Method and system for determining intraocular lens parameters for ophthalmic surgery using simulated finite element analysis model

Abstract

Certain aspects of the present disclosure provide techniques for performing an ophthalmic surgical procedure, such as a cataract procedure. An example method of determining one or more intraocular lens (IOL) parameters to be used in a cataract surgical procedure includes using a fused machine learning model based at least in part on a first predicted lens behavior of each of the one or more IOLs obtained by a simulated finite element analysis (EFEA) model and by an IOL degree calculator (IPC) A machine learning model obtains a second predicted lens behavior of each IOL of the one or more IOLs to generate a recommendation including one or more IOL parameters of the IOL to be used in the cataract procedure.

Description

Technical Field

[0001] Various aspects of the present disclosure relate to ophthalmic surgery, and more particularly to using a simulated finite element (FEA) model to determine intraocular lenses (IOL) parameters for use during cataract surgery on a patient. As defined herein, IOL parameters include at least one of the type, size, and power of the IOL to be implanted in the patient's eye during cataract surgery. Background Art

[0002] Ophthalmic surgery generally covers a variety of procedures performed on the human eye. These surgical procedures can include cataract surgery, among other procedures. Cataract surgery is a procedure in which the crystalline lens, or natural lens, of a person's eye is removed and replaced with a synthetic lens (also known as an IOL) to correct vision problems caused by clouding of the natural lens.

[0003] Selecting the appropriate IOL for the patient is the key to achieving the best postoperative visual outcome. IOL has various types, degrees and sizes, and can be selected based on the measurement results of the anatomical parameters of the patient's eyes. The anatomical parameters of the human eye usually affect the IOL parameter selection made when planning and performing cataract surgery for the patient, and these anatomical parameters such as axial length (i.e., the distance between the anterior cornea and the retina), corneal thickness, anterior chamber depth (i.e., the distance between the anterior cornea and the anterior lens surface) and white to white diameter (white-to-white diameter) (i.e., the distance between the cornea and the sclera boundary on both sides of the eye). As defined herein, planning and performing cataract surgery includes determining the appropriate IOL parameters for improving patient vision. As an example, based on the measurement results of the patient's anatomical parameters, the surgeon can try to determine the IOL parameters that are likely to restore the patient's vision. Then, the surgeon selects the IOL whose parameters match the determined IOL parameters from a group of IOLs. Subsequently, the surgeon places or implants the selected IOL in the lens capsule of the patient.

[0004] In many cases, the measurements of anatomical parameters for a particular patient may fall within a known distribution (e.g., between some lower and upper limits, with a certain set percentage of patients falling within the range, such as a normal distribution of two standard deviations centered about a global mean, with approximately 95% of patients having measurements falling within the range), and therefore, planning and performing cataract surgery for such a patient may be a relatively simple task. However, if one or more anatomical parameters of a particular patient deviate from the known distribution or are otherwise abnormal (hereinafter "abnormal"), planning and performing cataract surgery for such a patient may be a more complex task. In addition, there may be some situations where a combination of anatomical parameters makes the treatment of the eye a complex task.

[0005] Therefore, an IOL power calculator (IPC) can be used to determine the IOL parameters for each specific patient. The IPC can provide a prediction of postoperative refractive outcomes based on a model or formula that takes measurements of the patient's anatomical parameters as input, and provides (multiple) predicted postoperative refractive outcomes for the patient for one or more different IOLs. Using the predictions from the IPC, the surgeon can attempt to select the IOL that has the greatest likelihood of restoring the patient's vision, such as the IOL with the smallest predicted postoperative refractive error.

[0006] Such IPC models or formulas have been developed for non-accommodating IOLs, but they may not perform well for accommodating IOLs. Accommodating IOLs differ from standard "static" IOLs in that they are able to change focal length. Accommodating IOLs may be fluid-filled, allowing the lens to dynamically change shape. Accommodating IOLs have flexible "arms" called haptics that use the movement of the eye muscles to change focus from distance to near. This adjustment allows incoming light rays to be properly focused on the retina. The eye accommodates when the eye is looking at a near object. During accommodation, the eye's ciliary muscles contract, causing the zonules to relax, thickening the shape of the natural lens. A thicker lens has a steeper curvature and is better able to focus incoming light rays from near objects onto the retina. When an accommodating IOL is implanted in a patient's eye, when the ciliary muscles contract, this causes the flexible haptics of the accommodating IOL to bend, causing the focal zone of the accommodating IOL to move forward. The forward position of the focal zone provides additional capability for near focusing or "accommodation." Current cataract planning and IPC(s) may not capture these complex interactions of the accommodating IOL with the components of the patient's eye, which may lead the surgeon to select IOL parameters that will result in an unsuccessful surgical outcome.

[0007]

[0006] Therefore, there is a need for techniques for accurately determining parameters of a patient's IOL based at least in part on measurements of anatomical parameters of the patient's eye. Summary of the invention

[0008] Certain embodiments provide a method for determining one or more IOL parameters of an intraocular lens (IOL) to be used in a cataract surgery procedure. The method generally includes: using one or more ophthalmic imaging devices to generate a plurality of data points associated with measurements of a plurality of anatomical parameters of an eye to be treated; using a machine learning model trained as a simulation finite element analysis (FEA) model to generate a first predicted lens behavior based at least in part on the plurality of data points associated with the measurements of the plurality of anatomical parameters of the eye to be treated and one or more IOL parameters of each of one or more IOLs; using an IOL power calculator machine learning model to generate a second predicted lens behavior based at least in part on at least a subset of the plurality of data points associated with the measurements of the plurality of anatomical parameters of the eye to be treated and one or more IOL parameters of each of one or more IOLs; and using a fused machine learning model to generate a recommendation including one or more IOL parameters of an IOL to be used in the cataract surgery based at least in part on the first predicted lens behavior and the second predicted lens behavior.

[0009] Aspects of the present disclosure provide apparatuses, devices, processors, and computer-readable media for performing the methods described herein.

[0010] To accomplish the foregoing and related ends, one or more aspects include the features fully described below and specifically pointed out in the claims. The following description and the accompanying drawings set forth in detail certain illustrative features of one or more aspects. However, these features are only indicative of a few of the various ways in which the principles of the various aspects can be employed. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Certain aspects of one or more embodiments are depicted in the accompanying drawings, and therefore should not be considered to limit the scope of the disclosure.

[0012] Figure 1 An example environment is depicted in which a simulated finite element analysis (EFEA) model, an IPC machine learning (ML) model, and a fused ML model are trained and deployed to generate a recommendation, including recommended IOL parameters, for cataract surgery for a patient based at least on measurements of anatomical parameters of the patient in accordance with certain aspects described herein.

[0013] Figure 2 is a graph of the model eye and eye properties.

[0014] Figure 3 is a flow chart illustrating example operations for training an EFEA model according to certain aspects described herein.

[0015] Figure 4Demonstrates training of an ensemble ML model using an EFEA model and an IPC ML model according to certain aspects described herein.

[0016] Figure 5 Example operations are presented that may be performed by a computing device to generate and output a recommendation including IOL parameters for cataract surgery for a patient based on anatomical parameters of the patient in accordance with certain aspects described herein.

[0017] Figure 6 An example computing device is shown in which embodiments of the present disclosure may be performed in accordance with certain aspects described herein.

[0018] To facilitate understanding, identical reference numerals have been used, where possible, to designate identical elements that are common to the figures. It is contemplated that elements and features of one embodiment may be beneficially incorporated in other embodiments without further recitation. DETAILED DESCRIPTION

[0019] As discussed above, cataract surgery is a surgical procedure to replace a defective natural lens with an IOL. Typically, a defective natural lens is a lens that has developed cataracts, which are clouding of the natural lens that negatively affects the patient's vision (e.g., causing the patient to see faded colors, blurred vision or double vision, see halos around point light sources, or other negative effects). An IOL can be selected to replace the patient's natural lens in order to restore or at least improve the patient's vision. For an IOL to be implanted in a patient's eye, the determination of a set of IOL parameters is affected by the measurement results of the patient's anatomical parameters. More specifically, as discussed above, based on the characteristics and / or measurement results of the anatomical parameters of the patient's eye, the surgeon can try to determine a set of IOL parameters that have a high probability of restoring the patient's vision. For example, some IOLs may allow optimization of myopia or hyperopia, while other IOLs may be used to compensate for the patient's natural corneal astigmatism, and so on.

