An intraocular pressure measuring device and method

Abstract

The application relates to the technical field of medical devices, in particular to an intraocular pressure measuring device and a measuring method. The measuring method comprises the following steps: identifying and segmenting a square interference fringe area contained in a photo based on a YOLOv11-segment model; binarizing and extracting the square interference fringe area based on a deep convolutional neural network; extracting the skeleton of the binarized interference fringe, identifying the order of each fringe, and then calculating the central deflection of a Fabry-Perot microcavity after the square film is pressed; and obtaining the intraocular pressure according to the central deflection. The square interference fringe area contained in the photo is accurately identified, segmented and extracted based on the YOLOv11-segment model, the skeleton of the extracted fringe after binarization is identified, the order of each fringe is identified, and then the central deflection of the square film after being pressed is calculated. The pressure value of the sensor calibrated by the calibration device is one-to-one corresponding, and the photo containing the interference fringe image taken is converted into a pressure value.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to an intraocular pressure measuring device and method. Background Technology

[0002] Glaucoma is one of the three leading causes of blindness and poses a significant threat to human health. High intraocular pressure is considered a major risk factor for glaucoma; therefore, intraocular pressure is an important indicator in clinical practice for determining glaucoma treatment goals and assessing treatment effectiveness and prognosis.

[0003] Currently, the main method for measuring intraocular pressure (IOP) is through instruments that measure the patient's IOP in real time. These instruments primarily include applanation tonometers and aero-jet tonometers. Applanation tonometers, however, are complex and have drawbacks such as requiring topical anesthesia before measurement, needing to instill fluorescein sodium into the cornea during measurement, and the measurement being affected by the central corneal thickness. Aero-jet tonometers, compared to applanation tonometers, simplify the measurement process and do not require topical anesthesia or fluorescein sodium. However, aero-jet tonometers also have several drawbacks, including the impact of the airflow causing eye discomfort, high cost, and inconvenience.

[0004] In addition, there are many studies on miniature implantable intraocular pressure sensors based on different principles. These studies share the following characteristics: 1. The sensor is separate from the detection device, and the sensing method is non-contact; 2. The sensor area and volume are tiny, ranging from hundreds of micrometers to a few millimeters; 3. The sensor is in contact with the eye structure, either attached to the eyeball or implanted inside the eyeball.

[0005] These implantable intraocular pressure (IOP) sensors, classified according to their sensing principles, mainly fall into three categories: electro-sensing, microfluidic sensing, and optical sensing. Chen et al. designed an IOP sensor based on capacitance sensitivity to pressure, using a contact lens as a carrier. The frequency of the LC oscillator formed by the capacitor and inductor changes with pressure, and the reading device is a large network analyzer. Agaoglu et al. used a microfluidic chip to achieve IOP detection. They implanted an artificial lens integrating the microfluidic chip into the eyeball using cataract surgery. As IOP fluctuates, the liquid-gas interface of the artificial lens shifts, and the IOP value can be obtained by monitoring this interface position. Electro-sensing is limited by circuit structure and materials, making it difficult to achieve sub-millimeter size, and the reading device is large and expensive. Microfluidic sensing is limited by stringent requirements for airtightness and the indirect sensing principle of image reading, which has hindered miniaturization. Optical sensors are generally smaller than electro-sensors and microfluidic sensors. Therefore, MEMS IOP sensors based on optical sensors have become the main research direction for implantable sensors.

[0006] Current measurement schemes for MEMS intraocular pressure sensors based on optical sensing are all based on the Fabry-Pérot interferometer structure. By capturing the interference fringe pattern, the interference region is identified and extracted, and the interference fringe pattern is converted into pressure. However, existing methods for identifying and extracting the interference region usually involve manual image cropping. This manual image cropping makes it difficult to apply to real-time continuous demodulation processes, thus limiting the application of intraocular pressure measurement. Summary of the Invention

[0007] To address the problem that existing methods for identifying and extracting interference regions in MEMS intraocular pressure sensors are difficult to apply to real-time continuous demodulation processes, this invention provides an intraocular pressure measurement method. This method introduces the YOLOv11-seg model to accurately identify, segment, and extract interference fringe regions contained in images. This allows the method to identify and extract interference regions without manual image cropping, solving the problem that existing interference region identification and extraction methods typically rely on manual image cropping, making them difficult to apply to real-time continuous demodulation processes.

[0008] The technical solution adopted by this invention to solve its technical problem is:

[0009] A method for measuring intraocular pressure includes the following steps:

[0010] S1: Sensing changes in intraocular pressure via the Fabry-Perot microcavity on the intraocular pressure sensor;

[0011] S2: Light is emitted into the Fabry-Perot microcavity via the imaging module, and an image of the interference pattern is acquired via the imaging element;

[0012] S3: Based on the YOLOv11-segment model, the interference fringe regions contained in the photo are identified and segmented to obtain the interference fringe regions;

[0013] S4: The interference fringe region is binarized and extracted based on a deep convolutional neural network to obtain binary interference fringes;

[0014] S5: Extract the skeleton of the binarized interference fringes, identify the order of each fringe, and then calculate the central deflection of the film of the Fabry-Perot microcavity after being compressed;

[0015] S6: Obtain intraocular pressure based on the central deflection.

[0016] Optionally, identifying and segmenting the interference fringe regions in the photograph based on the YOLOv11-segment model includes:

[0017] S31: Acquire photographs of interference patterns under multiple environmental conditions;

[0018] S32: Manually annotate the interference fringe regions in the photograph using the Labelme plugin to obtain a txt file, which serves as the label set corresponding to the training set images;

[0019] S33: Use a portion of the images as the training set and a portion of the images as the validation set to train the model and obtain the interference fringe region.

