An AI-based automatic recording system and method for HESS screen

Abstract

The present invention discloses an automatic recording system and method for HESS screen based on artificial intelligence, which relates to the technical field of medical device HESS screen. The system includes: a data acquisition module, a data processing module, an offset analysis module, a clustering analysis module, a visualization generation module, and a result analysis module. The present invention uses a density clustering algorithm to analyze the offset data of multiple examinations, automatically identify the offset patterns of fixed strabismus and dynamic strabismus, and through clustering processing of the offset vector data between the green light spot and the red light-emitting point, it can effectively identify the stability problems of the patient's eyes, distinguish stable strabismus (fixed strabismus) and gradually changing dynamic strabismus. The clustering algorithm not only simplifies the analysis steps of doctors in multiple preliminary examination data but also significantly improves the accuracy and efficiency of data analysis, provides accurate diagnostic basis for the eye function of doctors, and provides more sensitive diagnostic support for the detection of early and mild strabismus.

Description

Technical Field

[0001] The present invention relates to the technical field of medical device HESS screen, and particularly to an automatic recording system and method for HESS screen based on artificial intelligence. Background Art

[0002] HESS screen examination is an ophthalmic examination mainly used to identify the type of strabismus in patients with strabismus, estimate the strabismus angle, monitor the changes in the condition of strabismus patients, find the paralytic eye, the extraocular muscles with insufficient or excessive function, and provide guidance for selecting the surgical muscles for strabismus correction by eye muscle surgery. The HESS screen is made according to the principle of visual projection. During the examination, the examinee sits upright in the middle position about 500 millimeters in front of the HESS screen, with both eyes at the same height as the central fixation point. The examinee wears a pair of filter glasses with one side red and the other side green. The eye wearing the red lens can only see the red visual target on the HESS screen in front, and the green light spot projected onto the screen by the green light spot projector can only be seen by the eye wearing the green lens, thus realizing binocular vision separation. The doctor sequentially lights up the red visual targets on the HESS screen, and the examinee holds the green light spot projector to project a green light spot to track the red light-emitting visual target on the screen and make the two overlap. The testee without a conscious strabismus angle can overlap the green light spot projected onto the HESS screen with the red light-emitting point, while the green light spot projected onto the HESS screen by the testee with a conscious strabismus angle has a certain positional deviation from the red light-emitting point, and the positional deviation reflects the size of the strabismus angle. The positions of the green light spots corresponding to each red visual target are sequentially marked in the HESS screen grid diagram on the recording paper and drawn into a cross grid, that is, a HESS screen examination report is generated.

[0003] In the prior art, the HESS screen examination mainly relies on manual recording or timing acquisition equipment to record and manage the examination data. However, the existing data recording methods are easily affected by manual operations, and the data during the examination is easily omitted or errors occur. When real-time and accurate analysis of a large number of test points is required, the manual input method is difficult to meet the high-precision requirements. During the examination, strabismus patients often unconsciously show compensatory head positions such as facial rotation, head tilt, mandibular elevation or adduction. The changes in head position and eye posture often affect the examination results, but the prior art often lacks real-time monitoring of these dynamic factors, resulting in inaccurate data and affecting the reliability of strabismus angle analysis.

[0004] The prior art still relies on manual operations to judge the abnormal state and deviation of the device. For the deviation monitoring of the device, traditional HESS systems often perform manual analysis only after inspection, making it difficult to detect abnormal states in a timely manner or accurately evaluate the functional state of the extraocular muscles, and unable to effectively assist doctors in judging the functional health status of the patient's extraocular muscles. When the prior art conducts multiple inspections, it is often difficult to scientifically analyze the deviation patterns of the multiple inspection data, such as identifying fixed (stable) strabismus or dynamic strabismus patterns, resulting in a lack of in-depth support for the diagnosis of the patient's ocular function and prone to misjudgment or missed judgment.

[0005] Existing HESS screen inspection technologies also lack intelligent error correction means. Usually, manual judgment is used to determine the possible deviations during the test process and manual adjustment is made. There is a lack of an efficient clustering analysis algorithm for the deviation situations in multiple inspections to judge common deviation patterns, and there is also a lack of adaptive learning ability based on historical data. Therefore, traditional HESS inspection systems have significant defects in terms of data accuracy, automation, and intelligence, and it is difficult to meet the increasingly complex ophthalmic diagnosis requirements. Summary of the Invention

[0006] An object of the present invention is to provide an AI-based automatic recording system and method for HESS screen, which provides more sensitive diagnostic support for the early and mild strabismus as well as the qualitative and quantitative detection of strabismus.