[0020] Accommodative IOLs differ from non-accommodative IOLs in that accommodative IOLs are able to dynamically change shape in order to change focal length to accommodate the movement of the patient's eye. In the human eye, circular ciliary muscles surround the lens. The ciliary zonules connect the ciliary muscles to the lens capsule that encapsulates the lens. When the eye looks at a distant object, the ciliary muscles relax, causing the ciliary zonules to become taut, which makes the curve of the lens flatter, thereby better able to focus incoming light from distant objects onto the retina. In young eyes, accommodation is essentially instantaneous and effortless. With age, the lens becomes less flexible, resulting in near vision loss, which is a sign of presbyopia in people over the age of forty, for example. Such patients may be good candidates for the use of accommodative lenses. In the case of an accommodative IOL implanted in the patient's eye, when the patient's ciliary muscles contract, this causes the flexible haptics of the accommodative IOL to bend, moving the focal zone of the accommodative IOL forward, thereby increasing the focusing power.

[0021] As discussed above, the interaction of an accommodating IOL with a patient's eye can be complex and not taken into account by existing lens fitting procedures and IPCs. Accordingly, aspects presented herein provide systems and methods in which the interaction of a patient's eye with an IOL is modeled using FEA to better predict IOL behavior, thereby predicting postoperative outcomes and recommending IOL parameters for the patient. In particular, FEA can be used to train EFEA to inform the selection of an accommodating IOL for a patient based on the patient's anatomical characteristics and / or measurements.

[0022] Finite element analysis can be used to form predictions for both static and accommodative IOLs. FEA uses a numerical technique known as the finite element method (FEM) to simulate physical phenomena (e.g., the interaction of an IOL in the human eye). FEA can involve creating a grid that is composed of up to millions of smaller elements (e.g., components of the eye) that together form the shape of a structure (e.g., a patient's eye). Calculations are performed on each element, and the individual results are combined to provide the final result of the structure. FEM uses partial differential equations (PDEs) and / or matrices to describe complex physical behaviors, such as structures, fluids, and thermal behaviors (e.g., the behavior of an accommodative IOL in a patient's eye). These PDEs or matrices can be solved to calculate relevant quantities in order to estimate the behavior of an IOL under various conditions.

[0023] FEA can be used to model how an accommodating IOL will behave after it is implanted in the patient's capsular bag to understand how different geometries of the lens capsule affect the dynamics of the accommodating IOL in order to predict the patient's postoperative outcome for a given IOL. According to an embodiment of the present disclosure, a FEA model is designed and solved using preoperative measurements of one or more specific patients and one or more lens designs to predict the behavior of one or more lenses after implantation in the patient's eye. The FEA model can be fine-tuned by comparing the predictions output by the FEA model with clinical data over a large design space of patient preoperative measurements and IOL designs to predict how the IOL power of different sized fluid-accommodating IOLs will change when they are implanted in different capsular bags with different geometries in different patients' eyes.

[0024] However, FEA models may be computationally expensive and may take, for example, up to twenty-four hours or even longer to run. In addition, FEA models may require experts to generate, set up, and run the FEA models.

[0025] Therefore, aspects of the present disclosure provide for the training and use of simulated FEA models. The use of simulated FEA models enables a faster and more efficient method of using FEA modeling to predict postoperative outcomes and recommend IOL parameters for patients. In some embodiments, the simulated FEA model is trained to match or closely match the output of the FEA model. Compared to the FEA model, the trained simulated FEA model can have lower computational costs at runtime (e.g., the model can run in seconds) and does not require an expert to operate. Therefore, the simulated FEA can be portable and easily deployed in a clinical environment for clinicians or surgeons to fit patients with IOLs suitable for implantation.

[0026] In some embodiments, an IPC is used in addition to a simulated FEA model to provide a patient with a predicted postoperative outcome for a given set of IOLs having different IOL powers, types, and / or sizes. The IPC can be a standard IPC (e.g., Bartlett IPC, Haigis formula IPC, Hill-RBF IPC, SRK (Sanders-Retzlaff, Kraff) IPC, etc.). In some other embodiments, the IOL power calculator is an IOL power calculator based on machine learning (ML). As discussed in more detail herein, an ML-based IPC can be trained on a large design space of patient preoperative measurements and IOL designs to predict postoperative refractive errors of various accommodative IOLs for each patient based on the patient's anatomical measurements and / or characteristics.

[0027] In some embodiments, a fusion model is generated and trained using a simulated FEA model and an ML-based IPC model. In some embodiments, the fusion model is also an ML-based model. The fusion model uses the outputs from both the simulated FEA model and the ML-based IPC model as inputs to the fusion model, and then outputs recommended IOL parameters. The fusion model can be trained with inputs from the simulated FEA model and the IOL power calculator to make recommendations for IOL parameters. These recommendations can be compared with clinical data to improve the fusion model.

[0028] Example computing environment for ophthalmic surgical procedure planning using the EFEA model

[0029] Figure 1 An example computing environment is shown in which models are trained and used to generate recommendations including IOL parameters for cataract surgery for a patient. Typically, these models can be trained using a training data corpus including records corresponding to historical patient data, and deployed for cataract surgery planning for a current patient, including generating IOL parameters. As defined herein, a new patient or current patient (hereinafter referred to as "current") is typically a patient undergoing cataract surgery to replace a defective natural lens. As discussed in further detail below, the recommended IOL parameters for the current patient can be generated by a fusion model trained to generate recommended IOL parameters.

[0030] The historical patient data for each historical patient may include demographic information of the patient, recorded data points associated with measurements of anatomical parameters of the patient, desired outcomes, actual treatment data (such as actual IOL parameters of an IOL implanted in the patient's eye), or other information about the historical patient's treatment and treatment outcome data (e.g., postoperative refractive error and parameters indicating whether the historical patient was satisfied or dissatisfied with the treatment (e.g., a patient satisfaction score)). Note that in this context, the actual treatment performed on the patient may be different from the recommended treatment.

[0031] By using these models, a large amount of historical patient data can be used to generate recommended IOL parameters for the current patient. To some extent, this large amount of historical patient data indicates the expertise and previous experience of other surgeons who have performed similar operations on similar patients. By using the systems and methods described herein, for the current patient, the surgeon can use this large amount of historical patient data to determine the IOL parameters that will optimize the surgical results of the current patient. Therefore, the technology herein improves the medical field of view by allowing the selection of better IOL parameters, thereby improving vision after the placement of the IOL (such as during cataract surgery).

[0032] A variety of techniques can be used to train and deploy a model that generates the IOL parameters for the current patient. Figure 1 An example deployment is shown in , however, other deployments are also contemplated. For example, Figure 1 A deployment is shown where the model is trained on a remote server 120 and deployed to a user console 130 used by a surgeon during cataract surgery planning. In another example deployment, the model can be trained and deployed on a remote server 120, which can be accessed, for example, by a computing system, an imaging device, and / or a surgical console. In another example deployment, the model can be trained on a remote server 120 and deployed to an imaging device 110 used preoperatively and / or intraoperatively. However, it should be appreciated that various other techniques for training and deploying a model that generates IOL parameters for a current patient are contemplated, and Figure 1 The deployments shown in are non-limiting illustrative examples.

[0033] Figure 1 An example computing environment 100 is shown in which one or more imaging devices 110, a server 120, a user console 130, and a historical patient data repository 140 are connected via a network to train one or more models to generate recommended IOL parameters for a current patient based at least in part on data points associated with measurements of anatomical parameters of the current patient as provided by the one or more imaging devices 110.