[0020] Optionally, the deep convolutional neural network includes a left path, a right path, an encoding path, and a decoding path.

[0021] Optionally, the loss function used in training the deep convolutional neural network model includes a focus loss function and a multi-scale structural similarity measure.

[0022] Optionally, the focus loss function is defined as follows:

[0023] L FL (p, y) = L FL (p t )=-0.5(1-p t ) r log(p t );

[0024] in,

[0025] y∈{1,0} represents the true label value of the sample, p∈[0,1] is the output of the model, that is, the probability that the true label of the sample is 1; r is the adjustment factor in the focus loss function.

[0026] Optionally, the multi-scale structural similarity measure is defined as follows:

[0027]

[0028] Where M is the number of scales, α M ,β j ,γ j They are l M (x,y),c j (x,y),s j The weights of (x,y), l M (x,y),c j (x,y),s j (x,y) represent the similarity of x and y at scales M,i,j in terms of brightness, contrast, and structure.

[0029] Optionally, the deflection distribution function of the Fabry-Perot microcavity after the thin film is compressed is as follows:

[0030]

[0031] Where (x, y) are the coordinates of a point on the diaphragm, with the center of the diaphragm as the origin, w0 is the central deflection of the diaphragm, α is the side length of the diaphragm, and c1 and c2 are two empirical parameters.

[0032] Optionally, obtaining intraocular pressure based on the central deflection includes: calibrating the central deflection of the Fabry-Perot microcavity membrane under different pressure values ​​using a calibration device, and then obtaining the intraocular pressure based on the calculated central deflection.

[0033] Another object of the present invention is to provide an intraocular pressure measuring device that measures intraocular pressure using the intraocular pressure measuring method described above.

[0034] Optionally, it includes an intraocular pressure sensor, an imaging module, and an imaging element; wherein,

[0035] The intraocular pressure sensor is provided with a Fabry-Perot microcavity;

[0036] The imaging module includes a housing, an optical path assembly disposed inside the housing, and a light source disposed outside the housing;

[0037] The housing is provided with a camera hole;

[0038] The light emitted by the light source is transmitted to the Fabry-Perot microcavity after passing through the optical path assembly, and an interference pattern is generated.

[0039] The imaging element includes a lens;

[0040] The imaging aperture is adapted to cooperate with the lens to acquire the interference pattern through the imaging element.

[0041] The beneficial effects of this invention are:

[0042] The intraocular pressure measurement method provided by this invention uses a deep learning method, based on the YOLOv11-seg model, to accurately identify, segment and extract the square interference fringe regions contained in the photograph. The skeleton of the extracted fringe is binarized, and the order of each fringe is identified. Then, the central deflection of the square film after being compressed is calculated. The central deflection of the sensor corresponds one-to-one with the pressure value of the sensor calibrated by the calibration device, so as to convert the photograph containing the interference fringe into pressure value. Attached Figure Description

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0044] Figure 1 This is a schematic diagram of the intraocular pressure measurement method in this invention;

[0045] Figure 2 This is a schematic diagram illustrating the specific implementation process of converting an image into pressure in this invention;

[0046] Figure 3 This is a schematic diagram of the training results of YOLOv11-seg in this invention;

[0047] Figure 4 This is a schematic diagram of the deep convolutional neural network model structure in this invention;

[0048] Figure 5 These are images from a portion of the dataset used during training in this invention;

[0049] Figure 6 This refers to the change in stripe pattern under different pressures in this invention;

[0050] Figure 7 This is a simplified structural diagram of the intraocular pressure measuring device in this invention;

[0051] Figure 8 This is an exploded view of the intraocular pressure sensor in this invention;

[0052] Figure 9 This is an exploded view of the imaging module in this invention;

[0053] Figure 10 This is a schematic diagram of the optical path component in this invention;

[0054] Figure 11 This is a schematic diagram illustrating the effect of the shooting angle on the integrity of the optical interference pattern in this invention;

[0055] Figure 12 This is a schematic diagram of the pressurization process in this invention.

[0056] Figure 13 This is a physical diagram of the pressurization device in this invention.

[0057] In the diagram: 1-Intraocular pressure sensor; 11-Sensor body; 12-Bracket; 121-Mounting end; 122-Fixed end; 1221-Wide section; 1222-Gradual section; 1223-Narrow section; 1224-Anti-slip structure; 123-Drainage groove; 1231-First groove structure; 1232-Second groove structure; 2-Imaging module; 21-Housing; 211-Imaging hole; 22-Optical path assembly; 221-Beam splitter cube; 222-Planar-convex lens; 223-Narrow band filter; 23-Light source; 24-Clamp; 241-C-shaped element; 2411-Mounting groove; 2412-Through hole; 242-Threaded connector; 3-Imaging element; 31-Lens. Detailed Implementation

[0058] The present invention will now be described in further detail. The embodiments described below are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0059] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0060] To address the problem that existing methods for interference region identification and extraction in MEMS intraocular pressure sensors are difficult to apply to real-time continuous demodulation processes, this invention provides an intraocular pressure measurement method. Taking a mobile phone as the imaging element as an example, see [link to relevant documentation]. Figure 1 , Figure 2 As shown, the measurement method includes the following steps:

[0061] S1: Sensing changes in intraocular pressure via the Fabry-Perot microcavity on the intraocular pressure sensor;

