with patient authorization, and adhering to privacy protection principles to ensure the reliability and legality of the data source. Several blockchain nodes are set up in the database, and the blockchain nodes are interconnected to form a blockchain network. Each blockchain node is linked to a data up-chain end, which is used to store the dynamic parameter set and reflex level of the patient generated by the reflex assessment module to the blockchain node and mark the up-chain time.

[0044] It should be further explained that, in the specific implementation process, the detection terminal includes:

[0045] Light source component: It adopts an adjustable intensity cold light source as the stimulation light source, and has a built-in light source driving circuit that can output standardized single / multiple light stimulation signals according to a preset program.

[0046] Imaging components include a high-definition miniature camera, a macro lens, and an infrared fill light unit. The camera has a frame rate of no less than 30 frames per second and can capture dynamic images of the pupil in real time during light source stimulation.

[0047] Pupil positioning component: Built-in distance sensor and angle calibration sensor are used to collect the distance and angle between the detection terminal and the eye being tested. When the distance or angle exceeds the preset threshold, an alert signal is issued to ensure that the imaging component is always aligned with the pupil area.

[0048] Data transmission unit: Using Bluetooth or WiFi modules, the acquired dynamic pupil images are transmitted to the cloud in real time.

[0049] It should be further explained that, in the specific implementation process, the personalized denoising process for each frame of grayscale image includes:

[0050] The sequence of colored pupil images is converted to grayscale frame by frame, and the noise level threshold of each grayscale image is obtained, including:

[0051] Each frame (1080p) of the color pupil image sequence (30 frames / second) is converted to grayscale to generate a grayscale image. The local variance (neighborhood of 3×3) of each pixel (i,j) in the grayscale image is obtained. Statistical analysis is performed on the local variance of each pixel in the grayscale image to obtain the mean and minimum local variance. Based on the mean local variance... and local variance minimum Obtain noise level threshold .

[0052] ;in, The value is empirically set to 0.15 for pupil images to ensure that the selected areas are flat regions with no structure and pure noise.

[0053] Then, iterate through all pixels in the current frame grayscale image. If the local variance of a pixel is less than the noise level threshold, the pixel is marked as a pure noise region pixel. If the local variance of a pixel is greater than or equal to the noise level threshold, the pixel is marked as a structural region pixel. Perform grayscale sampling and noise level analysis on the pure noise region pixels to obtain the global noise index. Then, based on the global noise index of the previous frame grayscale image, smooth the global noise index of the current frame grayscale image to obtain the global noise index of the current frame grayscale image after smoothing correction.

[0054] The process of performing noise level analysis and inter-frame smoothing correction on pixels in purely noisy regions includes:

[0055] ;in, This is a global noise index. This represents the total number of pixels in the pure noise region. Let be the grayscale value of the k-th pixel in the pure noise region;

[0056] Where s is the current frame number, =0.7, Let be the global noise index for the s-th frame. The global noise index for smoothing correction of the s-th frame.

[0057] Finally, an adaptive model is constructed, which compares the local variance of each pixel in the current frame grayscale image with the global noise index to obtain the variance-to-noise ratio of each pixel. The variance-to-noise ratio of each pixel is input into the adaptive model, and the Gaussian filter parameters of each pixel are output according to the adaptive model. The corresponding Gaussian filter is applied to each pixel according to the Gaussian filter parameters to process it and obtain the denoised current frame grayscale image.

[0058] The specific output rules of the adaptive model are as follows:

[0059] For a given pixel, the local variance LV(i,j) has a variance-to-noise ratio of R. If 0.8≤R≤1.2, the local variance is close to the noise level, and it is determined to be a noise-dominated region. Increasing the standard deviation σ of the Gaussian filter (σ=2.0) enhances the smoothing effect.

[0060] If R > 2, the local variance is much greater than the noise level, which is determined to be a signal-dominant region (edge / structure). Reduce σ (σ = 0.8) to reduce the blurring of the edge.

[0061] If 1.2 < R < 2, it is determined to be a transition region, and a medium σ (σ = 1.4) is adopted to balance the requirements of noise reduction and edge preservation.

[0062] It should be further explained that, in the specific implementation process, the process of dividing the rectangular sub-blocks based on the global gray-level mean of the denoised gray-level image, pre-labeling the preliminary type according to the gray-level mean and extreme value difference of the rectangular sub-blocks, constructing the gray-level histogram of the rectangular sub-blocks and extracting the three-dimensional core features, and obtaining the segmentation threshold of the rectangular sub-blocks by combining the gray-level feature benchmark and the three-dimensional core features includes:

[0063] Obtain the global grayscale mean of the current frame grayscale image after denoising. Set the local region size (16×16), and divide the current frame grayscale image into several rectangular sub-blocks based on the local region size. Let the current frame grayscale image size be W×H, and the sub-block size be B×B (B=16). Divide it horizontally into Nw=W / B blocks and vertically into Nh=H / B blocks. For sub-blocks with edges smaller than 16×16, use mirror filling to supplement pixels to avoid histogram statistical bias in the edge region.

