Parkinson's disease early diagnosis system based on eyeball movement characteristic analysis

Abstract

The invention discloses a Parkinson's disease early diagnosis system based on eyeball movement characteristic analysis, and relates to the technical field of Parkinson's disease diagnosis. In order to solve the problems that although eye and face movement parameters possibly change in Parkinson's disease, the change of the parameters also possibly occurs in other neurodegenerative diseases or normal aging processes, and high-precision eye and face movement capture data is complex, so that early diagnosis characteristics of Parkinson's disease cannot be accurately captured, and the Parkinson's disease cannot be accurately diagnosed. And the accuracy and efficiency of diagnosis are influenced. By accurately capturing and analyzing eyeball movement data and facial expression video images of a patient, the Parkinson's disease can be effectively distinguished from other diseases possibly causing similar symptoms, misdiagnosis and missed diagnosis are reduced, the standardization and comparability of data are ensured by combining pupil distance information and diopter information, the consistency and reliability of diagnosis are improved, and the diagnosis accuracy is improved. Therefore, the patient can be diagnosed at the early stage of the disease and is treated as soon as possible, and the treatment effect and life quality are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of Parkinson's disease diagnosis, and in particular to an early diagnosis system for Parkinson's disease based on eye movement feature analysis. Background Art

[0002] Traditional Parkinson's disease diagnosis methods mostly rely on clinical symptom observation and subjective judgment, lack of quantitative indicators, and are not conducive to early identification and evaluation of treatment effects. Regarding the early diagnosis of Parkinson's disease, the patent application with publication number: CN115251901A discloses an early diagnosis method for Parkinson's disease based on eye-face motion capture technology, including: 1. Collecting the eye-face movement parameters of the test subject, including the following parameters: (1) the frequency, speed, and amplitude of horizontal eye saccades, vertical saccades, and blinks; (2) the amplitude, speed, and frequency of contraction of the frontal muscle and cheek muscles; (3) the frequency, speed, and amplitude of contraction of the corners of the mouth and lips; (4) the amplitude and frequency of head movement; (5) the duration of facial muscles from static-dynamic-recovery to static.

[0003] Although the above patent can realize the diagnosis of patients with Parkinson's syndrome, although the eye and facial movement parameters may change in Parkinson's disease, the changes in these parameters may also occur in other neurodegenerative diseases or normal aging. In addition, the high-precision eye and facial motion capture data is complex and often cannot accurately capture the characteristics of early diagnosis of Parkinson's disease, affecting the accuracy and efficiency of diagnosis. Summary of the invention

[0004] The purpose of the present invention is to provide an early diagnosis system for Parkinson's disease based on eye movement feature analysis. By combining facial expression analysis, the system can more comprehensively evaluate the patient's symptoms and improve the accuracy of diagnosis. It can also combine pupil distance information and refractive power information to explain abnormalities in eye movement, thereby improving the consistency and reliability of diagnosis and solving the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: The early diagnosis system of Parkinson's disease based on eye movement feature analysis includes: A data acquisition unit, used to capture the patient's eye movement data in real time based on an eye tracker, and to collect the patient's facial expression and eye movement video data in real time based on a camera on the eye tracker, and to process the acquired eye movement data and facial expression and eye movement video data; A feature extraction unit, used for extracting eye movement features related to Parkinson's disease from the processed data, including eye movement speed, movement amplitude, eye fixation point stability and eye saccadic movement features; The early diagnosis unit is used to train and model the extracted eye movement features, establish an early diagnosis model for Parkinson's disease, input the extracted patient's eye movement features into the early diagnosis model for Parkinson's disease, obtain the early diagnosis results of Parkinson's disease, and generate an early diagnosis report for Parkinson's disease.

[0006] Furthermore, the data acquisition unit comprises: A motion stimulation module is used to establish a visual stimulation pattern according to the diagnostic requirements of Parkinson's disease, generate dynamic visual stimulation according to the preset visual stimulation pattern, and display the dynamic visual stimulation on the display screen in real time to generate a scanning task; A data acquisition module, used to record the patient's eye movement data in real time based on an eye tracker when the patient performs a scanning task, including the start and end time, speed, amplitude and path of the scanning; An image acquisition module, used to capture video images of the patient's eye movements and the patient's facial expressions during eye movements; The data processing module is used to pre-process the data collected by the data acquisition module and the image acquisition module and convert the data into a unified format.

[0007] Furthermore, the data processing module further includes: Based on the video image of the patient's eyeball movement captured by the image acquisition module, the patient's eyeball image data is extracted, and the distance between the two pupils of the patient and the shape ratio of the two eyeballs are calculated; The eye image data of at least one patient is calculated, and based on the calculation result, the distance between pupils is compared to see whether it is consistent, and whether the variation range of the distance between pupils is abnormal is determined; According to the judgment result, the average value of the calculated results is used as the distance between the two pupils of the patient to obtain the final pupil distance information of the patient; The refractive information of the patient is obtained by calculation according to the shape ratio of the two eyeballs of the patient.

