Machine learning supported pipeline for determining intraocular lens size

Abstract

The invention relates to a computer-implemented method for determining a parameter value of an intraocular lens to be inserted using a machine learning supported processing pipeline. The method comprises providing a scan result of an eye. The scan result is an image of an anatomical structure of the eye. The method further comprises determining biometric data of the eye from the scan result of the eye and determining a final position of the intraocular lens to be inserted using a first trained machine learning system, the ophthalmic data being used as input data for the first machine learning system. The method further comprises determining a first refractive power of the intraocular lens to be inserted, the determination being based on a physical model in which the determined final position of the intraocular lens and the determined biometric data are used as input variables for the physical model.

Description

TECHNICAL FIELD

[0001] The present invention relates to determining the power of an intraocular lens, and in particular, to a computer-implemented method of a machine learning supported processing pipeline for determining parameter values of an intraocular lens to be inserted, a corresponding system and a corresponding computer program product for performing the method. BACKGROUND

[0002] In recent years, in the field of ophthalmology, for example in the case of (age-related) refractive errors or in the case of cataracts, it has become increasingly common to replace the biological lens of the eye with an intraocular lens (IOL). In this process, the biological lens is detached from the capsular bag and removed by minimally invasive intervention. Subsequently, an artificial lens implant replaces the lens that has become cloudy in the case of cataracts. In this process, such an artificial lens implant or intraocular lens is inserted into the then empty capsular bag. The knowledge of the correct position and the necessary power of the intraocular lens are interdependent.

[0003] There are several problems with the currently utilized IOL calculation formulas. First, the position of the intraocular lens is calculated as effective lens position (ELP) in many formulas. Since this variable is not a true anatomical variable, it cannot be directly considered in a physical model for calculating the complex ophthalmic optics of a patient. The ELP is calculated and optimized for the respective formula, so there is no direct comparability between the ELPs of different formulas, and this model does not use an anatomically correct optical system.

[0004] A second aspect is that the current IOL formulas use models in the prediction that try to fine-tune the availability of data by some parameters. Since these are manually predefined by the developer, this is not necessarily the best representation in every case. New formulas like the Hill RBF formula circumvent this limitation by using machine learning methods that can be optimized independently on the basis of the availability of data. However, in this case, the prediction is based only on a large amount of data, that is, the system does not use any physical concept, so it is limited in its validity.

[0005] In general, the current methods do not show the best combination of all available information and existing models.

[0006] Starting from the disadvantages of known methods for approximately determining the correct power of an IOL to be inserted, the basic purpose of the concept presented here is to specify a method and system for an improved, comprehensive and fast IOL power prediction for well-extendable intraocular lenses. SUMMARY

[0007] The object of the present invention is achieved by the method, the corresponding system and the associated computer program product presented here.

[0008] According to an aspect of the present invention, a computer-implemented method of a machine learning supported processing pipeline for determining parameter values of an intraocular lens to be inserted is presented. The method can comprise providing a scan result of an eye. The scan result can represent an image of an anatomical structure of the eye.

[0009] The method can further comprise determining biometric data of the eye from the scan result of the eye and determining a final position of the intraocular lens to be inserted using a first trained machine learning system. In this case, ophthalmic data can be used as input data for the first machine learning system. Finally, the method can comprise determining a first refractive power of the intraocular lens to be inserted, the determination being based on a physical model in which the determined final position of the intraocular lens and the determined biometric data are used as input variables for the physical model.

[0010] According to another aspect of the present invention, a processing pipeline system of a machine learning supported processing pipeline for determining parameter values of an intraocular lens to be inserted is presented.

[0011] The processing pipeline system can comprise a receiving module configured to provide a scan result of an eye. In this case, the scan result can represent an image of an anatomical structure of the eye.

[0012] Further, the processing pipeline system can comprise a determining unit configured to determine biometric data of the eye from the scan result of the eye and a first trained machine learning system for determining a final position of the intraocular lens to be inserted. The ophthalmic data can be used as input data for the first machine learning system.

[0013] Finally, the processing pipeline system can comprise a determining unit configured to determine a first refractive power of the intraocular lens to be inserted, the determination being based on a physical model in which the determined final position of the intraocular lens and the determined biometric data are used as input variables for the physical model.

[0014] Further, embodiments can relate to a computer program product comprising program code accessible from a computer usable or computer readable medium, which causes a computer or other instruction processing system to perform the functions of the present invention. In the context of this specification, a computer usable or computer readable medium can be any device or means that can store, communicate, transmit or receive program code.

[0015] The computer-implemented method for determining the refractive power of an intraocular lens to be inserted has a number of advantages and technical effects, which can also apply accordingly to the associated system: The method presented here well addresses the known negative attributes of the above-mentioned disadvantages. In particular, the "ZAI" algorithm on which the method is based contributes to optimizing the calculation of the required refractive power of an intraocular lens to be inserted during cataract surgery. The presented algorithm allows to unify an anatomically correct prediction of the IOL position optimized by machine learning with a complex physical model and allows to refine the IOL calculation by machine learning. Thus, both the IOL position and the IOL refractive power determination can be determined in one process - or in other words: within a pipeline - without the need for a medium break.