[0007] To achieve the above object, the present invention provides an AI-based automatic recording system for HESS screen, including: a data acquisition module, a data processing module, a deviation analysis module, a clustering analysis module, a visualization generation module, and a result analysis module;

[0008] The data acquisition module is used to collect the position data on the HESS screen, the green light spot projector, and the filter glasses in real time;

[0009] The data processing module is used to process the position data to obtain a preliminary inspection data set;

[0010] The deviation analysis module is used to judge whether the inspection data in the preliminary inspection data set shows deviation;

[0011] The clustering analysis module is used to perform clustering analysis on the inspection data with deviation, identify the deviation pattern and generate a deviation pattern data set;

[0012] The visualization generation module is used to generate a real-time updated HESS grid map, and at the same time map the deviation pattern data set to the HESS grid map and generate error analysis data;

[0013] The result analysis module is used to generate an inspection result report based on the preliminary inspection data, the deviation pattern data set, and the error analysis data.

[0014] Preferably, the data acquisition module is connected to the HESS screen, the green light spot projector, and the filter glasses by wired or wireless means;

[0015] The data acquisition module respectively acquires the position data from the HESS screen, the green light spot projector, and the filter glasses, including: the position of the red light-emitting point, the projection position of the green light spot, and the head position coordinates and eye attitude parameters of the tested person, and constructs an initial parameter set P:

[0016] ,

[0017] Among them, represents the position of the red light-emitting point; represents the projection position of the green light spot; represents the head position coordinates; , both represent the eye attitude parameters.

[0018] Preferably, the offset analysis module analyzes the overlapping situation of the projection position of the green light spot and the position of the red light-emitting point through an artificial intelligence model. If , it is determined to be completely overlapped;

[0019] When it is detected that the projection position of the green light spot and the position of the red light-emitting point are not completely overlapped, calculate the offset vector and the offset distance :

[0020] ,

[0021] Among them, i represents the i-th detection;

[0022] Take the offset vector and the offset distance as the oblique angle estimation data and store them in the preliminary inspection data set.

[0023] Preferably, the clustering analysis module uses a density-based clustering algorithm to perform clustering analysis on a number of oblique angle estimation data , sets the minimum number of samples MinPts and the neighborhood radius ε, and divides the oblique angle estimation data into several clusters C k , and performs the recognition of the offset mode. The method includes:

[0024] For the offset vector of each data point, calculate its Euclidean distance from other data points, and judge whether it satisfies Whether the number of data points is not less than MinPts; classifying the data points that meet the conditions into the same cluster, where k is the cluster number.

[0025] Preferably, the visualization generation module establishes an initial coordinate system of the HESS grid map according to the projection position of the green light spot and the position of the red light-emitting point defines the standard reference position of each test point, and uses the position of the red light-emitting point as the reference point in the HESS grid map;

[0026] Maps the offset vectors, offset distances, and offset angles of all test points to the constructed HESS grid map in sequence, and sets the color gradient of the overlapping situation based on the offset distance;

[0027] Realtime updates the status of the test points in the HESS grid map to generate error analysis data.

[0028] Preferably, the result analysis module combines the error analysis data set and generates an inspection result report on the extraocular muscle function based on the offset vectors, offset distances, corrected offset values, and head position changes of each test point.

[0029] The present invention also provides an automatic recording method for a HESS screen based on artificial intelligence. The method is applied to the above system, and the steps include:

[0030] Realtime collects the position data on the HESS screen, the green light spot projector, and the filter glasses;

[0031] Processes the position data to obtain a preliminary inspection data set;

[0032] Judges whether the inspection data in the preliminary inspection data set shows an offset;

[0033] Performs cluster analysis on the inspection data with an offset to identify the offset pattern and generate an offset pattern data set;

[0034] Generates a realtime updated HESS grid map, and at the same time maps the offset pattern data set to the HESS grid map and generates error analysis data;

[0035] Generates an inspection result report based on the preliminary inspection data, the offset pattern data set, and the error analysis data.

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0037] The present invention uses a density clustering algorithm to analyze the offset data of multiple examinations, automatically identify the offset patterns of fixed strabismus and dynamic strabismus. By clustering the offset vector data between the green light spot and the red light-emitting point, it can effectively identify the stability problems of the patient's eyes, distinguish stable strabismus (fixed strabismus) from gradually changing dynamic strabismus. The clustering algorithm not only simplifies the analysis steps of doctors in multiple preliminary examination data but also significantly improves the accuracy and efficiency of data analysis, provides accurate diagnostic basis for the eye function of doctors, and provides more sensitive diagnostic support for the detection of early and mild strabismus.

