Intraocular lens selection based on multiple machine learning models

Abstract

A method and system for selecting an intraocular lens, the system having a controller having a processor and tangible non-transitory memory. A plurality of machine learning models are selectively executable by the controller. The controller is configured to receive at least one pre-operative image of the eye and extract a first data set via a first input machine learning model. The controller is configured to receive a plurality of biometric parameters of the eye and extract a second data set via a second input machine learning model. The first data set and the second data set are combined to produce a hybrid data set. The controller is configured to generate at least one output factor based on the hybrid data set via an output machine learning model. An intraocular lens is selected based in part on the at least one output factor.

Description

BACKGROUND

[0001] The present disclosure relates generally to a system and method for selecting an intraocular lens to be implanted into an eye using multiple machine learning models. The human lens is normally clear so that light can easily pass through it. However, many factors can cause regions in the lens to become cloudy and dense, negatively impacting the quality of vision. This condition can be corrected via cataract surgery, i.e., selecting an artificial lens to be implanted into the patient’s eye. In fact, cataract surgery is a common procedure performed worldwide. One important driver of cataract surgery clinical outcomes is the selection of the appropriate intraocular lens. Currently, there are several calculators that use various preoperative information about the patient’s eye to predict the lens power to be implanted. However, existing calculators do not use the complete preoperative images of the eye as input data. SUMMARY

[0002] Disclosed herein is a system and method for selecting an intraocular lens to be implanted into an eye, the system having a controller having a processor and a tangible non-transitory memory. The system employs a comprehensive, multi-pronged approach and includes a controller having a processor and a tangible non-transitory memory having instructions recorded thereon. The controller is configured to selectively execute a plurality of machine learning models, including a first input machine learning model, a second input machine learning model, and an output machine learning model. Each of the plurality of machine learning models can be a respective regression model. In one example, the output machine learning model includes a multi-layer perceptron network.

[0003] Execution of the instructions by the processor causes the controller to receive at least one preoperative image of the eye. The controller is configured to extract, via the first input machine learning model, a first data set based in part on the at least one preoperative image. In one example, the at least one preoperative image of the eye is an ultrasound biomicroscopy image. The first data set can include a plurality of preoperative dimensions of the eye. The plurality of preoperative dimensions can include one or more of an anterior chamber depth, a lens thickness, a lens diameter, a sulcus-to-sulcus diameter, a first equator position, a second equator position, a third equator position, an iris diameter, an axial length from a first surface of the cornea to a posterior surface of the preoperative lens, and a ciliary process diameter. Alternatively, the plurality of preoperative dimensions can include each of an anterior chamber depth, a lens thickness, a lens diameter, a sulcus-to-sulcus diameter, an iris diameter, an axial length from a first surface of the cornea to a posterior surface of the preoperative lens, and a ciliary process diameter.

[0004] The controller is further configured to receive a plurality of biometric parameters of the eye and extract, via the second input machine learning model, a second data set based in part on the plurality of biometric parameters. The plurality of biometric parameters can include a K flat factor and a K steep factor. The first data set and the second data set are combined to obtain a hybrid data set. In one example, the pre-operative image is obtained from a first imaging device and the plurality of biometric parameters are obtained from a second imaging device, the second imaging device being different from the first imaging device. For example, the first imaging device can be an ultrasound device and the second imaging device can be an optical coherence tomography device.

[0005] The controller is configured to generate, via the output machine learning model, at least one output factor based on the hybrid data set. An intraocular lens is selected based in part on the output factor. The output factor can be a manifest equivalent spherical power (MRSE). The plurality of machine learning models can include a third input machine learning model. Prior to generating the output factor, the controller can be configured to access a historical pair of respective pre-operative and post-operative images and extract, via the third input machine learning model, a third data set based in part on the historical pair. The third data set is added to the hybrid data set prior to generating the output factor.

[0006] The intraocular lens can include an optical zone that is contiguous with one or more support structures. The intraocular lens can include an inner cavity that is at least partially filled with a fluid. The fluid is configured to move within the inner cavity to change a refractive power of the intraocular lens. It will be appreciated that any type of intraocular lens available to those skilled in the art can be employed.