[0062] This step can be achieved by implanting a corresponding optical sensor-based MEMS intraocular pressure sensor into the anterior chamber of the eye, with the outer surface of the Fabry-Perot microcavity in contact with the intraocular fluid to sense changes in intraocular pressure;

[0063] S2: Light is emitted into the Fabry-Perot microcavity via the imaging module, and an image of the interference pattern is captured by the imaging element;

[0064] The light emitted by the imaging module is incident on the Fabry-Perot microcavity, enters the cavity of the Fabry-Perot microcavity, is reflected by multiple surfaces and interferes, and an interference pattern is obtained;

[0065] Specifically, the preferred imaging module of the present invention includes an optical path assembly and a light source; the light emitted by the light source is transmitted to the Fabry-Perot microcavity after passing through the optical path assembly, and an interference pattern is generated; the imaging element includes a lens to acquire the interference pattern through the imaging element;

[0066] The preferred light source for this invention is a white light source; the imaging element can be any existing digital camera, camcorder, or smartphone, including a CMOS image sensor; to further reduce the difficulty of shooting, the preferred imaging element for this invention is a mobile phone; the mobile phone is equipped with a CMOS image sensor to realize the acquisition and imaging of interference patterns;

[0067] During operation, the light emitted by the light source passes through the optical path assembly and is incident perpendicularly into the Fabry-Perot microcavity (FP resonant cavity) of the intraocular pressure sensor. Upon entering the cavity of the Fabry-Perot microcavity, the light is reflected by multiple surfaces and interferes, resulting in an interference pattern. The interference pattern is then transmitted through the optical path assembly to the lens of the imaging element, and a photograph of the interference pattern can be captured in real time by taking a picture through the imaging element.

[0068] S3: Based on the YOLOv11-segment model, the interference fringe regions in the photo are identified and segmented to obtain the interference fringe regions;

[0069] Existing methods for identifying and extracting interference regions typically involve manual image cropping or edge extraction to further segment the region's edges. Manual image cropping makes this approach unsuitable for real-time continuous demodulation. While classical edge extraction methods exist, the characteristics of the image edges vary under different experimental conditions due to varying illumination intensities, image sharpness, and other environmental factors, requiring manual parameter adjustments for each experiment. This also hinders real-time continuous measurement across multiple scenarios. Furthermore, the strong edge features inherent in interference fringes further increase the instability of traditional edge detection in this task.

[0070] Based on this, the present invention uses a deep learning method, based on the YOLOv11-seg model, to identify, segment and extract the square interference fringe regions contained in the image. For the present invention, this method accurately identifies, segments and extracts the square interference fringe regions contained in the image, avoiding manual adjustment of parameters during the extraction of interference fringes. This enables the intraocular pressure measurement method provided by the present invention to be applied to the real-time continuous demodulation process.

[0071] S4: The interference fringe region is binarized and extracted based on a deep convolutional neural network, avoiding interference caused by noise and uneven background brightness in the image, and obtaining binarized interference fringes.

[0072] Furthermore, in order to extract the phase information carried by the interference fringes, this invention uses a deep convolutional neural network to perform further binarization processing on the interference fringes in the interference region.

[0073] S5: Extract the skeleton of the binary interference fringes, identify the order of each fringe, and then calculate the central deflection of the Fabry-Perot microcavity after the film is compressed.

[0074] S6: Obtain intraocular pressure based on central deflection.

[0075] The intraocular pressure measurement method provided by this invention uses a deep learning method, based on the YOLOv11-seg model, to accurately identify, segment and extract the square interference fringe regions contained in the photograph. The skeleton of the extracted fringe is binarized, and the order of each fringe is identified. Then, the central deflection of the square film after being compressed is calculated. The central deflection of the sensor corresponds one-to-one with the pressure value of the sensor calibrated by the calibration device, so as to convert the photograph containing the interference fringe into pressure value.

[0076] Specifically, the present invention preferably identifies and segments the interference fringe regions in the photograph based on the YOLOv11-segment model, including:

[0077] S31: Acquire photographs of interference patterns under multiple environmental conditions;

[0078] In this step, we will take a mobile phone as an example to illustrate the shooting element. The mobile phone takes a sensor interference photo. The photo includes different sensors, different pressures, different angles, different light intensities, different mobile phone camera settings, and other different environmental conditions.

[0079] S32: Manually annotate the interference fringe regions in the photograph using the Labelme plugin to obtain a txt file, which serves as the label set corresponding to the training set images;

[0080] The Labelme plugin exports Label files in JSON format by default. By using rule conversion, the JSON file is converted into the TXT file format required by the YOLO model, which serves as the label set corresponding to the training set images. These images and labels are then further expanded by performing rotation, scaling, translation, and other processing.

[0081] S33: Use a portion of the images as the training set and a portion of the images as the validation set to train the model and obtain the interference fringe region;

[0082] During model training, 153 images were selected for the training set and 54 for the validation set. Pre-trained weights from YOLOv11n-seg were loaded, data augmentation was enabled during training, and the image input size was specified as 640x640x3. Performance metrics during training iterations were as follows: Figure 3 As shown, by the end of training, the mask accuracy had reached over 99%.