[0038] The present invention combines an adaptive centroid correction algorithm to automatically identify and correct the error of the green light spot position. By calculating the centroid of the green light spot positions of multiple samplings, the present invention can dynamically adjust the coordinates of the green light spot when there are slight offsets or sampling noises, eliminate the errors and ensure the accuracy of the coordinates. It not only reduces the risk of error accumulation but also effectively improves the coordinate accuracy through the adaptive learning algorithm, enabling the system to maintain high data consistency when there are slight changes in the head position and eye posture, and ensuring the reliability of the examination results.

[0039] The present invention uses a dynamically real-time updated HESS grid map generation algorithm to dynamically map the offset vectors, offset distances, and strabismus angles of all test points to the grid map, and clearly presents the overlapping situation and offset values through color gradients. Traditional HESS systems often use a fixed graphical display method and are difficult to reflect real-time dynamic changes. The grid map generation algorithm of the present invention can not only update the coordinates and offset states of each test point in real time but also automatically adjust the various values in the grid map according to the changes in the patient's head position and eye posture. Brief Description of the Drawings

[0040] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 It is a schematic diagram of the system structure of an embodiment of the present invention;

[0042] Figure 2 It is a schematic diagram of the working process of the visualization generation module of an embodiment of the present invention. Detailed Embodiments

[0043] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0044] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0045] Embodiment 1

[0046] As Figure 1 shown, it is a schematic diagram of the system structure of this embodiment, including: a data acquisition module, a data processing module, an offset analysis module, a clustering analysis module, a visualization generation module, and a result analysis module; the data acquisition module is used to collect position data on the HESS screen, the green light spot projector, and the filter glasses in real time; the data processing module is used to process the position data to obtain a preliminary inspection data set; the offset analysis module is used to determine whether the inspection data in the preliminary inspection data set has an offset; the clustering analysis module is used to perform clustering analysis on the inspection data with an offset, identify the offset pattern, and generate an offset pattern data set; the visualization generation module is used to generate a real-time updated HESS grid map, and at the same time map the offset pattern data set to the HESS grid map and generate error analysis data; the result analysis module is used to generate an inspection result report based on the preliminary inspection data, the offset pattern data set, and the error analysis data.

[0047] Next, in conjunction with this embodiment, it will be described in detail how the present invention solves the technical problems in actual work.

[0048] In this embodiment, the acquisition module is connected to the HESS screen, the green light spot projector, and the filter glasses by wired or wireless means. The data acquisition module respectively collects position data from the HESS screen, the green light spot projector, and the filter glasses. The position data includes: the position of the red light-emitting point, the projection position of the green light spot, and the head position coordinates and eye posture parameters of the tested person.

[0049] The data acquisition module captures and collects the position of the red light-emitting point from the HESS screen in real time , where x r represents the position of the red light-emitting point on the horizontal axis, and y r represents the position of the red light-emitting point on the vertical axis;

[0050] The green light spot projector projects and collects the projection position of the green light spot on the HESS screen in real time , where x g represents the position of the green light spot on the horizontal axis, and yg Indicates the position of the green light spot on the vertical axis;

[0051] The head position coordinates and eye posture parameters of the tested person are detected in real time by the sensors in the data acquisition module and , where x h , y h and z h respectively represent the positions of the head in the horizontal, vertical, and depth directions, represents the horizontal rotation angle of the eyeball, represents the vertical rotation angle of the eyeball.

[0052] Construct the initial parameter set P through the above data:

[0053] ,

[0054] where, represents the position of the red light-emitting point; represents the projection position of the green light spot; represents the head position coordinates; , both represent the eye posture parameters.

[0055] After that, the data processing module performs noise filtering, edge detection, and corner extraction on the collected initial parameter set, obtains the processed data, and stores the valid data in a structured manner in the database to form a preliminary inspection data set.

[0056] Specifically, the steps of noise filtering include: performing moving window median filtering and Kalman dynamic prediction filtering on the collected red and green light spot position signals; optimizing modal-level features, performing head position coordinate filtering and eye posture parameter filtering; cross-modal joint verification, including physiological motion coupling verification and light spot position consistency verification; adaptive parameter optimization. Noise filtering reduces the positioning error of the red and green light spots to ±0.5 pixels, the suppression rate of head coordinate drift exceeds 90%, and more than 95% of the eye posture transient noise is effectively eliminated, realizing high-precision denoising of the red and green light spot positions, head coordinates, and eye posture parameters.