[0008] Furthermore, the feature extraction unit comprises: An eye movement feature extraction module is used to extract the patient's eye movement speed feature and eye movement amplitude feature, and to correct the eye movement speed feature and eye movement amplitude feature based on the patient's final pupil distance information; The eye gaze feature extraction module is used to extract the patient's eye gaze point features and eye saccadic movement features, and evaluate the visual function and motor ability of the eye gaze point features and eye saccadic movement features based on the patient's refractive information.

[0009] Further, the early diagnosis unit includes: A diagnosis model building module is used to build an early diagnosis model for Parkinson's disease, and to train the early diagnosis model for Parkinson's disease based on the output data of the eye movement feature extraction module and the eye gaze feature extraction module; The eye movement abnormality feature extraction module is used to extract the asymmetry features of eye movement based on the visual function and motor ability assessment results, including: asymmetry of saccade latency: comparing the difference in reaction time between the left and right eyes in the saccade task, asymmetry of saccade peak speed: comparing the difference in maximum speed between the left and right eyes during the saccade process, asymmetry of saccade amplitude: and the difference in movement distance between the left and right eyes during the saccade process, and the degree of eye movement abnormality of the patient is assessed based on the asymmetry features; The Parkinson's disease diagnosis module is used to input the extracted eye movement abnormality features into the trained Parkinson's disease early diagnosis model for diagnosis, and output the diagnosis result including the asymmetry index and conclusion in combination with the evaluation result of the degree of eye movement abnormality of the patient; The clinical diagnosis result comparison module is used to compare the diagnosis result output by the Parkinson's disease diagnosis module with the diagnosis result of the clinician, and adjust the parameters of the Parkinson's disease early diagnosis model according to the comparison result.

[0010] Furthermore, the eye movement abnormality feature extraction module is specifically: identifying saccadic events in the eye movement data of the patient captured by an eye tracker based on saccadic eye movement features, and segmenting the continuous eye movement data into individual saccadic events based on the identification result; The latency, peak velocity and amplitude of each saccade event are calculated, and the differences between the left and right eyes are compared. The mean and standard deviation of the saccade latency, peak velocity and amplitude of the left and right eyes are calculated based on the comparison results. The asymmetry index of the latency, peak velocity and amplitude is obtained based on the calculation results. The asymmetry index of latency, peak velocity and amplitude is compared with the corresponding preset asymmetry index threshold to determine whether there is significant asymmetry between the left and right eyes, and to comprehensively evaluate the degree of eye movement abnormality of the patient.

[0011] Furthermore, the Parkinson's disease diagnosis module also includes determining whether the patient has Parkinson's disease based on the video image of the patient's facial expression during eye movement, specifically: Obtaining a plurality of facial expression sample data of the patient based on the video image; Extract the patient's blinking frequency, facial movement change frequency and amplitude changes from each facial expression sample data; Identify the extracted multiple parameters, obtain identification results, and determine whether there are typical facial movement features of Parkinson's disease; The abnormal coefficient of the patient is calculated according to the blinking frequency, facial movement change frequency and amplitude change of the patient in each facial expression sample data and the identification result of each parameter; Confirm whether the abnormal coefficient is greater than a preset coefficient. If so, obtain an abnormal video image segment of the patient based on the video image. Otherwise, preliminarily determine that the patient does not suffer from Parkinson's disease.

[0012] Furthermore, obtaining an abnormal video image segment of a patient based on the video image further includes: Extracting human eye feature factors in each frame of image based on abnormal video image segments; The typical features of Parkinson's disease are screened out based on the human eye characteristic factors in each frame of the image; The severity of the patient's symptoms is calculated based on the typical features of Parkinson's disease in the human eye feature factors in each frame of the image; Confirm whether the severity of the patient's symptoms is greater than or equal to a preset threshold. If so, it is confirmed that the patient suffers from Parkinson's disease. Otherwise, it is considered that the patient does not show typical symptoms of Parkinson's disease.

[0013] Furthermore, the data acquisition unit includes: Compensation modules for: Collect environmental images based on the camera on the eye tracker; Extracting brightness component information of each pixel in the environment image; determining a brightness histogram according to the brightness component information of each pixel in the environment image; determining a brightness interval according to the brightness histogram, determining a middle value of the brightness interval, and counting a first number of pixels greater than the middle value and a second number of pixels less than the middle value; when it is determined that the second number is greater than the first number, generating and executing a lighting compensation instruction;

[0014] in, is the grayscale value of the environment image G at (x, y) after illumination compensation; is the gray value of the environment image G at (x, y); is the target grayscale mean generated based on illumination compensation; The number of pixels included in the environment image; Adjustment modules for: The propagation direction of the laser beam is changed based on the rotation of the eye tracker by a first angle so that the laser light path passes through the face of the patient, and the first angular resolution of the facial expression of the patient is adjusted according to the first angle of rotation of the eye tracker;

[0015] in, is the first angular resolution; is an integer, , is the laser beam divergence angle; is the first angle between the eye tracker and the patient's face; is the maximum angle that the eye tracker can rotate; is the initial angular resolution of the eye tracker; The propagation direction of the laser beam is changed based on the rotation of the eye tracker at a second angle, so that the laser light path passes through the patient's eyeball, and the second angular resolution of the patient's eyeball is adjusted according to the second angle of rotation of the eye tracker;

[0016] in, is the second angular resolution; is the second angle between the eye tracker and the patient's eyeball; The camera on the eye tracker collects the patient's facial expression and eye movement video data based on the first angular resolution and the second angular resolution.