[0083] This invention treats stripe binarization as an image segmentation problem. A model separates bright and dark stripes into two distinct classes. This model is an improvement on the M-net network used in medical image segmentation and is a typical deep convolutional neural network. It achieves multi-scale feature fusion by using skip connections to connect feature maps of different sizes at different scales. The deep convolutional neural network contains four paths: a left path, a right path, an encoding path, and a decoding path. The left and right paths serve as deep supervision. The entire network involves convolutional layers, max-pooling layers, upsampling layers, batch normalization layers, ReLU layers, and sigmoid layers. Each path in the network contains feature maps of four sizes. The encoding layer adopts a typical CNN architecture. At each layer, two 3×3 Conv-BN-ReLU modules are first used to extract features, and then 2×2 max-pooling with a stride of 2 is used to reduce the feature map size by half. Therefore, the number of convolutional kernels in the next convolutional layer is twice that of the previous layer. The decoder path adopts the exact same structure as the encoder. In the decoder path, the structure and convolutional layer parameter settings of each stage are the same as those in the symmetrical encoder path. The difference lies in the fact that the decoder uses a 2×2 up-sampling layer as the inverse operation of the max pooling layer to progressively restore the feature map scale to the original image size. Finally, the output of the right path is concatenated with the output of the decoder along the channel dimension and fed into a 1×1 Conv-Sigmoid module to obtain the probability that each pixel is classified as a positive sample. Each layer has skip connections within the encoding and decoding paths, as well as between adjacent different paths, to achieve better segmentation results. The specific model structure is as follows: Figure 4 As shown;

[0084] Furthermore, the loss function used in the training of the deep convolutional neural network model in this invention preferably includes a focal loss function and a multi-scale structural similarity measure; the loss function used in the training of this model mainly consists of two parts, namely the focal loss function (FocalLoss, FL) and the multi-scale structural similarity measure (MS-SSIM):

[0085] L FL (p, y) = L FL (p t )=-0.5(1-p t ) r log(p t );

[0086] Specifically, during the experiment, pixels located at the edges of the stripes are easily misclassified, resulting in uneven stripe edges. Focus loss is an improved cross-entropy loss function used to address the severe foreground-background class imbalance problem during the training process of dense detectors in object detection tasks. The preferred definition of the focus loss function in this invention is as follows:

[0087] L FL (p, y) = L FL (p t )=-0.5(1-p t ) r log(p t );

[0088] in,

[0089] y∈{1,0} represents the true label value of the sample, p∈[0,1] is the output of the model, that is, the probability that the true label of the sample is predicted to be 1; r is the adjustment factor in the focus loss function, which is set to 2 in this invention.

[0090] Interference fringes possess strong structural features. The human visual system is adept at extracting structural information from image scenes, and structural similarity (SSIM) is a good estimate of human-perceived image quality. The formula for calculating the structural similarity of images x and y is:

[0091] SSIM(x, y) = [l(x, y)] α [c(x, y)] β [s(x, y)] γ ;

[0092] Where α, β, and γ are the factors for each weight; l(x,y), c(x,y), and s(x,y) represent the image similarity in brightness, contrast, and structure, respectively. However, SSIM is calculated at a fixed scale, so it is only applicable to images at a specific scale. Multi-scale structural similarity measurement improves this; MS-SSIM is more flexible than single-scale SSIM when the viewpoint changes. The definition of multi-scale structural similarity measurement in this invention is as follows:

[0093]

[0094] Where M is the number of scales, α M ,β j ,γ j They are l M (x,y),c j (x,y),s j The weights of (x,y), lM (x,y),c j (x,y),s j (x,y) represent the similarity of x and y at scales M,i,j in terms of brightness, contrast, and structure.

[0095] The value of MS-SSIM ranges from [0,1], and it takes the value 1 if and only if the two images are completely identical. Therefore, the loss function for multi-scale structural similarity measurement is defined as:

[0096] L MS-SSIM (p, y) = 1 - MS_SSIM(x, y).

[0097] Furthermore, the dataset establishment method for this deep convolutional network in this invention is as follows: 26 interference fringe images taken during the experiment were used as the training set. These images were first pre-processed using the FCM clustering algorithm for binarization, and then further refined using Adobe Photoshop. Since this dataset was too small, the number of samples was increased to 2600 by simultaneously performing random cropping, rotation, mirroring, perspective transformation, noise addition, and adjustments to brightness and contrast on both the training and label sets. In addition, 500 interference images and their corresponding binarized images were simulated using MATLAB 2024a, totaling 3100 images with a size of 496×496, which were used as the training dataset. The training device was equipped with a 12th Gen Intel(R) Core(TM) i7-12650H CPU with 32GB RAM and an NVIDIA GeForce RTX4060 Laptop GPU. During training, the SGD algorithm was used to optimize network parameters. The batch size was set to 4, the initial learning rate to 0.01, the Nesterov momentum to 0.75, and the learning rate to decrease by 0.00005 per iteration. Training was stopped early if the loss did not decrease after 5 epochs on the validation set. The maximum number of training epochs was set to 50, and the final training duration was 35 epochs, approximately 9 hours. Some training set images used in the training are shown below. Figure 5 As shown.

[0098] Furthermore, skeleton lines are extracted from the binarized interference fringe image, and the order of each skeleton line is identified, and its corresponding unwinding phase is calculated. The deflection distribution function of the Fabry-Perot microcavity film under pressure is as follows:

[0099]

[0100] Where (x, y) are the coordinates of a point on the diaphragm, with the center of the diaphragm as the origin, w0 is the central deflection of the diaphragm, α is the side length of the diaphragm, and c1 and c2 are two empirical parameters. During the demodulation calculation, only the central sectional line of the diaphragm needs to be used for fitting, i.e., y = 0. The deflection shape function of the square diaphragm transforms into the deflection distribution function of the central sectional line of the square diaphragm:

[0101]

[0102] This invention extracts the pixel coordinates of the horizontal skeleton line at the center of the binarized interference fringes and the corresponding unwinding phase value, fits the deflection distribution of the diaphragm center section, and then obtains the deflection value of the center position of the interference fringe image, i.e., the center of the square diaphragm.