[0007] The foregoing features and advantages of the present disclosure, as well as other features and advantages of the present disclosure, will be more fully understood and appreciated by reference to the following detailed description, taken in conjunction with the accompanying drawings of the best modes presently contemplated of carrying out the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1 is a schematic illustration of a system for selecting an intraocular lens to be implanted into an eye, the system having a controller;

[0009] Figure 2 is a schematic perspective view of an example intraocular lens;

[0010] Figure 3 is a schematic flowchart of a method that can be performed by Figure 1 the controller of

[0011] Figure 4 is a schematic partial cross-sectional view of an example pre-operative image of an eye;

[0012] Figure 5 is a schematic partial cross-sectional view of an example post-operative image of an eye;

[0013] Figure 6 is a schematic example of a multi-layer perceptron algorithm that can be executed by a controller of Figure 1

[0014] Figure 7 is a schematic example of a support vector regression (SVR) technique of a controller of Figure 1 DETAILED DESCRIPTION

[0015] Referring to the drawings, wherein like reference numbers refer to like components, Figure 1 A system 10 for selecting an intraocular lens to be implanted is schematically illustrated. Figure 2 An example of an intraocular lens 12 is shown. However, any type of intraocular lens available to those skilled in the art can be employed. The system 10 employs a robust, multi-pronged approach that utilizes multiple machine learning models. As described below, the system 10 utilizes both rich image data and numerical data to optimize the selection of the intraocular lens 12.

[0016] Referring to Figure 2 , the intraocular lens 12 includes an optical zone 14 that defines a first surface 16 and a second surface 18. The optical zone 14 can be contiguous with one or more support structures, such as a first support structure 20 and a second support structure 22, that are configured to support the positioning and retention of the intraocular lens 12. The intraocular lens 12 can define an internal cavity 24 that is at least partially filled with a fluid F. The fluid F is configured to be movable within the internal cavity 24 to change the thickness (and dioptric power) of the intraocular lens 12. It should be appreciated that the intraocular lens 12 can take many different forms and include multiple components and / or alternative components.

[0017] Referring to Figure 1 , the system 10 includes a controller C having at least one processor P and at least one memory M (or non-transitory, tangible computer-readable storage medium) having instructions recorded thereon for performing a method 100 of selecting an intraocular lens 12. The method 100 is shown and described below with reference to Figure 3 Figure 4 An example pre-operative image 200 of an eye E is shown. Figure 5 An example post-operative image 300 of an eye E is shown.

[0018] Reference is now made to Figure 1 ​​​Controller C can be configured to communicate with various entities, such as a first imaging device 30, via a short-range network 28. The first imaging device 30 can be an ultrasound machine, a magnetic resonance imaging machine, or other imaging device available to those skilled in the art. The short-range network 28 can be wireless or can include physical components. The short-range network 28 can be a bus implemented in various ways, such as a serial communication bus in the form of a local area network. The local area network can include, but is not limited to, a controller area network (CAN), a controller area network with flexible data rate (CAN-FD), Ethernet, Bluetooth, WIFI, and other forms of data connections. The short-range network 28 can be Bluetooth TM Connect, defined as a short-range radio technology (or wireless technology) that aims to simplify communication between Internet-enabled devices, as well as devices and the Internet. Bluetooth TM is an open wireless technology standard for short-range communications of fixed and mobile devices, and creates personal networks that operate in the 2.4 GHz band. Other types of connections can be employed.

[0019] Referring to Figure 1 , the controller C can communicate with a second imaging device 32, a display module and / or user interface 34, and a database 36. Additionally, the controller C can be configured to communicate with a remote server 40 and / or a cloud unit 42 via a long-range network 44. The remote server 40 can be a private or public source of information maintained by an organization such as a research institute, a company, a university, and / or a hospital. The cloud unit 42 can include one or more servers hosted on the Internet to store, manage, and process data. The long-range network 44 can be a wireless local area network (LAN) that links multiple devices using a wireless distribution method, a wireless metropolitan area network (MAN) that connects several wireless LANs, or a wireless wide area network (WAN) that covers a large area such as adjacent towns and cities. Other types of connections can be employed.

