Machine learning system and method for intraocular lens selection

Abstract

Certain embodiments disclosed herein provide an apparatus and corresponding method for guiding the selection of an IOL. A machine learning model is trained to generate output parameters based on patient attribute data including dimensions of a patient's eye and characteristics of an IOL. The output parameters can be an emmetropic IOL power (EIOL) corresponding to an IOL that, when implanted to replace an implanted IOL, would reduce postoperative refractive error to zero. The machine learning model is trained using data from past IOL implantation surgeries, including patient attribute data (dimensions of the eye and power of the implanted IOL), and output parameters derived by summing the power of the implanted IOL and postoperative refractive error scaled by a scaling factor.

Description

Technical Field

[0001] The present disclosure generally relates to a method for selecting the power of an intraocular lens (IOL) to minimize postoperative refractive errors after cataract surgery.

Background Art

[0002] Light entering the human eye passes through the transparent cornea that covers the iris and pupil of the eye. The light passes through the pupil and is focused by the lens located in a structure called the lens capsule behind the pupil. The light is focused by the lens onto the retina, which comprises rods and cones that can generate nerve impulses in response to light.

[0003] The lens may become cloudy due to aging or disease. This is a disease known as cataract. Cataracts and other diseases are easily treated by removing the lens and inserting an artificial lens known as an intraocular lens (IOL). The IOL can be fabricated to further correct the aberrations of the patient's eye, such as myopia, hyperopia, spherical aberration, cylindrical aberration, and astigmatism. Although the refractive errors of the patient's eye before surgery can be easily determined, it is extremely difficult to estimate the combined refractive errors of the patient's eye after implantation and healing and the IOL.

[0004] Currently, advanced IOL power calculation formulas and models are used to facilitate the selection of the IOL for a patient's eye. These formulas and models employ preoperative and / or intraoperative data related to the patient's eye as inputs and output an estimated value of the postoperative refractive error of the patient's eye for a given IOL power. However, although widely used and providing sufficient accuracy in many cases, existing IOL power calculation formulas and models are often not accurate enough.

[0005] Facilitating the selection of the refractive properties of the IOL to minimize the postoperative refractive error of the patient's eye after IOL implantation would be an advancement in the art.

Summary of the Invention

Means for Solving the Problems

[0006] The present disclosure generally relates to a system for selecting an intraocular lens to reduce postoperative refractive anomalies.

[0007] A machine learning model is trained using a plurality of training data entries. Each training data entry may include (a) past preoperative data for a patient's eye, the past preoperative data including biometric data for the patient's eye and characteristics of an IOL implanted in the patient's eye, and (b) actual output parameters derived from both the characteristics of the IOL and the postoperative refractive anomalies of the patient's eye after implantation of the IOL. The machine learning model is trained to generate predicted output parameters for input preoperative data corresponding to a future IOL implantation surgery.

[0008] The following description and the related drawings detail specific exemplary features of one or more embodiments.

[0009] The accompanying drawings illustrate specific aspects of one or more embodiments and, therefore, should not be considered as limiting the scope of the present disclosure.

Brief Description of the Drawings

[0010]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7A

Figure 7B

Figure 8

[0011] For purposes of facilitating understanding, the same reference numerals are used where possible to indicate common identical elements among the drawings. It is contemplated that elements and features of one embodiment may be beneficially incorporated into other embodiments without further recitation.

[0012] As described above, currently, many IOL power calculation formulas and models are widely used in the industry. However, existing IOL power calculation formulas and models are not accurate enough when estimating or predicting postoperative refractive anomalies for a given IOL power. For example, various machine learning models have been developed that take specific patient data as input and predict or recommend an IOL power that would minimize the patient's postoperative refractive anomaly. However, many of these machine learning models are trained to predict postoperative refractive anomalies, which means that the postoperative refractive anomaly itself is the output parameter. However, there are technical challenges in training a machine learning model based on the postoperative refractive anomaly as the output parameter. In particular, the technical challenges associated with such machine learning models are caused by the nature of the postoperative refractive anomaly, which, as a data point, generally has a very variable value. Therefore, training a machine learning model to predict postoperative refractive anomalies can directly result in overfitting, where the machine learning model simply reproduces the random variations of the postoperative refractive anomaly for the training data entries.

