Systems and methods for predicting contact lens compatibility using machine learning

Abstract

The present application relates to systems and methods for predicting contact lens compatibility using machine learning. Systems and methods for determining compatibility between a multifocal contact lens and a patient seeking presbyopic vision correction include receiving, from a first device associated with a first eye care professional, a request to select a contact lens for a consumer, wherein the request includes biometric / ophthalmic information associated with the consumer; obtaining a performance metric associated with the first ECP; determining, using the machine learning model and based on the performance metric, a customized compatibility index that indicates compatibility between a particular contact lens and the consumer for the first ECP; and presenting, on the first device, a report indicating the compatibility index. Additional systems, methods, and non-transitory machine-readable media are also provided. The idea is to replace or at least supplement the work of an ophthalmologist with an automated system.

Description

[0001] This application is a divisional application of Chinese patent application PCT / IB2020 / 055389, filed on June 8, 2020, with application number 202080046027.8 and application date of June 8, 2020, entitled "System and method for predicting contact lens compatibility using machine learning". Technical Field

[0002] This disclosure relates to the use of machine learning to predict lens compatibility for different users according to various embodiments of this disclosure. Background Technology

[0003] In recent years, the technology for manufacturing contact lenses has improved significantly, making most contact lenses easily fit most users. However, certain types of contact lenses (such as multifocal contact lenses) remain difficult to fit. Typically, fitting a user with contact lenses (e.g., multifocal contact lenses) is a lengthy process. An ophthalmologist (ECP) will recommend a particular type of contact lens (e.g., a specific prescription, specific power, specific shape (such as curvature), specific manufacturer, etc.) for a trial period (e.g., using the recommended type of contact lens for a period such as two weeks). The user may have a follow-up appointment after the trial to determine if the type of contact lens is a good fit, and / or to have another prescription to determine which type of contact lens is more suitable. Due to the difficulty in fitting lenses and because ECPs cannot accurately predict the outcome of using contact lenses (e.g., multifocal contact lenses) for users seeking vision correction (e.g., presbyopia correction), users may need to visit the ECP multiple times before determining that a particular type of contact lens is a good fit (or that no type of contact lens is suitable). As a result, the cost of prescribing contact lenses to users becomes unnecessarily high, and in some cases, ECPs become reluctant to recommend certain types of contact lenses and / or prescribe them to users who might benefit from such innovative vision correction solutions. Therefore, there is a need in the art for techniques to better predict outcomes for users of multifocal contact lenses. Summary of the Invention

[0004] According to some embodiments, a system includes one or more hardware processors. The one or more hardware processors are configured to: receive a request from a first device associated with a first ophthalmic care professional (ECP) to select a contact lens for a consumer, wherein the request includes biometric information associated with the consumer; obtain a performance metric associated with the first ECP; determine a customized compatibility index using the machine learning model and based on the performance metric, the customized compatibility index indicating compatibility between a particular contact lens and the consumer for the first ECP; and present a report indicating the compatibility index on the first device.

[0005] According to some embodiments, a method includes: receiving, by one or more hardware processors, a request from a first device associated with a first ophthalmic care professional (ECP) to determine compatibility between a particular lens and a patient, wherein the request includes attributes associated with the patient; comparing performance information associated with the first ECP with performance information associated with at least a second ECP by the one or more hardware processors; determining a performance metric associated with the first ECP based on the comparison by the one or more hardware processors; selecting, by the one or more hardware processors, a specific compatibility index representing the degree of compatibility between the particular lens and the patient from a plurality of compatibility indices and based on attributes associated with the patient, wherein the compatibility index is tailored for the first ECP at least in part based on the performance metric associated with the first ECP; and presenting a report indicating the specific compatibility index on the first device by the one or more hardware processors.

[0006] According to some embodiments, a non-transitory machine-readable medium includes multiple machine-readable instructions executable to cause a machine to perform operations including: receiving a request from a first device associated with a first ophthalmic care professional (ECP) to select a contact lens for a consumer, wherein the request includes biometric information associated with the consumer; obtaining a performance metric associated with the first ECP; using the machine learning model and based on the performance metric to determine a customized compatibility index, the customized compatibility index indicating compatibility between a specific contact lens and the consumer for the first ECP; and presenting a report indicating the compatibility index on the first device. Attached Figure Description

[0007] To gain a more comprehensive understanding of this technology, its features and advantages, please refer to the following description given in conjunction with the accompanying drawings.

[0008] Figure 1 This is a diagram of a system for determining lens compatibility according to some embodiments.

[0009] Figure 2 This is a block diagram of a prediction engine based on some embodiments.

[0010] Figure 3 The process for determining the compatibility between a lens and a patient, according to some embodiments, is illustrated.

[0011] Figure 4A Exemplary input interfaces according to some embodiments are shown.

[0012] Figure 4B An exemplary reporting interface according to some embodiments is shown.

[0013] Figure 4C An exemplary feedback interface according to some embodiments is shown.

[0014] Figure 4D Exemplary presentation interfaces according to some embodiments are shown.

[0015] Figure 4E Another exemplary presentation interface according to some embodiments is shown.

[0016] Figure 5 The process for determining performance metrics for ophthalmic care professionals, according to some embodiments, is illustrated.

[0017] Figure 6A and Figure 6B This is a diagram of a processing system according to some embodiments.

[0018] Figure 7 This is a diagram of a multilayer neural network according to some embodiments.

[0019] In the accompanying drawings, elements with the same reference numerals have the same or similar functions. Detailed Implementation

[0020] The descriptions and drawings illustrating aspects, embodiments, implementations, or modules of the invention should not be considered limiting—it is the claims that define the protected invention. Various mechanical, compositional, structural, electrical, and operational changes may be made without departing from the spirit and scope of this specification and the claims. In some cases, well-known circuits, structures, or techniques have not been shown or described in detail so as not to obscure the invention. Similar numbers in two or more figures denote the same or similar elements.

[0021] In this description, specific details of some embodiments consistent with this disclosure are set forth. Numerous specific details are set forth to provide a thorough understanding of the embodiments. However, it will be apparent to those skilled in the art that some embodiments may be practiced without some or all of these specific details. The specific embodiments disclosed herein are intended to be illustrative and not limiting. Those skilled in the art can implement other elements that, while not specifically described herein, are within the scope and spirit of this disclosure. Furthermore, to avoid unnecessary repetition, one or more features shown and described in association with one embodiment may be incorporated into other embodiments unless otherwise specifically stated or if such one or more features would render the embodiment inoperable.

[0022] The techniques described below relate to systems and methods for using machine learning to predict the compatibility of one or more types of lenses (e.g., one or more types of contact lenses) for a user. As discussed above, specific types of lenses (e.g., for a specific prescription, with a specific power, associated with a specific manufacturer, with a specific geometry (such as curvature), etc.) may have different fits for different users, and determining whether a particular type of lens is suitable for a patient is a challenge for ophthalmic care professionals (ECPs). This type of lens can be a contact lens (e.g., a multifocal contact lens, etc.) or any type of lens used by the patient. Thus, in some embodiments, a lens compatibility system can be provided to assist ECPs in recommending different types of lenses and / or prescribing different types of lenses to patients. In some embodiments, the lens compatibility system may provide a user interface (e.g., a graphical user interface (GUI), etc.) on the ECP's device. The user interface can be configured to receive compatibility requests from one or more ECPs for determining whether a certain type of lens is compatible with the patient. Compatibility requests can include patient attributes such as age, dry eye index, computer usage time and frequency, user preferences (e.g., comfort level with regular glasses), user motivation for using this type of lens, user arm length (indicating reading distance), whether the patient has red-eye, current prescription data, and eye measurements such as pupil diameter. Lens compatibility systems can use these patient attributes to determine compatibility with the patient's lens type.