[0020] As shown in Figure 1 , the controller C can be configured to receive wireless communications from and transmit wireless communications to the remote server 40 through a mobile application 46. The mobile application 46 can communicate with the controller C via the short-range network 28, such that data in the controller C can be accessed. In one example, the mobile application 46 is physically connected (e.g., wired) to the controller C. In another example, the mobile application 46 is embedded into the controller C. Circuitry and components of the remote server 40 and the mobile application 46 (“app”) available to those skilled in the art can be employed.

[0021] The controller C is specially programmed to selectively execute a plurality of machine learning models 48. The controller C can access the plurality of machine learning models 48 via the short-range network 28, the long-range network 44, and / or the mobile application 46. Alternatively, the plurality of machine learning models 48 can be embedded into the controller C. The plurality of machine learning models 48 can be configured to find parameters, weights, or structures that minimize a respective cost function. Each of the plurality of machine learning models 48 can be a respective regression model. In one example, referring to Figure 1 the plurality of machine learning models 48 includes a first input machine learning model 50, a second input machine learning model 52, a third input machine learning model 54, and an output machine learning model 56.

[0022] The plurality of machine learning models 48 can include neural network algorithms. As understood by those skilled in the art, neural networks are designed to recognize patterns and generally mimic the human brain. These patterns are recognized by the neural network from real-world data (e.g., images, sounds, text, time series, etc.) that is transformed or converted into numerical form and embedded into vectors or matrices. The neural network can use deep learning mappings to match an input vector x with an output vector y. In other words, each of the plurality of machine learning models 48 learns an activation function f that maps f(x) to y. The training process enables the neural network to associate the appropriate activation function f(x) to transform the input vector x to the output vector y. In the case of a simple linear regression model, two parameters are learned: a bias and a slope. The bias is the level of the output vector y when the input vector x is 0, and the slope is the rate of increase or decrease in the prediction of the output vector y for each unit increase in the input vector x. Once the plurality of machine learning models 48 are trained, respectively, an estimated value of the output vector y can be calculated with a given new value of the input vector x.

[0023] The plurality of machine learning models 48 can include a multi-layer perceptron network. Referring to Figure 6An example of a multilayer perceptron network 400 is shown. The multilayer perceptron network 400 is a feedforward artificial neural network with at least three layers of nodes N, including an input layer 402, one or more hidden layers 408 (e.g., a first hidden layer 404 and a second hidden layer 406), and an output layer 410. Each layer consists of nodes N configured to perform an affine transformation of the linear sum of the inputs. Node N is a neuron characterized by corresponding biases and corresponding weighted connections. Node N in the input layer 402 receives inputs, normalizes them, and forwards them to nodes N in the first hidden layer 404. Each node N in subsequent layers computes a linear combination of the outputs of the previous layer. A network with three layers will form an activation function f(x) = f(3)(f(2)(f(1)(x))). For the corresponding node N in the output layer 410, the activation function f can be linear. For the first hidden layer 404 and the second hidden layer 406, the activation function f can be a sigmoid function. A linear combination of sigmoid functions is used to approximate a continuous function representing the output vector y. Other types of neural networks can be used.

[0024] Multiple machine learning models 48 may include support vector regression (SVR) models. Figure 7 An example of a support vector regression model 500 for data point 502 is shown. The support vector regression model 500 is configured to find the function ( Figure 7 The hyperplane 504 in the function is such that data point 502 lies within the stripe 506 of the function, i.e., within the first boundary line 508 and the second boundary line 510. (See reference) Figure 7 Hyperplane 504 can be defined as the line that matches the input vector x with the output vector y (i.e., predicts the target value). Hyperplane 504 is customized to maximize stripe 506 and minimize a predefined error. If points exist outside stripe 506 (e.g., external point 512), a penalty can be introduced into the support vector regression model 500. Before determining hyperplane 504, the support vector regression model 500 can use a kernel function to map the lower-dimensional dataset to a higher-dimensional dataset. Other machine learning models available to those skilled in the art can be employed.

