Method for predicting arch height after intraocular lens implantation operation with lens eye
By training OCT image data with a generative adversarial network model, the problem of insufficient accuracy in predicting the arch height after PIOL surgery in existing technologies has been solved, achieving more accurate arch height prediction and personalized selection, and reducing surgical risks.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- THE EYE HOSPITAL OF WENZHOU MEDICAL UNIVERSITY
- Filing Date
- 2026-01-23
- Publication Date
- 2026-05-15
AI Technical Summary
Existing technologies for predicting the arch height after intraocular lens implantation in phakic eyes suffer from anatomical mismatch, insufficient utilization of imaging parameters, bias in AI model feature selection, and failure to fully utilize the three-dimensional morphological features in panoramic anterior segment images. These issues result in low prediction accuracy and the risk of requiring a second surgery.
A postoperative image prediction model was constructed using generative adversarial networks. The model was trained using OCT image data and the piol height was calculated by combining image features. Multimodal fusion technology was used for accurate alignment and collaborative analysis to output recommended piol diameter and piol height values.
It improves the accuracy of PIOL postoperative arch height prediction, reduces experience dependence, decreases the probability of secondary surgery, reduces the risk of complications, and provides a more reliable basis for personalized selection.
Smart Images
Figure CN122050702A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intraocular lens implantation technology, and more specifically, to a method for predicting the arch height after intraocular lens implantation in phakic eyes. Background Technology
[0002] Phakic intraocular lens (PIOL) implantation is a mainstream surgical procedure for correcting high myopia, and accurate prediction of the postoperative arch height is a core challenge to ensure surgical safety. Postoperative arch height is mainly related to the anterior segment structure and the PIOL diameter. Selecting an accurate PIOL diameter is crucial for obtaining a suitable arch height, thereby ensuring the success of PIOL implantation. Current clinical practice and research have the following significant limitations in their technical approaches to arch height prediction and PIOL size selection:
[0003] (1) Anatomical structure adaptation deviation of traditional empirical formula method
[0004] Current clinically widely used methods for selecting PIOL size primarily rely on online calculation systems provided by manufacturers, which recommend parameters based on white-to-white corneal diameter (WTW) and anterior chamber depth (ACD). However, due to the anatomical difference between WTW and the ciliary sulcus-to-sulcus diameter (STS), which truly reflects the PIOL implantation location, the accuracy of this method is not ideal (ideal postoperative scleral height ratio: 69%-78.31%), especially when there is a significant deviation between WTW and STS.
[0005] (2) Insufficient utilization of three-dimensional features and linearity defects in the application of imaging parameters
[0006] In recent years, sulcus-to-sulcus diameter (STS) measurements based on ultrasound biomicroscopy (UBM) and angle-to-angle diameter (ATA) analyses based on optical coherence tomography (OCT) have been introduced into clinical practice, attempting to improve predictive accuracy through anatomical parameter optimization. However, while UBM offers high penetration and can obtain structural images of the posterior chamber, its measurement accuracy is significantly affected by the operator's experience and suffers from issues such as contact operation and low repeatability. Although OCT can non-invasively acquire high-resolution anterior segment images, current methods only extract single parameters (such as ATA), failing to fully utilize three-dimensional morphological information such as corneal topography, iris configuration, and lens spatial relationships in panoramic images. Predictive formulas based on new imaging parameters rely only on a subset of variables selected through clinical experience and depend on linear regression, failing to accurately reflect actual spatial information and the complex process of multiple factors influencing the arch height.
[0007] (3) Limited data utilization and feature selection bias of AI models
[0008] With the deep application of artificial intelligence technology in ophthalmology, many AI models based on machine learning methods such as random forest models have emerged to predict iris height and PIOL diameter. However, the input of these models is limited to manually selected text parameters, which fails to effectively mine key morphological features such as the three-dimensional structure of the iris and the spatial relationship of the lens contained in OCT images, resulting in significant room for improvement in predictive performance.
[0009] (4) Common technical bottlenecks
[0010] Current methods for predicting the arch height after PIOL surgery generally suffer from feature selection bias. Whether it's traditional empirical formulas or existing AI models, they all rely on a limited set of textual parameters (such as WTW and ACD) selected by humans. This parameter selection mode based on prior clinical knowledge may lead to the systematic omission of key morphological information. Of particular note is that existing methods do not fully utilize the three-dimensional morphological features contained in panoramic anterior segment images, such as corneal topography, iris configuration, and lens spatial relationships. Summary of the Invention
[0011] The purpose of this invention is to provide a method for predicting the arch height after phakic intraocular lens implantation, thereby solving the above-mentioned problems.