[0016] In this case, both the physical calculation model and the machine learning concept based on clinical ophthalmological data can be linked within the pipeline to determine the position of the intraocular lens and to determine the refractive power of the intraocular lens in a comprehensive manner.

[0017] A machine learning system for determining the refractive power of an intraocular lens to be inserted based on available clinical ophthalmological data alone first requires a relatively long training time and secondly will not be able to take into account the known properties of the physical model.

[0018] In this case, the speed advantage that arises when a trained machine learning model is retrained by better or further training data is exploited in each case. This can significantly shorten the overall training time and thus significantly save computing power and thus better utilize the available computer capacity.

[0019] Furthermore, the use of the true physical position of the IOL allows the use of a model with any desired accuracy and ultimately also the use of an exact physical model. The presented method is therefore not limited to certain sizes of the model and the ultimately determined value is ultimately universal. This contrasts with the effective lens position (ELP) formula used previously, as this variable is not a true anatomical variable. It can therefore also not be directly considered in a physical model for calculating the complex ophthalmological optics of a patient.

[0020] The intraocular lens is calculated in many formulas as an effective lens position (ELP). As this variable is not a true anatomical variable, it cannot be directly considered in a physical model for calculating the complex ophthalmological optics of a patient. The ELP is calculated and optimized for the respective formula, so there is no direct comparability between the ELPs of different formulas and the model does not use an anatomically correct optical system.

[0021] Further exemplary embodiments are presented below, which are valid both in connection with the method and the corresponding system.

[0022] According to an advantageous exemplary embodiment, the method can additionally comprise determining a final refractive power of the intraocular lens by a second machine learning system, at least one variable from the biometric data and the first refractive power can be used as input variables. By way of example, the at least one variable can be the axial eye length of the eye. Thus, a step of transfer learning can be actually performed, which uses the knowledge present in the physical model as a basis in order to determine the refractive power more accurately. For this purpose, the second machine learning system should be trained using clinical ophthalmological data, that is to say data from earlier real patients. Such clinical ophthalmological data is usually annotated. In this way, no information is lost in the pipeline: both the theoretical data of the physical model and the actual experience data from clinical routine can be taken into account.

[0023] In this way, it is also possible to include certain properties of the clinic or its method of operation in the pipeline. Usually, the use of a physical model does not allow this or only allows this with the disadvantage of deviating from known standards.

[0024] According to a further exemplary embodiment of the method, the biometric data of the eye can comprise at least one selected from the group consisting of preoperative axial eye length, preoperative lens thickness, preoperative anterior chamber depth, and intraoperative anterior chamber depth. These can result from the method step of determining the biometric data of the eye from the scan results of the eye. This can be performed in the conventional sense; however, a machine learning system can also be used for this purpose, which determines the biometric data of the eye in a direct scanning method, wherein no manual steps are required. The recorded image data of the scan results can be used directly for determining the biometric parameters.

[0025] According to an advantageous exemplary embodiment of the method, a convolutional neural network, a graph attention network or a combination of both networks can be used in the first machine learning system. By way of example, a convolutional neural network can be used to recognize characteristic features in the recorded scan results and for compressing the generated image data. As a result of a graph attention network, known, annotated images or compressed representations thereof can be arranged in a graph. By means of a current image of a newly recorded patient eye, the required biometric data, for example the postoperative final position of the intraocular lens, can then be determined by means of a distance measurement to the images already present in the graph. This data can then be used directly in the ZAI pipeline.

[0026] According to an exemplary embodiment of the method, the second machine learning system can be trained in two stages, wherein a first training step can comprise - in particular by a computer - generating first training data for the machine learning system based on a first physical model for the intraocular lens power. Subsequently, the second machine learning system can then be trained by the generated first training data to form a corresponding learning model for determining the power. In this case, the hyperparameters of the machine learning system are defined by the design and selection of the machine learning system, while the internal parameters of the machine learning system are adjusted step by step by the training.

[0027] In a second training step, the machine learning system trained with the first training data can then be trained using clinical ophthalmological training data to form a second learning model for determining the power. In this case, the principle of transfer learning is used; that is, the knowledge already learned from the physical model is now further specified by using real clinical ophthalmological training data. In this way, the training process can be significantly accelerated and less clinical ophthalmological training data is required, since the basic structure has already been preset by training using data from the physical model.

[0028] According to an extended exemplary embodiment of the method, one variable from the biometric data can be the preoperative axial eye length. This variable can be well determined using known measurement methods, for example by OCT measurements, for example A-scan, B-scan, or frontal OCT measurements.