[0025] Now for reference Figure 3 This shows that it can be generated by Figure 1 The flowchart illustrates method 100 executed by controller C. Method 100 does not need to be applied in the specific order listed herein, and some boxes may be omitted. Memory M may store a set of controller executable instructions, and processor P may execute the set of controller executable instructions stored in memory M.

[0026] according to Figure 3 In frame 102, controller C is configured to receive at least one preoperative image of the eye. Figure 4An example pre-operative image 200 of an eye E is shown. Figure 4 Not drawn to scale. Figure 4 The superior surface 202 of the cornea 203, the inferior surface 204 of the cornea 203, the pre-operative lens 206, the iris 208, and the ciliary muscle 210 are shown. The pre-operative image 200 can be obtained via ultrasound biomicroscopy techniques. Ultrasound biomicroscopy techniques can employ a relatively high frequency transducer of between about 35 MHz and 100 MHz with a tissue penetration depth of between about 4 mm and 5 mm. Other imaging modalities can be employed, including but not limited to optical coherence tomography and magnetic resonance imaging. A single image or a series of images can be used to train the plurality of machine learning models 48.

[0027] According to Figure 3 block 104, the method 100 includes extracting, via the first input machine learning model 50, a first data set based in part on at least one pre-operative image, such as the example pre-operative image 200 shown. Figure 4 The first data set can be presented in the form of a three-dimensional matrix. This provides a technical advantage of leveraging rich image data. Referring to Figure 4 , the first data set can include a plurality of pre-operative dimensions, such as an anterior chamber depth 212, a lens thickness 214, a lens diameter 216, and a sulcus-to-sulcus diameter 218. The plurality of pre-operative dimensions can include a first equator plane location 220 (measured from the anterior lens pole), a second equator plane location 222 (measured relative to the anterior chamber depth 212), and a third equator plane location 224 (measured relative to the posterior lens pole). Referring to Figure 4 , the plurality of pre-operative dimensions can further include an iris diameter 226, an axial length 228 from the cornea 203 to the posterior surface of the pre-operative lens 206, and a ciliary process diameter 230.

[0028] According to Figure 3The controller C is configured to receive a plurality of biometric parameters, which can include pre-operative dimensions of the eye E, such as K flat factor, K steep factor, and mean K factor. The plurality of biometric parameters can further include anterior chamber depth 212, lens thickness 214, lens diameter 216, ciliary process diameter 230, and sulcus-to-sulcus diameter 218. The plurality of biometric parameters can further include parameters related to the intraocular lens 12, such as lens power and thickness. In one example, the pre-operative image 200 is obtained from the first imaging device 30, and the plurality of biometric parameters are obtained from the second imaging device 32, where the second imaging device 32 is different from the first imaging device 30. For example, the first imaging device 30 can be an ultrasound device, and the second imaging device 32 can be an optical coherence tomography device. It should be appreciated that other imaging modalities can be employed. In another example, the pre-operative image 200 and the plurality of biometric parameters are obtained by the same imaging modality.

[0029] According to Figure 3 block 108, the method 100 includes extracting, via the second input machine learning model 52, a second data set based in part on the plurality of biometric parameters. The second data set can be in the form of a three-dimensional vector. According to Figure 3 block 110, the controller C is configured to combine the first data set and the second data set to obtain a hybrid data set.

[0030] Optionally, according to block 112, the method 100 can include accessing a historical pair of respective pre-operative and post-operative images, such as the pre-operative image 200 and the post-operative image 300 shown in FIGS. 1A and IB, respectively. Figure 4 and Figure 5 Figure 5 The superior surface 302 of the cornea 303, the inferior surface 304 of the cornea 303, the iris 308, and the ciliary muscle 310 are shown. Figure 5 The implanted intraocular lens 12, the first surface 16, the second surface 18, the first support structure 20, and the second support structure 22 are also shown in FIG. IB. Figure 5 Not drawn to scale. The post-operative image 300 can be obtained via ultrasound biomicroscopy technology or other imaging modalities available to those skilled in the art. With reference to Figure 1 , the controller C can be configured to obtain the historical pair from the database 36 via the short-range network 28. The controller C can be configured to obtain the historical pair from the remote server 40 via the long-range network 44.