[0012] The above-mentioned technical objective of the present invention is achieved through the following technical solution:
[0013] The first aspect of this invention provides a method for predicting the arch height after phakic intraocular lens implantation, comprising the following steps:
[0014] After obtaining the trained postoperative image prediction model, the superimposed loop algorithm is used. After inputting the preoperative image, clinical parameters, and PIOL refractive parameters, the PIOL diameter is enumerated and input, and the postoperative anterior segment OCT prediction image is output.
[0015] Identify the features of the postoperative anterior segment OCT prediction image and calculate the arch height value using the image features;
[0016] Determine if the arch height is within the safe range. If it is, output the recommended PIOL diameter, the corresponding predicted image, and the predicted arch height.
[0017] In conjunction with the first aspect, the present invention is further configured such that: the method for obtaining the trained postoperative image prediction model is as follows: obtaining an OCT image data sample set; constructing a postoperative image prediction model using a generative adversarial network; training and evaluating the postoperative image prediction model using the OCT image data sample set; and obtaining a trained postoperative image prediction model.
[0018] In conjunction with the first aspect, the present invention is further configured such that: the method for obtaining the OCT image data sample set is: to obtain OCT image data, to preprocess the OCT images, and to obtain the OCT image data sample set.
[0019] In conjunction with the first aspect, the present invention is further configured such that the preprocessing includes noise reduction, contrast enhancement, light beam elimination, and corneal segmentation.
[0020] In conjunction with the first aspect, the present invention is further configured such that the generative adversarial network includes a generative network and a discriminative network.
[0021] In conjunction with the first aspect, the present invention is further configured such that: the generating network takes OCT images, the refractive state of the operated eye, eye structural parameters, PIOL refractive parameters and diameter parameters as inputs, and outputs the postoperative anterior segment OCT image as the prediction target.
[0022] In conjunction with the first aspect, the present invention is further configured such that: the discriminant network takes preoperative images and postoperative predicted images or postoperative real images as inputs to learn to distinguish between real and fake postoperative images.
[0023] A second aspect of the present invention also provides an apparatus / device / system for calculating the arch height after intraocular lens implantation, comprising a memory, a processor, and a computer program stored in the memory, characterized in that the processor executes the computer program to implement the steps of any of the above methods.
[0024] A third aspect of the present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, characterized in that the computer program / instructions, when executed by a processor, implement the steps of any of the above methods.
[0025] A fourth aspect of the present invention also provides a computer program product, including a computer program / instructions, characterized in that the computer program / instructions, when executed by a processor, implement the steps of any of the above methods.
[0026] In summary, the present invention has the following beneficial effects:
[0027] 1. By employing multimodal fusion technology, high-resolution OCT images are precisely aligned and collaboratively analyzed with relevant text information, providing a more reliable and comprehensive decision-making basis for arch height prediction and personalized selection of intraocular lens diameter.
[0028] 2. By using the "graph prediction graph" method, we can make full use of the morphological features of ocular images and overcome the shortcomings of the previous parameter screening mode based on clinical prior knowledge, which may lead to the systematic omission of key morphological information.
[0029] 3. Improve the accuracy of postoperative arch height prediction after PIOL surgery, reduce the reliance on experience in selecting the diameter of the intraocular lens, thereby increasing the postoperative arch height target rate, reducing the probability of secondary surgery, and reducing the risk of complications. Attached Figure Description
[0030] Figure 1 This is a flowchart of the method for predicting the arch height after intraocular lens implantation in an embodiment of the present invention;
[0031] Figure 2 This is a flowchart of the method for obtaining a trained postoperative image prediction model in an embodiment of the present invention;
[0032] Figure 3 This is a schematic diagram of OCT image preprocessing in an embodiment of the present invention;
[0033] Figure 4 This is a schematic diagram of postoperative predictive image generation in an embodiment of the present invention;
[0034] Figure 5 This is a schematic diagram of image annotation in an embodiment of the present invention;
[0035] Figure 6 This is an example diagram of the arch height calculation before scale conversion in an embodiment of the present invention. Detailed Implementation
[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0037] Example 1:
[0038] Clinically, after implanting an intraocular lens (IOL) into a patient's eye via minimally invasive surgery, the postoperative arch height is typically used to assess the suitability of the implanted IOL. If the postoperative arch height is too low, it can lead to mechanical friction between the implantable collamer lens (ICL) and the eye's lens, as well as obstructed circulation of aqueous humor on the anterior surface of the lens, potentially inducing cataracts. Conversely, if the postoperative arch height is too high, it may cause problems such as disseminated pigmentary syndrome, iris atrophy, and acute angle-closure glaucoma. Postoperative arch height generally refers to the height from the center of the posterior surface of the ICL's optical zone to the anterior surface of the eye's lens.