[0064] Extract the grayscale mean of each rectangular sub-block and grayscale extreme value difference Based on the global grayscale mean and the grayscale mean of each rectangular sub-block, a preliminary type pre-label is performed on each rectangular sub-block to obtain the preliminary type of each rectangular sub-block. The preliminary type includes pupil candidate sub-block, iris candidate sub-block, sclera candidate sub-block, and interference candidate region; if the grayscale mean of the sub-block... <0.3 If , then it is marked as a candidate pupil sub-block; if 0.3 ≤ ≤0.7 If the average gray value of the sub-block is less than 1, it is marked as a candidate sub-block for the iris; if the average gray value of the sub-block is less than 1, it is marked as a candidate sub-block for the iris. >0.7 If it is marked as a candidate sub-block of sclera, then it is considered a candidate sub-block of sclera. =1, then it is marked as a candidate region of interference;

[0065] Construct a grayscale histogram for each rectangular sub-block, with grayscale levels k (0≤k≤255). Count the number of pixels at each grayscale level (0-255) within the rectangular sub-block. Obtain the pixel count for each grayscale level and extract the 3D core features from the grayscale histogram of each rectangular sub-block. The 3D core features include: the dominant peak grayscale value. The gray level with the highest pixel frequency in the histogram reflects the dominant gray level of the sub-block; histogram width... A grayscale range where pixel accumulation accounts for ≥90% reflects the concentration of grayscale distribution; a bimodal criterion. If the histogram has two distinct peaks (peak difference ≥ 10% of maximum frequency, valley frequency ≤ 30% of maximum frequency), then =1 (indicates the existence of a boundary between two regions), otherwise =0. Based on clinical sample statistics, different preliminary grayscale feature benchmarks were obtained, including the peak grayscale, histogram width, and threshold ranges corresponding to the bimodal judgment index;

[0066] Pupil candidate sub-block: ≤50, ≤30, =0 (highly concentrated grayscale distribution);

[0067] Iris candidate sub-blocks: 50 < <180, ≥50, It can be 0 or 1 (the gray level inside the iris is uniform, or it is at the junction with the pupil / sclera).

[0068] Scleral candidate sub-blocks: ≥180, ≤40, =0 (highly concentrated grayscale distribution);

[0069] Interference candidate sub-blocks: =1, the main peak grayscale fluctuation is large (such as the simultaneous existence of peaks with grayscale values ​​of 20 and 200), which is likely to be eyelashes or reflective areas.

[0070] The three-dimensional core features of each rectangular sub-block are matched with the gray-scale feature benchmark of the preliminary type of each rectangular sub-block. If the three-dimensional core features of the rectangular sub-block match the gray-scale feature benchmark, the peak type of the rectangular sub-block is determined according to the three-dimensional core features of the rectangular sub-block. The segmentation threshold of the rectangular sub-block is obtained based on the peak type and the preliminary type of the rectangular sub-block.

[0071] It should be further explained that, in the specific implementation process, the segmentation threshold for obtaining rectangular sub-blocks includes:

[0072] for For a single peak with a value of 0, the threshold is calculated using the mean-standard deviation method. = - ,in, The threshold for segmenting rectangular sub-blocks. Here, α is the grayscale standard deviation of the rectangular sub-block, and α is an adaptive coefficient that is dynamically adjusted according to the pre-labeling type of the sub-block.

[0073] Pupil candidate sub-block: α=1.2, reduce the threshold to ensure that low grayscale pixels in the pupil region are completely segmented;

[0074] Scleral candidate sub-block: α=0.8, to increase the threshold and avoid scleral reflective points being misclassified as dark areas;

[0075] for For a bimodal structure with a value of 1, the valley gray level between the two peaks is used as the segmentation threshold. This threshold can accurately distinguish different regions on both sides of the boundary, avoiding the blurring of the boundary caused by a single threshold. For example, the histogram of the iris-pupil boundary sub-block has "low gray-level peak (pupil)" and "medium gray-level peak (iris)", and the valley gray level is the optimal segmentation boundary between the two.

[0076] For interfering candidate sub-blocks, the global Otsu threshold is directly used as the segmentation threshold for this type of sub-block to avoid local threshold failure caused by interference factors such as eyelashes and reflections.