[0034] The imaging device(s) 110 generally represent various devices that can generate data points associated with one or more measurements of anatomical parameters of a patient's eye.In this document, anatomical parameters of the eye may refer to a set of optical parameters. Figure 2 is a diagram of a model eye 200 illustrating various optical parameters according to certain aspects described herein. Figure 2As shown, the optical parameters may include an axial length of the eye (e.g., a distance from the anterior corneal surface 202A of the cornea 202 to the retina 204), a central corneal thickness (CT) measurement of the cornea 202, an anterior chamber depth (AD) of the anterior chamber 208 (e.g., a distance from the top of the posterior cornea 202P to the top of the anterior lens surface 206A of the lens 206), one or more lens characteristic dimensions (such as, but not limited to, a lens thickness (LT) of the lens 206, a lens diameter (LD) of the lens 206 (also known as a lens equatorial diameter), a lens volume of the lens 206, a lens surface area of ​​the lens 206, an anterior corneal surface area of ​​the lens 206, a lens ... The patient's eye may include, but is not limited to, the curvature of the anterior corneal surface 202A, the curvature of the posterior lens surface 206P, the white-to-white diameter (WD) (e.g., the distance between the corneal or scleral borders on each side of the eye), the anterior chamber depth (e.g., the distance between the anterior vertex of the cornea 202 and the anterior vertex of the lens 206) or other lens characteristic dimensions), the curvature and astigmatism of the anterior corneal surface 202A, the shape of the anterior corneal surface 202A, the depth of the vitreous humor 210, and other optical parameters associated with the patient's eye, including the patient's lens characteristic dimensions, lens capsule geometry, ciliary muscles 212, and ciliary zonules (e.g., but not limited to ciliary extent, ciliary processing torque, ciliary length, etc.).

[0035] These optical parameters are examples of anatomical parameters measured by the imaging device(s) 110 , however, the imaging device(s) 110 may measure additional anatomical parameters that may be used in the model.

[0036] In some embodiments, data generated by imaging device 110 for each patient can be stored in repository 140 as patient data, which can include: raw measurements of anatomical parameters of the patient, biometric information derived from the measurements, and / or imaging data including two-dimensional cross-sectional images showing the cornea, iris, lens, and retina; three-dimensional images of the eye; two-dimensional topographic maps of the eye; or other types of imaging data. In general, any number of imaging devices 110 can be included in computing environment 100 and can be used to generate different types of data that can be used as input to one or more models for generating recommended IOL parameters.

[0037] The (multiple) imaging devices 110 may include (1) preoperative measurement and / or imaging devices used in the clinic and / or (2) intraoperative measurement and / or imaging devices. In one example, one of the (multiple) imaging devices 110 may be an optical coherence tomography (OCT) device that can generate optical imaging. Various types of OCT devices may be used. As an example, the OCT device may be used to generate a two-dimensional cross-sectional image of the current patient's eye, from which measurement results of various anatomical parameters may be derived. The two-dimensional cross-sectional image may show the position of the cornea, lens, and retina on a two-dimensional plane (e.g., the cornea is on one side of the two-dimensional cross-section, and the back of the retina is on the other side of the two-dimensional cross-section). Based on the two-dimensional cross-sectional image of the current patient's eye, the OCT device may derive various measurement results.

[0038] For example, an OCT device can generate axial length measurements, central corneal thickness measurements, anterior chamber depth measurements, lens thickness measurements, and other related measurements from a cross-sectional image. In some aspects, an OCT device can generate one-dimensional data measurements (e.g., from a center point) or can generate three-dimensional measurements, from which additional information such as tissue thickness maps can be generated. Examples of OCT devices are further described in detail in U.S. Patent No. 9,618,322, which discloses "Process for Optical Coherence Tomography and Apparatus for Optical Coherence Tomography" and U.S. Patent Application Publication No. 2018 / 0104100, which discloses "Optical Coherence Tomography Cross View Image," both of which are hereby incorporated by reference in their entirety.

[0039] Another one of the (multiple) imaging devices 110 can be a keratometer. Typically, the keratometer can reflect a light pattern (such as a circle of illuminated dots) from the current patient's eye and capture the reflected light pattern. The keratometer can perform image analysis on the reflected pattern (relative to the pattern reflected from the current patient's eye output by the imaging device 110) to measure or otherwise determine the values ​​of various anatomical parameters. These anatomical parameters can include, for example, curvature information and astigmatism information of the anterior corneal surface of the current patient's eye. For example, the curvature information can be a general curvature measurement result, a maximum curvature and axial line information identifying an axis along which the maximum curvature occurs, and a minimum curvature and axial line information identifying an axis along which the minimum curvature occurs.

[0040] Yet another of the imaging devices 110 may be a topographic device that measures the topography of the anterior corneal shape. The topographic device may use analysis of reflected light patterns distributed over an area of ​​the cornea to generate a detailed surface profile map relative to a base contour. For example, since the cornea is generally spherical or approximately spherical, the surface profile map may show deviations from the base contour, with different colors representing the amount of deviation from the base contour at any discrete point along the cornea.

[0041] Another example imaging device 110 may be an intraoperative aberrometer. An example of an intraoperative aberrometer is a TM Ora TM (Alcon, Switzerland), which is described in more detail in part in co-owned U.S. Patent Nos. 7,883,505, disclosing "Integrated Surgical Microscope and Wavefront Sensor," and 8,784,443, disclosing "Real-Time Surgical Reference Indicium Apparatus and Methods for Astigmatism Correction," each of which is incorporated herein by reference in its entirety. Among other measurements, Ora TM Total eye refraction measurements that account for total eye astigmatism (including surgically induced astigmatism and posterior corneal astigmatism) as well as myopic post-PRK / LASIK and long-short eye measurements can be captured to provide guidance for adjusting lens selection and placement for all eye types.

[0042] The imaging device(s) 110 may also include a rotating camera (e.g., a Scheimpflug camera), a magnetic resonance imaging (MRI) device, an ophthalmoscope, an optical biometer, a three-dimensional stereo digital microscope (e.g., 3D visualization system (Alcon Inc., Switzerland).

[0043] The server 120 generally represents a single computing device or a cluster of computing devices on which a training data set can be generated and used to train one or more models to generate recommended IOL parameters. The server 120 is communicatively coupled to a historical patient data repository 140 (hereinafter referred to as "repository 140") that stores historical patient records. In some embodiments, the repository 140 can be or include a database server that receives information from the server 120, the user console 130, and / or the imaging device 110 and stores the information in a corresponding patient record in a structured and organized manner.

[0044] In certain aspects, each patient record in the repository 140 includes information such as the patient's demographic information, data points associated with the measurements of anatomical parameters, actual treatment data associated with the patient's cataract surgery, and treatment outcome data. For example, the demographic information of each patient includes the patient's age, gender, race, etc. The data points associated with the preoperative and / or intraoperative measurements of anatomical parameters may include raw data generated by an OCT device, a keratometer, a topography device, etc. or measurements obtained from or provided by the OCT device, a keratometer, a topography device, etc., as discussed above. As further described below, the actual treatment data includes the actual IOL parameters (e.g., IOL type, IOL power, IOL size) of the IOL used for the patient, as well as any additional relevant information related to the patient's treatment. For example, the actual treatment data may indicate the method of performing cataract surgery for the patient, the tools used for treatment, and other information about the specific procedures performed during the operation. Each patient record also includes treatment outcome data, which may include various data points indicating outcome parameters, such as patient satisfaction with treatment (e.g., a binary indication of satisfaction or dissatisfaction with the outcome of the surgery), measured vision level after treatment, etc.

[0045] Server 120 uses these historical patient records to generate data sets for training models that can recommend IOL parameters to surgeons to treat current patients. Figure 1 As shown, the server 120 includes one or more training data generators 122 (hereinafter referred to as "TDG 122") and one or more model trainers 124. The TDG 122 obtains data from the repository 140 to generate a data set for use by the (multiple) model trainers 124 to train the FEA model 125, the EFEA model 126, the IPC ML model 127 and / or the fusion model 128. It should be understood that the FEA model 125, the EFEA model 126, the IPC ML model 127 and the fusion model 128 can be trained with the same, overlapping or different data sets. Further, the FEA model 125, the EFEA model 126, the IPC ML model 127 and the fusion model 128 can be trained by different model trainers 124. Figure 3 The generation and training of the FEA model 125 and the EFEA model 126 are discussed in more detail. Figure 4 The generation and training of the IPC ML model 127 and the fusion model 128 are discussed in more detail.