[0103] Furthermore, for the central deflection of the sensor diaphragm demodulated in the photograph, the present invention correlates it with the pressure value of the narrow cavity in which the sensor is located through a calibration experiment; that is, obtaining intraocular pressure based on the central deflection includes: calibrating the central deflection of the Fabry-Perot microcavity membrane under different pressure values ​​using a calibration device, and then obtaining the intraocular pressure based on the calculated central deflection.

[0104] For details, see Figure 12 As shown, the cubic-shaped intraocular pressure sensor is fixed by an I-shaped bracket made of silicone. The sensor is mounted onto the I-shaped bracket using UV-cured adhesive, and then the I-shaped bracket is further bonded to the pressure chamber. See also Figure 13 As shown, the pressurization device in the experiment uses a syringe. At the beginning of the experiment, the pressure chamber and the bottom of the graduated cylinder need to be fixed at the same level to ensure that the pressure value in the pressure chamber is the liquid level height displayed in the graduated cylinder. In the experiment, the liquid level height of pure water in the graduated cylinder is controlled by pushing and pulling the syringe, thereby controlling the hydraulic pressure on the intraocular pressure sensor in the pressure chamber.

[0105] In the experiment, the syringe was controlled to apply pressure in three cycles at room temperature. The pressure was maintained at 5 cmH2O, 15 cmH2O, 25 cmH2O, 35 cmH2O, 45 cmH2O, and 55 cmH2O (i.e., 3.68 mmHg, 11.03 mmHg, 18.39 mmHg, 25.74 mmHg, 33.1 mmHg, and 40.46 mmHg, respectively) for 3 minutes to ensure the intraocular pressure sensor was in a stable hydraulic environment. The recorded fringe images are shown below. Figure 6 As shown, it can be observed that the stripe density increases with increasing hydraulic pressure.

[0106] The images recorded during the test were demodulated, and center deflection curves under different pressures were plotted. Linear regression analysis was performed on the six curves, and the coefficient of determination R of the three-round pressure rise and fall curves was calculated. 2All values ​​were above 0.99, and the six curves showed high repeatability. Fitting the data from three rounds of pressure increases and decreases yielded a final pressure sensitivity of 20.92 nm / mmHg.

[0107] Another object of the present invention is to provide an intraocular pressure measuring device that measures intraocular pressure using the intraocular pressure measuring method described above.

[0108] The intraocular pressure measurement device provided by this invention uses a deep learning method based on the YOLOv11-seg model to accurately identify, segment, and extract the square interference fringe regions contained in the photograph during the intraocular pressure measurement process. The skeleton of the extracted fringe is binarized, and the order of each fringe is identified. Then, the central deflection of the square film after being compressed is calculated. The central deflection of the sensor corresponds one-to-one with the pressure value of the sensor calibrated by the calibration device, so as to convert the photograph containing the interference fringe into pressure value.

[0109] For details, see Figure 7 As shown, the intraocular pressure measurement device of the present invention includes an intraocular pressure sensor 1, an imaging module 2, and an imaging element 3; it should be noted that the present invention Figure 7 To clearly illustrate the structure of the intraocular pressure measurement device, the size of the intraocular pressure sensor has been artificially enlarged; among them, intraocular pressure sensor 1 is a MEMS sensor, see [link / reference]. Figure 8 As shown, the intraocular pressure sensor 1 is equipped with a Fabry-Perot microcavity; see [link / reference] Figure 9 As shown, the imaging module 2 includes a housing 21, preferably made of polylactic acid (PLA) material and fabricated by extrusion 3D printing; an optical path assembly 22 disposed inside the housing 21, and a light source 23 disposed outside the housing 21; an imaging hole 211 is provided on the housing 21; the light emitted by the light source 23 is transmitted to the Fabry-Perot microcavity after passing through the optical path assembly 22, and generates an interference pattern; the imaging element 3 includes a lens 31; the imaging hole 211 is adapted to cooperate with the lens 31 to obtain the interference pattern through the imaging element 3.

[0110] The preferred light source 23 of this invention is a white light source.

[0111] During operation, the light emitted by the light source 23 passes through the optical path assembly 22 and is incident perpendicularly on the Fabry-Perot microcavity (FP resonant cavity) of the intraocular pressure sensor 1. Upon entering the cavity of the Fabry-Perot microcavity, the light is reflected by multiple surfaces and interferes to obtain an interference pattern. The obtained interference pattern is then transmitted through the optical path assembly 22 to the lens 31 of the imaging element 3. The interference pattern can be captured in real time by taking a picture through the imaging element 3, and the real-time intraocular pressure can be obtained based on the real-time captured interference pattern.

[0112] Specifically, the method described above can be used to obtain intraocular pressure based on interference patterns.