[0017] Furthermore, the image acquisition module includes: The first determination module is used to: Performing frame processing on the video image of the patient's facial expression during eye movement to obtain a number of frame sub-images; Performing grayscale processing on a plurality of frames of sub-images to obtain a plurality of frames of grayscale images; Traverse each pixel point on the grayscale image, add up the grayscale value of each pixel point, divide the total grayscale value by the total number of pixels, and obtain the average grayscale value; in the process of traversing each pixel point on the grayscale image, record the maximum grayscale value and the minimum grayscale value; determine the pixel point whose grayscale value is less than the average grayscale value, and calculate the first grayscale sum value; determine the pixel point whose grayscale value is greater than the average grayscale value, and calculate the second grayscale sum value; Calculate a characteristic value of the grayscale image based on the maximum grayscale value, the minimum grayscale value, the first grayscale sum value, and the second grayscale sum value;

[0018] in, is the eigenvalue of the grayscale image; is the second grayscale value; is the first grayscale value; is the maximum gray value; is the minimum grayscale value; comparing the characteristic value of the grayscale image with the preset characteristic threshold, screening out the grayscale images whose characteristic value is greater than the preset characteristic threshold, and determining them as the target image set; The second determination module is used for: Determine a first target image and a second target image of adjacent frames in a target image set; Acquire a first grayscale feature and a first texture feature of a first target image; acquire a first grayscale feature and a second texture feature of a second target image; Based on the first grayscale feature and the second grayscale feature, performing a first image registration on the first target image and the second target image, and determining first change information of the facial expression according to a result of the first image registration; Based on the first texture feature and the second texture feature, performing a second image registration on the first target image and the second target image, and determining second change information of the facial expression according to a result of the second image registration; The facial expression change information is determined according to the first change information and the second change information.

[0019] Compared with the prior art, the present invention has the following beneficial effects: By accurately capturing and analyzing patients' eye movement data and facial expression video images, it is possible to effectively distinguish Parkinson's disease from other diseases that may cause similar symptoms, reducing misdiagnosis and missed diagnosis. At the same time, combining pupil distance information and refractive power information ensures data standardization and comparability, improves diagnostic consistency and reliability, and enables patients to be diagnosed in the early stages of the disease, thereby receiving treatment as soon as possible, potentially slowing disease progression and improving treatment outcomes and quality of life. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a module diagram of the Parkinson's disease early diagnosis system based on eye movement feature analysis of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] In order to solve the technical problem that although eye and facial movement parameters may change in Parkinson's disease, changes in these parameters may also occur in other neurodegenerative diseases or normal aging, and high-precision eye and facial motion capture data is complex and often cannot accurately capture the characteristics of early diagnosis of Parkinson's disease, affecting the accuracy and efficiency of diagnosis, please refer to Figure 1 , this embodiment provides the following technical solutions: The early diagnosis system of Parkinson's disease based on eye movement feature analysis includes: A data acquisition unit, used to capture the patient's eye movement data in real time based on an eye tracker, and to collect the patient's facial expression and eye movement video data in real time based on a camera on the eye tracker, and to process the acquired eye movement data and facial expression and eye movement video data; The feature extraction unit is used to extract eye movement features related to Parkinson's disease from the processed data, including eye movement speed, movement amplitude, eye fixation point stability and eye saccadic movement features, including: An eye movement feature extraction module is used to extract the patient's eye movement speed feature and eye movement amplitude feature, and to correct the eye movement speed feature and eye movement amplitude feature based on the patient's final pupil distance information; The eye gaze feature extraction module is used to extract the patient's eye gaze point features and eye saccadic movement features, and evaluate the visual function and motor ability of the eye gaze point features and eye saccadic movement features based on the patient's refractive information.

[0023] In this embodiment, the device is calibrated according to the pupil distance information to ensure that the eye movement is measured in the correct coordinate system, and the corrected standardized data is obtained, so that when analyzing the eye movement data, the data of different patients are comparable. The patient's refractive information is used to help explain the abnormalities in the eye movement. For example, myopia or hyperopia may affect the performance of the eye tracking task. The visual function and motor ability assessment are performed to more accurately reflect the patient's actual visual ability. If the patient has significant saccadic asymmetry, combining the refractive information can help determine whether this is due to a visual problem or a motor symptom of Parkinson's disease. The early diagnosis unit is used to train and model the extracted eye movement features, establish an early diagnosis model for Parkinson's disease, input the extracted patient's eye movement features into the early diagnosis model for Parkinson's disease, obtain the early diagnosis results of Parkinson's disease, and generate an early diagnosis report for Parkinson's disease.

