Real-time dynamic tonometer

Abstract

The invention relates to the technical field of tonometer, in particular to a real-time dynamic tonometer which comprises a signal acquisition module, a signal processing module, an algorithm model module and a display module, the signal acquisition module is in wireless communication with the signal processing module through Bluetooth, and the display module is used for displaying real-time intraocular pressure signal data. The signal acquisition module comprises a plurality of flexible patches. According to the device, the three-axis acceleration sensor and the surface myoelectricity sensor are arranged through the flexible patch, eyelid tremor signals are detected through the three-axis acceleration sensor and the surface myoelectricity sensor, the tremor signals are processed through the signal processing module, the intraocular pressure value is calculated through the algorithm model module, measurement can be carried out without contact with the eye mask, and the measurement accuracy is improved. Meanwhile, the influence of the biomechanical characteristics of the cornea is avoided, the device is suitable for being used by a part of patients with cornea diseases, and a user can wear the device by himself / herself for measurement.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tonometer, in particular to a real-time dynamic tonometer. BACKGROUND

[0002] As an important indicator for assessing eye health, intraocular pressure measurement plays a key role in the diagnosis and treatment of eye diseases such as glaucoma. According to the different measurement principles and operation methods, modern tonometers can be mainly divided into two categories: contact and non-contact, each of which contains a variety of specific technical forms.

[0003] Among them, the contact tonometer measures intraocular pressure by directly or indirectly contacting the cornea. This kind of device usually needs to have physical contact with the patient's cornea during measurement. According to different working principles, it can be further divided into flattening, indentation and rebounding, etc. The advantage of contact tonometer is that the measurement result is relatively accurate and reliable. Non-contact tonometer indirectly measures intraocular pressure through air flow or other non-contact methods, avoiding direct contact with the cornea. The biggest advantage of this kind of device is that it is easy to operate, does not require anesthesia, and has no risk of infection, which is particularly suitable for large-scale screening and daily monitoring.

[0004] However, both methods are instantaneous measurements, which cannot capture 24-hour intraocular pressure fluctuations, and both are affected by the biomechanical properties of the cornea, such as hardness and thickness, which need to be combined with other examinations for comprehensive evaluation. At the same time, some patients with corneal diseases cannot use it, such as patients with severe dry eye and active inflammation who are not suitable for using contact tonometer for measurement, and patients with uneven cornea or severe ocular surface disease who are not suitable for using non-contact tonometer. SUMMARY

[0005] The purpose of the present application is to provide a real-time dynamic tonometer to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides the following technical scheme: The real-time dynamic tonometer comprises a signal acquisition module, a signal processing module, an algorithm model module and a display module. The signal acquisition module and the signal processing module communicate wirelessly through Bluetooth. The display module is used to display real-time intraocular pressure signal data. The signal acquisition module comprises a plurality of flexible patches. The flexible patches are provided with three-axis acceleration sensors and surface electromyography sensors. The flexible patches are provided with positioning mechanisms. The plurality of flexible patches are provided with a spectacle frame. The flexible patches are provided with Bluetooth transmitters and power supplies. The triaxial acceleration sensor can detect eyelid vibration, and the surface electromyography sensor can collect electrical activity of the orbicularis oculi muscle. The triaxial acceleration sensor and the surface electromyography sensor can transmit the collected signal data to the signal processing module through Bluetooth communication. The signal processing module can extract the intraocular pressure tremor feature. The signal processing module can transmit the intraocular pressure tremor signal data to the algorithm model module. The algorithm model module can output an estimated intraocular pressure value based on the calibration relationship between the tremor signal and the intraocular pressure. The algorithm model module can transmit the intraocular pressure value to the display module for display.

[0007] Further, the flexible patch is provided with a sponge block, and the sponge block is attached to the flexible patch. The positioning mechanism comprises a positioning frame, a plurality of plug-in sockets, a plurality of clamping grooves, a plurality of reset springs, and a plurality of sliding grooves. The positioning frame is arranged on the sponge block. The plurality of plug-in sockets are arranged on the positioning frame. The plurality of clamping grooves are respectively arranged on the plug-in sockets. The plurality of reset springs are respectively fixedly connected to the plug-in sockets. The plurality of sliding grooves are all arranged on the positioning frame, the plug-in socket is slidingly connected with the adjacent sliding groove, and the reset spring is fixedly connected with the adjacent sliding groove.