[0046] Model trainer(s) 124 include or refer to one or more algorithms configured to train FEA model 125, EFEA model 126, IPC ML model 127, and fusion model 128 using a training data set. In some embodiments, the trained model refers to a function for making IOL-related predictions, such as with weights and parameters. The IOL-related predictions may include predicted postoperative refractive error, predicted optimal IOL parameters, and IOL power per frame.

[0047] The model trainer(s) 124 may use one or more ML algorithms to train one or more of the EFEA model 126, the IPC ML model 127, and the fusion model 128. ML algorithms may generally include supervised learning algorithms, unsupervised learning algorithms, and / or semi-supervised learning algorithms. Unsupervised learning is an ML algorithm for making inferences from data sets that consist of input data without labeled responses. Supervised learning is an ML task of learning a function that maps inputs to outputs, for example, based on example input-output pairs. Supervised learning algorithms generally include regression algorithms, classification algorithms, decision trees, various types of neural networks, and the like.

[0048] In some embodiments, the model trainer(s) 124 may train deep learning models. These deep learning models may include, for example, convolutional neural networks (CNNs), adversarial learning algorithms, generative networks, or other deep learning algorithms that can learn relationships in a data set that may not be clearly defined in the data used to train such models. The deep learning model may, for example, map inputs to different neurons in one or more layers of the deep learning model (e.g., where the model is generated using a neural network), where each neuron in the model represents new features learned over time in the internal representation of the input. These neurons may then be mapped to outputs representing recommended IOL parameters, as discussed above.

[0049] After training the models, the model trainer(s) 124 may deploy one or more of the trained models to the user console 130 for use in predicting and recommending IOL parameters for the current patient. In some embodiments, the EFEA model 126, IPC ML model 127, and fusion model 128 are deployed on the user console 130, but the FEA model 125 is not deployed. For example, as discussed herein, the FEA model 125 may be computationally expensive and may require an expert to generate and run. In contrast, as described below with respect to Figure 3 As discussed in more detail, the FEA model 125 can be used to generate a computationally less expensive and operationally simpler EFEA model 126, which can then be deployed on a user console 130. As described above, Figure 1Only one example of how the EFEA model 126, IPC ML model 127, and fused model 128 may be deployed is shown. In other examples, the EFEA model 126, IPC ML model 127, and fused model 128 may be deployed or executed at the server 120. In yet other examples, the model trainer(s) 124 may deploy the EFEA model 126, IPC ML model 127, and fused model 128 to one or more of the imaging devices 110.

[0050] The user console 130 generally represents a computing device or system that is communicatively coupled to the server 120, the repository 140, and / or the imaging device(s) 110. In some embodiments, the user console 130 may include or be associated with a desktop computer, a laptop computer, a tablet computer, a smart phone, or other computing device(s). For example, the user console 130 may include or be associated with a computing system used in a surgeon's office or clinic. In another example, the user console 130 may be a surgical console used by a surgeon in an operating room to perform cataract surgery on a current patient.

[0051] exist Figure 1 In the example of FIG. 1 , the trained model is deployed by server 120 to user console 130 for use in predicting IOL parameters for the current patient that will optimize the patient's surgical outcome. As shown, user console 130 includes a treatment data recorder (TDR) 132 and an IOL recommendation generator (IRG) 134. Note that although in Figure 1 In the embodiment, TDR 132 and IRG 134 are executed on the user console 130, but in certain other embodiments, TDR 132 and IRG 134 may be executed on one or more other computing systems (such as server 120, imaging device 110 and / or any computing system capable of communicating with one or more of server 120, imaging device 110 and / or historical patient data repository 140).

[0052] IRG 134 generally refers to a software module or a set of software instructions or algorithms, including EFEA model 126, IPC ML model 127, and fusion model 128, that takes a set of inputs about a current patient and generates recommended IOL parameters as output. In some embodiments, IRG 134 is configured to receive the set of inputs from at least one of repository 140, imaging device(s) 110, a user interface of user console 130, and other computing devices that a medical team may use to record information about the current patient. In some embodiments, IRG 134 outputs the recommended IOL parameters to a display device communicatively coupled to user console 130, prints the recommended IOL parameters, generates and transmits one or more electronic messages (including the recommended IOL parameters) to a destination device (e.g., a connected device such as a tablet computer, a smart phone, a wearable device, etc.), etc.

[0053] The set of inputs includes data points associated with measurements of anatomical parameters of the new patient's eye as provided by the imaging device(s) 110. These data points are provided to the user console 130 for use as input to the model, and may also be provided to the server 120 and / or repository 140.

[0054] Prior to or during surgery, the user console 130 retrieves (e.g., from the repository 140 or from temporary storage at the user console 130) data points associated with preoperative and / or intraoperative measurements of anatomical parameters and other patient information associated with the current patient (e.g., demographic information of the current patient, etc.) to use as inputs to the models stored at the user console 130. These data points and other patient information may be provided to the IRG 134, which runs the inputs through the EFEA model 126, the IPC ML model 127, and the fusion model 128 to generate recommended IOL parameters. The user console 130 then outputs the recommendations generated by the one or more models.

[0055] The TDR 132 receives or generates treatment data regarding the treatment provided to the current patient. As described above, the treatment data may include the actual IOL parameters used for the current patient and any additional relevant information, such as the method of performing cataract surgery, etc. As previously defined, the actual IOL parameters refer to the type, power, and size information of the IOL actually implanted by the surgeon into the eye of the current patient. In the event that the surgeon does not follow the recommended IOL parameters, the actual IOL parameters will be different from the recommended IOL parameters. In some such cases, the TDR 132 may receive the treatment data as user input to a user interface of the user console 130. In the event that the surgeon follows the recommended IOL parameters, the actual IOL parameters will be the same as the predicted IOL parameters. In some such cases, the TDR 132 treats the recommended IOL parameters as the actual IOL parameters recorded as part of the treatment data. Figure 1 In an exemplary embodiment, TDR 132 transmits the actual IOL parameters to repository 140 and / or server 120. Repository 140 then augments the current patient's record with the actual IOL parameters used for treatment.

[0056] TDR 132 generally allows a user of user console 130 to provide post-operative information identifying the surgical outcome of the treatment. Although TDR 132 is illustrated as executing on user console 130, those skilled in the art will appreciate that TDR 132 may execute on a computing device separate from user console 130.

[0057] TDG 122 continues to enhance the training data set with information related to the patient for whom the deployed model provides recommended IOL parameters. In some aspects, TDG 122 augments the (multiple) data sets whenever information about a new (i.e., current) patient becomes available. In some other aspects, TDG 122 augments the (multiple) data sets with a batch of new patient records, which may be more resource efficient. TDG 122 may convert the records in repository 140 (including data points associated with the measurement results of one or more anatomical parameters of the current patient, actual treatment data, and treatment outcome data) into new samples in the training data set. (Multiple) model trainers 124 use the new training data set to retrain one or more models. More generally, whenever a new (i.e., current) patient is treated, information about the new patient may be saved in repository 140 for TDG 122 to supplement the (multiple) training data sets, and (multiple) model trainers 124 use (multiple) supplemented training data sets to retrain one or more models.

[0058] Example Method for Generating FEA Models for Simulation

[0059] Figure 3is a flow chart illustrating example operations 300 for training a finite element analysis model (e.g., EFEA model 126) for simulation according to certain aspects described herein. Operation 300 may be performed by Figure 1 One or more model trainers 124 shown in are executed.