[0013] Accordingly, embodiments herein provide a system and method for training one or more machine learning models for guiding the selection of an IOL, where the one or more machine learning models are trained using output parameters derived from postoperative refractive anomalies, thereby providing a technical solution to the above-described technical challenges. For example, the training data set can be used including training data entries, each including various patient data and an output parameter derived from the measured postoperative refractive anomaly. In particular, for each training data entry, the output parameter can be obtained from a combination of one or more parameters that describe the implanted IOL and the postoperative refractive anomaly. For example, the output parameter can be calculated for each training data entry by combining the postoperative refractive anomaly with the power of the IOL implanted in the patient's eye. Therefore, the trained machine learning model will provide a smoother and more reliable prediction with respect to variations in IOL power. In one example, the output parameter can be the emmetropic IOL (EOIL) power, which is calculated as described in more detail below.

[0014] Figure 1 shows an eye 100 including an outer layer, shown as the sclera 102. The cornea 104 is a clear layer at the front of the eye. The cornea 104, in cooperation with the lens 106, focuses light onto the retina 108, which includes photosensitive nerve cells. The lens 106 is housed within the capsular bag 110. The iris 112 is located between the cornea 104 and the lens 106.

[0015] Refractive abnormalities of the eye 100 are based on various dimensions of various optical components of the eye 100 and can be used according to the methods described herein to guide the selection of an IOL that minimizes postoperative refractive abnormalities. These dimensions can include some or all of the average power K of the cornea 104, the white-to-white distance (WTW), the anterior chamber depth (ACD), the lens thickness (LT), the axial length (AL), and / or other dimensions. WTW can be defined as the diameter of the opening in the region of the sclera 102 occupied by the cornea 104 and the iris 112. ACD can be defined as the distance between the cornea 104 and the anterior pole (outermost point) of the lens 106. LT can be defined as the thickness of the lens 106 along the optical axis of the eye 100. AL can be defined as the distance between the anterior pole of the cornea 104 and the retina 108, specifically the Bruch's membrane of the retina 108.

[0016] The dimensions of the eye 100 can be measured using one or more preoperative and / or intraoperative imaging devices such as an optical coherence tomography (OCT) device, a rotating camera (e.g., a Scheimpflug camera), a magnetic resonance imaging (MRI) device, a keratometer, a tonometer, an optical biometer, an intraoperative aberrometer, and / or any other imaging device or technique. The dimensions of the eye 100 can also be characterized by evaluating an image of the eye. For example, the image 100 can be an OCT image that can be processed using a machine learning model to derive an array of values representing the eye 100. The machine learning model can be embodied as a convolutional neural network (CNN), a deep neural network (DNN), or other types of machine learning models.

[0017] The machine learning model can be implemented as, for example, an autoencoder trained to generate an array of values. In some embodiments, the array of values is the output of an intermediate layer (i.e., not the first or last layer) of a machine learning model trained to perform tasks such as estimating refractive anomalies from images of the eye. OCT images, or anatomical labels of the eye 100 in the OCT images, can be additionally or alternatively processed by conversion means such as Fourier transform, Karhunen-Loeve (K-L) transform, etc. to obtain an array of values. OCT images, or anatomical labels of the eye 100 in the OCT images, can be additionally or alternatively processed to obtain a Gaussian degree and / or the anatomical principal curvature of the eye 100 to obtain an array of values. OCT images, or anatomical labels of the eye 100 in the OCT images, can be additionally or alternatively reduced using a task recognition network (CogNet) to obtain an array of values.