[0029] According to yet another extended exemplary embodiment of the method, the biometric data of the eye can be determined manually from images or by a machine learning system from provided scan results of the eye. At this point, the proposed method does not make a statement about which partial method is used to determine the biometric data of the eye. However, the machine learning-based biometric data determination itself is in line with the meaning of the pipeline concept.

[0030] According to yet another extended exemplary embodiment of the method, further parameters of the eye can be determined when determining the final position of the intraocular lens to be inserted. These further parameters can relate to the IOL position - in particular the expected final position of the IOL after the growth process - can be specified as a typical further parameter. In addition, a value for the IOL offset can also be used, which indicates an offset perpendicular to the optical axis. The beam path in the respectively selected model will change depending on the offset value.

[0031] Additionally or in a complementary way, also the IOL tilt value (i.e. the angle of tilt of the IOL with respect to the optical axis) can be used; in this case, also the beam path should be adjusted according to the variations. Also the IOL type - in particular the haptics, shape, etc. used - can be taken into account. It can determine the position of the lens through the haptics / shape, thus affecting the final quality of the surgery (insertion of the correct IOL).

[0032] Additionally, the forces of the capsular bag etc. on the IOL should also be specified as additional parameters. This allows to take into account long-term expected possible variations of the position. BRIEF DESCRIPTION OF DRAWINGS

[0033] It should be noted that the exemplary embodiments of the present application can be described with reference to different implementation categories. In particular, some exemplary embodiments are described with reference to methods, while others can be described in the context of corresponding devices. Regardless, a person skilled in the art can identify and combine possible combinations of features of the method and also of features with corresponding systems from the above and the following description, even if these features belong to different categories of claims, if not specifically stated otherwise.

[0034] The above and additional further specific embodiments of the present application become apparent from the described exemplary embodiments and from the description with reference to the accompanying drawings.

[0035] Preferred exemplary embodiments of the present application are described by way of example and with reference to the following drawings:

[0036] Figure 1 A representation of a similar flow chart of an exemplary embodiment of a computer implemented method for a machine learning supported processing pipeline for determining parameter values of an intraocular lens to be inserted.

[0037] Figure 2 A cross section of a part of an eye is depicted.

[0038] Figure 3 An eye as well as different biometric parameters of an eye are depicted.

[0039] Figure 4 A schematic structure of basic functional blocks of a machine learning supported pipeline for determining an intraocular lens size by a specified method is represented.

[0040] Figure 5 A diagram of a processing pipeline system for a machine learning supported processing pipeline for determining parameter values of an intraocular lens to be inserted according to the present application is shown.

[0041] Figure 6 A diagram of a computer system which additionally can comprise in whole or in part a system according to the present application is shown. Figure 5The processing pipeline system. Detailed Implementation

[0042] In the context of this specification, the conventions, terms and / or expressions should be understood as follows:

[0043] The term "machine learning-supported processing pipeline" describes the general concept of the method and system presented herein. Starting from recorded digital images, the final refractive power of the intraocular lens to be inserted can be determined without media interruption or intermediate manual parameter determination. In this case, even without manual parameter determination, the final postoperative IOL position is used as an intermediate result. At different points, the processing pipeline uses a machine learning system trained with real patient data. Alternatively, a physical model can be used. In this way, the final refractive power determination incorporates both the specialized knowledge of theoretical models and practical empirical values.

[0044] The term "artificial lens" describes an artificial lens that can be surgically inserted into a patient's eye to replace the natural biological lens.

[0045] The term "machine learning system" describes a system, often also assigned to methods, that learns from examples. For this purpose, annotated training data (i.e., training data that also includes metadata) is fed into the machine learning system to predict pre-set output values ​​(output classes in the case of classification systems). If the output class is correctly output with sufficient accuracy (i.e., a predetermined error rate), the machine learning system is said to be trained. Different machine learning systems are known. These include neural networks, convolutional neural networks (CNNs), or recurrent neural networks (RNNs).

[0046] In principle, the term "machine learning" comes from a fundamental term or function in the field of artificial intelligence, where, for example, statistical methods are used to endow computer systems with the ability to "learn." For instance, in this case, certain behavioral patterns within a specific task are optimized. The methods used give trained machine learning systems the ability to analyze data without requiring explicit programming for this purpose. Typically, NN (neural network) or CNN (convolutional neural network) are examples of systems used for machine learning, forming a network of nodes that act as artificial neurons, and artificial connections (so-called links) between these artificial neurons, where parameters (e.g., weighted parameters of the links) can be assigned to the artificial links. When training a neural network, the weight parameter values ​​are automatically adapted to the links based on the input signals to produce the desired result. In the case of supervised learning, the images provided as input values ​​(training data)—typically (input) data—are supplemented with the desired output data (annotations) to generate the desired output value (the desired class). Very generally, the mapping from input data to output data is learned.