[0023] In some embodiments, the lens compatibility system may use a machine learning model to determine the compatibility between a particular type of lens and a patient. For example, the machine learning model may be configured to provide an output (e.g., a compatibility score) indicating the likelihood that the patient is a suitable candidate for that type of lens, based on patient attributes and other data received via a user interface. The machine learning model can be trained using historical patient data from patients previously prescribed lenses of the type associated with the request. For example, the lens compatibility system may retrieve patient records from the devices of individual ECPs for patients who have been prescribed lenses of that type by those ECPs; these patient records include patient attributes and user outcomes (e.g., obtained through surveys provided to patients after a predetermined trial period for the lens). The retrieved patient records can then be used as training data for the machine learning model. The machine learning model may be implemented using one or more techniques, such as neural networks, regression models, or any other type of machine learning algorithm.

[0024] After receiving patient attributes and other data via a user interface, the lens compatibility system can provide these attributes and data to a machine learning model as input. Based on the patient's attributes, the machine learning model can generate an output indicating the likelihood that the patient is a suitable candidate for that type of lens. In some embodiments, the machine model can generate a compatibility score as output. The compatibility score can be expressed on a scale based on the patient's attributes, indicating the degree of compatibility of that type of lens with the patient, where one end of the scale represents very compatible and the other end represents very incompatible.

[0025] Different embodiments of the lens compatibility system may configure machine learning models differently to generate compatibility scores representing different aspects of patient compatibility. For example, in some embodiments, the compatibility score may represent the patient's predicted comfort while wearing the lens. In another embodiment, the compatibility score may represent the patient's predicted visual quality. In yet another embodiment, the compatibility score may represent the predicted number of lenses to be prescribed for the patient and / or the predicted number of follow-up visits required by the patient. In yet another embodiment, the compatibility score may represent any combination of the above factors.

[0026] In some embodiments, the lens compatibility system can provide a compatibility report on the interface of an ECP device based on the output generated by a machine learning model. For example, the compatibility report can indicate a compatibility score generated by the machine learning model, which indicates the likelihood that a patient is suitable for that type of lens. Because the machine learning model is pre-trained using patient data from multiple ECPs (which may or may not include the requesting ECP), the requesting ECP can now obtain patient compatibility information not only based on past fits of that type of lens for its patients, but also based on fits of that type of lens for patients of other ECPs. The lens compatibility system's ability to access a large sample size of patients from different ECPs means improved accuracy in patient compatibility. Furthermore, patient privacy is protected because the (requesting) ECP is only provided with a compatibility report generated by the lens compatibility system, and not with any patient records used to train the machine learning model.

[0027] In some embodiments, a compatibility report may include a compatibility score generated by a machine learning model. However, ECPs may find it difficult to translate the compatibility score into an actual recommendation level for prescribing that type of lens to patients. Therefore, to assist ECPs in making recommendations and / or prescribing, some embodiments of the lens compatibility system may derive additional data (e.g., meaning) from the compatibility score. For example, the lens compatibility system may determine multiple compatibility indices that represent different levels of recommendation for prescribing that type of lens to patients. In a particular, but not limiting, example, the lens compatibility system may determine three distinct compatibility indices representing highly recommended, recommended, and not recommended, respectively. The compatibility indices may be associated with different, non-overlapping ranges of compatibility scores. For example, if a compatibility score is generated by a machine learning model on a scale of 1-100, a first range 71-100 may correspond to a first compatibility index representing highly recommended, a second range 41-70 may correspond to a second compatibility index representing recommended, and a third range 1-40 may correspond to a third compatibility index representing not recommended. In some embodiments, these ranges can be determined based on specific results of previous prescriptions for that type of lens to patients. In one example, the lens compatibility system can determine the first range based on a predetermined percentage (e.g., 90%) of patients who were prescribed a lens of the type with a compatibility score within a first range (which includes a compatibility score between 71 and 100) and did not require further ECP visits. The lens compatibility system can determine the second range based on a predetermined percentage (e.g., 90%) of patients who were prescribed a lens of the type with a compatibility score within a second range (which includes a compatibility score between 41 and 70) and only required one ECP visit. Thus, the lens compatibility system can determine a patient's compatibility index based on which range the compatibility score falls within.

[0028] Therefore, in some embodiments, the lens compatibility system may present a compatibility index on the interface as an alternative to or supplement to the compatibility score to assist ECPs in recommending and / or prescribing lenses of that type to patients. In some embodiments, the lens compatibility system may use the same range of compatibility scores to determine the compatibility index for each ECP. However, different ECPs may have different preferences in how they recommend lenses or prescribe lenses to their patients. In some scenarios, differences in preference may be attributed to the demographics of the patients served by the ECP, the geographic region where the ECP is located, or the ECP's own personal preferences (e.g., their aggressiveness / conservatism in prescribing that type of lens to patients). Because the machine learning model is trained using historical patient data from different ECPs, and the model does not distinguish between different ECPs, the output of the machine learning model (e.g., the compatibility score) may be general to all patients rather than specific to a particular ECP. Thus, in some embodiments, the lens compatibility system may determine different ranges corresponding to different compatibility indices for different ECPs. For example, for a first ECP, the lens compatibility system can determine that a first compatibility index (e.g., highly recommended) corresponds to the range of 81-100, a second compatibility index (e.g., recommended) corresponds to the range of 51-80, and a third compatibility index (e.g., recommended) corresponds to the range of 1-50. For a second ECP, the lens compatibility system can determine that a first compatibility index (e.g., highly recommended) corresponds to the range of 61-100, a second compatibility index (e.g., recommended) corresponds to the range of 31-60, and a third compatibility index (e.g., recommended) corresponds to the range of 1-30.

[0029] Different ranges can be determined for different ECPs based on various factors, such as the geographic region of the ECP, the demographics of the patients with the ECP, or the performance of the ECP relative to the recommendation / prescription performance of other ECPs (e.g., peers of the ECP). Thus, in some embodiments, the lens compatibility system can determine the position of the requesting ECP relative to other ECPs based on the recommendation / prescription performance of multiple ECPs. For example, the lens compatibility system can determine the ratio of recommending / prescribing that type of lens to a patient for each ECP (e.g., the ratio of prescribing multifocal lenses to patients seeking presbyopia correction). The lens compatibility system can compare this ratio of the requesting ECP with the ratios of other ECPs and can adjust the range associated with the compatibility index based on this comparison. In a specific and non-limiting example, if the lens compatibility system determines that the submitted ECP has a lower-than-average ratio compared to other ECPs, the lens compatibility system may expand the range of (multiple) compatibility indices (representing highly recommended and recommended, respectively) with respect to the first (and / or second) compatibility indices, or conversely narrow the range with respect to the first (and / or second) compatibility indices.

[0030] In another example, the lens compatibility system can determine the number of lenses prescribed to a patient and / or the number of follow-up visits (e.g., average number) for each ECP (e.g., people who were prescribed that type of lens, generally, people who were prescribed lenses by that ECP, etc.). The lens compatibility system can compare the number of lenses prescribed to a patient and / or the number of follow-up visits associated with the requesting ECP with the number of lenses prescribed to a patient and / or the number of follow-up visits associated with other ECPs, and can adjust the range associated with the compatibility index based on this comparison. In a specific and non-limiting example, if the lens compatibility system determines that the number of lenses prescribed to a patient by the requesting ECP is below average and / or the number of follow-up visits is below average compared to other ECPs, the lens compatibility system can expand the range(s) corresponding to the first (and / or second) compatibility index (representing highly recommended and recommended, respectively), and conversely narrow the range.