[0077] It should be further explained that, in the specific implementation process, the process of cutting each rectangular sub-block based on the segmentation threshold to obtain the preliminary type segmentation region, performing region feature recognition on the preliminary type segmentation region, and dividing the preliminary type segmentation region into non-type regions and interference regions based on the recognition results includes:

[0078] Based on the preliminary type and segmentation threshold of each rectangular sub-block, the pixels of each rectangular sub-block are segmented to obtain the preliminary type segmentation region of each rectangular sub-block. Obtaining the preliminary type segmentation region of each rectangular sub-block includes:

[0079] Pupil candidate region determination: The grayscale values ​​in the rectangular sub-blocks are... Segmentation threshold less than or equal to rectangular sub-blocks The pixels are marked as pupil sub-blocks;

[0080] Iris candidate region determination: The gray values ​​in the rectangular sub-blocks are... Pixels that are greater than the lower limit of the segmentation threshold (segmentation threshold for pupil candidate regions) and less than the upper limit of the segmentation threshold are marked as iris sub-blocks;

[0081] Scleral candidate region determination: grayscale value in rectangular sub-blocks Pixels larger than the segmentation threshold are marked as scleral sub-blocks;

[0082] Extract the grayscale gradient magnitude and direction of each pixel in each preliminary segmentation region (using the Sobel operator to calculate the horizontal gradient Gx and vertical gradient Gy of each pixel, and then solve for the gradient magnitude G and gradient direction θ). A preset gradient magnitude threshold is used, and pixels with grayscale gradients greater than the threshold are marked as edge pixels. Based on the grayscale gradient direction of the edge pixels in each preliminary segmentation region, the gradient direction variance Var(θ) of each preliminary segmentation region is obtained. Based on clinical sample statistics, regional gradient feature benchmarks for different preliminary types are obtained. The gradient direction variance of each preliminary segmentation region is matched with the regional gradient feature benchmarks of the same preliminary type. If the gradient direction variance of the preliminary segmentation region matches the regional gradient feature benchmark, the preliminary segmentation region is marked as a region without a defined type. Regions without a defined type include pupil sub-blocks, iris sub-blocks, and sclera sub-blocks. If the gradient direction variance of the preliminary segmentation region does not match the regional gradient feature benchmark, the preliminary segmentation region is marked as an interference region.

[0083] Among them, regional gradient feature benchmark matching includes:

[0084] For the pupil edge: if Var(θ)≤15° (the edge direction is distributed in a ring shape with high consistency) and it is located inside the continuous ring edge, then the segmented region of the pupil candidate region is marked as an indisputable type region; otherwise, it is marked as an interference region.

[0085] For the pupil-iris / iris-sclera edge: if Var(θ)≤15° and it is located between the pupil and sclera, then the segmented region of the iris candidate region is marked as an indisputable type region; otherwise, it is marked as an interference region.

[0086] For the edge of the scleral region, if Var(θ)≤15° and it is located outside the continuous ring edge, the segmented region of the scleral candidate region is marked as an indisputable type region; otherwise, it is marked as an interference region.

[0087] If Var(θ) > 45° (the edge direction is messy and irregular, such as the edge of an eyelash), then the preliminary type segmentation region is marked as an interference region.

[0088] It should be further explained that, in the specific implementation process, the process of constructing the pupil diameter-time variation curve of the pupil sub-block coverage area in the undoubtedly type region, and obtaining the key judgment parameters of the patient's pupil diameter-time variation curve under different scenario information conditions includes:

[0089] Hough circle detection is used to locate the center of the pupil sub-block coverage area in the current frame grayscale image, and the center coordinates and diameter of the pupil sub-block coverage area are obtained.

[0090] The diameter of the pupil sub-block coverage area of ​​each grayscale image is extracted in real time and marked with the corresponding timestamp to generate a pupil diameter-time change curve;

[0091] Acquire patient contextual information, including emergency rapid screening mode, routine testing mode, anesthesia depth testing mode, and neurological follow-up mode (medical staff can switch between contextual information with one click through the testing terminal).

[0092] If the patient is in the standard testing mode, the static features of the pupil diameter-time change curve are extracted. These static features include: baseline pupil diameter: the average pupil diameter within 3 seconds before light stimulation, reflecting the pupil's resting state; minimum pupil diameter: the diameter when the pupil constricts to its minimum state during light stimulation; recovery pupil diameter: the diameter when the pupil dilates to 90% of the baseline diameter after light stimulation ends; constriction time: the time taken from the start of light stimulation until the pupil reaches its minimum diameter; and dilation time: the time taken from the pupil reaching its minimum diameter to recovering to its normal pupil diameter. Based on the static features and the pupil diameter-time change curve, a dynamic parameter set for the pupil diameter-time change curve is obtained, and the static features and dynamic parameter set are used as key judgment parameters.

[0093] If the patient is in the emergency rapid screening mode, the static features of the pupil diameter-time change curve are masked, and the maximum constriction rate and complete recovery time in the dynamic parameter set are extracted as key judgment parameters (subsequently, only the three levels of reflexes, "normal", "dull" and "disappeared", are distinguished).