[0047] The term "neural network" describes a network of nodes, each with one or more inputs and one or more outputs, used to perform computational operations. Here, selected nodes are interconnected by connections (so-called links or edges). These connections may have certain properties (such as weighted parameter values) that can influence the output values ​​of the preceding nodes.

[0048] Neural networks typically consist of multiple layers. At least an input layer, hidden layers, and an output layer are present. In a simple example, image data might be fed into the input layer, and the output layer might contain a classification result based on the image data. However, typical neural networks have a large number of hidden layers. The way nodes are connected by links depends on the type of neural network. In this example, the predicted value of the neural learning system could be the desired refractive power of an intraocular lens.

[0049] As an example of a classifier / classifier system, the term "convolutional neural network" (CNN) describes a class of artificial neural networks based on feedforward techniques. They are commonly used in image analysis, which uses an image or its pixels as input data. The main component of a CNN in this case is the convolutional layer (hence the name) that enables efficient evaluation through parameter sharing. In contrast, each pixel of a recorded image is typically associated with an artificial neuron in the neural network, serving as the input value in a regular neural network.

[0050] The term "Graph Attention Network" (GAT) describes a neural network that operates on graph-structured data. It exhibits better behavior than the older "Graph Convolutional Network" (GCN). In this process, masked self-referencing node layers are used, which improve upon known approximations in GCNs without requiring computationally intensive matrix operations. One can envision a "Graph Convolutional Network" (GCN) instead of GAT; a GCN is a neural network architecture that can also operate directly on graphs and utilize the structural information present there. Alternatively, the "GraphSage" framework is also available. It is well-suited for inductive representation learning within the context of large graphs. In this case, GraphSage can be used to generate low-dimensional vector representations of nodes, which is particularly useful for graphs with comprehensive node attribute information.

[0051] In the context of this paper, the term "transfer learning" (or curriculum learning) describes the retraining of a previously developed learning model—developed by training a machine learning system with training data from a physical model. Although it is trained a second time using relevant data, the source of this relevant data differs from that of the first training. This might consist of clinical ophthalmology data or a second physical model known for its more accurate results. As a result, a second learning model is produced, which itself unifies the physical model parameters and the real clinical data. Therefore, the "knowledge" of the corresponding first learning model is used as the basis or starting point for training the second learning model. Thus, the learning outcomes from the first training can be transferred to the learning outcomes from the second training. A significant advantage is that the second training can be performed relatively more efficiently, thereby saving computing resources and allowing for a faster and more targeted approach.

[0052] The term "parameter value" describes the geometric or biometric values, or ophthalmic data, of a patient's eye. Based on Figure 2 Examples of eye parameter values ​​will be discussed in more detail.

[0053] The term "scan result" describes, for example, digital data based on digital images / records, representing the results of an OCT (Optical Coherence Tomography) examination of a patient's eye.

[0054] The term "Optical Coherence Tomography" (OCT) describes a known ophthalmic imaging method for obtaining two-dimensional and three-dimensional (2-D or 3-D) records of scattering material (e.g., biological tissue) with micrometer resolution. The process essentially utilizes a light source, a beam splitter, and a sensor (e.g., in the form of a digital image sensor). In ophthalmology, OCT is used to detect spatial differences in the reflective behavior of individual retinal layers, allowing morphological structures to be represented at high resolution.

[0055] The term "A-scan" (also known as axial depth scan) describes a one-dimensional result of a scan of a patient's eye, which describes information about the geometry and location of the structures inside the eye.

[0056] The term "B-scan" describes the lateral overlap of multiple aforementioned A-scans to obtain a cross-section through the eye. A volumetric view can also be generated by combining multiple layers of the eye thus created.

[0057] In this context, the term "frontal OCT" describes a method for generating a transverse cross-sectional image of the eye (compared to a longitudinal cross-sectional image using the aforementioned A-scan or B-scan).

[0058] In this context, the term "image" or "digital image" (e.g., from a scan) describes an image representation of a physically existing object (e.g., the retina of the eye in this case), or the result of generating a certain amount of data in the form of pixel data from that object. More generally, a "digital image" can be understood as a two-dimensional signal matrix. The individual vectors of the matrix can be contiguous with each other, thus generating the input vectors for the layers of a CNN. A digital image can also be the individual frames of a video sequence. In this context, "image" and "digital image" can be understood as synonyms.

[0059] The term "clinical ophthalmology training data" describes data about a patient's eyes and the intraocular lenses previously inserted into those patients. Clinical ophthalmology training data can include defined ophthalmic parameter values, such as the refractive power and position of the inserted lens. This data is used to train a machine learning system that has previously been trained based on data from a physical model. Typically, clinical ophthalmology training data is annotated.

[0060] The term "physical model" involves relating various parameters of the eye to each other to determine a mathematical formula for refractive power. The known formula is the Haigis formula.

[0061] The term "refractive power of an intraocular lens" describes the refractive index of an IOL.