[0031] In some embodiments, a lens compatibility system may include one or more visual simulators associated with or communicatively coupled to different ECP offices. The visual simulator can provide a patient with a simulation of wearing and / or using this type of lens. For example, the visual simulator may include wearable devices (e.g., headbands, goggles, etc.) that provide the patient with simulated vision that the patient could have when using this type of lens in one or more environments (e.g., daytime driving, nighttime driving, reading, using electronic devices such as computers, mobile devices, etc.). In some embodiments, the visual simulator may be configured to obtain feedback from the patient (e.g., visual quality, comfort, etc.) based on the simulation. For example, the visual simulator may be configured to prompt the patient for feedback during various parts of the simulation (e.g., during simulations of different environments), enabling the patient to provide feedback for the different simulated environments. In some embodiments, the lens compatibility system may configure a machine learning model to receive feedback from the patient as input, among other attributes, and determine an output (e.g., a compatibility score) for that patient based on the patient's feedback on the visual simulation and these attributes. In some embodiments, the visual simulator may further include one or more sensors (e.g., a camera, heart rate monitor, humidity sensor, etc.) to detect and / or track patient-associated biometric data (e.g., eye movements, blink frequency, red eyes, facial expressions (such as frowning, squinting, etc.), heart rate, tearing, etc.) during the simulation, and may be configured to provide the detected biometric data to the lens compatibility system. Therefore, the lens compatibility system may also configure a machine learning model to also receive biometric data obtained from the visual simulator as input values ​​and determine the output based on the biometric data and other patient attributes.

[0032] Therefore, the lens selection system according to various embodiments of this disclosure can provide ECPs with a customizable interface to assist ECPs in recommending a certain type of lens and / or prescribing a certain type of lens to patients by using the patient experiences associated with different ECPs, without disclosing the patient's sensitive and / or private data.

[0033] Figure 1A system 100 for determining lens compatibility according to some embodiments is shown, within which a lens compatibility system can be implemented. System 100 includes a lens selection platform 102 coupled to one or more diagnostic training data sources 110 via a network 115. In some examples, network 115 may include one or more switching devices, routers, local area networks (e.g., Ethernet), wide area networks (e.g., the Internet), etc. Each diagnostic training data source 110 may be a database, data repository, etc., available through ECP practices, ophthalmology clinics, medical universities, electronic medical record (EMR) repositories, etc. Each diagnostic training data source 110 may provide training data to lens selection platform 105 in the form of patient records of patients who have been prescribed one or more types of lenses (e.g., multifocal lenses, etc.). Each patient record may include: attributes of the corresponding patient seeking vision correction associated with that type of lens (e.g., a patient seeking presbyopia vision correction, etc.) (such as age, dry eye index, duration and frequency of computer use, user preferences (e.g., patient comfort with conventional glasses, etc.), the user's motivation for using that type of lens, the user's arm length (which indicates the user's reading distance), whether the patient has red-eye, current prescription data, eye measurements (such as pupil diameter, etc.)), and the patient's outcome (which may indicate whether the patient was indeed prescribed that type of lens (e.g., multifocal contact lenses), and if the patient was indeed prescribed that type of lens, the quality of the prescription after the patient was prescribed that type of lens (e.g., comfort, visual quality level, number of lenses prescribed to the patient, number of follow-up visits, etc.)). The lens selection platform 102 may store training data in one or more databases 104, which may be configured to anonymize, encrypt, and / or otherwise secure the training data.

[0034] The lens selection platform 102 includes a prediction engine 106, which (as explained in more detail below) processes the received training data, performs raw data analysis on the training data, trains machine learning algorithms and / or models based on patient attributes to predict the compatibility between the type of lens and the patient, and iteratively refines the machine learning to optimize various models used to predict the compatibility between a certain type of lens and the patient, thereby assisting the ECP in recommending the type of lens to the patient and / or prescribing the type of lens.

[0035] Lens selection platform 102 is further coupled to one or more ECP devices, such as ECP devices 130, 140, and 150, via network 115. Each ECP device can be associated with an ECP. For example, ECP device 130 is associated with ECP 170. As shown, ECP device 130 includes a user interface application 132, which can be a web browser, desktop application, or mobile application provided by lens selection platform 102. User interface application 132 enables ECP device 130 to interact with lens selection platform 102. ECP device 130 may also include an ECP identifier 134, which can be implemented as an identifier associated with ECP 170, a device identifier associated with ECP device 130, or any other type of identifier that distinguishes ECP 170 from other ECPs.

[0036] In some embodiments, ECP device 130 may be communicatively coupled to visual simulator 160. Visual simulator 160 may include wearable devices (e.g., headbands, goggles, etc.) and is configured to provide patients with ECP 170 with a simulation of wearing and / or using lenses of this type in one or more environments (daytime driving, nighttime driving, reading, using electronic devices such as computers, mobile devices, etc.). In some embodiments, visual simulator 160 may also include one or more sensors (e.g., cameras, heart rate monitors, humidity sensors, etc.) for acquiring patient biometric data (e.g., eye movements, blink frequency, redness of the eyes, facial expressions (such as frowning, squinting, etc.), heart rate, tearing, etc.) during the simulation. Although not shown in the figures, each of ECP devices 140 and 150 may also include modules and / or components similar to those of ECP device 130. Furthermore, each of ECP devices 140 and 150 may also be coupled to a corresponding visual simulator similar to visual simulator 160.

[0037] Figure 2A block diagram of a prediction engine 106 according to some embodiments of the present disclosure is shown. The prediction engine 106 is communicatively coupled to various ECP devices, such as ECP devices 130, 140, and 150. The prediction engine 106 includes a prediction manager 202, a report generation module 204, a performance evaluation module 206, and a model configuration module 208. The model configuration module 208 can use training data stored in a database 104 to train a machine learning model (such as a lens compatibility model 212) to predict the compatibility of a certain type of lens for a patient based on information such as patient attributes, patient feedback from visual simulations, and biometric information obtained during visual simulations. The prediction manager 202 can provide a user interface (e.g., a GUI) to be presented on ECP devices 130-150. In some embodiments, an ECP (e.g., ECP 170) can use an ECP device (e.g., ECP device 130) to submit a compatibility request via a user interface. For example, when user interface application 132 receives an instruction from ECP to submit a compatibility request, user interface application 132 can present an input interface associated with prediction engine 106. This input interface may include multiple input fields that enable ECP to provide information associated with the request (e.g., lens type, patient attributes, etc.).

[0038] In some embodiments, upon receiving an instruction, the prediction manager 202 may also configure a visual simulator (e.g., visual simulator 160) associated with the ECP device to provide visual simulations to the patient based on the type of lens. The visual simulator 160 may acquire patient feedback and / or biometric data during the simulation and may provide this patient feedback and / or biometric data to the ECP device 130.

[0039] User interface application 132 can then send data associated with the compatibility request, such as information obtained via the input interface, patient feedback, and biometric data, to prediction engine 106 via network 115. Upon receiving the compatibility request and data (e.g., data 240), prediction manager 202 can use lens compatibility model 212 to determine the compatibility of that type of lens for the patient based on data 240. Lens compatibility model 212 can generate output (e.g., a compatibility score) based on data 240. Using the output from lens compatibility model 212, report generation module 204 can generate a compatibility report to be presented on the report interface of user interface application 132.