[0057] Apply an edge detection algorithm to process the images of the head position and eye posture, and identify the edge coordinates of the HESS matrix , where represents the position of the edge point on the horizontal axis, represents the position of the edge point on the vertical axis, and i is the serial number of the edge point.

[0058] Apply a corner extraction algorithm to perform feature recognition on the processed image, and extract the corner coordinates of the HESS matrix , where Indicates the position of the corner point on the horizontal axis, Indicates the position of the corner point on the vertical axis, where j is the serial number of the corner point.

[0059] The offset analysis module analyzes the overlapping situation of the projection position of the green light spot and the position of the red light-emitting point through an artificial intelligence model. If , it is determined to be completely overlapped and stored in the preliminary inspection data set. When it is detected that the projection position of the green light spot and the position of the red light-emitting point are not completely overlapped, the offset vector and the offset distance are calculated as follows:

[0060] ,

[0061] where i represents the i-th detection;

[0062] The offset vector and the offset distance are used as the oblique angle estimation data and stored in the preliminary inspection data set.

[0063] Meanwhile, the offset analysis module performs error analysis based on the offset values recorded in the preliminary inspection data set to determine the inspection points with large offset errors. The offset error threshold δ is a preset error limit. When , it is determined that there is a significant error at this inspection point.

[0064] The offset analysis module can also automatically trigger a retest for the inspection points with large errors, re-collect the projection position of the green light spot and the position of the red light-emitting point at this inspection point, and recalculate the offset vector and the offset distance as follows:

[0065] ,

[0066] .

[0067] Meanwhile, the functions of the offset analysis module also include correcting the accurate coordinates of the green light spot using the centroid calculation algorithm based on the error analysis. Suppose the current acquisition point set of the green light spot is , then the corrected accurate coordinates of the green light spot are expressed as:

[0068] ,

[0069] ,

[0070] where n is the number of samplings in the retest.

[0071] Update the accurate coordinates of the corrected green light spot to the preliminary inspection dataset and record the offset distances before and after the offset error adjustment and the corrected offset value .

[0072] The clustering analysis module uses a density-based clustering algorithm to perform clustering analysis on a number of skew angle estimation data , sets the minimum number of samples MinPts and the neighborhood radius ε, and divides the skew angle estimation data into several clusters C k , and performs the identification of the offset pattern. The methods include:

[0073] For the offset vector of each data point , calculate its Euclidean distance from other data points, and judge whether the number of data points satisfying is not less than MinPts; classify the data points that meet the conditions into the same cluster, where k is the number of the cluster.

[0074] After that, calculate the central offset vector k and the offset variance of each cluster C :

[0075] ,

[0076] where is the number of data points in cluster C k .

[0077] Store the central offset vector and the offset variance of each cluster into the offset pattern dataset, and associate it with the preliminary inspection dataset to generate a data report containing offset vectors, offset distances, and offset patterns for identifying the characteristics of different types of strabismus:

[0078] If the offset variance k of a certain cluster C is small and the central offset vector remains stable, it indicates the presence of fixed strabismus characteristics; if there are significant differences in the central offset vectors between clusters and the offset variance is large, it indicates the presence of dynamic strabismus characteristics.

[0079] As Figure 2 shown, it is the workflow of the visualization generation module, including: according to the positions of the red light-emitting points and the green light spot positions Establish the initial coordinate system of the HESS grid map, define the standard reference position of each test point, and use the position of the red light-emitting point as the reference point in the grid map.

[0080] Draw the overlapping situation in the HESS grid map according to the offset vector between the green light spot position and the red light-emitting point position of each test point. For the i-th test point, calculate the offset distance and the offset angle :

[0081] ,

[0082] ,

[0083] where arctan represents the offset angle, and the angle offset is used for the normalization of adjusting the direction, is the sign function, which is used to adjust the positive and negative values of the angle in the calculation of the offset angle so that the direction calculation can correctly identify in the four quadrants.

[0084] Map the offset vectors , offset distances and offset angles of all test points to the HESS grid map in sequence to display the relative positions and offset situations of each test point. The coordinates and relative overlapping situations of each test point are shown in the HESS grid map, enabling users to observe the overlapping accuracy on the graph.

[0085] Color-mark the HESS grid map and set the color gradient of the overlapping situation based on the offset distance : When is relatively small, display the test points in green; when is relatively large, display the test points in red to generate a visual prompt for the overlapping situation;

[0086] Update the status of the test points in the HESS grid map in real time. When the inspection data changes, the system automatically recalculates the relative position of the green light spot and the red light-emitting point and reflects the latest offset values and overlapping situations in the grid map.