[0113] During use, the intraocular pressure sensor 1 is implanted in the anterior chamber of the eye. The outer surface of the Fabry-Perot microcavity is in contact with the intraocular fluid to sense changes in intraocular pressure. When the intraocular pressure increases, the Fabry-Perot microcavity deforms, causing a change in the optical path of the reflected light, which in turn changes the interference pattern, causing the interference fringes to bend. By accurately identifying, segmenting, and extracting the interference fringe regions contained in the interference pattern, and then binarizing them to extract the fringe skeleton and identify the order of each fringe, the central deflection of the Fabry-Perot microcavity membrane after being compressed can be calculated. The central deflection of the Fabry-Perot microcavity membrane corresponds one-to-one with the pressure value of the Fabry-Perot microcavity membrane calibrated by the calibration device. This allows the captured image containing interference fringes to be converted into a pressure value, thus enabling the acquisition of intraocular pressure based on changes in the interference pattern.

[0114] The intraocular pressure measurement device provided by the present invention introduces an imaging module 2 that is compatible with the existing imaging element 3, so that the intraocular pressure can be detected by means of the existing imaging element 3. Furthermore, during the detection process, the positions of the imaging module 2 and the imaging element 3 can be flexibly adjusted according to the position of the Fabry-Perot microcavity in the intraocular pressure sensor 1, so that the incident beam meets the requirement of perpendicular incidence, thereby greatly reducing the difficulty of angle adjustment during the intraocular pressure detection process, making it more portable, efficient and easy to use.

[0115] To measure intraocular pressure, see [link to relevant documentation]. Figure 10 As shown, the preferred optical path assembly 22 of the present invention includes a beam splitter cube 221 and a plano-convex lens 222 sequentially disposed in the imaging aperture 211. The beam splitter cube 221 operates in the wavelength range of 450–650 nm. When light is incident at an incident angle of 45°, it can split the incident light into two beams with a ratio of approximately 50% transmission (T) and 50% reflection (R), with a tolerance of ±5% (T / R = 50%: 50% ± 5%). This allows the light emitted by the light source 23 to be reflected by the beam splitter cube 221 to adjust the optical path direction, and then converged by the plano-convex lens 222 to the target plane used to generate the optical interference pattern, i.e., the position of the Fabry-Perot microcavity. The reflected light of the interference pattern is then transmitted to the lens 31 of the imaging element 3 after passing through the plano-convex lens 222 and the beam splitter cube 221, thereby realizing the acquisition and imaging of the interference pattern. Preferably, the design wavelength of the plano-convex lens 222 is 350-700 nm, and the focal length is 20 mm.

[0116] Furthermore, the preferred optical path component 22 of the present invention further includes a narrowband filter 223 disposed between the light source 23 and the beam splitter 221, wherein the narrowband filter 223 is a 633nm narrowband filter, so that the light emitted by the light source 23 is filtered by the narrowband filter 223 to obtain monochromatic light with a center wavelength of 633nm and a bandwidth of ±10nm.

[0117] The imaging element 3 in this invention can be any existing digital camera, camcorder, or smartphone, including a CMOS image sensor; to further reduce the difficulty of shooting, the imaging element 3 is preferably a mobile phone; the mobile phone is equipped with a CMOS image sensor to realize the acquisition and imaging of interference patterns.

[0118] Existing optical pressure sensors require equal-inclination interference from the incident light beams when measuring intraocular pressure. For equal-inclination interference, the main requirement is that the incident angle and reflection angle (or refraction angle) of the two interfering beams are equal during reflection or refraction. In other words, the incident beams must be perpendicular to the ground to ensure complete interference fringes. When a miniature pressure sensor is implanted in a pressure detection environment, its small size makes it impossible to guarantee its horizontal orientation. When using a desktop microscope, the microscope can generally only maintain a vertically downward angle. Therefore, to obtain a complete optical interference pattern, one must rely on intuition to adjust the spatial angle of the object being measured, which is extremely difficult, especially when the object being measured is one whose spatial angle cannot be adjusted, such as an intraocular pressure sensor implanted in the eye. Based on this, this invention proposes an external camera module 2 adaptable to any smartphone. Compared to adjusting the uncertain spatial angle of the object being measured, adjusting the angle of the handheld phone is obviously much easier. Furthermore, we can determine the appropriate angle to tilt the phone based on the real-time image captured by the phone's camera. The optical interference pattern captured in real-time by a mobile phone camera is a square region. When the angle of incidence of light is not perpendicular to the interference plane, the square region is incomplete, appearing as a mixture of bright and dark areas. See also Figure 11 As shown, we can imagine the square interference region as a sealed "box" filled with water, and the bright part as a "bubble" inside the sealed space. The "box" should tilt in the direction the "bubble" is positioned within the square region, until the "bubble" moves to the center of the square region. The mobile phone is the "box." When the "bubble" moves to the center of the square region, the angle of incidence of the light is perpendicular to the interference plane, at which point the complete optical interference pattern is captured.

[0119] To facilitate connection with a mobile phone, the shooting module 2 of the present invention preferably also includes a clamp 24; one end of the clamp 24 is connected to the shooting element 3, i.e., the mobile phone, and the other end is connected to the housing 21.

[0120] The clamp 24 of this invention is preferably made of polylactic acid (PLA) material and is prepared by extrusion 3D printing process; and more preferably, the clamp 24 is a C-shaped clamp structure, including a C-shaped element 241 and a threaded connector 242; the C-shaped element 241 is connected to the housing 21; the C-shaped element 241 is connected to the shooting element 3 through the threaded connector 242; during use, the mobile phone is placed in the C-shaped element 241 and the connection with the mobile phone is achieved by tightening the threaded connector 242; the opening and closing range of the C-shaped element 241 is preferably 8-20mm, which can be adapted to the thickness of most smartphones on the market.