[0008] Further, the plug-in socket is fixedly connected with a stop block, the positioning mechanism is matched with an elastic band, and the elastic band is fixedly connected with a magic tape one and a magic tape two.

[0009] Further, when the signal processing module processes the signal data, a band-pass filter with a frequency of 0.5Hz to 35Hz is used to eliminate baseline drift and slow body movement through low-frequency cutoff, and to suppress high-frequency electronic noise through high-frequency cutoff.

[0010] Further, during the signal data processing process of the signal processing module, PCA analysis is performed on the triaxial acceleration data to extract the principal axis direction with the maximum vibration energy, and the data in this direction is taken as the main tremor signal for analysis.

[0011] Preferably, the signal processing module can use amplitude threshold and pattern recognition algorithm to automatically identify and eliminate invalid data segments of large amplitude voluntary blinking and sudden head turning.

[0012] Preferably, the signal processing module can extract frequency domain features, time domain features and nonlinear features from the processed signal through feature engineering, and then capture key information related to intraocular pressure in multiple dimensions, wherein the frequency domain features include main frequency, frequency spectrum energy distribution and frequency spectrum entropy feature, the time domain features include root mean square and zero-crossing rate feature, and the nonlinear feature is mainly the entropy value feature of the signal.

[0013] Preferably, in the process of establishing the calibration relationship between tremor frequency and intraocular pressure by the algorithm model module, standard data of the intraocular pressure of the patient is collected, and while the eyelid tremor data is collected by the signal acquisition module, the real and calibrated intraocular pressure value is measured and recorded simultaneously by using the flattening tonometer, forming the paired data set of features and real intraocular pressure, and based on the importance ranking of the tree model, the feature subset with the most relevant and least redundant features to the intraocular pressure is selected from all the features extracted above.

[0014] Further, in the process of establishing the paired data set of features and real intraocular pressure by the algorithm model module, multivariate linear regression is used as the starting model, and more complex gradient boosting trees are used to capture the potential nonlinear relationship between features and intraocular pressure, and the data set is divided into a training set and a test set, and the training set data is used to train the model and optimize its parameters.

[0015] Preferably, the algorithm model module constructs the model based on large sample training data, adjusts the model by combining the reference value and the user's physiological parameters for the first time, dynamically corrects the deviation for long-term use, and fine-tunes the general model according to the individual data.

[0016] Compared with the prior art, the present application has the following advantages: The three-axis acceleration sensor and the surface electromyography sensor are arranged on the flexible patch, the three-axis acceleration sensor and the surface electromyography sensor detect the eyelid tremor signal by being attached to the eyelid, the tremor signal is processed by the signal processing module, and the key information related to the intraocular pressure is extracted, the intraocular pressure value is calculated based on the calibration relationship between the tremor signal and the intraocular pressure established by the algorithm model module based on the multivariate regression model, and finally the real-time display of the intraocular pressure value is realized by the display module, so that the measurement can be realized without contacting the eye membrane, which is conducive to reducing the risk and discomfort during measurement, and at the same time, the measurement is not affected by the corneal biomechanical properties, and is suitable for patients with some corneal diseases, does not require professional operation, and can be measured by the user himself, is suitable for home monitoring, large-scale population screening and postoperative continuous monitoring, can be used after closing the eyes and resting, and can realize continuous recording of the intraocular pressure change, such as diurnal fluctuation, which is conducive to early detection of the night intraocular pressure peak.

[0017] The clamping grooves are arranged in the insertion seat, and under the action of the return spring, the clamping grooves on the positioning frame are clamped on the spectacle frame, so that the flexible patch and the sponge block are connected on the spectacle frame through the positioning mechanism, so that the flexible patch corresponds to the patient's eyelid after wearing the spectacle frame, and the wearing is relatively convenient, and the position of the positioning mechanism on the spectacle frame can be adjusted according to the interpupillary distance of the patient, so as to adapt to the wearing of patients with different interpupillary distances, and the elastic band can be selected to pass through the block and the positioning frame, and the flexible patch is positioned by being tied on the head, so that the patient can rest on the side at night. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is the overall block diagram of the real-time dynamic tonometer of the present application; Figure 2 is the schematic diagram of the flexible patch structure in the present application; Figure 3 is the schematic diagram of the positioning mechanism structure in the present application; Figure 4 is the schematic diagram of the internal structure of the positioning mechanism in the present application; Figure 5 is the schematic diagram of the elastic band structure in the present application; Figure 6 is the schematic diagram of the signal receiving master controller structure in the present application.