[0039] To ensure an appropriate postoperative arch height after intraocular lens (IOL) implantation, it is often necessary to predict the postoperative arch height preoperatively. After predicting a suitable target postoperative arch height, an IOL corresponding to that target arch height is selected and implanted into the patient's eye. To predict the postoperative arch height after implantation of a target IOL in a target eye, this invention provides a method for predicting the postoperative arch height after IOL implantation in phakic eyes, such as... Figure 1 It includes the following steps:
[0040] S100. Obtain the trained postoperative image prediction model, superimpose the loop algorithm, input the preoperative image, clinical parameters, and ICL refractive parameters, then enumerate the ICL diameter input, and output the postoperative anterior segment OCT (Anterior Segment Optical Coherence tomography, AS-OCT) predicted image.
[0041] S200. Identify the features of the postoperative anterior segment OCT prediction image and calculate the arch height value using the image features;
[0042] S300. Determine if the arch height value is within the safe range. If the arch height value is within the safe range, output the recommended ICL diameter and the corresponding predicted image and predicted arch height value.
[0043] In this embodiment, a trained postoperative image prediction model is obtained, such as... Figure 2 As shown, the specific method is as follows:
[0044] S110. Obtain the OCT image data sample set;
[0045] S120. A postoperative image prediction model is constructed using a generative adversarial network. The postoperative image prediction model is trained and evaluated using an OCT image data sample set to obtain a well-trained postoperative image prediction model.
[0046] In this embodiment, the method for obtaining an OCT image data sample set is as follows: obtain OCT image data, preprocess the OCT images, and obtain an OCT image data sample set.
[0047] Preprocessing such as Figure 3 As shown, the specific steps are as follows:
[0048] ① Obtain OCT image data, crop the original anterior segment OCT image data into a square according to the short side length, and then adjust it to 512×512 pixels. Perform simple noise reduction, contrast enhancement, and light pillar elimination through algorithms.
[0049] ②The upper boundary of the cornea is obtained by segmenting the cornea using a self-trained corneal segmentation algorithm. The highest point of the cornea is used as the vertex, and a certain rectangular range is reserved. The rest is filled with black squares.
[0050] ③ Pair the images of the same patient from the same angle before surgery and 3 months after surgery, and perform image registration and alignment.
[0051] ④ Select and mark regions of interest in the image.
[0052] In this embodiment, the postoperative image prediction model is constructed, trained, and evaluated using generative adversarial network (GAN) technology to predict anterior segment OCT images three months postoperatively. Figure 4 As shown.
[0053] Model construction: Generative adversarial networks are used to build the model, which consists of a generator network and a discriminator network.
[0054] 1) Generative Network: The network takes the preoperative anterior segment OCT image, the refractive state of the operated eye, eye structural parameters (corneal curvature, corneal thickness, anterior chamber depth, corneal transverse diameter, axial length), and ICL parameters (ICL lens diameter, spherical power, cylindrical power, cylindrical axis, implantation angle) as inputs and outputs the anterior segment OCT prediction image 3 months after surgery.
[0055] 2) Discriminant Network: By using preoperative images and predicted or real images as input, the discriminant network learns to distinguish between real and fake postoperative images.
[0056] Model Training: The data is randomly divided into training, validation, and test sets in an 8:1:1 ratio. Mean Squared Error (MSE) is primarily used as the loss function during GAN training. The true value is assumed to be 1, and the false value to be 0. In the generation phase, the MSE of the generated network's predicted image and the real image, as well as the MSE of the predicted image's output vector through the discriminator network compared to the true value, are calculated to minimize the loss. In the discrimination phase, the MSE of the real post-operative image's output vector through the discriminator network compared to the true value, and the MSE of the generated network's predicted image's output vector through the discriminator network compared to the false value, are calculated separately to minimize the loss.
[0057] Model evaluation: The quality of the generated network's predicted images is evaluated using metrics such as SSIM, PSNR, and FID.
[0058] In this embodiment, the arch height is calculated as follows:
[0059] ①For example Figure 5 As shown, Labelme software was used to manually annotate some postoperative label images to train the ICL, lens segmentation network and anterior chamber angle vertex recognition network, ensuring that the features (ICL, lens segmentation network and anterior chamber angle vertex) of the above-mentioned postoperative anterior segment OCT prediction images are accurately identified by the recognition network.