[0018] As described above, certain embodiments of the present specification provide one or more machine learning models trained using output parameters derived from postoperative refractive anomalies. To train the one or more machine learning models, one or more training data sets including patient data related to many different patients (e.g., hundreds, thousands, or more patients) can be used. Each of the one or more training data sets can include many training data records or entries, each related to a different patient.

[0019] Figure 2 shows an exemplary training data entry 200 including patient data that can be used for training a machine learning model to provide guidance for the selection of an IOL. The exemplary training data entry 200 can be provided for each eye that has been surgically treated. As shown, the training data input 200 includes patient attributes 204 including patient demographic data, preoperative biometric data, intraoperative biometric data, and the power of the implanted IOL, postoperative refractive anomaly 202, and derived output parameter 206.

[0020] Pre-operative biometric data may include any of the dimensions described above with respect to FIG. 1, an array of values derived from an image of the eye 100 as described above, and / or diagnostic / imaging data (or information derived therefrom) provided by a pre-operative diagnostic / imaging system and device. Intra-operative biometric data may include any of the dimensions described above with respect to FIG. 1, an array of values derived from an image of the eye 100 as described above, and / or diagnostic / imaging data (or information derived therefrom) provided by an intra-operative imaging system and device. Patient attributes 204 may also include the power and / or type of the IOL actually implanted during the surgery. The IOL power may be selected before the surgery is started or may be selected based on one or more measurements of refractive anomalies obtained during the surgery, such as after the removal of the lens. The IOL power may be measured in diopters. The IOL power may be enhanced by other attributes of the IOL, such as the characteristics of the IOL that correct astigmatism or spherical aberration.

[0021] Patient attributes 204 may further include demographic data. The patient's outcome may have some correlation with the patient's age, gender, ethnicity, race, and other demographic factors. Accordingly, some or all of these items of demographic data may be included in patient attributes 204. Also, other values such as the identifier of the surgeon, the clinic where the surgery was performed, the location (country, state, city, etc.) where the surgery was performed, etc. may also be used as patient attributes 204 that may correlate with the post-operative refractive anomaly 202.

[0022] Post-operative refractive anomaly 202 may include the spherical equivalent value (SE) of the patient's eye after implantation of an IOL at the power specified in the training data entry 200. Other metrics of post-operative refractive anomaly 202 may also include other post-operative refractive anomalies such as astigmatism and spherical aberration. Each training data input 200 may further include a derived output parameter 206 that is used as a desired output of the training data input for which the machine learning model is trained to generate. In certain embodiments, the output parameter is a function of the post-operative refractive anomaly 202 and one or more data points provided as part of the patient attributes 204, such as the IOL power. For example, the derived output parameter 206 may be the emmetropic equivalent IOL power (EIOL), as defined below.

[0023] Figure 3 shows an exemplary method 300 for training a machine learning model to guide the selection of an IOL based on a derived output parameter (e.g., EIOL). The machine learning model may be trained using a training data set that includes training data entries such as training data entry 200. Note that the training data set may be updated periodically. For example, information regarding new patients may be collected periodically and used to update the training data set. It should be noted that method 300 is performed separately for each type of IOL, or for each class of IOL types, and may result in a machine learning model trained for each type of IOL (e.g., monofocal, a set of specific monofocal IOL models, multifocal, toric, etc.).

[0024] Method 300 may include, at step 302, receiving the patient attributes 204 and the post-operative refractive anomaly 202 for each training data entry 200. In the following discussion, the training of the "machine learning model" will be described with the understanding that the machine learning model may be composed of multiple machine learning models that perform the tasks described in relation to, for example, FIGS. 4, 6, 7A, and 7B, and the corresponding descriptions, either individually or together.

[0025] Method 300 may include, for each set of patient attributes 204 and postoperative refractive anomalies 202 in corresponding data input 200, deriving output parameter 206 from one or more items of postoperative refractive anomalies 202 and patient attributes 204, such as the IOL power of the implanted IOL, at step 304. For example, the derived output parameter 206 may be EIOL, which is an approximation of the IOL power that would correct the postoperative refractive anomalies 202 of the corresponding patient if used instead of the IOL power of the implanted IOL.