[0019] In the figure: 10, flexible patch; 20, signal receiving master controller; 30, glasses frame; 40, positioning mechanism; 41, positioning frame; 42, plug-in seat; 43, clamping groove; 44, stop block; 45, return spring; 46, sliding groove; 50, sponge block; 60, elastic band; 61, magic tape I; 62, magic tape II. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the protection scope of the present application.

[0021] Please refer to Figures 1-6 In the embodiments of the present application, the real-time dynamic tonometer comprises a signal acquisition module, a signal processing module, an algorithm model module and a display module. The signal acquisition module and the signal processing module perform wireless communication through Bluetooth. The display module is used for displaying real-time eye pressure signal data. The signal acquisition module comprises a plurality of flexible patches 10. The flexible patch 10 is provided with a three-axis acceleration sensor and a surface electromyography sensor. The flexible patch 10 is provided with a positioning mechanism 40. The plurality of flexible patches 10 are provided with a glasses frame 30. The flexible patch 10 is provided with a Bluetooth transmitter and a power supply. The Bluetooth transmitter is used for sending detected signals. The power supply is used for supplying power to the three-axis acceleration sensor, the surface electromyography sensor and the Bluetooth transmitter. The flexible patch 10 is provided with a sponge block 50. The sponge block 50 is pasted on the flexible patch 10. The triaxial acceleration sensor can detect eyelid vibration, and the surface electromyography sensor can collect the electrical activity of the orbicularis oculi muscle. The triaxial acceleration sensor and the surface electromyography sensor can transmit the collected signal data to the signal processing module through Bluetooth communication. The signal processing module can extract the ocular pressure tremor characteristics. The signal processing module can transmit the ocular pressure tremor signal data to the algorithm model module. The algorithm model module can construct a model based on large sample training data. The algorithm model module can output an estimated intraocular pressure value based on the calibration relationship between the tremor signal and the intraocular pressure. The algorithm model module can transmit the intraocular pressure value to the display module for display.

[0022] Specifically, the signal processing module, the algorithm model module, and the display module can be integrated in the signal receiving host controller 20. A Bluetooth receiver is arranged in the signal receiving host controller 20. The signal acquisition module transmits signal data to the signal processing module in the signal receiving host controller 20 through Bluetooth communication. The signal processing module performs signal processing and analysis. The algorithm model module runs the model. The display module displays real-time and historical intraocular pressure trend graphs, provides abnormal value alarms, generates health reports, etc. The signal receiving host controller 20 can transmit data to the cloud platform through the Internet of Things. The cloud platform stores user long-term data. The core algorithm model is continuously optimized and iterated using larger-scale population data. The cloud platform provides a remote data access interface for doctors, facilitating doctors to view monitoring data in a timely manner. The triaxial acceleration sensor adopts a high-sensitivity, low-noise MEMS type sensor to collect original triaxial acceleration and angular velocity data at a sampling rate of at least 100 Hz, ensuring the capture of tremor signals at five to thirty hertz. When the flexible patch 10 is attached to the eyelid skin surface, the triaxial acceleration sensor records the amplitude (acceleration value), frequency, and waveform (such as a sine wave or a pulse wave) of the vibration, reflecting the mechanical characteristics of the vibration. The surface electromyography sensor (sEMG) can synchronously collect the electrical activity of the orbicularis oculi muscle, obtain the amplitude and frequency of the electromyography signal, correlate the muscle contraction intensity, and exclude non-muscle-derived vibrations such as external touch, thereby comprehensively characterizing the eyelid vibration and providing a rich data basis for accurate intraocular pressure measurement. Collecting signals through sensors, combining signal processing algorithms and machine learning models can improve the accuracy of measurement, thereby minimizing the error range to meet clinical needs. The measurement process does not require contact with the eye membrane, which can reduce the risk of infection and discomfort during measurement. At the same time, it is not affected by the biomechanical properties of the cornea and is suitable for use by patients with corneal diseases. It does not require professional operation and can be worn by users for self-measurement. It is suitable for home monitoring, large-scale population screening, and postoperative continuous monitoring. After wearing, the user can rest with their eyes closed, and the device can continuously record changes in intraocular pressure, such as diurnal fluctuations, which can help early detection of high intraocular pressure peaks at night.