[0060] As shown, operation 300 begins at box 302 by generating a FEA model. In some embodiments, the FEA model 125 is generated by the model trainer 124. Generating the FEA model can include generating a mathematical model of the interaction of components of a human eye (e.g., model eye 200) with an IOL (such as an accommodating lens). In some embodiments, the FEA model includes a set of PDEs and / or matrices that can be solved to calculate one or more quantities associated with the interaction of components of an accommodating IOL and a human eye. The variables of the mathematical model include preoperative anatomical parameters of the eye. For example, these variables can include any anatomical parameters measured by (multiple) imaging devices 110 discussed above. In particular, the anatomical parameters include, but are not limited to, the parameters associated with the eye discussed above, lens feature dimensions, and lens capsule geometry. The lens feature dimensions can be used to define the complete preoperative contour of the lens capsule. The use of additional anatomical parameters can increase the accuracy of the FEA model.

[0061] In some embodiments, the FEA model can be solved for one or more sets of anatomical parameters to generate data indicating how the IOL may behave in the lens capsule in view of the different parameters and characteristics of the IOL. The FEA model also takes as input one or more parameters of the IOL. For example, the IOL type, size, and / or label power can be input to the FEA model along with each set of anatomical parameters. As used herein, the IOL label power can be the preoperative power of the IOL.

[0062] For an input set of patient preoperative anatomical parameters and IOL parameters, the FEA model predicts the postoperative behavior of each input IOL after implantation in the patient's capsular bag. In some embodiments, the FEA model predicts the postoperative resting state and effective lens position of the IOL (e.g., the postoperative position of the IOL relative to the cornea and retina), and thereby predicts the refractive outcome of the IOL. For example, based on the postoperative effective lens position, the degree required to ensure that light is properly focused on the retina can be determined. The postoperative resting state of the IOL can be determined (e.g., by the haptics when the haptics bend in response to pressure), and the pressure applied to the IOL changes the shape and size of the IOL, which in turn affects the IOL power. Therefore, by predicting the postoperative resting state and effective lens position by the FEA model, the FEA model can predict the size adjustment factor of the IOL and the postoperative refractive outcome. As described with respect to Figure 3As discussed in more detail, the FEA model can output a predicted IOL power for each frame. As used herein, a frame refers to a measurement instance in time. Therefore, the FEA model can output predicted IOL powers at different time points, thereby modeling how an accommodating IOL interacts with a patient's capsular bag after being implanted in the patient's eye.

[0063] As shown, operation 300 can continue at box 304 by fine-tuning the FEA model with clinical data. As shown, at box 304, fine-tuning the FEA model with clinical data includes inputting historical patient anatomical parameters and one or more IOL parameters into the FEA model to predict IOL behavior based on the input at box 306. In some embodiments, the historical patient anatomical parameter inputs to the FEA model include values ​​for each anatomical parameter used to construct the FEA model. In some embodiments, the patient anatomical parameter inputs to the FEA model include values ​​for a subset of anatomical parameters used to construct the FEA model. In an illustrative embodiment, the historical patient anatomical parameters include preoperative lens feature dimensions of historical patients, and the IOL parameters are one or more IOL label powers. Based on these inputs, the FEA model outputs predicted IOL behavior. In an illustrative embodiment, the predicted IOL behavior includes the IOL power per frame.

[0064] As shown, at box 304, fine-tuning the FEA model with clinical data includes comparing the clinical data for the same anatomical parameters with the IOL behavior predicted by the FEA model at box 308. For example, in the case where the FEA model 125 outputs a predicted IOL power per frame for a set of anatomical parameters and IOL parameters input, the predicted IOL power per frame is compared with clinical results (e.g., treatment results in the repository 140), which may include actual IOL power per frame for patients with the same set of anatomical parameters and the same IOL parameters. At box 310, the FEA model is adjusted based on the comparison. In some embodiments, the FEA model is adjusted by changing one or more parameters of the FEA model. The parameters of the FEA model may include, but are not limited to, zonular tension, zonular dynamics, capsular elasticity, friction, and / or other parameters of the FEA model.

[0065] The FEA model 125 may be fine-tuned across many sets of anatomical parameters and IOL parameters to find an FEA model that accurately predicts IOL behavior. As used herein, an FEA model that accurately predicts IOL behavior may predict IOL behavior at a specified success threshold level. In some embodiments, the FEA model 125 is fine-tuned across the entire range of anatomical values ​​expected to be seen in the clinic, or a larger range.

[0066] Return to reference Figure 3After fine-tuning the FEA model, the operation 300 continues at block 312 by generating a simulated FEA model (e.g., the EFEA model 126). The EFEA model is generated to match (e.g., or closely match within a specified threshold) the FEA model. In some embodiments, the EFEA model is a machine learning model.

[0067] Operation 300 continues at block 314 by training the EFEA model to replicate the output of the FEA model. In some embodiments, the EFEA model 126 is composed of Figure 1 . In some embodiments, the model trainer 124 refers to an AI / ML learning algorithm or a combination of AI / ML learning algorithms for training an AI / ML model. Examples of AI / ML learning algorithms include various types of optimization algorithms, such as gradient descent, stochastic gradient descent, nonlinear conjugate gradient, etc. The EFEA model 126 can be a Gaussian process regression (GPR) model, a linear regression model, an autoregressive integrated moving average (ARIMA) model, a neural network, etc. For input values ​​that run the FEA model 125, the EFEA model 126 is expected to return the same output value with zero uncertainty. For input values ​​that do not run the FEA model 125, the EFEA model 126 predicts what the FEA model 125 will output and the estimated uncertainty.

[0068] like Figure 3 As shown in FIG. 1 , training the EFEA model at block 314 to replicate the output of the FEA model includes inputting historical patient anatomical parameters and one or more IOL parameters to the EFEA model at block 316 to predict IOL behavior based on the inputs. To train the EFEA model 126 to simulate the FEA model 125, the input parameters are the input parameters run by the FEA model 125.

[0069] Training the EFEA model to be consistent with the FEA model at box 314 further includes comparing the predicted IOL behavior of the EFEA model with the predicted IOL behavior of the FEA model with the same input at box 316. As discussed above, in some embodiments, the IOL behavior predicted by the FEA model 125 and the EFEA model 126 is the IOL power per frame. Therefore, in some embodiments, the EFEA model 126 can be trained to reproduce the IOL power curve per frame predicted by the FEA. Training the EFEA model to be consistent with the FEA model at box 314 further includes adjusting the EFEA model (e.g., adjusting the weights of the model) based on the comparison at box 318.

[0070] The TDG 122 may generate a training data set (or multiple training data sets) for training the EFEA model 126. The training data set for the EFEA model 126 may include mapping demographic information and / or data points associated with measurements of anatomical parameters of each of a plurality of historical patients to corresponding predictions of IOL behavior by the FEA model 125. The model trainer 124 trains the EFEA model 126 based on the (multiple) training data sets. For example, the model trainer 124 trains the EFEA model 126 to generate predicted IOL behavior (e.g., IOL power per frame) for given inputs (including anatomical parameters and (multiple) IOL parameters). The EFEA model 126 then provides predicted IOL behavior based at least in part on the inputs. The repository 140 may include records of IOL behavior predicted by the FEA model 125 for the same set of inputs. The predicted IOL behavior provided by the EFEA model 126 can be compared to a record of the outputs provided by the FEA model 125 for the same set of inputs to minimize the loss between the predictions provided by the EFEA model 126 and the outputs provided by the FEA model 125. Once the EFEA model 126 is trained to the point where it returns predictions that are the same or nearly the same (e.g., within an acceptable error margin) as the outputs provided by the FEA model 125 for the same set of inputs, the EFEA model 126 is able to extrapolate and provide predicted IOL behavior even for inputs that were not run through the FEA model 125. Thus, training the EFEA model 126 according to the embodiments described herein addresses technical deficiencies in predicting IOL behavior in the lens capsule using FEA models by significantly reducing the amount of resources (e.g., computational resources) that would otherwise have to be utilized to predict IOL behavior using the FEA model over a wide range of inputs.

[0071] As discussed above, the FEA model 125 can be computationally expensive and require an expert to generate, set up, and run. However, the EFEA model 126 can return a prediction in less than a second with minimal software and hardware requirements. Therefore, according to an embodiment of the present disclosure, the EFEA model 126 is used instead of the FEA model 125 as part of the IOL recommendation generator 134 to generate recommended IOL parameters. Therefore, after the EFEA model 126 is trained, the EFEA model 126 can be deployed to one or more server computers, user consoles, imaging devices having a computing device integrated therein, etc. For example, Figure 1 As shown in , the EFEA model 126 may be deployed to a user console 130 .