[0094] It should be further explained that, in the specific implementation process, the dynamic parameter set includes:

[0095] Constriction latency: the time from the onset of light stimulation to a significant decrease in pupil diameter (≥0.2 mm, reduced to 0.1 mm in children);

[0096] Maximum rate of contraction: The maximum slope of the pupil diameter-time curve during the pupil contraction phase, in mm / s, reflecting the speed of pupil contraction;

[0097] Constriction amplitude: The difference between the baseline pupil diameter and the minimum pupil diameter, in mm, reflecting the strength of pupil constriction;

[0098] Diastolic latency: The time from the end of light stimulation to the start of pupillary diameter increase (≥0.2 mm, reduced to 0.1 mm in children), reflecting the speed of pupillary diastolic initiation;

[0099] Maximum diastolic rate: The maximum slope of the pupil diameter-time curve during the pupillary dilation phase, in mm / s, reflecting the speed of pupillary dilation;

[0100] Full recovery time: The time required for the pupil diameter to recover to 90% of its baseline diameter from the end of light stimulation, measured in seconds, reflecting the pupillary reflex recovery ability;

[0101] Coefficient of variation: The ratio of the standard deviation to the mean of each dynamic parameter under multiple pulse stimulations, reflecting the stability of the pupillary reflex.

[0102] It should be further explained that, in the specific implementation process, the process of additionally generating the safety threshold range of key judgment parameters under the anesthesia depth detection mode includes:

[0103] If the patient is in anesthesia depth detection mode, a preset anesthesia depth-pupil parameter safety threshold mapping library is used. This library includes safety thresholds for key judgment parameters corresponding to different anesthesia depths and population types. The dynamic parameter set of pupil diameter and pupil diameter-time change curve is used as key judgment parameters. The patient's current anesthesia depth (light, moderate, and deep) and population type are input into the anesthesia depth-pupil parameter safety threshold mapping library. The library outputs the safety threshold ranges for key judgment parameters corresponding to the population type and current anesthesia depth. The key judgment parameters are compared with their corresponding safety threshold ranges. If any key judgment parameter exceeds its corresponding safety threshold range, an audible and visual warning is immediately triggered. At the same time, any key judgment parameter is marked as an abnormal parameter, and the abnormal parameter is compared with its safety threshold range to generate the degree of deviation. The abnormal parameter and its degree of deviation are fed back to the detection terminal to prompt medical staff to adjust the dosage of anesthetic drugs.

[0104] It should be further explained that, in the specific implementation process, the process of generating personalized derived parameters for key judgment parameters under the neurological follow-up model includes:

[0105] If the patient is in neurological follow-up mode, the patient's full historical data is extracted from the blockchain node. This full historical data includes the patient's population type (age, underlying diseases, previous surgical history, medication history), individual key judgment parameter baseline data (baseline values ​​of all parameters in the static and dynamic parameter sets established during the initial test, including bilateral individual baselines and bilateral average baselines), previous and all follow-up test data (all parameters in the static and dynamic parameter sets), individual baseline threshold ranges (parameter safety threshold ranges defined for this patient), and the weights of neurological function indicators corresponding to the disease type (e.g., for patients with cerebrovascular diseases, focus on contraction latency and bilateral symmetry parameters). The static features and dynamic parameter set of the current pupil diameter-time change curve are used as key judgment parameters. The absolute difference and relative rate of change between the current key judgment parameter and the individual key judgment parameter baseline data and the previous follow-up test data are obtained (rate of change = (current value - comparison value) / comparison value × 100%). For example, "the maximum contraction rate of the left eye this time is 1..." The current rate of contraction is 0.1 mm / s, compared to 0.9 mm / s previously, with a baseline value of 1.0 mm / s. The absolute difference is +0.2 mm / s, the relative rate of change is +22.2%, and the difference from the baseline is +0.1 mm / s. Based on the current key judgment parameters and previous follow-up data, a linear regression algorithm is used to obtain the changing trends (continuous increase / continuous decrease / stable / fluctuating) and rates of change of the current key judgment parameters (e.g., "the maximum contraction rate has been continuously increasing in the last three follow-up visits, with an average monthly increase of 0.12 mm / s"). Targeting the core characteristics of neurological diseases (such as unilateral brainstem injury and optic neuropathy), the static characteristics and dynamic parameter sets of the pupil diameter-time change curves of the left and right pupils of the patient are compared to obtain the bilateral parameter symmetry (difference). The bilateral parameter symmetry is then compared with the previous follow-up test data to obtain the asymmetry change coefficient (e.g., "the left eye contraction latency is 0.08s longer than the right eye, the previous comparison difference was 0.05s, the difference increased by 0.03s, indicating that the unilateral nerve function recovery is slower").