[0040] The ECP can recommend and / or prescribe the appropriate type of lens to the patient based on a report presented on the user interface application 132. After the compatibility report is presented on the ECP device, the prediction manager can present a feedback interface on the user interface application 132, allowing the ECP to provide information related to the outcome of prescribing the appropriate type of lens to the patient. For example, the ECP can provide information on the feedback interface regarding the patient's continued use of the lens for a predetermined trial period, such as comfort and visual quality levels. The ECP can also provide information such as the number of lenses prescribed to the patient and / or the number of follow-up visits after the patient was prescribed the appropriate type of lens. The prediction manager 202 can receive feedback information associated with prescribing the appropriate type of lens to the patient via the feedback interface. The prediction manager 202 can use the feedback information and patient attributes to update (e.g., train) the lens compatibility model 212.

[0041] The prediction manager 202 can provide compatibility reports for patients with different ECPs and receives corresponding feedback data through different ECP devices (e.g., ECP devices 130-150). In some embodiments, the prediction manager 202 can use the performance evaluation module 206 to evaluate the performance of each ECP. For example, the performance can be based on the rate of prescriptions for that type of lens to patients, the number of lenses prescribed to patients, and / or the number of follow-up visits after a prescription for that type of lens to patients. In some embodiments, the prediction manager 202 can present the performance evaluation report to the ECP via a user interface. Furthermore, the report generation module 204 can adjust the configuration of the compatibility report for each ECP based on the performance of the ECP relative to the performance of other ECPs (e.g., modifying the compatibility score range corresponding to the compatibility index).

[0042] Figure 3A process 300 for determining the compatibility of a certain type of lens for a patient, according to one embodiment of this disclosure, is illustrated. In some embodiments, process 300 may be performed by a prediction engine 106 of a lens selection platform 102. Process 300 begins by receiving (at step 305) a compatibility request from an ECP device and obtaining (at step 310) attributes associated with the patient. For example, when prediction manager 202 receives an instruction from ECP 170 to submit a compatibility request via user interface application 132 of ECP device 130, prediction manager 202 may provide an input interface to be displayed on user interface application 132. The input interface may include multiple input fields that enable ECP 170 to provide information related to the compatibility request. Input fields may correspond to information such as the type of lens that ECP 170 wants to determine whether to recommend and / or prescribe to the patient, patient attributes (such as age, dry eye index, computer usage time and frequency, user preferences (e.g., patient comfort with regular glasses), user motivation for using this type of lens (e.g., multifocal contact lenses), user arm length (which indicates the user's reading distance), whether the patient has red-eye, current prescription data, eye measurements (such as pupil diameter), etc.

[0043] Figure 4A An example input interface 400 according to one embodiment of this disclosure is shown. As shown, the input interface 400 includes an input field 402 indicating the identity of ECP 170. This identity can be represented by an ECP identifier associated with ECP 170, a device identifier associated with ECP device 130, or any other type of identifier that can distinguish ECP 170 from other ECPs. In this example, the identifier '1471' is displayed on the input field 402. This identifier can be automatically retrieved from ECP device 130 by the user interface (e.g., a cookie or registry stored on ECP device 130). The input interface 400 also includes a field 404 indicating the patient's age. The input interface 400 also includes fields 406-410 indicating current prescription data, such as spherical data, cylindrical data, and lens power data. The input interface 400 also includes fields 412-418 representing patient lifestyle data, such as daily digital device use, daily computer use, whether the patient is a first-time user of this type of lens (e.g., multifocal contact lenses), and preferred reading distance. The input interface 400 also includes fields 420 and 422 representing patient biometric information, such as whether the patient has congestion and the pupil diameter for photopic vision. The ECP 170 can input the corresponding patient information in fields 402-422.

[0044] In some embodiments, because different ECPs may prioritize certain patient attributes while neglecting others when recommending and / or prescribing this type of lens to a patient, an ECP may select a subset of attributes from the attribute set to determine the compatibility of this type of lens for the patient. Prediction engine 106 may store each ECP's preference data in a database, such as database 104. Therefore, prediction engine 106 can configure input interface 400 for that ECP based on their preferences (e.g., based on an ECP identifier). For example, upon receiving an indication that an ECP wants to submit a compatibility request, prediction engine 106 may configure input interface 400 to present only input fields corresponding to the subset of attributes selected by the ECP, based on the ECP's identity (e.g., ECP 170).

[0045] In some embodiments, an ECP device (e.g., ECP device 130) may be communicatively coupled to one or more biometric instruments in the ECP's office. When the ECP 170 uses a biometric instrument to determine a patient's biometric information (e.g., pupil diameter, etc.), the biometric instrument may automatically provide the biometric information to the input interface 400 (e.g., pre-populate some input fields on the input interface 400) without requiring the ECP 170 to manually transfer the biometric information from the biometric instrument to the input interface 400. The ECP 170 may submit a compatibility request by selecting a compatibility request button 424, using the information provided on the input interface 400. Once the compatibility request button 424 is selected, the input interface 400 may automatically transmit the information obtained from the input fields to the prediction engine 106.

[0046] In some embodiments, the lens selection platform 102 and / or ECP device (e.g., ECP device 130) may be communicatively coupled to a visual simulator (e.g., visual simulator 160) associated with an ECP office of ECP 170. Visual simulator 160 may provide a patient with a simulation of wearing and / or using this type of lens. For example, visual simulator 160 may include wearable devices (e.g., headbands, goggles, etc.) that provide the patient with simulated vision that the patient could have when using this type of lens in one or more environments (e.g., daytime driving, nighttime driving, reading, using electronic devices such as computers, mobile devices, etc.). In some embodiments, visual simulator 160 may be configured to obtain feedback from the patient (e.g., visual quality, comfort, etc.) based on the simulation. For example, visual simulator 160 may be configured to prompt the patient for feedback during various parts of the simulation (e.g., during different environmental simulations), enabling the patient to provide feedback for the different simulated environments. In some embodiments, the model configuration module 208 may configure the lens compatibility model 212 to receive feedback from the patient as input, among other attributes, and to determine an output (e.g., a compatibility score) for the patient based on the patient feedback to the visual simulation and these attributes. In some embodiments, the visual simulator 160 may also include one or more sensors (e.g., a camera, a heart rate monitor, a humidity sensor, etc.) to detect and / or track patient-associated biometric data (e.g., eye movements, blinking frequency, red eyes, facial expressions (such as frowning, squinting, etc.), heart rate, tearing, etc.), and may be configured to provide the detected biometric data to the prediction engine 106. Therefore, the model configuration module 208 may also configure the lens compatibility model 212 to also receive biometric data obtained from the visual simulator 160 as input values ​​and to determine an output based on the biometric data and other patient attributes.

[0047] Process 300 then obtains (at step 320) a performance metric associated with the ECP. For example, performance evaluation module 206 may generate a performance metric associated with the ECP based on a comparison of the ECP's performance with that of other ECPs. Figure 5A process 500 for generating performance metrics for ECP is illustrated. In some embodiments, process 500 may be performed by performance evaluation module 206 of prediction engine 106 and / or model configuration model 208. Process 500 begins by obtaining (at step 505) patient attribute and outcome data associated with patients with different ECPs. For example, performance evaluation module 206 may obtain patient records retrieved from diagnostic training data sources 110, which contain patient attributes and outcomes. In some embodiments, each patient record may include: attributes of the corresponding patient seeking vision correction associated with that type of lens (e.g., a patient seeking presbyopia vision correction, etc.) such as age, dry eye index, duration and frequency of computer use, user preferences (e.g., patient comfort with regular glasses, etc.), user arm length (which indicates the user's reading distance), whether the patient has red-eye, current prescription data, eye measurements (such as pupil diameter, etc.) and the patient's outcome (which may indicate whether the patient was indeed prescribed that type of lens (e.g., multifocal contact lens), and if the patient was indeed prescribed that type of lens, the quality of the prescription after the patient was prescribed that type of lens (e.g., comfort, visual quality level, number of lenses prescribed to the patient, number of follow-up visits, etc.)).