[0121] Furthermore, the present invention preferably connects the housing 21 and the clamp 24 by means of a snap-fit. Specifically, the C-shaped element 241 of the clamp 24 preferably has a mounting groove 2411 that is adapted to the housing 21, and the mounting groove 2411 has a concave point. The outer side of the housing 21 has a protrusion that is adapted to the concave point. The two parts can be easily assembled or disassembled by the cooperation of the protrusion and the concave point.

[0122] Furthermore, the mounting slot 2411 is provided with a through hole 2412 that matches the shooting hole 211, so as to avoid the C-shaped element 241 affecting the light transmission.

[0123] This invention provides an external camera module 2 compatible with any smartphone. The housing 21 of the external camera module 2 is made of environmentally friendly polylactic acid (PLA) material and is manufactured through a precisely controlled extrusion 3D printing process, ensuring consistent structural strength and quality. The camera module 2 integrates an optimized, customized optical path design and high-performance optical components, combined with real-time image capture, to achieve stable capture of high-quality images. Compared to the first-generation desktop microscope-style image capture method, this module not only guarantees image clarity and optical imaging quality but also significantly reduces the complexity of user operation and the impact of camera shake during shooting, making the device more portable, more efficient, and improving both user experience and applicability.

[0124] The intraocular pressure sensor 1 in this invention can be any existing intraocular pressure sensor with a Fabry-Perot microcavity. Since the intraocular pressure monitoring device provided by this invention is based on optical sensing to detect intraocular pressure, as mentioned above, during the detection process, the incident light needs to satisfy equal inclination interference, and equal inclination interference requires the incident beam to meet the requirement of perpendicular incidence. Therefore, in order to ensure the clarity of the detection image, the position of the intraocular pressure sensor 1 must remain fixed and not move during the detection process. To avoid the intraocular pressure sensor 1 moving during the detection process, this invention preferably includes a sensor body 11 and a bracket 12 connected to the sensor body 11, so as to fix the intraocular pressure sensor body 11 through the bracket 12, reduce the detection difficulty, and improve the clarity of the detection image.

[0125] Existing stents used to fix intraocular implants are mostly cylindrical structures. However, in the intraocular pressure measurement device provided by this invention, if the position of the intraocular pressure sensor body 11 moves slightly during the detection process, the incident light beam will not be able to be incident perpendicularly, and the shooting angle of the imaging element 3 needs to be readjusted. Therefore, in order to ensure the stability of the position of the intraocular pressure sensor body 11 during the detection process, this invention preferably uses a plate-shaped structure for the stent 12 to increase the contact area between the stent 12 and the inside of the eye, and to prevent the intraocular pressure sensor body 11 from moving.

[0126] Specifically, the preferred bracket 12 of the present invention includes an installation end 121 and a fixing end 122 connected to the installation end 121; wherein the installation end 121 is used to connect to the intraocular pressure sensor body 11, and the intraocular pressure sensor body 11 and the installation end 121 in the present invention can be connected by high temperature bonding, adhesive bonding of compatible materials, etc.; the size of the installation end 121 is determined according to the size of the intraocular pressure sensor body 11.

[0127] Since the eyeball has a certain curvature, in order to improve the fit between the bracket 12 and the eyeball, improve the stability of the intraocular pressure sensor body 11, and improve the patient's comfort, the present invention preferably has an arc-shaped structure for the fixed end 122, the curvature of which is determined according to the curvature of the eyeball.

[0128] To ensure the stability of the intraocular pressure sensor body 11 while improving comfort, the fixed end 122 of the present invention preferably includes a wide segment 1221, a gradient segment 1222, and a narrow segment 1223 connected in sequence; wherein the widths of the wide segment 1221, the gradient segment 1222, and the narrow segment 1223 decrease sequentially.

[0129] It should be noted that, in the bracket 12 of the present invention, the length direction in which the wide segment 1221, the gradient segment 1222 and the narrow segment 1223 are distributed is the width direction on the plane of the plate-like structure of the bracket 12, which is perpendicular to the length direction.

[0130] Specifically, the present invention preferably has a rectangular plate structure for the wide section 1221 to ensure the contact area between the support 12 and the eyeball; preferably, the width of the gradient section 1222 decreases sequentially, the width of the end connected to the wide section 1221 is the same as the width of the wide section 1221, and the width of the end connected to the narrow section 1223 is the same as the width of the narrow section 1223.

[0131] To balance comfort and the stability of the position of the intraocular pressure sensor body 11, the present invention preferably has a length ratio of (1.4~1.7):(1.1~1.4):(0.7~1) for the wide segment 1221, the gradient segment 1222 and the narrow segment 1223.

[0132] To further improve the stability of the intraocular pressure sensor body 11 after implantation, the present invention preferably provides an anti-slip structure 1224 on the outer side of the narrow segment 1223, and specifically preferably the anti-slip structure 1224 is a protruding structure extending outward along the narrow segment 1223.

[0133] To balance the stability and comfort of the position of the intraocular pressure sensor body 11, the present invention further preferably has at least two sets of protrusion structures, each set including two protrusions of the same size, symmetrically arranged on both sides of the narrow section 1223; and the size of the protrusions gradually decreases in the direction away from the gradient section 1222.

[0134] In a further preferred embodiment, the fixed end 122 is provided with a drainage groove 123 so that the intraocular pressure measuring device provided by the present invention has a certain drainage function while having an intraocular pressure measuring function.