[0072] Example method for training an ensemble model

[0073] According to an embodiment of the present disclosure, in addition to an IPC ML based model (eg, IPC ML model 127 ), an EFEA (eg, EFEA model 126 ) may be used to train a fusion model to generate predicted IOL behaviors and / or recommended IOL parameters.

[0074] Figure 4 The use of the EFEA model 126 and the IPC ML model 127 to train the fusion model 128 according to certain aspects described herein is shown. Figure 3 Training of the EFEA model 126 is discussed. To train the fusion model 128, the EFEA model 126 can be run, and the output of the EFEA model 126 can be used as an input to the fusion model 128. The EFEA model 126 uses input 402 to generate predicted IOL behavior. The input 402 of the EFEA model 126 includes patient anatomical parameters 404 and IOL parameters 406. As discussed above, the patient anatomical parameters 404 that can be input to the EFEA model 126 can include all or a subset of the anatomical parameters used to generate the FEA model 125. The IOL parameters 406 can include IOL type, IOL size, and / or IOL label power. Based on the input, the EFEA model 126 outputs predicted IOL behavior. As discussed above, the predicted IOL behavior can include predicted postoperative refractive error, predicted postoperative effective lens position, and / or predicted IOL power frame (which can include predicted preoperative, intraoperative, and postoperative IOL powers). In the illustrative embodiment, the patient anatomical parameters 604 include lens characteristic dimensions; the IOL parameters 406 include at least an IOL label power for one or more IOLs; and the predicted IOL behavior includes at least a predicted per-frame IOL power for each of the one or more IOLs.

[0075] In addition to the output from the EFEA model 126, the output from the IPC is also used as an input to the fusion model 128. As discussed above, the IPC can be a standard IPC (e.g., Bartlett IPC, Haigis IPC, Hill-RBF IPC, SRKIPC, etc.) In some embodiments, an IPC ML model 127 is used.

[0076] The TDG 122 may generate a training data set (or multiple training data sets) for training the IPC ML model 127. The training data set for the IPC ML model 127 may include demographic information and / or data points associated with measurements of anatomical parameters for each of a plurality of historical patients mapped to corresponding clinical outcomes of postoperative outcomes. In some embodiments, the training data set for the IPC ML model 127 involves the same preoperative anatomical parameters as the preoperative anatomical parameters used for the EFEA model 126. In some embodiments, the training data set for the IPC ML model 127 uses a subset of anatomical parameters as the EFEA model 126. That is, in some embodiments, the FEA model 125 and the EFEA model 126 may use all of the anatomical parameters of the IPC ML model 127 as well as additional modeling anatomical parameters.

[0077] The model trainer 124 trains the IPC ML model 127 based on the training data set(s). For example, the model trainer 124 trains the IPC ML model 127 to generate predicted postoperative outcomes for given inputs 408, including patient anatomical parameters 410 and (multiple) IOL parameters 412. In some embodiments, the patient anatomical parameters 410 of the input 408 of the IPC ML model 127 are the same patient anatomical parameters 404. In some embodiments, the patient anatomical parameters 410 of the input 408 of the IPC ML model 127 are a subset of the patient anatomical parameters 404. The IPC ML model 127 outputs a predicted postoperative outcome based at least in part on the inputs. In some embodiments, the predicted postoperative outcome is a predicted postoperative refractive error, a predicted postoperative effective lens position, and / or a predicted IOL power. In some embodiments, the IPC ML model 127 outputs a predicted IOL per IOL per frame based at least in part on the inputs.

[0078] To train and refine the IPC ML model 127, the postoperative outcomes predicted by the IPC ML model 127 can be compared with historical clinical data records in the repository 140 of treatment outcomes for the same set of inputs to determine whether the predictions made by the IPC ML model 127 provide satisfactory results. Based on the comparison, the model trainer 124 can adjust the IPC ML model 127 to improve the model. The IPC ML model 127 can be trained with a large set of anatomical values ​​and IOL parameters.

[0079] After training, the IPC ML model 127 can be deployed to one or more server computers, user consoles, imaging devices in which a computing device is integrated, etc. Figure 1 As shown in , the IPC ML model 127 can be deployed to the user console 130.

[0080] The fused model 128 may be another ML-based model that is trained with inputs from both the EFEA model 126 and the IPC ML model 127 to generate recommended IOL parameters. In some embodiments, the fused model 128 may be trained to weight the inputs from the EFEA model 126 and the inputs from the IPC ML model 127 differently. In some embodiments, the fused model 128 may weight the inputs from the EFEA model 126 and / or the IPC ML model 127 differently for different data points.

[0081] The output 414 from the fusion model 128 includes at least one or more recommended IOL parameters 416. As discussed herein, the one or more recommended IOL parameters 416 may include a recommended IOL type, IOL power, and / or IOL size.

[0082] In some other embodiments, the output 414 from the fusion model 128 may include one or more predicted post-operative outcomes, such as a predicted post-operative refractive error for each of the one or more IOLs.

[0083] Prior to deployment, the fused model 128 may undergo a training cycle that involves the model trainer 124 adjusting the weights of the fused model 128 to reduce a loss calculated between outputs 414, which may be indicative of a predicted postoperative refractive error. The fused model 128 may be trained to recommend IOL parameters that result in the lowest predicted postoperative refractive error. The treatment outcome data may indicate an actual postoperative refractive error measured postoperatively for a patient across a large set of inputs. The treatment outcome data, which may be provided as part of each historical patient record, may typically indicate an actual IOL used for the patient, postoperative measurements (e.g., actual postoperative refractive error measured postoperatively for the patient), and / or a postoperative customer satisfaction score.

[0084] Once trained, the fusion model 128 can be deployed to one or more server computers, user consoles, imaging devices in which the computing device is integrated, etc. As shown in FIG. Figure 1 As shown in , the fused model 128 can be deployed to the user console 130. After deployment, the fused model 128 can continue to be similarly trained using the treatment outcome data recorded by the TDR 132 for the current / new patient and added to the repository 140.

[0085] Example method for performing ophthalmic surgery based on recommendations generated using a fusion model

[0086] Figure 5Example operations 500 are shown that may be performed by a computing device to generate and output a recommendation including IOL parameters for cataract surgery for a patient based on anatomical parameters of the patient in accordance with certain aspects described herein. Operations 500 may be performed by a clinical device, such as user console 130 .

[0087] Operation 500 may begin at block 502 by generating a first predicted lens behavior (e.g., IOL power per frame) for each IOL in a set of IOLs using an EFEA model (e.g., EFEA model 126) given first preoperative anatomical parameters of the patient's eye (e.g., patient anatomical parameters 404) and one or more parameters of each IOL in one or more IOLs (e.g., IOL parameters 406). In an illustrative embodiment, for a current patient in the clinic, a clinician inputs the patient's preoperative anatomical parameters, including the patient's lens feature dimensions, into a user console 130. In some embodiments, the user console 130 itself measures the patient's anatomical parameters. In some embodiments, the imaging device 110 measures the patient's anatomical parameters, and the imaging device 110 provides the measured anatomical parameters to the user console 130 or displays the measured anatomical parameters to the clinician, who manually inputs the anatomical parameters into the user console 130. In some embodiments, the patient's anatomical parameters are measured and stored in advance, and the user console 130 retrieves the stored anatomical parameters, or the clinician retrieves the stored anatomical parameters and manually inputs the anatomical parameters into the user console 130. In some embodiments, the one or more IOLs include a set of accommodating IOLs, such as a set of IOLs in stock or available to a clinician. The set of IOLs may be IOLs covering a range of IOL powers at a particular IOL power step size.

[0088] The EFEA model outputs a first predicted lens behavior for each of the one or more IOLs. In an illustrative example, based on the patient's input preoperative anatomical parameters, the EFEA model 126 outputs a plurality of frames of predicted IOL powers for each input IOL and preoperative anatomical parameters. The plurality of frames may include at least frames from implantation of the IOL in the patient's eye until the IOL reaches a resting state.