[0048] In some embodiments, after the prediction engine 106 provides a recommendation to the ECP (e.g., in the form of a compatibility score and / or compatibility index, etc.), the prediction engine 106 can track the patient's progress and / or outcomes. For example, the prediction engine 106 can obtain information from the ECP indicating whether the ECP actually prescribed that type of lens for the patient, and if the ECP did prescribe that type of lens, the prediction engine 106 can obtain outcome data such as the number of lenses prescribed to the patient, the number of follow-up visits by the patient, the comfort level of that type of lens for the patient, the patient's visual quality after using that type of lens (e.g., overall visual quality, distance visual quality, intermediate visual quality, etc.), and whether the patient is still wearing that type of lens after a predetermined period of time (e.g., 6 months). The prediction engine 106 can update 104 based on the patient's attributes and outcome data.

[0049] The process then uses (at step 510) the patient's attribute and outcome data to train a machine learning model. In some embodiments, prediction engine 106 may use a general machine learning model (e.g., lens compatibility model 212) to determine and / or predict the compatibility between the patient and that type of lens (e.g., using the same machine learning model to perform predictions for each ECP). Thus, model configuration module 208 may train a general model (e.g., lens compatibility model 212) based on the obtained attribute and outcome data. However, as discussed herein, when recommending and / or prescribing that type of lens to a patient, different ECPs may prioritize certain patient attributes while neglecting others. In some embodiments, prediction engine 106 may enable ECPs to select a subset of attributes from the attribute set for determining the compatibility of that type of lens with the patient, thereby allowing the input interface 400 to be customized for the ECP to prompt and obtain information only corresponding to the selected subset of attributes. Therefore, in some embodiments, the model configuration module 208 can configure a customized machine learning model for ECP (e.g., where the lens compatibility model 212 is a customized model), such that the lens compatibility model 212 customized for ECP 170 is configured to use only a subset of attributes selected by ECP 170 as input data to determine and / or predict the compatibility between the patient and that type of lens.

[0050] Furthermore, different embodiments of the prediction engine 106 can configure the lens compatibility model 212 differently to generate compatibility scores representing different aspects of patient compatibility. For example, in some embodiments, the compatibility score can represent the patient's predicted comfort while wearing the lenses. In another embodiment, the compatibility score can represent the patient's predicted visual quality. In yet another embodiment, the compatibility score can represent the predicted number of lenses prescribed for the patient and / or the predicted number of follow-up visits required by the patient. In yet another embodiment, the compatibility score can represent any combination of the above factors. Reference will be made below. Figure 7 The process of training the lens compatibility model 212 is described in further detail.

[0051] Process 500 then determines (at step 515) performance information associated with different ECPs. For example, performance assessment module 206 may determine performance information associated with an ECP based on the obtained attribute and outcome data associated with the patient. In some embodiments, the performance information determined by performance assessment module 206 may include the rate or percentage of patients prescribed that type of lens. For example, performance assessment module 206 may determine the number of patients prescribed that type of lens by an ECP out of the total number of patients seeking vision correction associated with an ECP and that type of lens. Performance assessment module 206 may also determine multiple rates or percentages, each corresponding to a specific recommendation (e.g., a specific compatibility index). Thus, performance assessment module 206 may determine the rate or percentage of patients prescribed that type of lens (for which the compatibility index has been determined) for each compatibility index (e.g., a highly recommended compatibility index, a recommended compatibility index, a non-recommended compatibility index, etc.).

[0052] For patients prescribed this type of lens, performance information may also include outcome data, such as the number of lenses prescribed, the number of follow-up visits, comfort quality, and visual quality. The process (at step 520) then compares the performance information of a specific ECP with that of other ECPs and generates a performance metric for that specific ECP based on this comparison (at step 525). For example, performance assessment module 206 may compare the performance information of ECP 170 with that of other ECPs and generate a performance metric for ECP 170. In some embodiments, performance assessment module 206 may rank ECPs based on one or more criteria, and the performance metric may include the ranking of ECP 170 relative to other ECPs. Performance assessment module 206 may also rank ECPs for different criteria. Therefore, performance assessment module 206 may generate a first ranking of ECPs based on the percentage of patients prescribed this type of lens, a second ranking based on patient comfort, a third ranking based on patient visual quality, a fourth ranking based on the number of lenses prescribed and / or the number of follow-up visits, and so on. In some embodiments, performance information may also indicate how the performance of an ECP (e.g., ECP 170) changes (e.g., improves) over time. For example, performance information may indicate an improvement in the patient's quality of comfort and / or visual quality at ECP 170 over a period of time. Performance information may also indicate a reduction in the number of lenses prescribed to patients by ECP 170 and / or the number of follow-up visits required by their patients over a period of time.

[0053] Back to reference Figure 3This process uses a machine learning model to determine (at step 325) the compatibility between the patient and a certain type of lens. For example, the prediction manager 202 can use the lens compatibility model 212 (in... Figure 5 Step 510 (trained) determines the compatibility between a certain type of lens and the patient. For example, the lens compatibility model 212 can be configured to provide an output (e.g., a compatibility score) indicating the likelihood that the patient is suitable for that type of lens based on the patient's attributes received via the user interface 400.

[0054] The process then generates (at step 330) a customized report indicating the determined compatibility for ECP, wherein the report is customized based on ECP performance metrics. In some embodiments, the report generation module 204 of the prediction engine 106 may provide a compatibility report on an interface (e.g., a reporting interface) displayed on the ECP device 130 based on the output generated by the lens compatibility model 212. Figure 4B An exemplary implementation of a report interface 430 generated by the report generation module 204 is shown. In this example, the report interface 430 includes a compatibility score 432 (e.g., 90%) determined by the lens compatibility model 212. As discussed herein, the compatibility score 432 indicates the likelihood that a patient is suitable for that type of lens. Since the lens compatibility model 212 is trained using patient data from multiple ECPs (which may or may not include the requesting ECP 170) (at step 510), the requesting ECP 170 can now obtain patient compatibility information not only based on past fit of that type of lens for that ECP 170's patients, but also based on fit of that type of lens for patients of other ECPs. The ability of the prediction engine 106 to access a large sample size of patients from different ECPs means that the accuracy of patient compatibility is improved. Furthermore, since ECP 170 (which submitted the request) is only provided with a compatibility report generated by prediction engine 106, and not with any patient records used to train lens compatibility model 212 (ECP 170 cannot access patient records stored in database 104), patient privacy can be protected.

