A method for measuring an eyeball and eyeball measuring device

Abstract

A method for measuring an eyeball comprises irradiating a laser beam on a point on a surface of an eyeball with a laser diode of an SMI sensor, capturing a time-resolved signal indicative of a position of the point with the SMI sensor, and calculating a model of the eyeball on the basis of the time resolved signal.

Description

[0001] A METHOD FOR MEASURING AN EYEBALL AND EYEBALL MEASURING

[0002] DEVICE

[0003] DESCRIPTION

[0004] The present invention relates to a method for measuring an eyeball and to an eyeball measuring device .

[0005] This patent application claims the priority of German patent application 10 2024 100 355 . 3 , the disclosure content of which is hereby incorporated by reference .

[0006] Corneal topography is known in the state of the art . Established methods are described in M . Corbett , Corneal Topography, Principals and Applications , second edition, Springer, 2019 , for example . Sel f-mixing interferometry ( SMI ) sensors are also known in the state of the art .

[0007] It is an obj ect of the present invention to provide a method for measuring an eyeball . It is a further obj ect of the present invention to provide an eyeball measuring device . These obj ectives are achieved by a method for measuring an eyeball and by an eyeball measuring device according to the independent claims . Further variants are disclosed in the dependent claims .

[0008] A method for measuring an eyeball comprises irradiating a laser beam on a point on a surface of an eyeball with a laser diode of an SMI sensor, capturing a time-resolved signal indicative of a position of the point with the SMI sensor, and calculating a model of the eyeball on the basis of the time- resolved signal .

[0009] This method allows for determining a model of the eyeball that describes a topography of the eyeball . Advantageously, the method does not require any interaction or cooperation of the user . The user does not even need to be aware of the measuring . The method can be carried out in any kind of setting and does not require presence of a medical expert .

[0010] In a variant of the method, the SMI sensor is operated in FMCW ( frequency modulated continuous wave ) mode . This allows the SMI sensor to determine both an absolute distance between the SMI sensor and the point on the surface of the eyeball and a relative displacement of the point during eye motion .

[0011] In a variant of the method, the time-resolved signal is pre- processed using a detrending routine , a bandpass filter, a scaling routine , or a temporal smoothing routine . Advantageously, applying such standard pre-processing to the time- resolved signal may improve the data quality of the time-re- solved signal and may allow to calculate a model of the eyeball with increased accuracy .

[0012] In a variant of the method, an input feature is extracted from the time-resolved signal . The model of the eyeball is calculated on the basis of the input feature . The input feature may represent an information that is contained in the time-resolved signal and that is signi ficant for the topography of the eyeball . Calculating the model of the eyeball on the basis of the input feature may simpli fy the calculation of the model .

[0013] In a variant of the method, the time-resolved signal is trans formed into the frequency domain before extracting the input feature . The trans formation into the frequency domain may be carried out by applying a FFT ( fast Fourier trans formation) , for example . The frequency domain may be particularly suitable for extracting the input feature .

[0014] In a variant of the method, the input feature is a distance between the point and the laser diode or a displacement of the point relative to the laser diode during eye motion . The displacement of the point may contain information about a local curvature of the surface of the eyeball in the vicinity of the point . Advantageously, both the distance between the point and the laser diode and the displacement of the point may help to determine a topography of the surface of the eyeball .

[0015] In a variant of the method, a power spectral density is calculated from the time-resolved signal . The input feature is extracted from the power spectral density . The power spectral density may be a useful representation for extracting the input feature .

[0016] In a variant of the method, the input feature is a multi-dimensional vector . In this variant , the input feature is extracted using a trained neural network . The extracted input feature may contain a high amount of useful information that describes the topography of the surface of the eyeball .

[0017] In a variant of the method, the trained neural network comprises convolutional layers with residual blocks and squeeze- and-excitation blocks . Advantageously, a neural network with such an architecture had proven to be suitable for extracting a useful input feature from the time-resolved signal .

[0018] In a variant of the method, the model of the eyeball is calculated using a numerical optimi zation routine . This may include an application of standard mathematical optimi zation routines and may advantageously be carried out in an easy and computationally cheap way .

[0019] In another variant of the method, the model of the eyeball is calculated using a trained machine learning model . Such a machine learning model can be trained in a simple way using data obtained on an arti ficial eyeball , for example .

[0020] In a variant of the method, the machine learning model is a deep learning model . In this way, the machine learning model may take advantage of many aspects of the information contained in the time-resolved signal . In a variant of the method, the model of the eyeball includes a parameterised 3D-model . This may allow to describe the eyeball with a low number of parameters .

[0021] In another variant of the method, the model of the eyeball includes a parameterised topographical model of the surface of the eyeball . This may also allow to describe the eyeball with a relatively low number of parameters and may still allow to model even complex shapes of the surface of the eyeball .

[0022] In another variant of the method, the model of the eyeball includes a numerical topographical map of the surface of the eyeball . Advantageously, using a numerical topographical map of the surface of the eyeball allows for a high flexibility of the model of the eyeball .