[0026] In one example, IOL D is taken as the IOL power in diopters and R is taken as the postoperative refractive anomaly 202 in the form of a spherical equivalent value. Then, EIOL can be calculated as IOL D + p*R, where p is a scaling factor. The value of p can be a fixed predetermined amount. For example, p can be a value between 0.64 and 0.72, between 0.66 and 0.7, or between 0.675 and 0.685. In experiments conducted by the inventors, it has been found that for most applications, p = 0.68 is suitable. The value of p is an adjustable parameter and can be adjusted to improve the results. The value of p can be IOL the result of a more complex function of one or both of D and R, and other patient attributes 204. Once EIOL is calculated, it is then added to the corresponding patient data record.

[0027] The postoperative values of astigmatism or spherical aberration (as opposed to the spherical equivalent value that represents both values), or any other type of refractive anomaly, can be processed in a similar manner, i.e., the output parameter can be derived as the IOL characteristic that compensates for the type of refractive anomaly, combined (e.g., summed) with the product of the postoperative value for the type of refractive anomaly and the scaling factor. The scaling factor can be selected using a known relationship between the type of refractive anomaly and the IOL characteristic that compensates for that type of refractive anomaly, or can be determined experimentally. In the following description, the spherical equivalent value is considered with the understanding that other types of refractive anomalies can be substituted and the IOL characteristics for compensating that type of refractive anomaly can be selected using the embodiments disclosed herein.

[0028] Each training data input 200 may include a set of patient attributes 204 and an EIOL as the desired output, where the EIOL is derived from the patient attributes 204 and the corresponding postoperative refractive anomaly 202. To train the machine learning model, thousands of training data entries may be used. The machine learning model can then be trained in step 306 by processing the patient attributes 204 using the machine learning model to obtain an estimated EIOL for each training data entry. The estimated EIOL is compared to the EIOL of the corresponding training data entry, and one or more weights of the machine learning model can be modified by the training algorithm according to the difference between the actual EIOL and the estimated EIOL of the training data entry. As will be described in connection with FIG. 4, once the machine learning model is trained, the machine learning model can be deployed and used such that the difference between the estimated EIOL and the EIOL of the training data entry is minimized or approaches zero.

[0029] FIG. 4 shows an exemplary method 400 for utilizing a machine learning model trained according to method 300 to assist in selecting an IOL power for a future surgery on a patient's eye. Method 400 may include, at step 402, receiving patient attributes 204 of the patient's eye that do not include the IOL power. Method 400 may include, at step 404, selecting an initial IOL power. The initial IOL power can be selected using any technique known in the art, including any number of base selection criteria such as Barrett-style, refractive convergence-style, etc., based on preoperative and / or intraoperative biometric data.

[0030] Method 400 may include, at step 406, generating an IOL set. The IOL set may include a range of IOL powers, including IOL powers that are smaller or larger than the initial IOL power. If sufficient computing power and memory are available, step 404 may be omitted and the IOL set may simply be the entire set of available IOL powers. The IOL powers of the IOL set may be restricted to commercially available IOL powers that are typically available in 0.5 diopter increments.

[0031] Method 400 may include, at step 408, predicting, for each IOL power in the IOL set, a postoperative refractive anomaly based on the EIOL predicted by a machine learning model. The input to the machine learning model at step 408 may include the patient attributes 204 from step 402 and each of the set of IOL powers. For example, the predicted postoperative refractive anomaly may be calculated as (EIOL - D IOL ) / p, where D IOL is the IOL power from the IOL, and EIOL is the value predicted by the machine learning model. The predicted postoperative refractive anomaly may then be displayed, at step 410, along with the corresponding IOL power, such as on a display of a computing device. If the IOL set is very large, step 410 may include displaying only the N IOL powers from the IOL set that have the lowest N postoperative refractive anomalies, where N is an integer such as a value from 3 to 20.