[0024] In this embodiment, by comprehensively analyzing the eye movement characteristics, Parkinson's disease can be distinguished from other diseases that may cause similar eye movement abnormalities, thereby reducing misdiagnosis. Early diagnosis allows patients to receive treatment earlier, which may slow the progression of the disease and improve the treatment effect. By calibrating the equipment according to the pupil distance information, the standardization of the data is ensured, making the data of different patients comparable, improving the consistency and reliability of the diagnosis, and combining the patient's refractive information, it is possible to better explain the abnormalities in the eye movements and distinguish whether they are due to visual problems or motor symptoms of Parkinson's disease. The detailed diagnostic report provides doctors with an objective and quantitative basis for diagnosis, which helps doctors to formulate personalized treatment plans based on the specific circumstances of the patients.

[0025] In this embodiment, the data acquisition unit includes: The motion stimulation module is used to establish visual stimulation patterns according to the diagnostic requirements of Parkinson's disease, such as random point displacement, target tracking, etc., generate dynamic visual stimulation according to the preset visual stimulation pattern, and display the dynamic visual stimulation on the display screen in real time to generate a scanning task; A data acquisition module, used to record the patient's eye movement data in real time based on an eye tracker when the patient performs a scanning task, including the start and end time, speed, amplitude and path of the scanning; An image acquisition module, used to capture video images of the patient's eye movements and the patient's facial expressions during eye movements; The data processing module is used to pre-process the data collected by the data acquisition module and the image acquisition module and convert the data into a unified format; The data processing module further includes: Based on the video image of the patient's eyeball movement captured by the image acquisition module, the patient's eyeball image data is extracted, and the distance between the two pupils of the patient and the shape ratio of the two eyeballs are calculated; The eye image data of at least one patient is calculated, and based on the calculation result, the distance between pupils is compared to see whether it is consistent, and whether the variation range of the distance between pupils is abnormal is determined; According to the judgment result, the average value of the calculated results is used as the distance between the two pupils of the patient to obtain the final pupil distance information of the patient; The refractive information of the patient is obtained by calculation according to the shape ratio of the two eyeballs of the patient.

[0026] In this embodiment, the motion excitation module can establish a specific visual stimulation pattern according to the diagnostic requirements of Parkinson's disease, ensuring the pertinence and effectiveness of eye movement data. The data acquisition module can record detailed information of eye movements in real time, including the start and end time, speed, amplitude and path of the saccade, providing a rich data source for subsequent analysis. The image acquisition module not only captures eye movements, but also records facial expressions, providing additional visual information for diagnosis, which is conducive to more comprehensive symptom analysis. The data processing module converts the data into a unified format through preprocessing, which is convenient for subsequent feature extraction and model analysis, and improves the availability of data. Furthermore, by calculating the ratio of the interpupillary distance and the shape of the eyeball, the patient's pupil distance information and refractive power information can be accurately obtained. By comparing the range of changes in the interpupillary distance, abnormal conditions can be detected. For example, inconsistent pupil distance may indicate a neurological disease, which is extremely important for understanding the patient's visual function and the causes of abnormal eye movements. Parkinson's disease can be diagnosed more accurately and misdiagnosis and missed diagnosis can be reduced.

[0027] In this embodiment, the early diagnosis unit includes: A diagnosis model building module is used to build an early diagnosis model for Parkinson's disease, and to train the early diagnosis model for Parkinson's disease based on the output data of the eye movement feature extraction module and the eye gaze feature extraction module; The eye movement abnormality feature extraction module is used to extract the asymmetric features of eye movement based on the visual function and motor ability assessment results, including: the asymmetry of the saccade latency: comparing the reaction time difference between the left and right eyes in the saccade task, the asymmetry of the saccade peak speed: comparing the maximum speed difference between the left and right eyes during the saccade process, the asymmetry of the saccade amplitude: and the difference in the moving distance of the left and right eyes during the saccade process. Based on the asymmetry features, the degree of eye movement abnormality of the patient is assessed, specifically: identifying saccadic events in the eye movement data of the patient captured by an eye tracker based on saccadic eye movement features, and segmenting the continuous eye movement data into individual saccadic events based on the identification result; The latency, peak velocity and amplitude of each saccade event are calculated, and the differences between the left and right eyes are compared. The mean and standard deviation of the saccade latency, peak velocity and amplitude of the left and right eyes are calculated based on the comparison results. The asymmetry index of the latency, peak velocity and amplitude is obtained based on the calculation results. Compare the asymmetry index of latency, peak velocity and amplitude with the corresponding preset asymmetry index threshold to determine whether there is significant asymmetry between the left and right eyes, and comprehensively evaluate the degree of eye movement abnormality of the patient; The Parkinson's disease diagnosis module is used to input the extracted eye movement abnormality features into the trained Parkinson's disease early diagnosis model for diagnosis, and output the diagnosis result including the asymmetry index and conclusion in combination with the evaluation result of the degree of eye movement abnormality of the patient; The clinical diagnosis result comparison module is used to compare the diagnosis results output by the Parkinson's disease diagnosis module with the diagnosis results of clinicians to verify the accuracy and reliability of the system, and adjust the parameters of the Parkinson's disease early diagnosis model according to the comparison results to improve the performance of the diagnosis model.