[0087] The final result analysis module directly calls the green light spot position , red light-emitting point position and offset vector and offset distance from the preliminary inspection data set and the grid map to display the overlapping situations of each test point, and update the relative positions and offset situations of each test point in combination with the HESS grid map; at the same time, it also calls the head position coordinates and the initial head position , calculate the movement amount of the head :

[0088] ,

[0089] Among them, represents the displacement of the head at the i-th test point, which is used to reflect the influence of the head on the inspection data during the inspection process.

[0090] Based on the offset vector , the head position change and the offset angle in the preliminary inspection dataset , show the offset direction angle between the green light spot and the red light-emitting point. The offset angle represents the offset direction of the green light spot relative to the red light-emitting point.

[0091] Finally, combined with the error analysis dataset, based on the offset vector at each test point, the offset distance , the corrected offset value and the head position change , the system automatically generates a report on the extraocular muscle function inspection results, analyzes the potential functional abnormalities of the extraocular muscles, and displays the corrected strabismus angle estimation and offset error at each test point in the data visualization interface.

[0092] The present invention also discloses a surgical advice generation module, which is used to generate inspection results in the form of data according to the offset data, offset angle, and head position change in the extraocular muscle function analysis report, and the doctor marks the advice for strabismus correction in the inspection report.

[0093] Embodiment 2

[0094] This embodiment also provides an AI-based automatic recording method for the HESS screen. The steps include:

[0095] S1. Real-time collect the position data on the HESS screen, the green light spot projector, and the filter glasses.

[0096] Collect the position data from the HESS screen, the green light spot projector, and the filter glasses respectively. The position data includes: the position of the red light-emitting point, the projection position of the green light spot, and the head position coordinates and eye posture parameters of the tested person.

[0097] Specifically, capture and collect the position of the red light-emitting point from the HESS screen in real time , where x r represents the position of the red light-emitting point on the horizontal axis, and y r represents the position of the red light-emitting point on the vertical axis;

[0098] The green light spot projector projects onto the HESS screen in real time and the projection position of the green light spot is collected , where x g represents the position of the green light spot on the horizontal axis, and y g represents the position of the green light spot on the vertical axis;

[0099] The head position coordinates and eye posture parameters of the subject are detected in real time by the sensors in the data acquisition module and , where x h , y h and z h respectively represent the positions of the head in the horizontal, vertical and depth directions, represents the horizontal rotation angle of the eye, represents the vertical rotation angle of the eye.

[0100] Construct the initial parameter set P from the above data:

[0101] ,

[0102] where, represents the position of the red light-emitting point; represents the projection position of the green light spot; represents the head position coordinates; , both represent the eye posture parameters.

[0103] S2. Process the position data to obtain a preliminary inspection data set.

[0104] Perform noise filtering, edge detection and corner extraction on the collected initial parameter set, obtain the processed data and store the valid data in a structured manner in the database to form a preliminary inspection data set.

[0105] Specifically, the steps of noise filtering include: performing moving window median filtering and Kalman dynamic prediction filtering on the collected red and green light spot position signals; optimizing the modal-level features, performing head position coordinate filtering and eye posture parameter filtering; cross-modal joint verification, including physiological motion coupling verification and light spot position consistency verification; adaptive parameter optimization. The noise filtering reduces the positioning error of the red and green light spots to ±0.5 pixels, the head coordinate drift suppression rate exceeds 90%, and more than 95% of the eye posture transient noise is effectively eliminated, realizing high-precision denoising of the red and green light spot positions, head coordinates and eye posture parameters.

[0106] Specifically, apply the edge detection algorithm to process the images of the head position and eye posture, and identify the edge coordinates of the HESS matrix , where represents the position of the edge point on the horizontal axis, Indicates the position of the edge point on the vertical axis, where i is the serial number of the edge point.

[0107] Apply the corner extraction algorithm to perform feature recognition on the processed image and extract the corner coordinates of the HESS matrix , where Indicates the position of the corner on the horizontal axis, Indicates the position of the corner on the vertical axis, and j is the serial number of the corner.

[0108] S3. Determine whether the inspection data in the preliminary inspection dataset has shifted.

[0109] Through the artificial intelligence model, analyze the overlapping situation of the projection position of the green light spot and the position of the red light-emitting point . If , it is determined to be completely overlapped and stored in the preliminary inspection dataset. When it is detected that the projection position of the green light spot and the position of the red light-emitting point are not completely overlapped, calculate the offset vector and the offset distance :

[0110] ,

[0111] where i represents the i-th detection;

[0112] Take the offset vector and the offset distance as the oblique angle estimation data and store it in the preliminary inspection dataset.