[0031] ​According to box 114, controller C is configured to extract a third dataset via a third-input machine learning model 54, partially based on comparisons of historical pairs. This third dataset is added to the mixed dataset. In one example, the third-input machine learning model 54 is a deep learning neural network configured to classify preoperative measurements (x) in preoperative image 200 to determine the proposed lens refractive power (f(x)) and subsequently determine the estimation error that might result from using the proposed intraocular lens refractive power. The third-input machine learning model 54 can be configured to minimize a cost function defined as the mean squared error between the predicted manifest equivalent spherical power (based on preoperative image 200) and the postoperative manifest equivalent spherical power (based on postoperative image 300).

[0032] Historical comparisons may require tracking changes in specific parameters between preoperative image 200 and postoperative image 300. For example, the comparison may include assessments of… Figure 4 The first distance d1 shown is Figure 5 The difference between the second distance d2 shown. The first distance d1 is the distance between the center 240 of the preoperative lens 206 and the reference point 242 on the upper surface 202 of the cornea 203 in the preoperative image 200. The second distance d2 is the distance between the center 340 of the implanted intraocular lens 12 and the reference point 342 on the upper surface 302 of the cornea 303 in the postoperative image 300. Other parameters may also be used.

[0033] according to Figure 3 Box 116, method 100 includes generating at least one output factor based on a mixed dataset via an output machine learning model 56. (Reference) Figure 6 The output machine learning model 56 can be a fully connected perceptron model, where the parameters of each node N are independent of those of other nodes; that is, each node N has a unique set of weights as its features. (Reference) Figure 6 The machine learning model 56 can generate multiple outputs, such as a first output factor 412 and a second output factor 414. The first output factor 412 can be the manifest equivalent spherical power (MRSE). The second output factor 414 can be the uncorrected distance visual acuity (UCDVA).

[0034] Optionally, prior to generating the output factor in accordance with block 116, the controller C can be configured to obtain one or more estimated post-operative variables based in part on the plurality of pre-operative dimensions. The estimated post-operative variables can include post-operative lens thickness and post-operative lens position. The estimated post-operative variables are added to the mixed dataset and treated as additional inputs to the output machine learning model 56 to generate the output factor in block 116. The estimated post-operative variables can be obtained from geometric models or intraocular lens power calculation formulas available to those skilled in the art, such as the SRK / T formula, Holladay formula, Hoffer Q formula, Olsen formula, and Haigis formula. The estimated post-operative variables can be obtained from other estimation methods available to those skilled in the art.

[0035] According to Figure 3 Block 118, the method 100 includes selecting the intraocular lens 12 based in part on the at least one output factor generated in block 116. In the case of multiple output factors, the controller C can be configured to use a weighted average of the multiple output factors or other statistical methods (e.g., neural networks) to determine the correct power of the intraocular lens 12 to be implanted.

[0036] In summary, the system 10 and method 100 optimize the selection process of the intraocular lens 12 and enable greater predictive success rates, especially for eyes with irregular biometric characteristics. The system 10 and method 100 can be applied to a wide range of imaging modalities during the model training and model execution processes.

[0037] Figure 1 The controller C includes computer-readable media (also referred to as processor-readable media) including non-transitory (e.g., tangible) media that participate in providing data (e.g., instructions) that can be read by a computer (e.g., by a processor of a computer). Such media can take many forms, including but not limited to non-volatile media and volatile media. Non-volatile media can include, for example, optical or magnetic disks and other persistent memory. Volatile media can include, for example, dynamic random access memory (DRAM), which can constitute a main memory. Such instructions can be transmitted by one or more transmission media including coaxial cables; copper wire and fiber optic cables; as well as infrared, laser, and radio frequency (RF) links. Some forms of computer-readable media include the following: a floppy disk; a flexible disk; a hard disk; magnetic tape; other magnetic media; a CD-ROM; DVDs; other optical media; punch cards; paper tape; other physical media with patterns of holes; a RAM; a PROM; an EPROM; a FLASH-EPROM; any other memory chip or cartridge; a carrier wave transported over a computer network, wireless communication channel, or other transmission medium; or any other medium from which a computer can read.