[0028] In this embodiment, by combining the constructed Parkinson's disease early diagnosis model with the abnormal eye movement characteristics, the early signs of Parkinson's disease can be more accurately identified, thereby improving the accuracy of diagnosis. The abnormal eye movement feature extraction module evaluates the degree of abnormal eye movement through a quantitative method, providing doctors with objective evaluation indicators. By comparing with the clinical diagnosis results, possible misdiagnosis and missed diagnosis of the system can be discovered and corrected in time, thereby improving the diagnostic efficiency. Through early diagnosis, patients can receive treatment earlier, which may slow the progression of the disease and improve the quality of life.

[0029] In this embodiment, the Parkinson's disease diagnosis module further includes determining whether the patient has Parkinson's disease based on the video image of the patient's facial expression during eye movement, specifically: Obtaining a plurality of facial expression sample data of the patient based on the video image; Extract the patient's blinking frequency, facial movement change frequency and amplitude changes from each facial expression sample data; Identify the extracted multiple parameters, obtain identification results, and determine whether there are typical facial movement features of Parkinson's disease; The abnormal coefficient of the patient is calculated according to the blinking frequency, facial movement change frequency and amplitude change of the patient in each facial expression sample data and the identification result of each parameter; confirming whether the abnormal coefficient is greater than a preset coefficient, if so, obtaining an abnormal video image segment of the patient based on the video image, otherwise, preliminarily determining that the patient does not suffer from Parkinson's disease; Extract eye feature factors in each frame based on abnormal video image segments, such as eye wrinkles, eyelid closure status, etc. The typical features of Parkinson's disease, such as mask face (reduced facial expression), tremor, etc., are screened out based on the human eye feature factors in each frame of the image; The severity of the patient's symptoms is calculated based on the typical features of Parkinson's disease in the human eye feature factors in each frame of the image; Confirm whether the severity of the patient's symptoms is greater than or equal to a preset threshold. If so, it is confirmed that the patient suffers from Parkinson's disease. Otherwise, it is considered that the patient does not show typical symptoms of Parkinson's disease.

[0030] In this embodiment, the combination of eye movement and facial expression analysis can more comprehensively evaluate the patient's symptoms. Combined with eye movement characteristics, signs of Parkinson's disease can be identified earlier, improving the accuracy of diagnosis. Through sophisticated video image analysis, slight facial movement and blinking abnormalities can be identified, reducing the risk of misdiagnosis and missed diagnosis that may be caused by a single indicator. By quantifying the blinking frequency, the frequency and amplitude of facial movement changes, and human eye characteristic factors, the abnormality coefficient and symptom severity are calculated, reducing the error of subjective judgment, improving the efficiency of diagnosis, helping doctors assess the severity of the disease, and providing a basis for the formulation of treatment plans. As an auxiliary tool for clinicians to diagnose Parkinson's disease, it improves the consistency and reliability of diagnosis.

[0031] Furthermore, the data acquisition unit includes: Compensation modules for: Collect environmental images based on the camera on the eye tracker; Extracting brightness component information of each pixel in the environment image; determining a brightness histogram according to the brightness component information of each pixel in the environment image; determining a brightness interval according to the brightness histogram, determining a middle value of the brightness interval, and counting a first number of pixels greater than the middle value and a second number of pixels less than the middle value; when it is determined that the second number is greater than the first number, generating and executing a lighting compensation instruction;

[0032] in, is the grayscale value of the environment image G at (x, y) after illumination compensation; is the gray value of the environment image G at (x, y); is the target grayscale mean generated based on illumination compensation; The number of pixels included in the environment image; Adjustment modules for: The propagation direction of the laser beam is changed based on the rotation of the eye tracker by a first angle so that the laser light path passes through the face of the patient, and the first angular resolution of the facial expression of the patient is adjusted according to the first angle of rotation of the eye tracker;

[0033] in, is the first angular resolution; is an integer, , is the laser beam divergence angle; is the first angle between the eye tracker and the patient's face; is the maximum angle that the eye tracker can rotate; is the initial angular resolution of the eye tracker; The propagation direction of the laser beam is changed based on the rotation of the eye tracker at a second angle, so that the laser light path passes through the patient's eyeball, and the second angular resolution of the patient's eyeball is adjusted according to the second angle of rotation of the eye tracker;

[0034] in, is the second angular resolution; It is the second angle between the eye tracker and the patient's eyeball; the camera on the eye tracker collects the patient's facial expression and eye movement video data based on the first angular resolution and the second angular resolution.