[0062] A detailed description of the accompanying drawings is given below. It should be understood that all details and information in the drawings are shown schematically in this context. Initially, a block diagram of an exemplary embodiment of a computer-implemented method for determining parameter values ​​of an intraocular lens to be inserted, supported by a machine learning pipeline, is shown. Further exemplary embodiments, or corresponding exemplary embodiments of the system, are described below:

[0063] Figure 1A similar flowchart representation of an exemplary embodiment of a computer-implemented method 100, according to the invention, for determining parameter values ​​for an intraocular lens to be inserted (particularly into a patient's eye), is shown. In this case, method 100 includes providing 102 scans of the eye, representing images of the eye's anatomy. This can be implemented via OCT. Another method (albeit less accurate) is based on ultrasound.

[0064] Method 100 also includes determining 104 the biometric data of the eye (in a conventional manner or with the aid of a machine learning system) based on the eye scan results, and using 106 a first trained machine learning system to determine the final position of the intraocular lens (IOL) to be inserted into the eye. In this case, the long-term postoperative position of the IOL is understood to refer to the final position. The determination based on the trained machine learning system can directly determine the long-term postoperative position from one (or more) recorded images of the patient's eye; intermediate manual steps can be omitted in this process. Alternatively, ophthalmic data—particularly data from the previous step or data determined by “direct scanning”—can be used as input data for the first trained machine learning system.

[0065] Finally, method 100 includes determining a first refractive power of the intraocular lens to be inserted 108, the determination being based on a physical model in which the determined final position of the intraocular lens and determined biometric data are used as input variables. In this case, the physical model is a mathematical deterministic model.

[0066] Optionally, the determination of the final refractive power can be refined or improved by a second machine learning system. In this case, the first refractive power and at least one variable from the biometric data—e.g., axial length—are used as input data for the second trained machine learning system.

[0067] Figure 2A symbolic representation of a cross-section of the eye 200 is shown. An inserted intraocular lens 202 can be seen, which has been operatively inserted into the capsular bag 204 after the removal of the natural lens. The lateral structures 206 on the intraocular lens 202 are intended to ensure that the intraocular lens 202 is truly and stably fixed within the capsular bag 204. However, until now, it has been practically impossible to predict the precise position of the intraocular lens 202 after a relatively long growth phase, such as several weeks. This is especially true because the capsular bag 204 is much larger than the inserted intraocular lens 202, as it previously encompassed the entire natural but now removed lens. The tendons and muscles 208 that hold the capsular bag 204 in place within the eye or on the skull change after such surgery, resulting in changes to the size, shape, and position of the capsular bag 204, and consequently, the position of the inserted intraocular lens 202. Consequently, the distance between the inserted intraocular lens 202 and the retina, located further posterior to the eye, also changes. However, optimal postoperative results can only be achieved by optimally matching the refractive power (refractive index) of the inserted intraocular lens 202 with its distance to the retina. Since the refractive power of the inserted intraocular lens 202 is generally not subsequently changeable, it is highly desirable to predict the position of the inserted intraocular lens 202.

[0068] Figure 3 The eye 300 and its various biometric parameters are described. Specifically, the following parameters are represented: axial length 302 (AL), anterior chamber depth 304 (ACD), corneal curvature value 306 (K, radius), lens refractive power, lens thickness 308 (LT), central corneal thickness 310 (CCT), white-to-white distance 312 (WTW), pupil size 314 (PS), posterior chamber depth 316 (PCD), and retinal thickness 318 (RT). At least one of these parameters is included in both ophthalmic training data and patient ophthalmic data, each of which is encompassed within the scope of the concepts presented herein.

[0069] In other words, the machine learning system model, which incorporates prior physical knowledge, is initially created with the help of a physical model. This can be implemented, for example, by using a machine learning system pre-trained with simulated data or by training itself, which may include physical constraints (constraint-based training). Subsequently, the learning model is adapted to real anatomical variations using real clinical ophthalmological data. In this case, the chosen approach helps the entire machine learning system to self-learn and optimize for any data availability (e.g., after LASIK surgery). In this case, adjustments can be made explicitly for each physician or each clinic. Then, in the application phase of the machine learning system, real biometric data is used as input values ​​to determine or predict the optimized intraocular lens refractive power.

[0070] The physical model is formalized into a purely parametric form for the neural network. This parametric form can then independently and as adaptably as possible to real-world data structures during a second training phase. Thus, any amount of training data can be generated using an optical physical model. This data includes parameters of the eye model and the associated IOL refractive power as so-called ground truth. Using the concept of "transfer learning," the model trained in this way can be transferred to more complex physical models that generate training data based on the same concepts. Therefore, the neural network already possesses pre-trained artificial neurons, allowing it to adapt to stronger or better physical models more quickly and easily. This learning can be applied to models of any intensity (e.g., ray tracing models).