[0055] However, ECP 170 may struggle to translate compatibility scores into a practical recommendation level for prescribing that type of lens to a patient. Therefore, to assist ECP 170 in making recommendations and / or prescribing, the prediction manager 202 in some embodiments can derive additional data (e.g., meaning) of the compatibility scores. For example, prediction manager 202 can determine multiple compatibility indices that represent different levels of recommendation for prescribing that type of lens to a patient. In a particular, but not limiting, example, prediction manager 202 can determine three distinct compatibility indices representing highly recommended, recommended, and not recommended, respectively. In other embodiments, different compatibility indices can be used to categorize compatibility scores. Compatibility indices can be associated with different, non-overlapping ranges of compatibility scores. For example, if the lens compatibility model 212 generates a compatibility score on a scale of 1-100, then a first range 71-100 may correspond to a first compatibility index indicating a high recommendation, a second range 41-70 may correspond to a second compatibility index indicating a recommendation, and a third range 1-40 may correspond to a third compatibility index indicating a non-recommendation. In some embodiments, these ranges may be determined based on the specific results of the ECP prescribing that type of lens to previous patients. In one example, the prediction manager 202 may determine the first range based on a predetermined percentage (e.g., 90% or any other predetermined percentage) of patients who were prescribed lenses of the type with a compatibility score within the first range (which includes compatibility scores between 71-100) and who no longer require glasses and / or no longer need to see the ECP for a follow-up visit. The prediction manager 202 can determine the second range based on a predetermined percentage (e.g., 90% or any other predetermined percentage) of patients who have been prescribed lenses of a type with a compatibility score within the second range (which includes a compatibility score between 41 and 70) and who only need one more eye exam and / or only one follow-up visit to the ECP. The prediction engine 202 can then determine the patient's compatibility index based on which range the compatibility score falls into.

[0056] Therefore, in some embodiments, the report generation module 204 may configure the report interface 430 to present a compatibility index on the interface as an alternative to or supplement to the compatibility score, to assist the ECP in recommending and / or prescribing that type of lens to patients. In some embodiments, the prediction engine 106 may use the same range of compatibility scores to determine the compatibility index for each ECP. However, different ECPs may have different preferences in how they recommend lenses or prescribe lenses to their patients. In some scenarios, differences in preference may be attributed to the demographics of the patients served by the ECP, the geographic region where the ECP is located, or the ECP's own personal preferences (e.g., their aggressiveness / conservatism in prescribing that type of lens to patients). Since the lens compatibility model 212 is trained using historical patient data from different ECPs, and the model 212 does not distinguish between different ECPs, the output of the lens compatibility model 212 (e.g., the compatibility score) may be general to all patients rather than specific to a single ECP (e.g., ECP 170). Thus, in some embodiments, the prediction engine 106 can determine different ranges corresponding to different compatibility indices for different ECPs. For example, for ECP 170, the prediction manager 202 can determine that a first compatibility index (e.g., highly recommended) corresponds to a range of 81-100, a second compatibility index (e.g., recommended) corresponds to a range of 51-80, and a third compatibility index (e.g., recommended) corresponds to a range of 1-50. For the second ECP, the lens compatibility system can determine that a first compatibility index (e.g., highly recommended) corresponds to a range of 61-100, a second compatibility index (e.g., recommended) corresponds to a range of 31-60, and a third compatibility index (e.g., recommended) corresponds to a range of 1-30.

[0057] Different ranges can be determined for different ECPs based on various factors, such as the geographic region of the ECP, the demographics of the patients of the ECP, or the performance of an ECP (e.g., ECP 170) relative to the recommendation / prescription performance of other ECPs (e.g., its peers). Thus, in some embodiments, prediction manager 202 can determine the position of the submitting ECP 170 relative to other ECPs based on the recommendation / prescription performance of multiple ECPs. For example, prediction manager 202 can use performance metrics associated with different ECPs (including ECP 170) (determined at step 320) to determine different ranges for different compatibility indices. As discussed herein, performance metrics associated with ECP 170 may include the rate or percentage of patients prescribed that type of lens, the number of lenses prescribed to patients seeking that type of lens, the number of follow-up visits by patients prescribed that type of lens, the comfort quality level of patients prescribed that type of lens, the visual quality level of patients prescribed that type of lens, and other items described herein. Performance metrics may also include the ranking of ECP 170 relative to other ECPs based on these performance metrics. Prediction manager 202 can then determine different ranges for different compatibility indices based on the performance metrics and / or ranking of ECP 170. In a specific and non-limiting example, if prediction manager 202 determines that ECP 170 has a below-average rate compared to other ECPs (a lower ranking for the performance criteria), prediction manager 202 may expand the range(s) corresponding to the first (and / or second) compatibility indices (representing highly recommended and recommended, respectively), and conversely narrow the range. In another example, if prediction manager 202 determines that ECP 170 prescribes fewer lenses and / or has fewer follow-up visits compared to other ECPs (a lower ranking for the performance criteria), the lens compatibility system may expand the range(s) corresponding to the first (and / or second) compatibility indices (representing highly recommended and recommended, respectively), and conversely narrow the range.

[0058] like Figure 4BAs shown, the reporting interface 430 includes a graphical representation 434 of the compatibility index determined for patients with ECP 170. In this example, the graphical representation 434 is implemented in the form of a gauge divided into three sections 436, 438, and 440. Each of sections 436-440 corresponds to a different compatibility index determined by the prediction engine 106. For example, section 436 corresponds to a third compatibility index indicating a "not recommended" level of recommendation, section 438 corresponds to a second compatibility index indicating a "recommended" level of recommendation, and section 440 corresponds to a first compatibility index indicating a "highly recommended" level of recommendation. Since 90% of the compatibility scores fall within the score range (81-100) corresponding to the first compatibility index, the graphical representation 434 includes a pointer 442 pointing to the section 440 of the gauge corresponding to the "highly recommended" level of recommendation.

[0059] In addition to compatibility scores and compatibility indices, the report generation module 204 in some embodiments may configure the report interface 430 to include other analytical information that can assist the ECP 170 in providing recommendations and / or prescriptions to patients. The analytical information may be determined based on compatibility scores and / or compatibility indices previously determined for different ECPs (which may or may not include ECP 170). For example, the prediction manager 202 may determine the number of patients assigned a first compatibility index, the number of patients assigned a second compatibility index, and the number of patients assigned a third compatibility index. Figure 4B As shown, the report interface 430 can be configured by the report generation module 204 to include a representation indicating the distribution of patients under three compatibility indices (e.g., graph 444).

[0060] After the compatibility report is presented to ECP 170 via reporting interface 430, ECP 170 can determine whether to recommend and / or prescribe that type of lens to the patient based on the compatibility report. In some embodiments, after providing the compatibility report to ECP 170, prediction manager 202 can generate and provide a feedback interface to ECP 170 to obtain outcome feedback regarding the patient. For example, through the feedback interface, ECP 170 can instruct prediction engine 106 whether ECP 170 recommended and / or prescribed that type of lens to the patient. Furthermore, if ECP 170 indicates that it has recommended and / or prescribed that type of lens to the patient, ECP 170 can also provide feedback information indicating the prescription outcome. For example, ECP 170 can indicate the number of lenses actually prescribed to the patient and / or the number of follow-up visits actually performed by the patient after recommending and / or prescribing that type of lens. ECP170 can also indicate patient comfort and visual quality when using this type of lens (e.g., through assessment and / or survey).