[0035] The working principle of the above technical solution is as follows: the data acquisition unit is further refined to include a compensation module and an adjustment module, which are responsible for processing ambient light compensation and angular resolution adjustment of the eye tracker data respectively. In this embodiment, the compensation module collects the ambient image based on the camera on the eye tracker, and the ambient image is in the YUV color space, and extracts the brightness component information of each pixel in the ambient image, that is, the Y component. The brightness histogram is a graphical representation that shows the number of pixels with different brightness levels in the image. A brightness histogram can be generated by dividing the brightness value into a series of intervals and calculating the number of pixels in each interval. The brightness histogram is analyzed to determine the range of brightness values, and the median (or "median") of this interval is calculated, which is a center point of the brightness distribution. A first number of pixels greater than the median and a second number of pixels less than the median are counted; when it is determined that the second number is greater than the first number, it indicates that the image is dark as a whole, and a lighting compensation instruction is generated to increase the brightness of the image, thereby improving the visual quality of the image.

[0036] In this embodiment, the adjustment module adjusts the angular resolution of the eye tracker to collect data to meet different collection requirements. The first angular resolution for collecting the patient's facial expression is smaller than the second angular resolution for collecting the patient's eyeballs, and higher requirements are required for collecting the patient's eyeball information.

[0037] The beneficial effect of the above technical solution is that through the work of the compensation module and the adjustment module, the data acquisition unit can ensure that high-quality and high-resolution video data can be acquired under different lighting conditions and acquisition requirements.

[0038] Furthermore, the image acquisition module includes: The first determination module is used to: Performing frame processing on the video image of the patient's facial expression during eye movement to obtain a number of frame sub-images; Performing grayscale processing on a plurality of frames of sub-images to obtain a plurality of frames of grayscale images; Traverse each pixel point on the grayscale image, add up the grayscale value of each pixel point, divide the total grayscale value by the total number of pixels, and obtain the average grayscale value; in the process of traversing each pixel point on the grayscale image, record the maximum grayscale value and the minimum grayscale value; determine the pixel point whose grayscale value is less than the average grayscale value, and calculate the first grayscale sum value; determine the pixel point whose grayscale value is greater than the average grayscale value, and calculate the second grayscale sum value; Calculate a characteristic value of the grayscale image based on the maximum grayscale value, the minimum grayscale value, the first grayscale sum value, and the second grayscale sum value;

[0039] in, is the eigenvalue of the grayscale image; is the second grayscale value; is the first grayscale value; is the maximum gray value; is the minimum grayscale value; comparing the characteristic value of the grayscale image with the preset characteristic threshold, screening out the grayscale images whose characteristic value is greater than the preset characteristic threshold, and determining them as the target image set; The second determination module is used for: Determine a first target image and a second target image of adjacent frames in a target image set; Acquire a first grayscale feature and a first texture feature of a first target image; acquire a first grayscale feature and a second texture feature of a second target image; Based on the first grayscale feature and the second grayscale feature, performing a first image registration on the first target image and the second target image, and determining first change information of the facial expression according to a result of the first image registration; Based on the first texture feature and the second texture feature, performing a second image registration on the first target image and the second target image, and determining second change information of the facial expression according to a result of the second image registration; The facial expression change information is determined according to the first change information and the second change information.

[0040] The working principle of the above technical solution: In this embodiment, the video image of the patient's facial expression during eye movement is framed to obtain several frame sub-images, which are processed to determine the characteristic value of the grayscale image; the formula takes into account the distribution range of the grayscale value and the distribution difference of the grayscale value on both sides of the average grayscale value. The characteristic value of the grayscale image is compared with the preset characteristic threshold, and the grayscale images with characteristic values ​​greater than the preset characteristic threshold are screened out and determined as the target image set; the target image set contains images that meet the image quality requirements, and images with poor image quality are eliminated.

[0041] In this embodiment, two adjacent frames of images are identified and marked as the first target image and the second target image, respectively. These two frames of images will be used for subsequent image registration and change information extraction. For the first target image and the second target image, their grayscale features and texture features are extracted respectively. Grayscale features usually refer to the brightness information of the image, which can be obtained by calculating the grayscale histogram, grayscale co-occurrence matrix, etc. of the image. Texture features describe the arrangement and distribution pattern of pixels in the image, which can be obtained by calculating the local binary pattern (LBP) of the image, the statistics of the grayscale co-occurrence matrix, Gabor filter, etc. The grayscale features of the first target image and the second target image are used for the first image registration. Image registration is an image processing technology used to align two or more images acquired at different times, different viewing angles or different sensors so that they have a consistent geometric relationship in the same coordinate system. Find the optimal transformation parameters (such as translation, rotation, scaling, etc.) between the two images so that they have the smallest grayscale difference after alignment. According to the results of the first image registration, the first change information of the facial expression can be determined, such as the slight movement of facial muscles, the degree of opening and closing of the eyes, etc. The second image registration is performed using the texture features of the first target image and the second target image. Since the texture features contain more image detail information, the second image registration may more accurately capture the subtle changes in facial expressions. Based on the results of the second image registration, the second change information of the facial expression may be determined, which may include changes in skin texture, the appearance or disappearance of wrinkles, etc. The second determination module accurately captures and analyzes the facial expression change information between adjacent frame images.