[0106] The current key judgment parameters are compared with the individual baseline threshold range, and key judgment parameters that exceed the individual baseline threshold range are marked as abnormal parameters.

[0107] It should be further explained that, in the specific implementation process, the process of constructing the clinical assessment checklist includes:

[0108] Using different population types and contextual information as cluster centers, the static features and dynamic parameter sets of several different population types and contextual information in the database are clustered to obtain the static features and dynamic parameter sets of different population types under different contextual information conditions. Distribution type tests are performed on each group of dynamic parameters in the static features and dynamic parameter sets of different population types. The null hypothesis H0 is set as follows: a certain dynamic parameter of a certain population type (such as the maximum contraction rate of healthy adults) follows a normal distribution. The p-value (the probability of randomly collecting the current sample data under the premise that H0 is true) is calculated using the Shapiro-Wilk test, and the significance level of the test is set to α=0.05.

[0109] If the p-value of the dynamic parameter is greater than 0.05, it is determined that the dynamic parameter conforms to a normal distribution (such as the maximum contraction rate, contraction amplitude, and maximum diastolic rate of a healthy adult).

[0110] If the P-value of the dynamic parameter is less than or equal to 0.05, the dynamic parameter is determined to be skewed (e.g., the systolic latency, diastolic latency, and complete recovery time in children / elderly individuals are prone to shift to one side due to individual physiological development / degenerative changes).

[0111] Obtain the static features and distribution type (normal / skewed distribution) of each set of dynamic parameters for different population types under different contextual information conditions. Based on the distribution type, obtain the general reference range for the static features and dynamic parameters of different population types under different contextual information conditions, specifically:

[0112] Normal distribution parameters: mean ± 1.96 times standard deviation (μ ± 1.96σ), where μ is the parameter mean, σ is the parameter standard deviation, and 1.96 is the critical value of the 95% confidence interval in a normal distribution, representing coverage of the middle 95% of individuals, excluding the extreme values ​​of the two 2.5%. Example: For the maximum contraction rate in a healthy adult group, statistically, μ = 1.6 mm / s and σ = 0.2 mm / s, then the general reference range = 1.6 ± 1.96 × 0.2 = 1.2 ~ 2.0 mm / s.

[0113] Skewed distribution parameters: Sort all dynamic parameters of the current group from smallest to largest, take the 2.5th percentile (P2.5) as the lower limit and the 97.5th percentile (P97.5) as the upper limit, directly covering the middle 95% of individuals. For example, for the elderly group, after sorting, P2.5 = 4.0s and P97.5 = 8.0s, then the general reference range is 4.0~8.0s.

[0114] Obtain the clinical classification (excellent value parameter / inferior value parameter) of static features and each group of dynamic parameters, and obtain the critical interval of the reflex level (sensitive / normal / sluggish / absent) of static features and each group of dynamic parameters based on the clinical classification and general reference range.

[0115] A clinical assessment comparison table was constructed based on the static characteristics of different population types under different contextual information conditions and the critical intervals of the reflex levels of each group of dynamic parameters.

[0116] It should be further explained that, in the specific implementation process, the process of obtaining the four-level critical intervals for each group of dynamic parameters based on the clinical classification and general reference range of each group of dynamic parameters includes:

[0117] Based on the clinical significance of each dynamic parameter, they are divided into "good value parameters" (the higher the value, the better) and "bad value parameters" (the lower the value, the better), and a classification logic is formulated for each. The critical points are all derived based on the boundary values ​​of the general reference range.

[0118] (1) Excellent parameters (maximum systolic rate, maximum diastolic rate, amplitude of contraction):

[0119] Sensitivity: Parameter value ≥ upper limit of normal reference range;

[0120] Normal: The parameter value falls within the normal reference range;

[0121] Sluggishness: The parameter value is less than the lower limit of the normal reference range and greater than the disappearance threshold;

[0122] Disappearance: Maximum systolic / diastolic rate < 0.4 mm / s, systolic amplitude < 0.2 mm (the core threshold for clinically determining no reflex, verified by a large number of clinical trials).

[0123] (2) Inferior parameters (systolic latency, diastolic latency, and complete recovery time):

[0124] Sensitivity: Parameter value ≤ lower limit of normal reference range;

[0125] Normal: The parameter value falls within the normal reference range;

[0126] Sluggish: Parameter value > upper limit of normal reference range, and < disappearance threshold;

[0127] Disappearance: systolic / diastolic latency > 1.0s, complete recovery time > 15s (clinically determined to be no reflex, pupil does not respond to light stimulation).

[0128] (3) Stability parameter (coefficient of variation):

[0129] There is no distinction between good and bad values ​​for the coefficient of variation; it is directly classified according to the stability of the reflection, and the classification logic is consistent for all basic populations.