[0023] Ten minutes before collection, avoid vigorous exercise, blink too frequently, reduce the impact of eyelid muscle fatigue on the signal, during the measurement process, the user remains seated, closed eyes, to avoid eye rotation, facial expression muscle activity.

[0024] Embodiment one

[0025] As Figures 2-4 shown, in this embodiment, the positioning mechanism 40 includes a positioning frame 41, a plurality of sockets 42, a plurality of clamping grooves 43, a plurality of reset springs 45, and a plurality of sliding grooves 46. The positioning frame 41 is arranged on the sponge block 50, the plurality of sockets 42 are arranged on the positioning frame 41, the plurality of clamping grooves 43 are respectively formed on the sockets 42, the plurality of reset springs 45 are respectively fixedly connected to the sockets 42, the plurality of sliding grooves 46 are all formed on the positioning frame 41, the socket 42 is slidingly connected with the adjacent sliding groove 46, and the reset spring 45 is fixedly connected with the adjacent sliding groove 46, so that the socket 42 can be connected to the positioning frame 41 through the reset spring 45.

[0026] In specific implementation, the two sockets 42 can be moved away from each other first, so that the clamping grooves 43 correspond to the spectacle frame 30. After the sockets 42 are loosened, the clamping grooves 43 on the positioning frame 41 can be clamped on the spectacle frame 30 under the action of the reset spring 45, so that the flexible patch 10 and the sponge block 50 are connected to the spectacle frame 30 through the positioning mechanism 40. In this way, after wearing the spectacle frame 30, the flexible patch 10 can correspond to the patient's eyelid, and wearing is relatively convenient. In addition, the position of the positioning mechanism 40 on the spectacle frame 30 can be adjusted according to the interpupillary distance of the patient, so as to facilitate the wearing of patients with different interpupillary distances.

[0027] As Figure 1 shown, in this embodiment, when the signal processing module processes the signal data, a band-pass filter of 0.5Hz to 35Hz is used to eliminate baseline drift and slow body movement through low-frequency cutoff, and to suppress high-frequency electronic noise through high-frequency cutoff. During the signal data processing process of the signal processing module, the principal axis direction with the maximum vibration energy is extracted by performing PCA analysis on the three-axis acceleration data, and the data in this direction is taken as the main tremor signal for analysis, so as to reduce the influence of the random direction of the sensor. The signal processing module can use amplitude threshold and pattern recognition algorithm to automatically identify and exclude invalid data segments of large amplitude active blinking and sudden head rotation.

[0028] In specific implementation, by adopting a band-pass filter of 0.5 Hz to 35 Hz, the main frequency range of the eyelid vibration (blinking frequency of about 0.5-2 Hz) is retained, and low frequencies are filtered to eliminate baseline drift, slow body movement, and 0.2-0.5 Hz respiratory interference noise, while high frequency electromagnetic interference noise greater than 100 Hz is filtered. The signal processing module can introduce independent component analysis (ICA) to separate the orbicularis oculi muscle electromyographic signal from other muscles (such as the temporal muscle) from the mixed signal, exclude non-target muscle interference, and effectively filter out interference signals such as breathing and other facial muscle activity by adopting band-pass filtering and independent component analysis technology, to more accurately separate the orbicularis oculi muscle signal related to intraocular pressure, and improve the purity and reliability of the signal.

[0029] As shown in Figure 1 In the present embodiment, the signal processing module can extract frequency domain features, time domain features and nonlinear features from the processed signals through feature engineering, and further capture key information related to intraocular pressure in multiple dimensions, wherein the frequency domain features include dominant frequency, spectral energy distribution, spectral entropy feature, the time domain features include root mean square, zero-crossing rate feature, and the nonlinear feature is mainly the entropy value feature of the signal.