[0071] In the final step, the learning model is then "fine-tuned" using real biometric data from the patient's eyes, with the actual IOL refractive power used serving as the ground truth. Thus, the trained model can predict the final required IOL refractive power during the prediction phase. In practice, the more real-world data (clinical ophthalmology data) available, the better the machine learning system optimizes for that data. Therefore, the learning model can be continuously developed based on data availability, adapting to various real-world data records.

[0072] In principle, this pipeline uses a machine learning model to optimize the prediction of the anatomically correct position of the intraocular lens (IOL) using input data from OCT measurements of the patient's eye. This position is then used in a physical model, which can be any realistic model (e.g., a normal mathematical physics model or ray tracing) due to the known location of the IOL. The physical model calculates the required IOL refractive power for the eye, and the results are subsequently further refined using machine learning to correct for relatively small model errors in the physical model. To optimize the use of information, both ground-based ground-based IOL refractive power data and IOL location data are used for training.

[0073] in this regard, Figure 4 A schematic structure of the basic functional block 400 for a machine learning-supported pipeline for determining intraocular lens (IOL) size using the methods described above (including scan results / images 402 from an eye scan) is shown. These results—particularly in the form of at least one digital image—can be used for routine extraction of biometric data 404. At least some of this biometric data, along with the scan results themselves, is fed as input data to a graph-based neural network 406 to directly determine the final postoperative IOL location 408.

[0074] Next, the refractive power 410 is determined manually, based on a formula, using a mathematical physics model. Both the extracted biometric data (or a portion thereof) and the final IOL position 408 are used as input values ​​for the refractive power determination 410. Alternatively, another machine learning system 412 can be used to optimize the refractive power determination, using both the initially determined refractive power of the intraocular lens (as a result of the refractive power determination 410) and the previously determined biometric data 404 (or a portion thereof) as input data. The trained machine learning system 412 then provides the final refractive power 414 based on an appropriate machine learning model.

[0075] Since both the final postoperative IOL position (408) and the final optimized IOL refractive power can be determined through a comprehensive process with a gridded component, therefore... Figure 4 The representation of each step or functional unit specified in the document directly and clearly discloses the term pipeline.

[0076] Figure 5 For completeness, a preferred exemplary embodiment of components of a processing pipeline system 500, a machine learning-supported processing pipeline for determining parameter values ​​for an intraocular lens to be inserted, is shown. The processing pipeline system 500 includes a receiving module 502 configured to provide scan results of the eye, representing at least one image of the eye's anatomical structures.

[0077] In addition, the pipeline processing system 500 includes a determination unit 504 configured to determine biometric data of the eye based on eye scans, and a first trained machine learning system 506 for determining the final position of the intraocular lens to be inserted (see also...). Figure 4 The graph-based neural network 406), ophthalmological data was used as input data for the first machine learning system.

[0078] In addition, the processing pipeline system 500 includes a determining unit 508 configured to determine a first refractive power of the intraocular lens to be inserted (see also...). Figure 4 Functional block 410), the determination is based on a physical model in which the determined final position of the intraocular lens and the determined biometric data are used as input variables of the physical model.

[0079] Additionally, another machine learning system, 510, can be used to improve the prediction of IOL refractive power (see [link]). Figure 4 Function block 412).

[0080] It is explicitly stated that modules and units—specifically, receiving module 502, determining unit 504, the first trained machine learning system 506, and determining unit 508 for determining the first refractive power—can be connected via electrical signal lines or via the system's internal bus system 512 to transmit appropriate signals and / or data from one module (unit) to another. Furthermore, additional modules or functional units may optionally be connected to the system's internal bus system 512.

[0081] If a classification system is used as a machine learning system, the predicted refractive power is generated based on the predicted category predicted with the highest probability. Alternatively, the final refractive power of the IOL can also be implemented using a regression system as a machine learning system with numerical output variables.

[0082] In addition, system 500 may include an output unit (not depicted here) adapted to output or display the predicted final IOL refractive power, and optionally also to display the predicted IOL location.

[0083] Figure 6 A block diagram of a computer system is shown, which may have at least some parts of a system for determining refractive power. Embodiments of the concepts presented herein can, in principle, be used with virtually any type of computer, regardless of the platform used to store and / or execute program code. Figure 6 A computer system 600 is illustrated by way of example, which is adapted to execute program code according to the methods presented herein, and may also contain a prediction system in whole or in part.

[0084] Computer system 600 has multiple general-purpose functions. In this context, the computer system can be a tablet computer, a laptop / notebook computer, another portable or mobile electronic device, a microprocessor system, a microprocessor-based system, a smartphone, a computer system with specially configured special functions, or a component of a microscope system. Computer system 600 can be configured to execute computer system executable instructions (such as program modules), which can be executed to implement the functions of the concepts presented herein. For this purpose, program modules can include routines, programs, objects, components, logic, data structures, etc., to implement specific tasks or specific abstract data types.