[0130] Stable (corresponding to sensitivity / normal): ≤15%;

[0131] Basically stable (corresponding to normal): 15%~25%;

[0132] Unstable (corresponding to sluggishness): 25%~35%;

[0133] Extremely unstable (corresponding disappearance): >35% (parameters fluctuate greatly and irregularly under repeated stimulation, clinically judged to be severely impaired reflex function).

[0134] It should be further explained that, in the specific implementation process, the key judgment parameters of patients in the emergency rapid screening mode, routine testing mode, or anesthesia depth testing mode are evaluated according to the clinical assessment checklist, and the process of obtaining the reflex level corresponding to the key judgment parameters includes:

[0135] The key parameters of the pupil diameter-time change curve and the patient's population type are used as evaluation indicators. The indicator weights of the evaluation indicators are set (determined based on expert experience to reduce uncertainty in the fuzzy comprehensive evaluation process). The clinical assessment checklist is used as the fuzzy evaluation rule table. The membership matrix of the evaluation indicators for different reflex levels is obtained through fuzzy comprehensive evaluation. The reflex level corresponding to the evaluation indicator is obtained based on the membership matrix and indicator weights.

[0136] It should be further explained that, in the specific implementation process, the process of obtaining the reflection level of the evaluation index based on the membership matrix and index weights includes:

[0137] The evaluation index weights and membership matrices are fused by formula to obtain the fuzzy comprehensive evaluation matrix of the evaluation index. The membership degree of the evaluation index for different reflection levels is obtained based on the fuzzy comprehensive evaluation matrix. The reflection level with the highest membership degree corresponding to the evaluation index is selected and taken as the reflection level of the evaluation index.

[0138] The formula is as follows:

[0139] ;

[0140] in, The fuzzy comprehensive evaluation matrix for the evaluation indicators. To evaluate the indicator weights, For the membership matrix, "This indicates that the elements at corresponding positions in the weight matrix and membership matrix of the evaluation index are multiplied together." and The weighting parameter is used to balance the weight matrix and membership matrix in the fuzzy comprehensive evaluation matrix used to control the evaluation index.

[0141] Based on the reflex level and population type, the system automatically connects to the built-in clinical knowledge base to provide possible etiological clues and treatment suggestions, such as "prolonged contraction latency may indicate brainstem functional damage, and further head CT examination is recommended," providing medical staff with diagnostic and treatment references, especially suitable for primary healthcare institutions or young physicians.

[0142] It should be further explained that, in the specific implementation process, the process of generating standardized diagnosis and treatment suggestions and clinical early warning signals for patients based on reflex levels, and the process of generating personalized neurological function assessment conclusions for patients based on personalized derived parameters from key judgment parameters, includes:

[0143] A pre-defined inline clinical knowledge base is provided, which includes standardized treatment suggestions and clinical warning signals corresponding to different reflex levels and population types. If the patient is in the emergency rapid screening mode, routine testing mode, or anesthesia depth testing mode, the patient's population type and reflex level are input into the inline clinical knowledge base, and the standardized treatment suggestions and clinical warning signals for the patient are output.

[0144] If the patient is under neurological follow-up mode, and pre-set quantitative judgment rules for neurological function grading are established, then a personalized neurological function assessment conclusion is generated based on these rules, the absolute difference, relative rate of change, trend of change, rate of change, coefficient of asymmetric change, and the presence of abnormal parameters. Specifically, this includes:

[0145] The neurological function grading and quantitative judgment rules divide the neurological function status into three categories, clearly indicating the judgment criteria: ① Recovery status: The parameters of this test are significantly improved compared with the previous test. The key judgment parameters (such as contraction rate, contraction amplitude, bilateral symmetry) must meet the following requirements: the absolute difference is ≥ the system detection accuracy (0.1), and the relative change rate is ≥ 15%, and the change trend is continuously increasing / continuously decreasing (such as increasing contraction rate and shortening latency), and the change rate is stable (without sudden increases or decreases), and there are no new abnormal parameters, and the asymmetry change coefficient is less than the preset threshold (0.05s). For example, "the maximum contraction rate of this test is 22.2% higher than the previous test, the contraction latency is shortened by 18.5%, and the bilateral symmetry difference is reduced by 0.03s, which is judged as a significant recovery of neurological function, indicating that the current treatment plan is effective."

[0146] ② Stable state: Compared with the previous test, the absolute difference of the parameters in this test is less than the system detection accuracy (0.1), or the relative change rate is less than 10%, and all parameters are within the normal fluctuation range of the individual, and there are no new abnormal parameters. For example, "the difference of each parameter from the previous test is less than 0.1 mm / s, the bilateral symmetry is stable, and the neurological function is determined to be stable. It is recommended to maintain the current treatment plan."