[0030] In specific implementation, by performing fast Fourier transform on the effective signal segment, the peak frequency of the power spectral density in the range of five to thirty hertz is found, the dominant frequency is obtained, and the weighted average frequency of the spectrum in the tremor frequency band is calculated to reflect the concentration point of the tremor energy, and then the spectral energy distribution is obtained. The spectral entropy can represent the regularity and complexity of the tremor frequency distribution, which is related to the stability of the intraocular pressure; The root mean square value of the time domain feature can represent the overall energy or amplitude of the tremor, and the number of times the signal crosses zero in a unit time is recorded as the zero-crossing rate, which is a simple time-domain estimation of frequency; Sample entropy or approximate entropy can represent the complexity and unpredictability of the quantized signal sequence, and is used to capture subtle changes in the neural control pattern caused by changes in intraocular pressure.

[0031] As shown in Figure 1As shown, in this embodiment, in the process of establishing the calibration relationship between tremor frequency and intraocular pressure by the algorithm model module, standard data collection is performed on the intraocular pressure of the patient, the eyelid tremor data is collected by the signal collection module, and the real and calibrated intraocular pressure value is measured and recorded simultaneously by using the flattening tonometer to form the paired data set of features and real intraocular pressure. Based on the importance ranking of the tree model, the most relevant and least redundant feature subset is selected from all the features extracted above. In the process of establishing the paired data set of features and real intraocular pressure by the algorithm model module, multivariate linear regression is used as the starting model, and more complex gradient boosting trees are used to capture the potential nonlinear relationship between features and intraocular pressure. The data set is divided into a training set and a test set, the training set data is used to train the model, and the parameters are optimized to establish the mapping relationship between the eyelid tremor features and the intraocular pressure value.

[0032] In specific implementation, the Goldmann tonometer or dynamic contour tonometer is used to measure the real and calibrated gold standard intraocular pressure value. Multivariate linear regression, support vector regression or random forest regression can be used as the starting model. Gradient boosting trees or shallow neural networks can be used as advanced models. The data set is divided into a training set and a test set, and the training set data is used to train the model. The goal is to minimize the root mean square error between the predicted intraocular pressure and the real intraocular pressure. On the test set independent of the training set, the consistency between the predicted value of the algorithm model module and the Goldman measurement value is evaluated, and statistical indicators such as the intraclass correlation coefficient and the 95% consistency limit of the Bland-Altman graph are calculated to verify the accuracy, and the performance of the product in different body positions (sitting, lying), different activity states (resting, light activity) and different user groups (different ages, different corneal thicknesses) can be tested to test the robustness.

[0033] Embodiment Two

[0034] Based on Embodiment One, as shown in Figure 5 In this embodiment, the socket 42 is fixedly connected with a stop block 44, and the positioning mechanism 40 is matched with an elastic band 60, and the elastic band 60 is fixedly connected with a magic tape one 61 and a magic tape two 62.

[0035] In specific implementation, the positioning mechanism 40 can be detached from the glasses frame 30, and the elastic band 60 can be passed between the stop block 44 and the positioning frame 41. By binding the elastic band 60 on the head, the magic tape one 61 and the magic tape two 62 are adhered to each other, and the flexible patch 10 is positioned by the elastic band 60, which is convenient for the patient to rest on the side at night.

[0036] As shown in Figure 1As shown, in the embodiment, the algorithm model module constructs a model based on large sample training data, adjusts the model by combining the reference value and the user physiological parameters for the first use, dynamically corrects the deviation for long-term use, and fine-tunes the general model according to the personal data.

[0037] In specific implementation, a multivariate regression model or a deep learning model is constructed based on large sample training data, individual calibration is performed, the model is adjusted by combining the gold standard intraocular pressure value and the user physiological parameters for the first use, the deviation is dynamically corrected for long-term use, and thus the universality of the model is improved, and the model is suitable for use by different people.

[0038] It is apparent for those skilled in the art that the present application is not limited to the details of the foregoing exemplary embodiments, and the present application can be implemented in other concrete forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered in all aspects as illustrative and not restrictive, and the scope of the present application is defined by the appended claims rather than the foregoing description, and it is intended to embrace all changes falling within the meaning and scope of the equivalent elements of the claims. Any reference signs in the claims should not be considered as limiting the claims involved.

[0039] In addition, it should be understood that, although the present specification is described in terms of embodiments, not every embodiment contains only one independent technical solution, and the specification is described in this way only for the sake of clarity, and those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be properly combined to form other embodiments that those skilled in the art can understand.