[0032] In this way, a surgeon or other expert can easily observe whether the available IOL powers, for example, will result in the lowest postoperative refractive anomalies. Experiments conducted by the inventors have shown that using the postoperative refractive anomaly simply as the desired output when training the machine learning model is not effective due to a large amount of noise within the measured postoperative refractive anomalies. Therefore, improved results have been obtained by using derived output parameters, such as EIOL derived from the characteristics of the IOL and the postoperative refractive anomaly.

[0033] FIG. 5 shows an exemplary interface 500 for utilizing the machine learning model described in connection with FIGS. 3 and 4. The interface 500 may include some or all of a field 502 for entering a patient identifier, a field 504 for entering the eye (right or left) to be operated on, a field 506 for entering the IOL type, a field 508 for selecting a machine learning model from among a plurality of available machine learning models, one or more fields 510 for entering biometric measurement data of the patient's eye (optionally, selecting the dimensions or other biometric measurement data to be used), and a field 512 for entering a target postoperative refractive anomaly. Fields for entering other patient attributes such as age, gender, ethnicity, eye identifier (right or left), surgeon name, clinic name, and / or location may also be provided. Additionally, other fields for entering intraoperative data parameters, including intraoperative measurements of the aphakic eye, may be provided. Data obtained from other systems or devices (e.g., imaging systems, clinic servers, health databases, electronic medical record (ERM) systems, surgical consoles, digital microscopes, etc.) may be manually entered or automatically added to the fields of the interface 500.

[0034] The interface 500 may include a result section 514 that presents an IOL power 516 and a predicted postoperative refractive anomaly 518 for each IOL power within the IOL set. The predicted refractive anomaly 518 may be measured in diopters such as a value (E IOL - D IOL ) / p.

[0035] Referring to FIGS. 6, 7A, and 7B, the machine learning model trained according to method 300 and utilized according to method 400 may have various forms such as those shown in FIGS. 6, 7A, and 7B. However, these are merely exemplary, and any machine learning model or artificial intelligence model known in the art may be trained to perform the described tasks.

[0036] FIG. 6 shows a method 600 for using clustering to improve the accuracy of the machine learning model described in connection with FIGS. 3 and 4. Method 600 may include, at step 602, clustering of training data entries based on values for one or more of patient attributes 204, postoperative refractive anomalies 202, and derived output parameters 206 (e.g., EIOL) in each data entry. The clustering may include using a k-nearest neighbor (KNN) algorithm. Other clustering techniques may be used, such as k-means, Gaussian mixture models, centroid-based clustering, density-based clustering, distribution-based clustering, hierarchical clustering, etc.

[0037] Method 600 may then include, at step 604, receiving patient attributes 204 for a patient's eye for which an IOL is selected for a future surgery. Patient attributes 204 may include an IOL power selected from an IOL lens set as described above with respect to FIG. 4. The method may include, at step 606, identifying a cluster associated with patient attributes 204. Patient attributes 204 may then be processed, at step 608, using a machine learning model specific to the cluster identified at step 606 to generate a predicted EIOL (or another derived output parameter). By using a machine learning model trained using only the training data entries within the cluster associated with patient attributes 204, the prediction accuracy of the machine learning model may be improved relative to a machine learning model for different clusters or a machine learning model trained on a larger, less specific set of training data entries.

[0038] For example, step 606 may include processing the patient attribute 204 according to KNN, which is itself a machine learning model, to identify the cluster (i.e., K nearest neighbors) of the training data entries in the training data set. Then, in step 608, it may include processing the patient attribute 204 according to a second machine learning model trained using that cluster of training data entries to output the predicted EIOL. The second machine learning model may include a multiple linear regression (MLR) model (KNN+MLR), or a random sample consensus (RANSAC) regression model (KNN+RAN). The second machine learning model may also include any machine learning model known in the art that is trained using the cluster of training data entries, such as a DNN, CNN, multiple polynomial regression (MPR) (quadratic, cubic, or higher order), support vector regression model (SVM), or SVM radial bias function (SVM-RBF).