[0135] Existing intraocular pressure sensors typically only have the function of detecting intraocular pressure and cannot achieve drainage. When the intraocular pressure is high, it is necessary to introduce a corresponding drainage device to achieve the therapeutic effect. Based on this, the present invention preferably provides a drainage groove 123 on the stent 12 so that the aqueous humor can be diffused to the periocular tissue through the drainage groove 23. This allows the intraocular pressure sensor to have both intraocular pressure detection and drainage functions, so that intraocular pressure detection and drainage can be achieved with a single implantation without increasing the number of implantations, increasing surgical damage, or increasing patient pain.

[0136] The drainage channel 123 includes a groove-shaped structure distributed longitudinally along the support 12 and passing through the mounting end 121 and the fixing end 122 in sequence, referred to as the first groove-shaped structure 1231; to further improve the drainage effect, the drainage channel 123 also includes a second groove-shaped structure 1232 obliquely distributed on the wide section 1221, and the second groove-shaped structure 1232 is connected to the first groove-shaped structure 1231.

[0137] The intraocular pressure sensor provided by this invention can be implanted into the eye via injection, significantly reducing surgical trauma. Through its groundbreaking design, this invention provides an intraocular pressure sensor that can be implanted into the eye via minimally invasive injection. The system can achieve continuous 24 / 7 monitoring of intraocular pressure without the need for electronic components or electromagnetic energy supply, and the intraocular pressure measurement accuracy reaches ±1 mmHg.

[0138] The intraocular pressure sensor in this invention establishes a functional relationship between the Fabry-Perot microcavity deflection in the sensor's central region and intraocular pressure based on the change in the spacing of the sensor's interference fringes caused by changes in intraocular pressure. By integrating advanced YOLO target detection instance segmentation and M-net deep convolutional neural network algorithms into a mobile app, it automatically focuses, identifies, and crops the sensor's interference fringes region and performs real-time intraocular pressure demodulation, enabling patients to self-monitor their intraocular pressure at home using their mobile phones. The signal transmission is independent of electromagnetic energy supply, effectively avoiding signal loss caused by external factors.

[0139] The intraocular pressure sensor provided by this invention has stronger compatibility and universality, and can be used in conjunction with any existing ophthalmic implantable device, which helps to further explore the feasibility of new clinical technologies that integrate glaucoma monitoring and diagnosis.

[0140] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for measuring intraocular pressure, characterized in that, Includes the following steps: S1: Sensing changes in intraocular pressure via the Fabry-Perot microcavity on the intraocular pressure sensor; S2: Light is emitted into the Fabry-Perot microcavity via the imaging module, and an image of the interference pattern is acquired via the imaging element; S3: Based on the YOLOv11-segment model, the interference fringe regions contained in the photo are identified and segmented to obtain the interference fringe regions; S4: The interference fringe region is binarized and extracted based on a deep convolutional neural network to obtain binary interference fringes; S5: Extract the skeleton of the binarized interference fringes, identify the order of each fringe, and then calculate the central deflection of the film of the Fabry-Perot microcavity after being compressed; S6: Obtain intraocular pressure based on the central deflection; The step S3, which involves identifying and segmenting the interference fringe regions in the photograph based on the YOLOv11-segment model, includes: S31: Acquire photographs of interference patterns under multiple environmental conditions; S32: Manually annotate the interference fringe regions in the photograph using the Labelme plugin to obtain a txt file, which serves as the label set corresponding to the training set images; S33: Use a portion of the images as the training set and a portion of the images as the validation set to train the model and obtain the interference fringe region; The deflection distribution function of the Fabry-Perot microcavity film under pressure is as follows: ； in These are the coordinates of points on the diaphragm, with the center of the diaphragm as the origin. The central deflection of the diaphragm. Let be the side length of the diaphragm. For two empirical parameters; Obtaining intraocular pressure based on the central deflection includes: calibrating the central deflection of the Fabry-Perot microcavity membrane under different pressure values ​​using a calibration device, and then obtaining the intraocular pressure based on the calculated central deflection.

2. The intraocular pressure measurement method as described in claim 1, characterized in that, The deep convolutional neural network includes a left path, a right path, an encoding path, and a decoding path.

3. The intraocular pressure measurement method as described in claim 2, characterized in that, The loss functions used in training the deep convolutional neural network model include the focal loss function and the multi-scale structural similarity measure.

4. The intraocular pressure measurement method as described in claim 3, characterized in that, The definition of the focus loss function is as follows: ； in, ; This represents the true label value of the sample. This is the output of the model, which is the probability that the true label of the predicted sample is 1; This is the adjustment factor in the focus loss function.

5. The intraocular pressure measurement method as described in claim 3, characterized in that, The definition of the multi-scale structural similarity measure is as follows: ； in, For the number of scales, They are respectively The weight, x and y at scales respectively The similarity in terms of brightness, contrast, and structure.

6. An intraocular pressure measuring device, characterized in that, Intraocular pressure is measured using the intraocular pressure measurement method as described in any one of claims 1-5.

7. The intraocular pressure measuring device as described in claim 6, characterized in that, It includes an intraocular pressure sensor (1), an imaging module (2), and an imaging element (3); among which, The intraocular pressure sensor (1) is provided with a Fabry-Perot microcavity; The shooting module (2) includes a housing (21), an optical path component (22) disposed inside the housing (21), and a light source (23) disposed outside the housing (21). The housing (21) is provided with a shooting hole (211); The light emitted by the light source (23) is transmitted to the Fabry-Perot microcavity after passing through the optical path assembly (22), and an interference pattern is generated; The imaging element (3) includes a lens (31); The imaging aperture (211) is adapted to cooperate with the lens (31) to obtain the interference pattern through the imaging element (3).

Images