Claims

1. A real-time dynamic tonometer, comprising a signal acquisition module, a signal processing module, an algorithm model module, and a display module, wherein the signal acquisition module and the signal processing module communicate wirelessly via Bluetooth, and the display module is used to display real-time intraocular pressure signal data, characterized in that, The signal acquisition module includes several flexible patches (10), each flexible patch (10) is equipped with a triaxial accelerometer and a surface electromyography sensor, each flexible patch (10) is equipped with a positioning mechanism (40), and each flexible patch (10) is equipped with an eyeglass frame (30). The triaxial accelerometer can detect eyelid tremors, and the surface electromyography (SEMG) sensor can collect the electrical activity of the orbicularis oculi muscle. The triaxial accelerometer and SEMG sensor can transmit the collected signal data to the signal processing module via Bluetooth communication. The signal processing module can extract intraocular pressure tremor features and transmit the intraocular pressure tremor signal data to the algorithm model module. The algorithm model module can output an estimated intraocular pressure value based on the calibration relationship between the tremor signal and intraocular pressure. The algorithm model module can transmit the intraocular pressure value to the display module for display.

2. The real-time dynamic tonometer according to claim 1, characterized in that, The flexible patch (10) is provided with a sponge block (50); The positioning mechanism (40) includes: The positioning frame (41) is set on the sponge block (50); Several plug-in sockets (42) are set on the positioning frame (41); Several slots (43) are respectively opened on the socket (42); Several return springs (45) are fixedly connected to the plug socket (42); Several sliding grooves (46) are provided on the positioning frame (41). The plug-in seat (42) is slidably connected to the adjacent sliding groove (46), and the reset spring (45) is fixedly connected to the adjacent sliding groove (46).

3. The real-time dynamic tonometer according to claim 2, characterized in that, The plug-in socket (42) is fixedly connected to a stop block (44), and the positioning mechanism (40) is equipped with an elastic band (60). The elastic band (60) is fixedly connected to Velcro 1 (61) and Velcro 2 (62).

4. The real-time dynamic tonometer according to any one of claims 1-3, characterized in that, When processing signal data, the signal processing module uses a bandpass filter from 0.5Hz to 35Hz to eliminate baseline drift and slow body movements through low-frequency cutoff and suppress high-frequency electronic noise through high-frequency cutoff.

5. The real-time dynamic tonometer according to claim 4, characterized in that, During the signal data processing, the signal processing module performs PCA analysis on the triaxial acceleration data to extract the principal axis direction with the greatest vibration energy, and uses the data in this direction as the main vibration signal for analysis.

6. The real-time dynamic tonometer according to claim 4, characterized in that, The signal processing module can use amplitude thresholding and pattern recognition algorithms to automatically identify and remove invalid data segments such as large-amplitude active blinking and sudden head rotation.

7. The real-time dynamic tonometer according to claim 4, characterized in that, The signal processing module can extract frequency domain features, time domain features, and nonlinear features from the processed signal through feature engineering, thereby capturing key information related to intraocular pressure from multiple dimensions. The frequency domain features include dominant frequency, spectral energy distribution, and spectral entropy features, while the time domain features include root mean square and zero crossover rate features. The nonlinear characteristics are mainly the entropy characteristics of the signal.

8. The real-time dynamic tonometer according to any one of claims 1-3, characterized in that, In the process of establishing the calibration relationship between tremor frequency and intraocular pressure in the algorithm model module, standard data collection of patient intraocular pressure is carried out. While the signal acquisition module collects eyelid tremor data, a flattening tonometer is used to simultaneously measure and record the real, calibrated intraocular pressure value, forming a paired dataset of features and real intraocular pressure. Based on the importance ranking of the tree model, the feature subset with the most relevant intraocular pressure and the least redundancy is selected from all the features extracted above.

9. The real-time dynamic tonometer according to claim 8, characterized in that, In the process of establishing the paired dataset of features and true intraocular pressure, the algorithm model module uses multiple linear regression as the starting model and a more complex gradient boosting tree to capture the potential nonlinear relationship between features and intraocular pressure. By dividing the dataset into training and test sets, the model is trained using the training set data to optimize its parameters.

10. The real-time dynamic tonometer according to claim 8, characterized in that, The algorithm model module is built based on large sample training data. The model is initially adjusted by combining the benchmark value and the user's physiological parameters. In the long term, the deviation is dynamically corrected, and the general model is fine-tuned based on individual data.

Images