[0039] Referring to FIG. 7A, the machine learning model trained in step 306 may include the illustrated ensemble machine learning model 700a. The ensemble machine learning model 700a is a composite of a plurality of machine learning models 702a-702e. Each machine learning model 702a-702e may be of a different type. Non-limiting examples of these types include KNN, MLR, KNN+MLR, KNN+RAN, DNN, multiple polynomial regression (MPR) (quadratic, cubic, or higher order), support vector regression model (SVM), SVM radial bias function (SVM-RBF).

[0040] Each of the machine learning models 702a - 702e can be trained individually according to method 300. The outputs of the machine learning models 702a - 702e, along with the input (patient attribute 204), are passed to the blending algorithm 704. The blending algorithm 704 can be any type of machine learning model, such as any of the types of machine learning models discussed herein. The blending algorithm 704 can be trained to either (a) select from among the outputs of the machine learning models 702a - 702e or (b) combine the outputs of the machine learning models 702a - 702e to generate a final result, such as a predicted EIOL.

[0041] FIG. 7B similarly shows a gradient - boosted machine learning model 700b composed of a plurality of different types of machine learning models 702a - 702e, such as any machine learning model type referred to herein. The first machine learning model 702a takes the patient attribute 204 as input and generates a predicted EIOL along with a residual anomaly 706a. Each of the other machine learning models 702b - 702e takes the patient attribute 204 and the residual anomalies 706a - 706d from the preceding stage as input. Then, the output of the final machine learning model 702e can be the final EIOL prediction 708.

[0042] The ensemble machine learning model 700a and the gradient - boosted machine learning model 700b can provide limited benefits over any one of the machine learning models 702a - 702e alone. In experiments conducted by the inventors, it has been found that the ensemble machine learning model 700a or the gradient - boosted machine learning model 700b can increase the percentage of predictions that are accurate within 0.5 diopters by up to about 2 percent at most.

[0043] Tables 1 and 2 list the results for various individual machine learning models in the form of the percentage of predicted EIOLs within 0.5 diopters of the actual EIOLs as provided in the data entries. The results in Tables 1 and 2 were obtained using patient attributes 204 and postoperative refractive abnormalities 202 of patients receiving the Acrysof single - focus intraocular lens. The results in Table 1 were obtained using data from 17,500 patients, and the data included three biometric variables (AL, K, and WTW) for each patient. The results in Table 2 were obtained using data from 5,277 patients. For the entries labeled "(Ex)", ACD and LT were additionally used for each patient. Another metric that characterizes the accuracy of the machine learning model is the root mean square (RMS) of the anomalies (predicted EIOL - EIOL from the training data entry).

[0044]

Table 1

[0045]

Table 2

[0046] FIG. 8 shows an exemplary computing system 800 that at least partially implements one or more of the functions described herein in response to input to interface 500. Computing system 800 may also implement methods 300, 400, and / or method 600.

[0047] As shown, computing system 800 includes a central processing unit (CPU) 802, one or more I / O device interfaces 804 that can connect various I / O devices 814 (e.g., keyboard, display, mouse device, pen input, etc.) to computing system 800, a network interface 806 through which computing system 800 is connected to a network 890 (which can be a local network, intranet, Internet, or any other group of computing devices communicatively connected to each other), a memory 808, a storage 810, and an interconnect 812.

[0048] CPU 802 can obtain and execute programming instructions stored in memory 808. Similarly, CPU 802 can read and store application data within memory 808. Interconnect 812 transmits programming instructions and application data among CPU 802, I / O device interface 804, network interface 806, memory 808, and storage 810. CPU 802 is included as representative of a single CPU, multiple CPUs, a single CPU with multiple processing cores, and the like.