[0089] Operation 500 may continue at block 504 by generating a second predicted lens behavior (e.g., postoperative refractive outcome) for each IOL in the set of IOLs using an IPC ML model (e.g., IPC ML model 127) given a second one or more preoperative anatomical parameters of the patient's eye (e.g., patient anatomical parameters 410) and one or more parameters of each of the one or more IOLs (e.g., IOL parameters 412). In some embodiments, the anatomical parameters input to the IPC ML model 127 are the same as or a subset of the anatomical parameters input to the EFEA model 126 at block 502. The IPC ML model outputs a second predicted postoperative outcome for each of the one or more IOLs. In some embodiments, based on the input preoperative anatomical parameters and IOL parameters, the IPC ML model 127 outputs a predicted postoperative refractive error for each of the one or more IOLs, a predicted IOL power for each of the one or more IOLs, a predicted IOL power per frame for each of the one or more IOLs, or a combination thereof.

[0090] Operation 500 may continue at block 506 by generating one or more recommended IOL parameters (e.g., recommended IOL parameters 416) using a fusion model (e.g., fusion model 128) given the first predicted lens behavior and the second predicted lens behavior of each of the one or more IOLs. In some embodiments, fusion model 128 outputs a recommended IOL type, power, and / or size based on input from EFEA model 126 and IPC ML model 127.

[0091] Example system for performing ophthalmic surgery based on recommendations generated using a fusion model

[0092] Figure 6 An example computing device 600 is shown that uses an EFEA model (e.g., EFEA model 126) to assist in performing an ophthalmic procedure such as cataract surgery, in accordance with certain aspects described herein. For example, computing device 600 may be Figure 1 The user console 130 shown in FIG.

[0093] As shown, computing device 600 includes a user interface 602 , one or more input / output (I / O) interfaces 604 , a network interface 606 , and a control module 608 .

[0094] User interface 602 may be a graphical user interface (GUI) through which a user interacts with computing device 600. I / O interface 604 may allow various I / O devices (e.g., keyboard, display, mouse device, pen input, etc.) to be connected to computing device 600. Network interface 606 may connect computing device 600 to a network (which may be a local area network, an intranet, the Internet, or any other group of computing devices communicatively connected to each other).

[0095] Control module 608 includes memory 610, central processing unit(s) 622, and storage device 624. CPU(s) 622 can retrieve and execute programmed instructions stored in memory 610. CPU(s) 622 represent a single CPU, multiple CPUs, a single CPU with multiple processing cores, etc.

[0096] The memory 610 represents a volatile memory such as a random access memory and / or a non-volatile memory such as a non-volatile random access memory, a phase change random access memory, etc. The memory 610 may include input parameters 612 , an EFEA model 614 , an ML IPC model 616 , a fusion model 618 , and treatment data 620 .

[0097] The computing device 600 may receive input from the user interface 602 and / or the I / O device interface(s) 604 and store the input parameters 612 in the memory 610. The CPU(s) 622 may take the input parameters 612 and execute the EFEA model 614 and the ML IPC model 616 using the input parameters 612 to generate a first predicted postoperative outcome and a second predicted postoperative outcome. The CPU(s) 622 may execute the fused model 618 with the first predicted postoperative outcome and the second predicted postoperative outcome to generate recommended IOL parameters or for use in cataract surgery for the patient. The computing device 600 may receive treatment data from the user interface 602 and / or the I / O device interface(s) 604 and store the treatment data 620 in the memory 610.

[0098] Additional considerations

[0099] The foregoing description is provided to enable any person skilled in the art to practice the various embodiments described herein. Various modifications to these embodiments are clear to those skilled in the art, and the general principles defined herein may be applied to other embodiments. For example, without departing from the scope of the present disclosure, the functions and arrangements of the elements discussed may be changed. Various examples may appropriately omit, replace or add various programs or components. In addition, the features described with respect to some examples may be combined in some other examples. For example, any number of aspects set forth herein may be used to implement a device or practice method. In addition, the scope of the present disclosure is intended to cover such devices or methods practiced using other structures, functions, or structures and functions in addition to or different from the various aspects of the present disclosure set forth herein. It should be understood that any aspect of the present disclosure disclosed herein may be embodied by one or more elements of a claim.

[0100] As used herein, the phrase "at least one of" a list of items refers to any combination of those items, including single members. For example, "at least one of a, b, or c" is intended to cover any combination of a, b, c, ab, ac, bc, and abc, as well as multiples of the same element (e.g., aa, aaa, aab, aac, abb, acc, bb, bbb, bbc, cc, and ccc or any other order of a, b, and c).

[0101] As used herein, the term "determining" encompasses a wide variety of actions. For example, "determining" may include calculating, computing, processing, deriving, investigating, searching (e.g., searching in a table, database, or other data structure), ascertaining, etc. Also, "determining" may include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory), etc. Also, "determining" may include resolving, selecting, choosing, establishing, etc.

[0102] The method disclosed herein includes one or more steps or actions for implementing the method. Without departing from the scope of the claims, the method steps and / or actions can be interchangeable with each other. In other words, unless a specific step or action sequence is specified, the order and / or use of specific steps and / or actions can be modified without departing from the scope of the claims. Further, the various operations of the above-mentioned method can be performed by any suitable device that can perform the corresponding function. These devices may include various hardware and / or software components and / or modules, including but not limited to circuits, application specific integrated circuits (ASICs) or processors. Generally, in the case of operations shown in the figure, those operations can have corresponding corresponding devices with similar numbers plus functional components.

[0103] The various illustrative logical blocks, modules, and circuits described in conjunction with the present disclosure may be implemented or executed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device (PLD), discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. The general purpose processor may be a microprocessor, but in an alternative, the processor may be any commercially available processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, for example, a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.

[0104] The processing system can be implemented with a bus architecture. Depending on the specific application and overall design constraints of the processing system, the bus may include any number of interconnecting buses and bridges. The bus may link together various circuits including a processor, a machine-readable medium, and an input / output device. A user interface (e.g., a keypad, a display, a mouse, a joystick, etc.) may also be connected to the bus. The bus may also link various other circuits such as timing sources, peripherals, voltage regulators, power management circuits, etc., which are well known in the art and therefore will not be described further. The processor may be implemented with one or more general and / or special processors. Examples include microprocessors, microcontrollers, DSP processors, and other circuit systems that can execute software. Those skilled in the art will recognize how to best implement the described functions of the processing system according to the specific application and the overall design constraints imposed on the entire system.

[0105] If implemented in software, the function may be stored as one or more instructions or codes on a computer-readable medium or transmitted via a computer-readable medium. Software should be broadly interpreted as instructions, data, or any combination thereof, whether referred to as software, firmware, middleware, microcode, hardware description language, or other. Computer-readable media include both computer storage media and communication media (such as any medium that facilitates the transfer of computer programs from one place to another). The processor may be responsible for managing the bus and general processing, including the execution of software modules stored on a computer-readable storage medium. A computer-readable storage medium may be connected to a processor so that the processor can read information from the storage medium and write information to the storage medium. In an alternative, the storage medium may be integrated into the processor. For example, a computer-readable medium may include a transmission line, a carrier modulated by data, and / or a computer-readable storage medium on which instructions separated from a wireless node are stored, all of which may be accessed by a processor via a bus interface. Alternatively or in addition, a computer-readable medium or any portion thereof may be integrated into a processor, such as a case where a cache and / or a general register file may be provided. For example, examples of machine-readable storage media may include RAM (random access memory), flash memory, ROM (read-only memory), PROM (programmable read-only memory), EPROM (erasable programmable read-only memory), EEPROM (electrically erasable programmable read-only memory), registers, magnetic disks, optical disks, hard drives, or any other suitable storage media, or any combination thereof. Machine-readable media may be embodied in a computer program product.