[0113] Based on the offset values recorded in the preliminary inspection dataset, perform error analysis to determine the inspection points with large offset errors. The offset error threshold δ is a preset error limit. When , it is determined that there is a significant error at this inspection point.

[0114] Automatically trigger a retest for the inspection points with large errors, and re-collect the projection position of the green light spot and the position of the red light-emitting point at this inspection point, and calculate the offset vector and the offset distance again:

[0115] ,

[0116] .

[0117] Based on the error analysis, use the centroid calculation algorithm to correct the accurate coordinates of the green light spot. Let the current acquisition point set of the green light spot be , then the accurate coordinates of the green light spot after correction Expressed as:

[0118] ,

[0119] ,

[0120] where n is the number of samplings in the retest.

[0121] Update the accurate coordinates of the corrected green light spot to the preliminary inspection dataset and record the offset distances before and after the offset error adjustment and the corrected offset value .

[0122] S4. Perform clustering analysis on the inspection data with offsets, identify the offset patterns, and generate an offset pattern dataset.

[0123] Use a density-based clustering algorithm to perform clustering analysis on a number of oblique angle estimation data Set the minimum number of samples MinPts and the neighborhood radius ε, and divide the oblique angle estimation data into several clusters C k , and perform the identification of the offset pattern. The methods include:[[]]

[0124] For the offset vector of each data point , calculate its Euclidean distance from other data points, and determine whether the number of data points satisfying is not less than MinPts; classify the data points that meet the conditions into the same cluster, where k is the cluster number.

[0125] After that, calculate the central offset vector k and the offset variance of each cluster C :[[]]

[0126] ,

[0127] where is the number of data points in cluster C k .

[0128] Store the central offset vector and the offset variance of each cluster into the offset pattern dataset, and associate it with the preliminary inspection dataset to generate a data report containing offset vectors, offset distances, and offset patterns for identifying the characteristics of different types of strabismus:[[]]

[0129] If the offset variance k of a certain cluster C is small and the central offset vector Remain stable, indicating the presence of fixed strabismus features; if the inter-cluster center offset vector has significant differences and a large offset variance, it indicates the presence of dynamic strabismus features.

[0130] S5. Generate a real-time updated HESS grid map, and at the same time map the offset pattern dataset to the HESS grid map and generate error analysis data.

[0131] According to the positions of the red luminous points in the preliminary inspection dataset and the positions of the green light spots Establish the initial coordinate system of the HESS grid map, define the standard reference position of each test point, and use the position of the red luminous point as the reference point in the grid map.

[0132] Draw the overlapping situation in the HESS grid map according to the offset vectors between the green light spot positions and the red luminous point positions of each test point. For the i-th test point, calculate the offset distance and the offset angle :

[0133] ,

[0134] ,

[0135] where arctan represents the offset angle, and the angular offset is used for the normalization of adjusting the direction, is the sign function, which is used to adjust the positive and negative values of the angle in the calculation of the offset angle so that the direction calculation can be correctly identified in the four quadrants.

[0136] Map the offset vectors , offset distances and offset angles of all test points to the HESS grid map in sequence to display the relative positions and offset situations of each test point. The coordinates and relative overlapping situations of each test point are shown in the HESS grid map, enabling users to observe the overlapping accuracy on the graph.

[0137] Color-mark the HESS grid map, and set the color gradient of the overlapping situation based on the offset distance : When is small, display the test points in green; when is large, display the test points in red to generate a visual overlapping situation prompt;

[0138] Real-time update the status of the test points on the HESS grid map. When the inspection data changes, the system automatically recalculates the relative positions of the green light spot and the red luminous point and reflects the latest offset values and overlapping situations in the grid map.

[0139] S6. Generate an inspection result report based on the preliminary inspection data, offset pattern dataset, and error analysis data.

[0140] Directly call the green light spot position from the preliminary inspection dataset and the grid map , the red light-emitting point position and the offset vector and the offset distance , display the overlap of each test point, and update the relative position and offset of each test point in combination with the HESS grid map; at the same time, also call the head position coordinates and the initial head position from the preliminary inspection dataset, and calculate the movement amount of the head :

[0141] ,

[0142] wherein, represents the displacement of the head at the i-th test point, which is used to reflect the influence of the head on the inspection data during the inspection process.

[0143] Based on the offset vector , the head position change and the offset angle in the preliminary inspection dataset , display the offset direction angle between the green light spot and the red light-emitting point, and the offset angle represents the offset direction of the green light spot relative to the red light-emitting point.