[0085] The components of a computer system may include one or more processors or processing units 602, a storage system 604, and a bus system 606 connecting various system components, including the storage system 604, to the processor 602. The computer system 600 typically has multiple volatile or non-volatile storage media accessible to the computer system 600. The storage system 604 may store data and / or instructions (commands) from the storage media in volatile form, such as in RAM (Random Access Memory) 608, for execution by the processor 602. These data and instructions implement one or more functions and / or steps of the concepts presented herein. Other components of the storage system 604 may be persistent memory (ROM) 610 and long-term memory 612, in which program modules and data (reference numeral 616) and workflows may be stored.

[0086] The computer system includes several dedicated devices for communication purposes (keyboard 618, mouse / pointing device (not shown), visual display unit 620, etc.). These dedicated devices can also be integrated into a touch-sensitive display. A separately provided I / O controller 614 ensures frictionless data exchange with external devices. A network adapter 622 can be used for communication over a local network or a global network (LAN, WAN, e.g., via the Internet). Other components of the computer system 600 can access the network adapter via the bus system 606. In this case, although not shown, it should be understood that other devices can also be connected to the computer system 600.

[0087] System 500 for determining the refractive power of an IOL (see Figure 5 At least some parts of it can also be connected to bus system 606.

[0088] Various exemplary embodiments of the invention have been described for purposes of better understanding; however, this description is not intended to limit the concept of the invention directly to these exemplary embodiments. Further modifications and variations will be developed by those skilled in the art. The terminology used herein has been chosen to best describe the basic principles of the exemplary embodiments and to make them readily understandable to those skilled in the art.

[0089] The principles presented herein can be implemented as systems, methods, combinations thereof, and / or computer program products. In this case, the computer program product may include one or more computer-readable storage media having computer-readable program instructions to enable a processor or control system to implement various aspects of the invention.

[0090] As a medium, electronic media, magnetic media, optical media, electromagnetic media, infrared media, or semiconductor systems are used as forwarding media; for example, SSDs (Solid State Devices / Drives as Solid State Storage), RAM (Random Access Memory) and / or ROM (Read-Only Memory), EEPROM (Electrically Erasable ROM), or any combination thereof. Suitable forwarding media also include propagating electromagnetic waves, electromagnetic waves in waveguides or other transmission media (e.g., optical pulses in optical cables), or electrical signals transmitted in wires.

[0091] Computer-readable storage media can be embodied devices that retain or store instructions for use by an instruction execution device. The computer-readable program instructions described herein can also be downloaded to a corresponding computer system, for example, as a (smartphone) app, via a cable-based connection or mobile radio network from a service provider.

[0092] Computer-readable program instructions used to perform the operations of the invention described herein can be machine-dependent or machine-independent instructions, microcode, firmware, state definition data, or any source code or object code written in a conventional programming language such as C++, Java, or, for example, the programming language "C" or similar programming languages. The computer-readable program instructions can be executed entirely by a computer system. In some exemplary embodiments, the computer-readable program instructions can also be executed by electronic circuitry such as programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs) using state information of the computer-readable program instructions to configure or individualize electronic circuitry according to aspects of the invention.

[0093] The invention presented herein is further illustrated with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to exemplary embodiments of the invention. It should be noted that any block in the flowchart illustrations and / or block diagrams can actually be implemented as computer-readable program instructions.

[0094] Computer-readable program instructions can be used in a general-purpose computer, special-purpose computer, or data processing system that can be programmed in some other way to produce a machine such that the instructions, which are executable by a processor or computer or other programmable data processing device, can generate means for implementing the functions or processes shown in the flowcharts and / or block diagrams. These computer-readable program instructions may also be stored accordingly on a computer-readable storage medium.

[0095] In this sense, any box in the flowchart or block diagram shown can represent a module, an instruction segment, or an instruction portion of multiple executable instructions for implementing a specific logical function. In some exemplary embodiments, the functions shown in the various boxes can be implemented in different orders, or optionally in parallel.

[0096] All structures, materials, sequences, and equivalents of the methods and / or steps with associated functions shown in the appended claims are intended to apply all structures, materials, or sequences expressed in the claims.