[0061] Figure 4CAn exemplary implementation of a feedback interface 450 according to an embodiment of this disclosure is shown. As shown, the feedback interface 450 includes an input field 452 indicating the patient's identity. When a compatibility request is submitted to the prediction engine 106 via the input interface 400, a patient identity can be generated for the patient. In this example, the input field 452 includes a patient identifier '1492'. The feedback interface 450 also includes a field 454 corresponding to the number of lenses prescribed for the patient and / or the number of follow-up visits. The feedback interface 450 also includes fields 456-462 representing visual quality, including distance visual quality 456, near visual quality 458, intermediate visual quality 460, and overall visual quality 462. Furthermore, the feedback interface 450 includes a field 464 indicating whether the patient intends to purchase this type of lens. Therefore, the ECP 170 can provide information corresponding to fields 454-464 regarding the results of recommending this type of lens to the patient and / or prescribing this type of lens. Once feedback information is received from the ECP device 130 via the feedback interface 450, the prediction engine 106 can update the database 106 based on the patient's attributes (obtained via the input interface 400 and the visual simulator 160) and the feedback information (obtained via the feedback interface 450). For example, the prediction engine 106 can generate new patient records based on the patient's attributes and feedback information, and can add the new patient records to the database 104. In some embodiments, the model configuration module 208 can use the updated database 104 to train (or retrain) the lens compatibility model 212, allowing the lens compatibility model 212 to be continuously modified and / or updated based on new cases (e.g., new patient records). In some embodiments, the performance evaluation module 206 can also use the updated database 104 to determine updated performance metrics for ECP 170 and / or other ECPs.

[0062] In addition, the prediction manager 202 can provide the ECP 170 with information about the patient's performance relative to other patients who have been prescribed this type of lens. Figure 4D A performance interface 470 is shown, representing the performance of a patient with ECP 170 relative to other patients prescribed that type of lens. For example, the performance interface 470 includes a histogram 472 showing the patient's overall visual quality relative to other patients, a histogram 474 showing the patient's distance visual quality relative to other patients, and a histogram 476 showing the patient's intermediate visual quality relative to other patients. Other graphs or representations showing these performance criteria (or other performance criteria) may also be included in the performance interface 470.

[0063] Figure 4EAnother exemplary implementation of a performance interface 480 is shown, which presents performance information about ECP 170 relative to other ECPs that can be provided by the prediction manager 202. As shown, performance interface 480 includes a performance histogram 482 that represents ECP 170's position relative to other ECPs in terms of the rate of patients prescribed this type of lens. Performance interface 480 also includes a graph 484 that represents the trend of the number of prescriptions for this type of lens issued by ECP 170 over a period of time.

[0064] While the above description uses examples of contact lenses, and more specifically multifocal contact lenses, to illustrate the process for determining compatibility, it is conceivable that the process described herein for determining custom compatibility for ECP can be used to determine the compatibility of other lenses, such as other types of contact lenses.

[0065] Figure 6A and Figure 6B This is a diagram of a processing system according to some embodiments. Although Figure 6A and Figure 6B Two embodiments are shown, but those skilled in the art will readily understand that other system embodiments are possible. According to some embodiments, Figure 6A and / or Figure 6B The processing system may include one or more of the following computing systems: lens selection platform 102, ECP devices 130, 140 and 150, (multiple) diagnostic training data sources 110, visual simulator 160, etc.

[0066] Figure 6AA computing system 600 is illustrated, wherein the components of system 600 are electrically connected to each other via a bus 605. System 600 includes a processor 610 and a system bus 605 that couples various system components (including memories in the form of read-only memory (ROM) 620, random access memory (RAM) 625, etc. (e.g., PROM, EPROM, FLASH-EPROM, and / or any other memory chip or cartridge)) to processor 610. System 600 may further include a cache 612 of high-speed memory directly connected to, adjacent to, or integrated into processor 610. System 600 can access data stored in ROM 620, RAM 625, and / or one or more storage devices 630 via cache 612 for high-speed access by processor 610. In some examples, cache 612 can provide performance improvements to avoid latency when processor 610 accesses data previously stored in cache 612 from memory 615, ROM 620, RAM 625, and / or the one or more storage devices 630. In some examples, the one or more storage devices 630 store one or more software modules (e.g., software modules 632, 634, 636, etc.). Software modules 632, 634, and / or 636 can control and / or be configured to control processor 610 to perform various actions, such as the processes of methods 300 and / or 500. And although system 600 is shown to have only one processor 610, it should be understood that processor 610 can represent one or more central processing units (CPUs), multi-core processors, microprocessors, microcontrollers, digital signal processors (DSPs), field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), graphics processing units (GPUs), tensor processing units (TPUs), etc. In some examples, system 400 can be implemented as a stand-alone subsystem, and / or as a board added to a computing device, or as a virtual machine.

[0067] To enable users to interact with system 600, system 600 includes one or more communication interfaces 640 and / or one or more input / output (I / O) devices 645. In some examples, the one or more communication interfaces 640 may include one or more network interfaces, network interface cards, etc., to provide communication according to one or more network and / or communication bus standards. In some examples, the one or more communication interfaces 640 may include interfaces for communicating with system 600 via a network (such as network 115). In some examples, the one or more I / O devices 645 may include one or more user interface devices (e.g., keyboard, pointing / selection devices (e.g., mouse, touchpad, scroll wheel, trackball, touchscreen, etc.), audio devices (e.g., microphone and / or speaker), sensors, actuators, display devices, etc.).

[0068] Each of the one or more storage devices 630 may include a nontransitory, non-volatile storage device, such as a hard disk, optical media, or a solid-state drive. In some examples, each of the one or more storage devices 630 may be located in the same location as system 600 (e.g., a local storage device) and / or remote from system 600 (e.g., a cloud storage device).

[0069] Figure 6B A chipset-based computing system 650 is illustrated, which can be used to execute any of the methods described herein (e.g., methods 300 and / or 510). System 650 may include a processor 655, which represents any number of physically and / or logically distinct resources capable of performing software, firmware, and / or other computations, such as one or more CPUs, multi-core processors, microprocessors, microcontrollers, DSPs, FPGAs, ASICs, GPUs, TPUs, etc. As shown, processor 655 is assisted by one or more chipsets 660, which may also include one or more CPUs, multi-core processors, microprocessors, microcontrollers, DSPs, FPGAs, ASICs, GPUs, TPUs, coprocessors, encoder-decoders (CODECs), etc. As shown, the one or more chipsets 660 interface processor 655 with one or more of one or more I / O devices 665, one or more storage devices 670, memory 675, bridges 680, and / or one or more communication interfaces 690. In some examples, the one or more I / O devices 665, one or more storage devices 670, memory, and / or one or more communication interfaces 690 may correspond to Figure 6A The counterpart with a similar name in System 600.

[0070] In some examples, bridge 680 may provide additional interfaces to provide system 650 with access to one or more user interface (UI) components, such as one or more keyboards, pointing / selection devices (e.g., mouse, touchpad, scroll wheel, trackball, touchscreen, etc.), audio devices (e.g., microphone and / or speaker), display devices, etc. According to some embodiments, system 600 and / or 650 may provide a graphical user interface (GUI) suitable for assisting users (e.g., surgeons and / or other medical personnel) in performing the processes of methods 300 and / or 500.

[0071] Figure 7This is a diagram of a multilayer neural network 700 according to some embodiments. In some embodiments, neural network 700 may represent at least some of the neural networks used to implement prediction engine 106, including lens compatibility model 212. While neural network 700 for implementing machine learning models is shown in this example, other machine learning techniques, such as regression models, may also be used to implement one or more of these machine learning models. Neural network 700 uses input layer 720 to process input data 710. In some examples, input data 710 may correspond to input data provided to the one or more models, and / or training data provided to the one or more models during a training process (e.g., step 510 of process 500). Input layer 720 includes a plurality of neurons for scaling, range limiting, and / or similarly adjusting the input data 710. Each neuron in input layer 720 generates an output that is fed to the input of hidden layer 731. Hidden layer 731 includes a plurality of neurons that process the output from input layer 720. In some examples, each neuron in hidden layer 731 generates an output, which is then propagated through one or more additional hidden layers (ending in hidden layer 739). Hidden layer 739 includes multiple neurons that process the output from the previous hidden layer. The output of hidden layer 739 is fed to output layer 740. Output layer 740 includes one or more neurons for modulating the output from hidden layer 739 by scaling, range limiting, and / or similar methods. It should be understood that the architecture of neural network 700 is only representative, and other architectures are possible, including neural networks with only one hidden layer, neural networks without input and / or output layers, neural networks with recurrent layers, etc.