[0049] Memory 808 represents volatile memory such as random access memory and / or non-volatile memory such as non-volatile random access memory or phase change random access memory. As shown, memory 808 can store a data preparation module 816 for calculating output parameters (e.g., EIOL) derived based on patient attribute data, and a training algorithm 818 for training a machine learning model to predict the derived output parameters. Memory 808 can store a prediction module 820 that uses a machine learning model to predict output parameters derived based on patient attributes and IOL power.

[0050] Storage 810 can be non-volatile memory such as a disk drive, a solid state drive, or a collection of storage devices distributed across multiple storage systems. Storage 810 can optionally store one or more machine learning models 822 trained as described above, and training data 824 for training the machine learning models 822.

[0051] Additional Considerations The foregoing description has been provided to enable a person skilled in the art to practice the various embodiments described herein. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments. For example, changes may be made to the functions and arrangements of the elements discussed without departing from the scope of the disclosure. In various examples, various procedures or components may be omitted, replaced, or added as necessary. Features described in connection with some examples may be combined in other examples. For example, an apparatus may be implemented or a method may be practiced using any number of the aspects described herein. Further, the scope of the disclosure is intended to cover such apparatus or methods implemented using other structures, functions, or structures and functions in addition to or other than the various aspects of the disclosure described herein. It should be understood that any aspect of the disclosure disclosed herein may be embodied by one or more of the elements recited in the claims.

[0052] As used herein, the phrase "at least one of" a list of items refers to any combination of those items including a single element. By way of example, "at least one of a, b, or c" is intended to cover a, b, c, a - b, a - c, b - c, and a - b - c, as well as any combination of multiple like elements (e.g., a - a, a - a - a, a - a - b, a - a - c, a - b - b, a - c - c, b - b, b - b - b, b - b - c, c - c, and c - c - c or any other arbitrary order of a, b, and c).

[0053] As used herein, the term "determine" encompasses a variety of operations. For example, "determine" can include calculating, computing, processing, deriving, investigating, searching (e.g., searching a table, database, or other data structure), ascertaining, etc. "Determine" can also include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory), etc. "Determine" can further include solving, selecting, choosing, specifying, etc.

[0054] The methods disclosed herein include one or more steps or operations for implementing the methods. The method steps and / or operations may be interchangeable with each other without departing from the scope of the claims. In other words, the specific order and / or use of the specific steps and / or operations may be changed without departing from the scope of the claims, unless the specific order of the steps or operations is specified. Further, the various operations of the above methods may be performed by any suitable means capable of performing the corresponding functions. These means 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, where there are operations shown in the drawings, those operations may include corresponding equivalent means-plus-function components with similar numbers.

[0055] The various illustrative logical blocks, modules, and circuits described in connection with the present disclosure can be implemented or performed 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 gates or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the 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, such as 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.

[0056] The processing system can be implemented with a bus architecture. The bus may include any number of interconnected buses and bridges depending on the specific application of the processing system and the overall design constraints. The bus can link together various circuits including, among other things, the processor, the machine readable medium, and the input / output devices. A user interface (e.g., keypad, display, mouse, joystick, etc.) can also be connected to the bus. The bus can also link together various other circuits well known in the art (and thus not described further herein) such as a timing source, peripherals, voltage regulators, power management circuits, and the like. The processor can be implemented with one or more general purpose and / or special purpose processors. By way of example, microprocessors, microcontrollers, DSP processors, and other circuits capable of executing software are included. Those of skill in the art will recognize how best to implement the described functionality for the processing system in view of the particular application and overall design constraints imposed on the overall system.