[0147] ③ Abnormal state: The parameters detected this time are significantly worse than the previous one, with an absolute difference ≥ the system detection accuracy (0.1), a relative change rate ≤ -15%, or the appearance of new abnormal parameters, or the asymmetry change coefficient is greater than or equal to the preset threshold. For example, "the maximum contraction rate decreased by 25.0% compared to the previous time, and the bilateral symmetry difference increased by 0.06s, which is judged as abnormal neurological function, suggesting that there may be a worsening of the condition or inappropriate medication."

[0148] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An objective evaluation system for the dynamic characteristics of pupillary light reflection, characterized in that, It includes cloud computing, detection terminal, segmentation threshold determination module, data processing module, reflection evaluation module, and evaluation and early warning module; The cloud communicates with the detection terminals within a preset range. The detection terminals are used to upload the collected color pupil image sequence and the patient's population type to the detection terminals. The cloud is equipped with a segmentation threshold determination module, a data processing module, a reflex assessment module, an assessment and early warning module, and a database. The database is equipped with several blockchain nodes, which are linked to each other to form a blockchain network. Each blockchain node is linked to a data uplink end, which is used to store the dynamic parameter set and reflex level of the patient generated by the reflex assessment module to the blockchain node and mark the uplink time. The segmentation threshold determination module converts the color pupil image sequence to grayscale frame by frame, performs personalized denoising on each grayscale image, divides the denoised grayscale image into several rectangular sub-blocks, obtains the segmentation threshold of each rectangular sub-block, and performs regional feature recognition on each rectangular sub-block based on the segmentation threshold to obtain the non-type regions and interference regions. The non-type regions include pupil sub-blocks, iris sub-blocks and sclera sub-blocks. The data processing module is used to locate the center of the pupil sub-block coverage area in the current frame grayscale image and obtain the center coordinates and diameter of the pupil sub-block coverage area; The diameter of the pupil sub-block coverage area of ​​each grayscale image frame is extracted in real time and marked with the corresponding timestamp, generating a pupil diameter-time variation curve. Obtain patient contextual information, including emergency rapid screening mode, routine testing mode, anesthesia depth testing mode, and neurological follow-up mode; If the patient is in the routine testing mode, the static features of the pupil diameter-time change curve are extracted. Based on the static features and the pupil diameter-time change curve, the dynamic parameter set of the pupil diameter-time change curve is obtained. The static features and the dynamic parameter set are used as key judgment parameters. If the patient is in the emergency rapid screening mode, the static features of the pupil diameter-time change curve are masked, and the maximum contraction rate and complete recovery time in the dynamic parameter set are extracted as key judgment parameters. In addition, the safety threshold range of key judgment parameters in the anesthesia depth detection mode and the personalized derivative parameters of key judgment parameters in the neurological follow-up mode are generated. The reflex assessment module is used to evaluate key judgment parameters for patients in emergency rapid screening mode, routine testing mode, or anesthesia depth testing mode based on a clinical assessment checklist, and to obtain the reflex level corresponding to the key judgment parameters, including: The key parameters of the pupil diameter-time change curve and the patient population type are used as evaluation indicators. The indicator weights of the evaluation indicators are set. The clinical assessment control table is used as the fuzzy evaluation rule table. The membership matrix of the evaluation indicators for different reflex levels is obtained through fuzzy comprehensive evaluation. The reflex level corresponding to the evaluation indicator is obtained according to the membership matrix and indicator weights. The assessment and early warning module is used to generate standardized diagnosis and treatment suggestions and clinical early warning signals for patients based on reflex levels, and to generate personalized assessment conclusions of patients' neurological functions based on personalized derived parameters derived from key judgment parameters.

2. The objective evaluation system for the dynamic characteristics of pupillary light reflection according to claim 1, characterized in that, The process of obtaining the unambiguous type region and the interference region includes: Set the local region size, divide the current frame grayscale image into several rectangular sub-blocks according to the local region size, and obtain the preliminary type of each rectangular sub-block based on the grayscale mean and grayscale extreme value difference of each rectangular sub-block; Obtain the 3D core features of each rectangular sub-block, obtain grayscale feature benchmarks of different preliminary types, determine whether the 3D core features of each rectangular sub-block match the grayscale feature benchmarks of each preliminary type of rectangular sub-block, if they match, determine the peak type of the rectangular sub-block based on the 3D core features of the rectangular sub-block, and obtain the segmentation threshold of the rectangular sub-block based on the peak type and preliminary type of the rectangular sub-block. Based on the preliminary type and segmentation threshold of each rectangular sub-block, the preliminary type segmentation region of each rectangular sub-block is obtained, the edge pixels in the preliminary type segmentation region are obtained, and the gradient direction variance of the preliminary type segmentation region is obtained based on the gray-level gradient direction of the edge pixels. It is determined whether the gradient direction variance of each preliminary type segmentation region matches the gradient feature benchmark of the region with the same preliminary type. If they match, the preliminary type segmentation region is marked as a region without a distinct type. If they do not match, the preliminary type segmentation region is marked as an interference region.