[0060] ②The features of the image obtained above, such as Figure 6 As shown, the distance between the coordinates of the two intersection points of the perpendicular bisector of the line connecting the vertices of the left and right anterior chamber corners and the posterior surface of the ICL and the anterior surface of the natural lens, respectively, can be calculated by multiplying the distance by the scaling scale.
[0061] Assuming the coordinates of vertex A of the left anterior chamber corner are (x1, y1) and the coordinates of vertex B of the right anterior chamber corner are (x2, y2), then the equation of the perpendicular bisector of line segment AB is:
[0062]
[0063] If y1 = y2, then the equation of the perpendicular bisector is:
[0064]
[0065] The perpendicular bisector intersects the posterior surface of the ICL and the anterior surface of the natural lens, resulting in two intersection points, C and D, with coordinates (x3, y3) and (x4, y4), respectively.
[0066] The pixel distance of the central arch height (line segment CD) is vault_pixel = .
[0067] Actual central arch height
[0068]
[0069] Wherein, Vault_actual represents the actual arch height value, in micrometers;
[0070] Vault_pixel represents the pixel camber value measured on the standard resolution image, in pixels;
[0071] W represents the scanning width of the anterior segment region by the OCT device, in micrometers;
[0072] F represents the number of original sampling points of the OCT device in the direction of the scan width, in pixels.
[0073] In this embodiment, the recommended method for determining the ICL diameter is as follows:
[0074] A recurrent algorithm is superimposed on the basic model. After inputting preoperative images, clinical parameters, and ICL parameters (except for diameter), the model enumerates the ICL diameter (12.1mm, 12.6mm, 13.2mm, 13.7mm) and sequentially runs an adversarial generative network to generate postoperative images and corresponding arch height values, which are then output in tabular form.
[0075] By determining whether the arch height is within the safe range of 250-750mm, the recommended ICL diameter and corresponding predicted image and predicted arch height are output.
[0076] The present invention also provides a device / equipment / system for raising the arch of an intraocular lens after implantation, comprising a memory, a processor, and a computer program stored in the memory, characterized in that the processor executes the computer program to implement the steps of any of the above methods.
[0077] The present invention also provides a computer-readable storage medium having a computer program / instructions stored thereon, characterized in that the computer program / instructions, when executed by a processor, implement the steps of any of the above methods.
[0078] The present invention also provides a computer program product, including a computer program / instructions, characterized in that the computer program / instructions, when executed by a processor, implement the steps of any of the above methods.
[0079] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for predicting the arch height after intraocular lens implantation in phakic eyes, characterized by: Includes the following steps: After obtaining the trained postoperative image prediction model, the superimposed loop algorithm is used. After inputting the preoperative image, clinical parameters, and PIOL refractive parameters, the PIOL diameter is enumerated and input, and the postoperative anterior segment OCT prediction image is output. Identify the features of the postoperative anterior segment OCT prediction image and calculate the arch height value using the image features; Determine if the arch height is within the safe range. If it is, output the recommended PIOL diameter, the corresponding predicted image, and the predicted arch height.
2. The prediction method according to claim 1, characterized in that: The method for obtaining the trained postoperative image prediction model is as follows: Obtain a sample set of OCT image data; construct a postoperative image prediction model using deep learning algorithms such as generative adversarial networks; train and evaluate the postoperative image prediction model using the sample set of OCT image data to obtain a well-trained postoperative image prediction model.
3. The prediction method according to claim 2, characterized in that: The method for obtaining the OCT image data sample set is as follows: obtain OCT image data, preprocess the OCT images, and obtain the OCT image data sample set.
4. The prediction method according to claim 3, characterized in that: The preprocessing includes noise reduction, contrast enhancement, light beam elimination, and corneal segmentation.
5. The prediction method according to claim 2, characterized in that: The generative adversarial network includes a generator network and a discriminator network.
6. The prediction method according to claim 5, characterized in that: The generative network takes OCT images, the refractive state of the operated eye, eye structural parameters, and PIOL parameters (PIOL refractive parameters, diameter parameters, etc.) as inputs, and outputs the postoperative anterior segment OCT image as the prediction target.
7. The prediction method according to claim 5, characterized in that: The discriminant network takes preoperative images and predicted or real images as input to learn how to distinguish between real and fake postoperative images.
8. A device / equipment / system for raising the arch of an intraocular lens after implantation, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-7.
9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that... When the computer program / instruction is executed by the processor, it implements the steps of the method described in any one of claims 1-7.
10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1-7.