[0144] Finally, in combination with the error analysis dataset, based on the offset vector , the offset distance , the corrected offset value and the head position change , the system automatically generates an extraocular muscle function inspection result report, analyzes the potential functional abnormalities of the extraocular muscles, and displays the corrected strabismus angle estimation and offset error of each test point in the data visualization interface.

[0145] The present invention also discloses a surgical recommendation generation module, which is used to generate inspection results in the form of data based on the offset data, offset angle, and head position change in the extraocular muscle function analysis report, and the doctor indicates the recommendations for strabismus correction in the inspection report.

[0146] Example Three

[0147] At 14:30 on the afternoon of August 12, 2023, a patient named Ms. Li, 34 years old, was admitted to an ophthalmic hospital in a certain city. The main symptom was double vision. To further evaluate the extraocular muscle function, the doctor needed to perform a HESS screen examination and decided to use the HESS screen automatic recording system of the present invention to assist in the examination.

[0148] At 14:35, in the examination room, Ms. Li put on the filter glasses and sat in front of the HESS screen, with her head placed on the instrument's chin rest. The hospital's detection team used this system for the examination. The data acquisition module of the system was first initialized and connected to the HESS screen, the green light spot projector, and the filter glasses. During the detection, Ms. Li was required to stare at the red light-emitting point generated by the system on the HESS screen, and the system generated a green light spot at each test point to overlap with it to evaluate the visual alignment.

[0149] At 14:37, the system automatically started to collect data of the test points. The sensor recorded that the head position coordinates of Ms. Li were (320, 180, 500) millimeters, and the attitude angles of the eyeballs were 15° and 10° respectively. The position of the red light-emitting point of the first test point collected by the system was (100, 150) pixels, and the position of the green light spot projected by the green light spot projector was (105, 153) pixels.

[0150] At 14:38, during the first-round data processing, the system detected that the offset distance between the green light spot and the red light-emitting point was [X] pixels, exceeding the set tolerance threshold of 3.0 pixels. The system triggered the offset analysis module and adjusted the light spot position through the centroid calculation algorithm to reduce the offset value to [Y] pixels. During this process, Ms. Li maintained a relatively static head posture, and the data acquisition module continuously recorded multiple samples to ensure that the corrected offset value reached an acceptable range.

[0151] At 14:48, the clustering analysis module took over the data analysis work and clustered the offset data of Ms. Li at nine test points. The system detected that the offset of the first seven test points was stable at about 3 pixels, while the offset of the eighth and ninth test points exceeded 5 pixels, showing dynamic change characteristics. According to the results of the density clustering algorithm, the system automatically identified that the patient's right eye had significant dynamic strabismus characteristics and marked them in the analysis report.

[0152] At 14:49, the visualization module of the system generated a real-time updated HESS grid map, creating a tic-tac-toe grid. The color gradient clearly showed the offset and the change in the area of the tic-tac-toe grid. At this time, the grid map showed that the area of the tic-tac-toe grid in Ms. Li's left eye visual field was reduced, significantly smaller than that of the right eye, and it shifted upward and adducted inward below. The test points in the left eye showed a red color, indicating a large offset, and the reduction in the area of the tic-tac-toe grid was shown as dark blue. The test points above the tic-tac-toe grid in the right eye visual field showed a green color, indicating a small offset; the remaining test points showed a red color, indicating a large offset, and it shifted downward, with the area of the tic-tac-toe grid expanding and showing a light pink color. The doctor observed on the screen that the tic-tac-toe grid in the left eye visual field was dark blue and that in the right eye visual field was light pink, and judged it to be non-comitant strabismus, with the left eye being the paretic muscle. The left eye visual field adducted inward below, and it was judged that the function of the left inferior rectus muscle was insufficient.

[0153] The result analysis module generated an extraocular muscle function analysis report based on the data of the offset vector, head position, and eye posture. The surgical recommendation module automatically generated a surgical recommendation plan according to the offset pattern and the abnormal parameters of the extraocular muscle function, and recommended correcting the left inferior rectus muscle to improve the symptoms.

[0154] During the detection process, this system automatically recorded and processed the data of each test point of Ms. Li, generated a clear and intuitive HESS grid map, and showed the overlapping situation and the change in the offset value of each test point in real time, intuitively showing the offset and the change in the area of the tic-tac-toe grid in the visual field. In the traditional method, due to relying on manual operations and fixed acquisition devices, the offset data could not be corrected in real time, often resulting in inaccurate detection data. Through the system of the present invention, the doctor does not need to manually adjust the overlapping situation of each test point, and the system has automatically corrected and recorded it. In addition, the surgical recommendations generated by this system, based on the offset pattern and extraocular muscle data after cluster analysis, have greatly improved the scientific nature of the diagnosis.