[0057] When implemented in software, the functions may be stored or transmitted as one or more instructions or codes on a computer-readable medium. Software is broadly construed to mean instructions, data, or any combination thereof, whether called software, firmware, middleware, microcode, or hardware description language, etc. The computer-readable medium includes both computer storage media and communication media such as any medium that facilitates transfer of a computer program from one place to another. The processor may be in charge of the management of a bus and general processing, including the execution of software modules stored in a computer-readable storage medium. The computer-readable storage medium may be coupled to the processor so that the processor can read information from, and write information to, the storage medium. Alternatively, the storage medium may be integrated with the processor. By way of example, the computer-readable medium may include a computer-readable storage medium having instructions stored thereon separate from a transmission line, a carrier wave modulated by data, and / or a wireless node, all of which may be accessed by the processor via a bus interface. Alternatively or additionally, the computer-readable medium or any portion thereof may be integrated with the processor, such as in cases where it may be accompanied by a cache and / or a general register file. Examples of machine-readable storage media include, by way of example, 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 medium or any combination thereof. The machine-readable medium can be embodied in a computer program product.

[0058] A software module can include a single instruction or many instructions and can be distributed across several different code segments, between different programs, and across multiple storage media. A computer-readable medium can include several software modules. A software module includes instructions that, when executed by an apparatus such as a processor, cause a processing system to perform various functions. A software module can include a sending module and a receiving module. Each software module can reside on a single storage device or can be distributed across multiple storage devices. As an example, when a trigger event occurs, a software module can be read from a hard drive into RAM. During execution of a software module, the processor can read some of the instructions into a cache to increase access speed. Next, one or more cache lines can be read into a general-purpose register file for execution by the processor. When referring to the functions of a software module, it will be understood that such functions are implemented by the processor when executing instructions from that software module.

[0059] The following claims are not intended to be limited to the embodiments shown in this specification, and the full scope corresponding to the language of the claims should be recognized. In the claims, a reference to an element in the singular is not intended to mean "one and only one" unless specifically so defined, but rather "one or more." Unless specifically stated otherwise, the term "some" refers to one or more. No element of a claim is to be construed under the provisions of 35 U.S.C. § 112(f) unless the element is expressly recited using the phrase "means for" or, in the case of a method claim, the element is recited using the phrase "steps for." All structural and functional equivalents to the elements of the various aspects described throughout this disclosure that are known or later come to be known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be encompassed by the claims. Further, nothing disclosed herein is intended to be dedicated to the public regardless of whether such disclosure is explicitly recited in the claims.

Claims

1. A system for guiding the selection of an intraocular lens (IOL), comprising: one or more processing devices; and one or more memory devices operably coupled to the one or more processing devices, which, when executed by the one or more processing devices, cause the one or more processing devices to: receive patient attribute data; receive a set of one or more IOL characteristics describing one or more IOLs; for each IOL characteristic of the one or more IOL characteristics: use a machine learning model to process the patient attribute data and the respective IOL characteristic to obtain an output parameter; calculate a refractive anomaly predicted from the combination of the respective IOL characteristic and the output parameter; output the predicted refractive anomaly for each IOL of the one or more IOLs; and one or more memory devices storing executable code. A system comprising the above.

2. The system according to claim 14, wherein the combination of each IOL characteristic and the output parameter includes a difference between the output parameter and the IOL characteristic.

3. The system according to claim 15, wherein the combination of each IOL characteristic and the output parameter includes a difference between the output parameter and the IOL characteristic, and the difference is scaled by a parameter between 0.66 and 0.

7.

4. The system according to claim 15, wherein the patient attribute data includes the dimensions of the patient's eye.

5. The system according to claim 17, wherein the dimensions of the patient's eye are any one of white-to-white distance (WTW), axial length (AL), average corneal power (K), lens thickness (LT), and anterior chamber depth (ACD).

6. The set of one or more IOL characteristics for the one or more IOLs includes a plurality of IOL characteristics for a plurality of IOLs, and when the executable code is executed by the one or more processing devices, the one or more processing devices are further caused to: calculate initial IOL characteristics based on the dimensions of the patient's eye; and select the plurality of IOL characteristics based on the initial IOL characteristics. The system according to claim 18.

7. The machine learning model is: a K-nearest neighbor model; a multiple linear regression model; a random sample consensus model; a multiple polynomial regression model; a support vector machine-radial bias function; and a deep neural network model. ​ One or more machine learning models selected from a group including The system according to claim 14.

Images