[0038] The lookup tables, databases, data repositories, or other data stores described herein can include various kinds of mechanisms for storing, accessing, and retrieving various kinds of data, including a hierarchical database, a set of files in a file system, an application database in a proprietary format, a relational database management system (RDBMS), etc. Each such data store can be included within a computing device employing a computer operating system such as one of those mentioned above, and can be accessed via a network in one or more manners such as those described above. A file system can be accessed from the computer operating system, and can include files stored in various formats. An RDBMS can employ the Structured Query Language (SQL), and a language for creating, storing, editing, and executing stored procedures such as the PL / SQL language mentioned above.

[0039] DETAILED DESCRIPTION AND DRAWINGS OR FIGURES The present disclosure is supportive and descriptive of the present disclosure, but the scope of the present disclosure is limited only by the claims. While some of the best modes and other embodiments for carrying out the claimed disclosure have been described in detail, various alternative designs and embodiments exist for practicing the disclosure defined in the appended claims. Also, the features of the embodiments shown in the drawings or mentioned in the specification can not necessarily be understood as independent embodiments. Rather, each feature described in one example of an embodiment can be combined with one or more other desired features from other embodiments, resulting in other embodiments that are not described or claimed but that are within the scope of the appended claims. Therefore, such other embodiments fall within the scope of the claims.

Claims

1. A system for selecting an intraocular lens to be implanted into an eye, the system comprising: a controller having a processor and a tangible non-transitory memory having instructions recorded thereon; wherein the controller is configured to selectively execute a plurality of machine learning models, including a first input machine learning model, a second input machine learning model, and an output machine learning model; wherein execution of the instructions by the processor causes the controller to: receive at least one pre-operative image of the eye and extract, via the first input machine learning model, a first data set based in part on the at least one pre-operative image; receive a plurality of biometric parameters of the eye and extract, via the second input machine learning model, a second data set based in part on the plurality of biometric parameters; combine the first data set and the second data set to obtain a hybrid data set; generate, via the output machine learning model, at least one output factor based on the hybrid data set; and select the intraocular lens based in part on the at least one output factor.

2. The system of claim 1, wherein: the at least one output factor is manifest refraction spherical equivalent (MRSE).

3. The system of claim 1, wherein: the at least one pre-operative image is obtained from a first imaging device and the plurality of biometric parameters are obtained from a second imaging device, the first imaging device being different from the second imaging device.

4. The system of claim 1, wherein: the plurality of machine learning models includes a third input machine learning model, and prior to generating the at least one output factor, the controller is configured to: access a historical pair of respective pre-operative and post-operative images; extract, via the third input machine learning model, a third data set based in part on the historical pair; and add the third data set to the hybrid data set prior to generating the at least one output factor.

5. The system of claim 1, wherein: the intraocular lens includes an optical zone contiguous with one or more support structures; and the intraocular lens includes an inner cavity at least partially filled with a fluid configured to move within the inner cavity to change a refractive power of the intraocular lens.

6. The system of claim 1, wherein: the at least one pre-operative image is an ultrasound biomicroscopy image.

7. The system of claim 1, wherein: each of the plurality of machine learning models is a respective regression model; and the output machine learning model includes a multi-layer perceptron network.

8. The system of claim 1, wherein: the plurality of biometric parameters includes a K flat factor and a K steep factor.

9. The system of claim 1, wherein: the first data set includes a plurality of pre-operative dimensions of the eye; and the second data set includes a plurality of pre-operative biometric parameters of the eye. The plurality of pre-operative dimensions includes one or more of an anterior chamber depth, a lens thickness, a lens diameter, a sulcus-to-sulcus diameter, a first equator position, a second equator position, a third equator position, an iris diameter, an axial length from a first surface of a cornea to a posterior surface of a pre-operative lens, and a ciliary process diameter.

10. The system of claim 9, wherein, Prior to generating the at least one output factor, the controller is configured to: obtain one or more estimated post-operative variables based in part on the plurality of pre-operative dimensions, the one or more estimated post-operative variables including a post-operative lens thickness and a post-operative lens position; and add the one or more estimated post-operative variables to the mixed data set prior to generating the at least one output factor.