[0042] The beneficial effects of the above technical solution are as follows: the work of the first determination module and the second determination module enables the image acquisition module to screen out a set of target images with significant features and high quality, and accurately determine the change information of facial expressions, thereby ensuring the accuracy of information acquisition.

[0043] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An early diagnosis system for Parkinson's disease based on eye movement feature analysis, characterized in that: include: A data acquisition unit, used to capture the patient's eye movement data in real time based on an eye tracker, and to collect the patient's facial expression and eye movement video data in real time based on a camera on the eye tracker, and to process the acquired eye movement data and facial expression and eye movement video data; A feature extraction unit, used for extracting eye movement features related to Parkinson's disease from the processed data, including eye movement speed, movement amplitude, eye fixation point stability and eye saccadic movement features; The early diagnosis unit is used to train and model the extracted eye movement features, establish an early diagnosis model for Parkinson's disease, input the extracted patient's eye movement features into the early diagnosis model for Parkinson's disease, obtain the early diagnosis results of Parkinson's disease, and generate an early diagnosis report for Parkinson's disease.

2. The early diagnosis system for Parkinson's disease based on eye movement feature analysis as claimed in claim 1, characterized in that: Data acquisition unit, including: A motion stimulation module is used to establish a visual stimulation pattern according to the diagnostic requirements of Parkinson's disease, generate dynamic visual stimulation according to the preset visual stimulation pattern, and display the dynamic visual stimulation on the display screen in real time to generate a scanning task; A data acquisition module, used to record the patient's eye movement data in real time based on an eye tracker when the patient performs a scanning task, including the start and end time, speed, amplitude and path of the scanning; An image acquisition module, used to capture video images of the patient's eye movements and the patient's facial expressions during eye movements; The data processing module is used to pre-process the data collected by the data acquisition module and the image acquisition module and convert the data into a unified format.

3. The early diagnosis system for Parkinson's disease based on eye movement feature analysis as claimed in claim 2, characterized in that: The data processing module further includes: Based on the video image of the patient's eyeball movement captured by the image acquisition module, the patient's eyeball image data is extracted, and the distance between the two pupils of the patient and the shape ratio of the two eyeballs are calculated; The eye image data of at least one patient is calculated, and based on the calculation result, the distance between pupils is compared to see whether it is consistent, and whether the variation range of the distance between pupils is abnormal is determined; According to the judgment result, the average value of the calculated results is used as the distance between the two pupils of the patient to obtain the final pupil distance information of the patient; The refractive information of the patient is obtained by calculation according to the shape ratio of the two eyeballs of the patient.

4. The early diagnosis system for Parkinson's disease based on eye movement feature analysis as claimed in claim 3, characterized in that: Feature extraction unit, including: An eye movement feature extraction module is used to extract the patient's eye movement speed feature and eye movement amplitude feature, and to correct the eye movement speed feature and eye movement amplitude feature based on the patient's final pupil distance information; The eye gaze feature extraction module is used to extract the patient's eye gaze point features and eye saccadic movement features, and evaluate the visual function and motor ability of the eye gaze point features and eye saccadic movement features based on the patient's refractive information.

5. The early diagnosis system for Parkinson's disease based on eye movement feature analysis as claimed in claim 4, characterized in that: Early diagnosis unit, including: A diagnosis model building module is used to build an early diagnosis model for Parkinson's disease, and to train the early diagnosis model for Parkinson's disease based on the output data of the eye movement feature extraction module and the eye gaze feature extraction module; An eye movement abnormality feature extraction module is used to extract asymmetric features of eye movement based on the visual function and motor ability evaluation results, and to evaluate the degree of eye movement abnormality of the patient based on the asymmetric features; The Parkinson's disease diagnosis module is used to input the extracted eye movement abnormality features into the trained Parkinson's disease early diagnosis model for diagnosis, and output the diagnosis result including the asymmetry index and conclusion in combination with the evaluation result of the degree of eye movement abnormality of the patient; The clinical diagnosis result comparison module is used to compare the diagnosis result output by the Parkinson's disease diagnosis module with the diagnosis result of the clinician, and adjust the parameters of the Parkinson's disease early diagnosis model according to the comparison result.

6. The early diagnosis system for Parkinson's disease based on eye movement feature analysis as claimed in claim 5, characterized in that: The eye movement abnormality feature extraction module is specifically: identifying saccadic events in the eye movement data of the patient captured by an eye tracker based on saccadic eye movement features, and segmenting the continuous eye movement data into individual saccadic events based on the identification result; The latency, peak velocity and amplitude of each saccade event are calculated, and the differences between the left and right eyes are compared. The mean and standard deviation of the saccade latency, peak velocity and amplitude of the left and right eyes are calculated based on the comparison results. The asymmetry index of the latency, peak velocity and amplitude is obtained based on the calculation results. The asymmetry index of latency, peak velocity and amplitude is compared with the corresponding preset asymmetry index threshold to determine whether there is significant asymmetry between the left and right eyes, and to comprehensively evaluate the degree of eye movement abnormality of the patient.