3. The objective evaluation system for the dynamic characteristics of pupillary light reflection according to claim 2, characterized in that, The process of generating additional safety threshold ranges for key decision parameters in the anesthesia depth detection mode includes: If the patient is in anesthesia depth detection mode, a preset anesthesia depth-pupil parameter safety threshold mapping library is used. This library includes safety thresholds for key judgment parameters corresponding to different anesthesia depths and population types. The dynamic parameter set of pupil diameter and pupil diameter-time change curve is used as key judgment parameters. The patient's current anesthesia depth and population type are input into the anesthesia depth-pupil parameter safety threshold mapping library, which outputs the safety threshold range of key judgment parameters corresponding to the population type and current anesthesia depth. The key judgment parameters are compared with their corresponding safety threshold ranges. If any key judgment parameter exceeds its corresponding safety threshold range, an audible and visual warning is immediately triggered. At the same time, any key judgment parameter is marked as an abnormal parameter, and the abnormal parameter is compared with its safety threshold range to generate the degree of deviation. The abnormal parameter and its degree of deviation are then fed back to the detection terminal.

4. The objective evaluation system for the dynamic characteristics of pupillary light reflection according to claim 3, characterized in that, The process of generating personalized derived parameters for key diagnostic parameters in the neurological follow-up model includes: If the patient is in neurological follow-up mode, the patient's full historical data is extracted from the blockchain node. The full historical data includes the patient's population type, individual key judgment parameter baseline data, previous and previous follow-up test data, individual baseline threshold range, and the weight of neurological function indicators corresponding to the disease type. The static features and dynamic parameter set of the current pupil diameter-time change curve are used as key judgment parameters. The absolute difference and relative change rate between the current key judgment parameter and the individual key judgment parameter baseline data, as well as the absolute difference and relative change rate between the current key judgment parameter and the previous follow-up test data are obtained. Based on the current key judgment parameters and previous follow-up test data, a linear regression algorithm is used to obtain the change trend and change rate of each key judgment parameter. At the same time, the static features and dynamic parameter sets of the pupil diameter-time change curves of the patient's left and right pupils are compared to obtain the bilateral parameter symmetry. The bilateral parameter symmetry is compared with the previous follow-up test data to obtain the asymmetry change coefficient. The current key judgment parameters are compared with the individual baseline threshold range, and key judgment parameters that exceed the individual baseline threshold range are marked as abnormal parameters.

5. The objective evaluation system for the dynamic characteristics of pupillary light reflection according to claim 4, characterized in that, The process of constructing a clinical assessment checklist includes: Using different population types and contextual information as cluster centers, the static features and dynamic parameter sets of several different population types and contextual information in the database are clustered to obtain the static features and dynamic parameter sets of different population types under different contextual information conditions. The distribution type test is performed on each group of dynamic parameters in the static features and dynamic parameter sets of different population types to obtain the distribution type of the static features and each group of dynamic parameters of different population types under different contextual information conditions. Based on the distribution type, the general reference range of the static features and each group of dynamic parameters of different population types under different contextual information conditions is obtained. The clinical classification of the static features and each group of dynamic parameters is obtained. Based on the clinical classification and general reference range of the static features and each group of dynamic parameters, the critical interval of the reflex level of the static features and each group of dynamic parameters is obtained. A clinical assessment control table was constructed based on the static characteristics of different population types under different contextual information conditions and the critical intervals of the reflex levels of each group of dynamic parameters.

6. The objective evaluation system for the dynamic characteristics of pupillary light reflection according to claim 5, characterized in that, The process of generating standardized treatment recommendations and clinical warning signals for patients based on reflex levels, and generating personalized neurological function assessment conclusions for patients based on personalized derived parameters from key judgment parameters, includes: A pre-defined inline clinical knowledge base is provided, which includes standardized treatment suggestions and clinical warning signals corresponding to different reflex levels and population types. If the patient is in the emergency rapid screening mode, routine testing mode, or anesthesia depth testing mode, the patient's population type and reflex level are input into the inline clinical knowledge base, and the standardized treatment suggestions and clinical warning signals for the patient are output. If the patient is in neurological follow-up mode, and the preset neurological function grading and quantitative judgment rules are used, then the patient's personalized neurological function assessment conclusion will be generated based on the neurological function grading and quantitative judgment rules, the absolute difference, relative rate of change, trend of change, rate of change, asymmetric change coefficient, and whether there are abnormal parameters.

Images