11. The system of claim 1, wherein: the first data set includes a plurality of pre-operative dimensions of the eye; and the plurality of pre-operative dimensions includes each of an anterior chamber depth, a lens thickness, a lens diameter, a sulcus-to-sulcus diameter, an iris diameter, an axial length from a first surface of a cornea to a posterior surface of a pre-operative lens, a ciliary process diameter, a first equator position, a second equator position, and a third equator position.

12. A method for selecting an intraocular lens to be implanted into an eye, the method comprising: receiving, via a controller having a processor and a tangible non-transitory memory, at least one pre-operative image of the eye; selectively executing, via the controller, a plurality of machine learning models, the plurality of machine learning models including a first input machine learning model, a second input machine learning model, and an output machine learning model; extracting, via the first input machine learning model, a first data set based in part on the at least one pre-operative image; receiving, via the controller, a plurality of biometric parameters of the eye; extracting, via the second input machine learning model, a second data set based in part on the plurality of biometric parameters; combining, via the controller, the first data set and second data set to obtain a mixed data set; generating, via the output machine learning model, at least one output factor based on the mixed data set; and selecting the intraocular lens based in part on the at least one output factor.

13. The method of claim 12, further comprising, prior to generating the at least one output factor: accessing, via the controller, a historical pair of respective pre-operative and post-operative images; including a third input machine learning model in the plurality of machine learning models; extracting, via the third input machine learning model, a third data set based in part on a comparison of the historical pair; and adding, prior to generating the at least one output factor, the third data set to the mixed data set.

14. The method of claim 12, wherein: the intraocular lens includes an optical zone contiguous with one or more support structures; and the intraocular lens includes an inner cavity at least partially filled with a fluid, the fluid configured to move within the inner cavity to change a refractive power of the intraocular lens.

15. The method of claim 12, wherein: each of the plurality of machine learning models is a respective regression model; and the output machine learning model comprises a multilayer perceptron network.

16. The method of claim 12, wherein: the plurality of biometric parameters comprises a K-flat factor and a K-steep factor.

17. The method of claim 12, wherein: the first data set comprises a plurality of preoperative dimensions of the eye; and the plurality of preoperative dimensions comprises one or more of an anterior chamber depth, a lens thickness, a lens diameter, a sulcus-to-sulcus diameter, a first equator location, a second equator location, a third equator location, an iris diameter, an axial length from a first surface of the cornea to a posterior surface of the preoperative lens, and a ciliary process diameter.

18. The method of claim 12, wherein: the first data set comprises a plurality of preoperative dimensions of the eye; and the plurality of preoperative dimensions comprises each of an anterior chamber depth, a lens thickness, a lens diameter, a sulcus-to-sulcus diameter, an iris diameter, an axial length from a first surface of the cornea to a posterior surface of the preoperative lens, and a ciliary process diameter.

19. The method of claim 12, further comprising: obtaining the at least one preoperative image from a first imaging device, and obtaining the plurality of biometric parameters from a second imaging device, the first imaging device being different from the second imaging device.

20. A system for selecting an intraocular lens to be implanted into an eye, the system comprising: a controller having a processor and a tangible non-transitory memory, the memory having instructions recorded thereon; wherein the controller is configured to selectively execute a plurality of machine learning models, including a first input machine learning model, a second input machine learning model, a third input machine learning model, and an output machine learning model; wherein execution of the instructions by the processor causes the controller to: receive at least one preoperative image of the eye, and extract, via the first input machine learning model, a first data set based in part on the at least one preoperative image; receive a plurality of biometric parameters of the eye, and extract, via the second input machine learning model, a second data set based in part on the plurality of biometric parameters; access a historical pair of respective preoperative and postoperative images, and extract, via the third input machine learning model, a third data set based in part on the historical pair; combine the first data set, the second data set, and the third data set to obtain a hybrid data set; generate, via the output machine learning model, at least one output factor based on the hybrid data set; and select the intraocular lens based in part on the at least one output factor; and wherein the at least one preoperative image is obtained from a first imaging device, and the plurality of biometric parameters is obtained from a second imaging device, the first imaging device being different from the second imaging device.

Images