7. The early diagnosis system for Parkinson's disease based on eye movement feature analysis as claimed in claim 6, characterized in that: The Parkinson's disease diagnosis module also includes judging whether the patient has Parkinson's disease based on the video image of the patient's facial expression during eye movement, specifically: Obtaining a plurality of facial expression sample data of the patient based on the video image; Extract the patient's blinking frequency, facial movement change frequency and amplitude changes from each facial expression sample data; Identify the extracted multiple parameters, obtain identification results, and determine whether there are typical facial movement features of Parkinson's disease; The abnormal coefficient of the patient is calculated according to the blinking frequency, facial movement change frequency and amplitude change of the patient in each facial expression sample data and the identification result of each parameter; Confirm whether the abnormal coefficient is greater than a preset coefficient. If so, obtain an abnormal video image segment of the patient based on the video image. Otherwise, preliminarily determine that the patient does not suffer from Parkinson's disease.

8. The early diagnosis system for Parkinson's disease based on eye movement feature analysis as claimed in claim 7, characterized in that: Obtaining an abnormal video image segment of the patient based on the video image also includes: Extracting human eye feature factors in each frame of image based on abnormal video image segments; The typical features of Parkinson's disease are screened out based on the human eye characteristic factors in each frame of the image; The severity of the patient's symptoms is calculated based on the typical features of Parkinson's disease in the human eye feature factors in each frame of the image; Confirm whether the severity of the patient's symptoms is greater than or equal to a preset threshold. If so, it is confirmed that the patient suffers from Parkinson's disease. Otherwise, it is considered that the patient does not show typical symptoms of Parkinson's disease.

9. The early diagnosis system for Parkinson's disease based on eye movement feature analysis as claimed in claim 1, characterized in that: The data acquisition unit comprises: Compensation modules for: Collect environmental images based on the camera on the eye tracker; Extracting brightness component information of each pixel in the environment image; determining a brightness histogram according to the brightness component information of each pixel in the environment image; determining a brightness interval according to the brightness histogram, determining a middle value of the brightness interval, and counting a first number of pixels greater than the middle value and a second number of pixels less than the middle value; when it is determined that the second number is greater than the first number, generating and executing a lighting compensation instruction; in, is the grayscale value of the environment image G at (x, y) after illumination compensation; is the gray value of the environment image G at (x, y); is the target grayscale mean generated based on illumination compensation; The number of pixels included in the environment image; Adjustment modules for: The propagation direction of the laser beam is changed based on the rotation of the eye tracker by a first angle so that the laser light path passes through the face of the patient, and the first angular resolution of the facial expression of the patient is adjusted according to the first angle of rotation of the eye tracker; in, is the first angular resolution; is an integer, , is the laser beam divergence angle; is the first angle between the eye tracker and the patient's face; is the maximum angle that the eye tracker can rotate; is the initial angular resolution of the eye tracker; The propagation direction of the laser beam is changed based on the rotation of the eye tracker at a second angle, so that the laser light path passes through the patient's eyeball, and the second angular resolution of the patient's eyeball is adjusted according to the second angle of rotation of the eye tracker; in, is the second angular resolution; is the second angle between the eye tracker and the patient's eyeball; The camera on the eye tracker collects the patient's facial expression and eye movement video data based on the first angular resolution and the second angular resolution.

10. The early diagnosis system for Parkinson's disease based on eye movement feature analysis as claimed in claim 2, characterized in that: The image acquisition module comprises: The first determination module is used to: Performing frame processing on the video image of the patient's facial expression during eye movement to obtain a plurality of frame sub-images; Performing grayscale processing on a plurality of frames of sub-images to obtain a plurality of frames of grayscale images; Traverse each pixel point on the grayscale image, add up the grayscale value of each pixel point, divide the total grayscale value by the total number of pixels, and obtain the average grayscale value; in the process of traversing each pixel point on the grayscale image, record the maximum grayscale value and the minimum grayscale value; determine the pixel point whose grayscale value is less than the average grayscale value, and calculate the first grayscale sum value; determine the pixel point whose grayscale value is greater than the average grayscale value, and calculate the second grayscale sum value; Calculate a characteristic value of the grayscale image based on the maximum grayscale value, the minimum grayscale value, the first grayscale sum value, and the second grayscale sum value; in, is the eigenvalue of the grayscale image; is the second grayscale value; is the first grayscale value; is the maximum gray value; is the minimum gray value; Compare the characteristic value of the grayscale image with a preset characteristic threshold, filter out the grayscale images whose characteristic value is greater than the preset characteristic threshold, and determine them as the target image set; The second determination module is used for: Determine a first target image and a second target image of adjacent frames in a target image set; Acquire a first grayscale feature and a first texture feature of a first target image; acquire a first grayscale feature and a second texture feature of a second target image; Based on the first grayscale feature and the second grayscale feature, performing a first image registration on the first target image and the second target image, and determining first change information of the facial expression according to a result of the first image registration; Based on the first texture feature and the second texture feature, performing a second image registration on the first target image and the second target image, and determining second change information of the facial expression according to a result of the second image registration; The facial expression change information is determined according to the first change information and the second change information.

Images