[0106] A software module may include a single instruction or multiple instructions, and may be distributed over several different code segments, between different programs, and across multiple storage media. A computer-readable medium may include multiple software modules. A software module includes instructions that, when executed by a device such as a processor, cause a processing system to perform various functions. A software module may include a transmission module and a reception module. Each software module may exist in a single storage device, or may be distributed across multiple storage devices. For example, when a triggering event occurs, a software module may be loaded from a hard drive into a RAM. During the execution of a software module, a processor may load some instructions into a cache to increase access speed. Then, one or more cache lines may be loaded into a general register file for execution by the processor. When referring to the function of a software module, it should be understood that such function is implemented by the processor when executing instructions from the software module.

[0107] The following claims are not intended to be limited to the embodiments shown herein, but are given the full scope consistent with the language of the claims. In the claims, unless otherwise specified, reference to a singular element is not intended to mean "one and only one", but "one or more". Unless otherwise specifically stated, the term "some" refers to one or more. According to 35 U.S.C. § 112 (f), the elements of any claim will not be interpreted unless the phrase "device for..." is used to expressly describe these elements, or in the case of a method claim, the phrase "step for..." is used to describe these elements. All structural and functional equivalents of the elements of the various aspects described throughout the present disclosure that are known or will be known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be covered by the claims. In addition, whether or not such disclosure is explicitly described in the claims, the content disclosed herein is not intended to be dedicated to the public.

Claims

1. A method of determining one or more intraocular lens (IOL) parameters of an intraocular lens (IOL) to be used in a cataract surgery procedure, the method comprising: using one or more ophthalmic imaging devices to generate a plurality of data points associated with measurements of a plurality of anatomical parameters of the eye to be treated; generating a first predicted lens behavior based at least in part on the plurality of data points associated with the measurements of the plurality of anatomical parameters and one or more IOL parameters of each of the one or more IOLs using a machine learning model trained to a simulation finite element analysis (FEA) model; generating, using an IOL power calculator machine learning model, a second predicted lens behavior based at least in part on at least a subset of the plurality of data points associated with the measurements of the plurality of anatomical parameters of the eye to be treated and one or more IOL parameters of each of the one or more IOLs; as well as A recommendation including one or more IOL parameters of the IOL to be used in the cataract surgery is generated based at least in part on the first predicted lens behavior and the second predicted lens behavior using the fused machine learning model.

2. The method of claim 1, wherein: The plurality of anatomical parameters include one or more characteristic dimensions of the lens of the eye.

3. The method of claim 1, wherein: The one or more IOL parameters include at least one of: IOL type, IOL size, or IOL power.

4. The method of claim 1, further comprising: generating a FEA model based on a set of data points associated with at least one historical patient anatomical parameter and at least one set of one or more IOL parameters using a finite element method (FEM); generating a predicted lens behavior using the FEA model based at least in part on the set of data points associated with the at least one historical patient anatomical parameter and the at least one set of one or more IOL parameters; as well as The FEA model is adjusted based on a comparison of predicted lens behavior and observed lens behavior for an IOL having the at least one set of IOL parameters implanted in a historical patient eye having the set of data points associated with the anatomical parameters.

5. The method of claim 1, further comprising training the machine learning model to simulate the FEA model by: generating a predicted lens behavior using the machine learning model based at least in part on a set of data points associated with at least one historical patient anatomical parameter and at least one set of one or more IOL parameters; and Weights associated with the machine learning model are adjusted based on a comparison of the predicted lens behavior output by the machine learning model for the set of data points associated with the anatomical parameters and the at least one set of one or more IOL parameters with the predicted lens behavior output by the FEA model.

6. The method of claim 5, wherein: The predicted lens behavior includes the IOL power for each frame.

7. The method of claim 1, further comprising training the IOL power calculator machine learning model by: generating a predicted lens behavior using the IOL power calculator machine learning model based at least in part on a set of data points associated with at least one historical patient anatomical parameter and at least one set of one or more IOL parameters; and The IOL power calculator machine learning model is adjusted based on a comparison of predicted lens behavior and observed lens behavior for an IOL having the at least one set of one or more IOL parameters implanted in a historical patient eye having the set of data points.

8. The method of claim 7, wherein: The predicted lens behavior includes a predicted postoperative refractive outcome.

9. The method of claim 1, further comprising training the fusion model by: generating one or more recommended IOL parameters for the historical patient using the fused model based at least in part on a third predicted lens behavior obtained by a machine learning model trained to simulate the FEA model and a fourth predicted lens behavior obtained by the IOL power calculator machine learning model; and The fusion model is adjusted based on a comparison of the one or more recommended IOL parameters of an IOL having the one or more recommended IOL parameters implanted in the historical patient eye with treatment outcome data.

10. A system for determining one or more intraocular lens (IOL) parameters of an IOL to be used in a cataract surgery procedure, the system comprising: one or more ophthalmic imaging devices configured to generate a plurality of data points associated with measurements of a plurality of anatomical parameters of the eye to be treated; a memory storing a machine learning model trained to simulate a finite element analysis (FEA) model, an IOL power calculator machine learning model, and a fused machine learning model; as well as at least one processor, the at least one processor being coupled to the memory, the at least one processor being configured to: generating a first predicted lens behavior based at least in part on the plurality of data points associated with the measurements of the plurality of anatomical parameters and one or more IOL parameters of each of the one or more IOLs using a machine learning model trained to simulate the FEA model; generating, using the IOL power calculator machine learning model, a second predicted lens behavior based at least in part on at least a subset of the plurality of data points associated with the measurements of the plurality of anatomical parameters of the eye to be treated and one or more IOL parameters of each of the one or more IOLs; as well as A recommendation including one or more IOL parameters of the IOL to be used in the cataract surgery is generated using the fused machine learning model based at least in part on the first predicted lens behavior and the second predicted lens behavior.

11. The system of claim 10, wherein: The plurality of anatomical parameters include one or more characteristic dimensions of the lens of the eye.

12. The system of claim 11, wherein: The one or more IOL parameters include at least one of: IOL type, IOL size, or IOL power.

13. The system of claim 11, wherein: The at least one processor is further configured to: obtaining a FEA model generated using a finite element method (FEM) based on a set of data points associated with at least one historical patient's anatomical parameters and at least one set of one or more IOL parameters; generating a predicted lens behavior using the FEA model based at least in part on the set of data points associated with the at least one historical patient anatomical parameter and the at least one set of one or more IOL parameters; as well as The FEA model is adjusted based on a comparison of predicted and observed lens behavior for an IOL having the at least one set of IOL parameters implanted in a historical patient eye having the set of data points associated with the anatomical parameters.

14. The system of claim 11, wherein: The at least one processor is further configured to train the machine learning model to simulate the FEA model by: generating a predicted lens behavior using the machine learning model based at least in part on a set of data points associated with at least one historical patient anatomical parameter and at least one set of one or more IOL parameters; as well as Weights associated with the machine learning model are adjusted based on a comparison of the predicted lens behavior output by the machine learning model for the set of data points associated with the anatomical parameters and the at least one set of one or more IOL parameters with the predicted lens behavior output by the FEA model.

15. The system of claim 14, wherein: The predicted lens behavior includes the IOL power for each frame.

16. The system of claim 11, wherein: The at least one processor is further configured to train the IOL power calculator machine learning model by: generating a predicted lens behavior using the IOL power calculator machine learning model based at least in part on a set of data points associated with at least one historical patient anatomical parameter and at least one set of one or more IOL parameters; as well as The IOL power calculator machine learning model is adjusted based on a comparison of predicted lens behavior and observed lens behavior for an IOL having the at least one set of one or more IOL parameters implanted in a historical patient eye having the set of data points.

17. The system of claim 16, wherein: The predicted lens behavior includes a predicted postoperative refractive outcome.

18. The system of claim 11, wherein: The at least one processor is further configured to train the fusion model by: generating one or more recommended IOL parameters for the historical patient using the fused model based at least in part on a third predicted lens behavior obtained by a machine learning model trained to simulate the FEA model and a fourth predicted lens behavior obtained by the IOL power calculator machine learning model; as well as The fusion model is adjusted based on a comparison of the one or more recommended IOL parameters of an IOL having the one or more recommended IOL parameters implanted in the historical patient eye with treatment outcome data.

Images