[0097] Figure Labels

[0098] 100 Context-Sensitive White Balance Methods

[0099] 102 100 Method Steps

[0100] 104 100 Method and Steps

[0101] 106 100 Method Steps

[0102] 108 100 Method Steps

[0103] 110 100 Method Steps

[0104] 200 Eyes with artificial lenses

[0105] 202 Intraocular Lens

[0106] 204 Pouch

[0107] 206 Horizontal Structure

[0108] 208 Tendons and Muscles

[0109] 300 eyes

[0110] 302 Axial length

[0111] 304 anterior chamber thickness

[0112] 306 corneal curvature value

[0113] 308 Lens thickness

[0114] 310 Central corneal thickness

[0115] 312 White-to-white distance

[0116] 314 Pupil size

[0117] 316 Rear Chamber Depth

[0118] 318 Retinal thickness

[0119] 400 Functional blocks for implementing the method

[0120] 402 scan results

[0121] 404 Biometric Data

[0122] 406 Neural Network

[0123] 408 IOL Location

[0124] 410 Determination of refractive power based on formula

[0125] 412 Machine Learning System

[0126] 414 Final refractive power

[0127] 500 Systems for Predicting Refractive Power

[0128] 502 Production Module

[0129] 504 First Training Module

[0130] 506 Second Training Module

[0131] 508 Receiver Module

[0132] 510 Another machine learning system

[0133] 512 bus system

[0134] 600 Prediction System

[0135] 600 Computer Systems

[0136] 602 processor

[0137] 604 Storage System

[0138] 606 bus system

[0139] 608 RAM

[0140] 610 ROM

[0141] 612 Long-Term Memory

[0142] 614 I / O Controller

[0143] 616 Program Module, Potential Data

[0144] 618 Keyboard

[0145] 620 screen

[0146] 622 Network adapter.

Claims

1. A computer-implemented method (100) for a machine learning-supported processing pipeline for determining parameter values ​​of an intraocular lens (202) to be inserted, the method (100) comprising the following steps - Provide (102) a scan result (402) of the eye (300), wherein the scan result (402) represents an image of the anatomical structure of the eye (300), - Based on the eye scan results (402), determine (104) the biometric data (404) of the eye (300). - A first machine learning system (406) trained (106) is used to determine the final position (408) of the intraocular lens (202) to be inserted, wherein biometric data (404) is used as input data for the first machine learning system (406). - Based on the physical model (410), the first refractive power (414) of the intraocular lens (202) to be inserted is determined (108). In this physical model, the determined final position (408) of the intraocular lens (202) and the determined biometric data (404) are used as input variables of the physical model (410). - The final refractive power of the intraocular lens (202) is determined by a second machine learning system (412), wherein at least one variable from the biometric data (404) and the first refractive power (414) are used as input variables. The second machine learning system (412) is trained in two phases, in which - The first training step includes: - A first physical model based on the refractive power of the intraocular lens (202) generates first training data for the second machine learning system (412). - The second machine learning system (412) is trained using the generated first training data to form a first learning model for determining refractive power, and - The second training step includes: - Use clinical ophthalmology training data to train the second machine learning system (412) trained with the first training data to form a second learning model for determining refractive power.

2. The method (100) according to claim 1, wherein, The biometric data of the eye include at least one of the following groups: preoperative axial length (302), preoperative lens thickness, preoperative anterior chamber depth (304), and intraoperative anterior chamber depth.

3. The method (100) according to claim 1 or 2, wherein, The first machine learning system (406) is a convolutional neural network, a graph attention network, or a combination of the two.

4. The method (100) according to claim 1, wherein, One variable derived from this biometric data (404) is the preoperative axial length (302).

5. The method (100) according to claim 1 or 2, wherein, The biometric data (404) of the eye is determined manually from the image or by a machine learning system from the scan results (402) provided by the eye (300).

6. The method (100) according to claim 1 or 2, wherein, When the final position (408) of the intraocular lens (202) to be inserted is determined, other parameters of the eye (300) are determined, wherein the other parameters of the eye (300) are the final position of the intraocular lens (202) after the growth process or the intraocular lens offset perpendicular to the optical axis of the eye (300).

7. A machine learning-supported processing pipeline system (500) for determining parameter values ​​of an intraocular lens (202) to be inserted, the system (500) comprising: - A receiving module (502) for providing scan results (402) of an eye (300), wherein the scan results (402) represent an image of the anatomical structure of the eye (300), - Determination unit (504) for determining the biometric data (404) of the eye (300) based on the scanning results (402) of the eye (300). - A first machine learning system (406) trained to determine the final position (408) of the intraocular lens (202) to be inserted, wherein biometric data (404) is used as input data for the first machine learning system (406). - A determining unit (508) for determining the first refractive power of the intraocular lens to be inserted based on a physical model, in which the determined final position of the intraocular lens and determined biometric data are used as input variables of the physical model, and - A determining unit for determining the final refractive power of the intraocular lens via a second machine learning system, wherein at least one variable from the biometric data (404) and the first refractive power (414) are used as input variables. The second machine learning system (412) is trained in two phases, in which - The first training step includes: - First training data is generated for the second machine learning system based on a first physical model of the refractive power of intraocular lenses. - The second machine learning system is trained using the generated first training data to form a first learning model for determining refractive power, and - The second training step includes: - Use clinical ophthalmology training data to train the second machine learning system trained with the first training data to form a second learning model for determining refractive power.

8. A machine learning-supported processing pipeline computer program product for determining parameter values ​​for an intraocular lens to be inserted, wherein, The computer program product has a computer-readable storage medium having program instructions stored thereon, wherein the program instructions are executable by one or more computers or control units and cause the one or more computers or control units to perform the method according to any one of claims 1 to 6.

Images