[0072] In some examples, each of the input layer 720, hidden layers 731-739, and / or output layer 740 includes one or more neurons. In some examples, each of the input layer 720, hidden layers 731-739, and / or output layer 740 may include the same or different numbers of neurons. In some examples, each neuron combines its input x (e.g., a weighted sum obtained using a trainable weighted matrix W), adds an optional trainable deviation rate b, and applies an activation function f to generate output a, as shown in Equation 1 below. In some examples, the activation function f may be a linear activation function, an activation function with an upper and / or lower bound, a log-sigmoid function, a hyperbolic tangent function, a modified linear unit function, etc. In some examples, each neuron may have the same or different activation functions.

[0073] a = f(Wx + b)…………………………………………Equation (1)

[0074] In some examples, supervised learning can be used to train neural network 700, where the combination of training data includes a combination of input data and standard ground truth (e.g., expected) output data. Neural network 700 is differentiated between the outputs generated using the input data as input data 710, and the output data 750 generated by neural network 700 is compared with the standard ground truth output data. The difference between the generated output data 750 and the standard ground truth output data can then be fed back into neural network 700 to correct for individual trainable weights and biases. In some examples, these differences can be fed back using backpropagation techniques such as stochastic gradient descent. In some examples, a large set of training data combinations can be presented to neural network 700 multiple times until the total loss function (e.g., the mean squared error based on the difference of each training combination) converges to an acceptable level.

[0075] The methods according to the above embodiments can be implemented as executable instructions stored on a non-transitory tangible machine-readable medium. These executable instructions, when run by one or more processors (e.g., processor 610 and / or processor 655), can cause the one or more processors to perform one or more processes within methods 200 and / or 210. Some common forms of machine-readable media that can include processes of methods 200 and / or 210 are, for example, floppy disks, hard disks, magnetic tapes, any other magnetic media, CD-ROMs, any other optical media, punched cards, paper tapes, any other physical media with a perforated pattern, RAM, PROMs, EPROMs, FLASH-EPROMs, any other memory chips or cartridges, and / or any other media suitable for a processor or computer to read from.

[0076] Devices implementing the methods according to these disclosures may include hardware, firmware, and / or software, and may take the form of any of a variety of form factors. Typical examples of such form factors include laptop computers, smartphones, minicomputers, personal digital assistants, etc. A portion of the functionality described herein may also be embodied in peripheral devices and / or add-on cards. By further example, such functionality may also be implemented on a circuit board between different chips within a single device or between different processes performed therein.

[0077] While illustrative embodiments have been shown and described, various modifications, alterations, and substitutions are contemplated in the foregoing disclosure, and in some cases, some features of the embodiments may be employed without correspondingly using other features. Many variations, alternatives, and modifications will be recognized by those skilled in the art. Therefore, the scope of the invention should be limited only by the claims, and it is appropriate that the claims be interpreted broadly in accordance with the scope of the embodiments disclosed herein.

Claims

1. A method for determining the compatibility of a contact lens, comprising: One or more hardware processors receive a request from a first device associated with a first ophthalmic care professional (ECP) to determine compatibility between a particular lens and a patient, wherein the request includes attributes associated with the patient; The one or more hardware processors compare the performance information associated with the first ECP with the performance information associated with at least the second ECP. The one or more hardware processors determine a performance metric associated with the first ECP based on the comparison, wherein the performance metric represents compatibility with respect to the contact lens and at least the second ECP with respect to the first consumer, and compatibility with the contact lens and the first ECP with respect to the first consumer. The one or more hardware processors select a specific compatibility index representing the degree of compatibility between the particular lens and the patient from a plurality of compatibility indices and based on attributes associated with the patient, wherein the compatibility index is tailored to the first ECP based at least in part on a performance metric associated with the first ECP; and A report indicating the specific compatibility index is presented on the first device by the one or more hardware processors.

2. The method of claim 1, further comprising providing a visual simulation to the patient, wherein, The visual simulation simulates vision using the specific lens.

3. The method of claim 2, further comprising obtaining the patient's biometric data while providing the visual simulation to the patient, wherein, The specific compatibility index is further selected based on biometric data obtained from the patient.

4. The method of claim 1, wherein, Performance information associated with the first ECP includes visual quality data of the patient for whom the first ECP prescribed the specific lens.

5. The method of claim 1, wherein, The attributes associated with the patient include at least one of the following: age, dry eye index, computer usage frequency, the patient's arm length, the patient's motivation, whether the patient has conjunctivitis, current prescription data, or eye measurement data.

6. A non-transitory machine-readable medium having stored thereon machine-readable instructions, the machine-readable instructions being executable to cause a machine to perform operations, the operations including: A request for selecting a contact lens for a consumer is received from a first device associated with a first ophthalmic care professional (ECP), wherein the request includes biometric information associated with the consumer; Obtain a performance metric associated with the first ECP, wherein the performance metric represents the compatibility between the contact lens and at least the second ECP with respect to the second consumer, and the compatibility between the contact lens and the first ECP with the first consumer; A compatibility index is determined using a machine learning model and based on the performance metric, the compatibility index indicating the compatibility between a specific contact lens and the consumer for the first ECP; and A report indicating the compatibility index is presented on the first device.

7. The non-transitory machine-readable medium as described in claim 6, wherein, The operation further includes: Obtain visual quality data of a patient using the specific contact lens; and The biometric information and the visual quality data are used to further train the machine learning model.

8. The non-transitory machine-readable medium as described in claim 6, wherein, The operation further includes: At least from a second device associated with the second ECP, biometric data and visual quality data associated with multiple consumers are obtained; and The obtained biometric data and visual quality data are used to train the machine learning model.

9. The non-transitory machine-readable medium as described in claim 6, wherein, The compatibility between the contact lens and at least the second consumer associated with the second ECP, and the compatibility between the contact lens and the first consumer associated with the first ECP, are determined based on at least one of the following: Did the corresponding patient receive a prescription for a contact lens recommended to that patient? Recommendations for contact lens comfort for the appropriate patient. Recommended visual quality level of contact lenses for the corresponding patient. The number of lenses prescribed for the corresponding patient, or The number of follow-up visits for the corresponding patients.

10. The non-transitory machine-readable medium of claim 6, wherein, The determination of the compatibility index includes: Using the machine learning model, a compatibility score is generated based on the biometric information; and Based on the compatibility score and the performance metric associated with the first ECP, a compatibility index for the first ECP is identified from a plurality of compatibility indices.

11. The non-transitory machine-readable medium of claim 10, wherein, The plurality of compatibility indices correspond to different ranges of compatibility scores, wherein determining the compatibility indices further includes: The different compatibility score ranges are adjusted based on the performance metrics to generate a customized compatibility score range for the first ECP, wherein the identifier is further based on the customized compatibility score range corresponding to the plurality of compatibility indices.

12. The non-transitory machine-readable medium of claim 10, wherein, The compatibility score represents the predicted comfort level of the consumer using the contact lens.

Images