[0155] The embodiments described above are only descriptions of the preferred embodiments of the present invention and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. An AI-based automatic recording system for HESS screens, characterized in that, Including: A data acquisition module, a data processing module, an offset analysis module, a clustering analysis module, a visualization generation module, and a result analysis module; The data acquisition module is used to collect the position data on the HESS screen, the green light spot projector, and the filter glasses in real time; The data processing module is used to process the position data to obtain a preliminary inspection data set; The offset analysis module is used to determine whether the inspection data in the preliminary inspection data set shows an offset; The clustering analysis module is used to perform clustering analysis on the inspection data with offsets, identify offset patterns and generate an offset pattern data set; the clustering analysis module uses a density-based clustering algorithm to perform clustering analysis on a number of oblique viewing angle estimation data to perform clustering analysis, set the minimum number of samples MinPts and the neighborhood radius ε, and divide the oblique viewing angle estimation data into several clusters C k , and identify the offset pattern. The method includes: Offset vector for each data point , calculate its Euclidean distance from other data points, and determine whether there are at least MinPts data points that meet the neighborhood radius ε within the neighborhood of each data point; classify the data points that meet the conditions into the same cluster, where k is the cluster number; After that, calculate the center offset vector of each cluster C k and the offset variance as follows: : ， Among them, is the number of data points in cluster C k ; The center offset vector of each cluster and the offset variance are stored in the offset pattern data set and associated with the preliminary inspection data set to generate a data report containing the offset vector, offset distance, and offset pattern for identifying the characteristics of different types of strabismus: If a certain cluster C k has an offset variance less than a preset value and the center offset vector remains stable, it indicates the existence of fixed strabismus features; if the center offset vectors between clusters are different and the offset variance is greater than the preset value, it indicates the existence of dynamic strabismus features; The visualization generation module is used to generate a real-time updated HESS grid map, and at the same time map the offset pattern data set to the HESS grid map and generate error analysis data; The result analysis module is used to generate an inspection result report based on the preliminary inspection data, the offset pattern data set, and the error analysis data.

2. The AI-based HESS screen automatic recording system according to claim 1, wherein The data acquisition module is connected to the HESS screen, the green light spot projector, and the filter glasses by wired or wireless means; The data acquisition module collects the position data from the HESS screen, the green light spot projector, and the filter glasses respectively, including: the position of the red light-emitting point, the projection position of the green light spot, and the head position coordinates and eye movement parameters of the tested person, and constructs an initial parameter set P: ， Among them, represents the position of the red light-emitting point; represents the projection position of the green light spot; represents the head position coordinates; and both represent the eyeball attitude parameters.

3. The AI-based HESS screen automatic recording system according to claim 2, wherein The offset analysis module analyzes the overlapping situation of the projection position of the green light spot and the position of the red light-emitting point through an artificial intelligence model. If , it is determined to be completely overlapped; When it is detected that the projection position of the green light spot does not completely overlap with the position of the red light-emitting point, calculate the offset vector between the two and the offset distance : ， Where i represents the i-th detection; The offset vector and the offset distance are used as the oblique angle estimation data and stored in the preliminary inspection data set.

4. The AI-based HESS screen automatic recording system according to claim 1, characterized in that The visualization generation module establishes an initial coordinate system of the HESS grid map according to the projection position of the green light spot and the position of the red light-emitting point defines the standard reference position of each test point, and uses the position of the red light-emitting point as the reference point in the HESS grid map; Map the offset vectors, offset distances, and offset angles of all test points to the constructed HESS grid map in sequence, and set the color gradient of the overlapping situation based on the offset distance; Real-time update the test point status of the HESS grid map to generate error analysis data.

5. The AI-based HESS screen automatic recording system according to claim 4, characterized in that, The result analysis module combines the error analysis data set and generates an inspection result report on the extraocular muscle function based on the offset vectors, offset distances, corrected offset values, and head position changes of each test point.

6. An automatic recording method for the HESS screen based on artificial intelligence, the method being applied to the system according to any one of claims 1-5, characterized in that the steps Including: Collect the position data on the HESS screen, the green light spot projector, and the filter glasses in real time; Process the position data to obtain a preliminary inspection data set; Determine whether the inspection data in the preliminary inspection data set shows an offset; Perform clustering analysis on the inspection data with an offset to identify the offset pattern and generate an offset pattern data set; Generate a real-time updated HESS grid map, and at the same time map the offset pattern data set to the HESS grid map and generate error analysis data; Generate an inspection result report based on the preliminary inspection data, the offset pattern data set, and the error analysis data.

Images