[0023] In a variant of the method, a time series of time-resolved signals is captured successively with the SMI sensor . The model of the eyeball is calculated on the basis of the time series of time-resolved signals . In this variant , data describing the eyeball is collected over an extended period of time before the model of the eyeball is calculated on the basis of the collected data . Advantageously, this may allow for calculating an accurate model of the eyeball .

[0024] In a variant of the method, several time-resolved signals indicative of positions of several points on the surface of the eyeball are captured with several SMI sensors . The model of the eyeball is calculated on the basis of the several time- resolved signals . Using information related to several points on the surface of the eyeball advantageously allows for calculating a model of the eyeball with high accuracy .

[0025] An eyeball measuring device comprises an SMI sensor and a processing unit . The SMI sensor comprises a laser diode adapted for irradiating a laser beam on a point on a surface of an eyeball . The SMI sensor is capable of capturing a time- resolved signal indicative of a position of the point . The processing unit is adapted for calculating a model of the eyeball on the basis of the time-resolved signal .

[0026] This device allows for determining a model of the eyeball that describes a topography of the eyeball . Advantageously, the device does not require any interaction or cooperation of the user . The user does not even need to be aware of the measuring . The device can be used in any kind of setting and does not require presence of a medical expert .

[0027] A variant of the eyeball measuring device comprises a plurality of SMI sensors . Each SMI sensor comprises a respective laser diode adapted for irradiating a laser beam on a respective point on the surface of the eyeball . Each SMI sensor is capable of capturing a time-resolved signal indicative of a position of the respective point . The processing unit is adapted for calculating the model of the eyeball on the basis of the plurality of time-resolved signals . Using information related to several points on the surface of the eyeball advantageously allows for calculating a model of the eyeball with high accuracy .

[0028] A variant of the eyeball measuring device is integrated into a pair of glasses . This allows a user to wear the eyeball measuring device over an extended period of time without being disturbed or limited by the eyeball measuring device .

[0029] The above-described properties , features , and advantages of the invention, as well as the way in which they are achieved, will become more clearly and comprehensively understandable in connection with the following description of exemplary variants , which will be explained in more detail in connection with the drawings in which, in schematic representation :

[0030] Fig . 1 shows an eyeball measuring device ; Fig . 2 shows a flow diagram of a method for measuring an eyeball ;

[0031] Fig . 3 shows a time-resolved SMI signal ;

[0032] Fig . 4 shows a parameterised 3D-model of an eyeball ;

[0033] Fig . 5 shows a parameterised topographical model of an eyeball ; and

[0034] Fig . 6 shows a numerical topographical map of a surface of an eyeball .

[0035] Figure 1 shows a schematic depiction of an eyeball measuring device 10 that can be used to measure an eyeball 110 of a person 100 . To this end, the eyeball measuring device 10 is adapted for executing a measuring method 400 that is schematically depicted in figure 2 . The eyeball measuring device 10 may be integrated into a pair of glasses 11 that is worn by the person 100 , for example . Alternatively, the eyeball measuring device 10 may be integrated into a tabletop or a bench- top station, for example .

[0036] The eyeball measuring device 10 comprises a plurality of SMI ( sel f-mixing interferometry) sensors 200 . In the example depicted in figure 1 , the eyeball measuring device 10 comprises two SMI sensors 200 . In other variants , however, the eyeball measuring device 10 may comprise only one SMI sensor 200 or more than two SMI sensors 200 .

[0037] Each SMI sensor 200 comprises a laser diode 210 that is adapted for emitting a laser beam 220 and irradiating the laser beam 220 on a respective point 120 on a surface of the eyeball 110 of the person 100 . The laser beam 220 may have a wavelength from the infrared ( IR) spectral range , for example . The laser diode 210 may be a VCSEL (vertical cavity surface emitting laser ) , for example . Each SMI sensor 200 may comprise an optical element 230 that is arranged in the beam path of the laser beam 220 of the respective SMI sensor 200 between the respective laser diode 210 and the respective point 120 on the surface of the eyeball 110 such that the laser beam 220 is irradiated on the eyeball 110 via the optical element 230 . The optical element 230 may be a transmissive , a reflective or a di f fractive optical element , for example . The optical element 230 may serve to shape the laser beam 220 . The optical element 230 may reduce a divergence of the laser beam 220 , for example . In this case , the optical element 230 may collimate the laser beam 220 or may focus the laser beam 220 on the respective point 120 on the surface of the eyeball 110 . The optical element 230 may comprise one or more optical components such as mirrors , lenses , and di f fractive elements . The optical element 230 may be omitted for some or all the SMI sensors 200 , however .

[0038] A part of the light of each laser beam 220 is reflected at the respective point 120 on the surface of the eyeball 110 and reaches back into the respective laser diode 210 via the respective optical element 230 , where it interferes with the laser light generated by the laser diode 210 . Depending on a position 130 of the respective point 120 and a resulting optical distance between the respective point 120 and the respective laser diode 210 , constructive or destructive interference occurs . This can be detected as an SMI signal 520 ( figure 3 ) by the respective SMI sensor 200 , for example by measuring a j unction voltage of the respective laser diode 210 or with the help of a photodiode .

[0039] The eyeball measuring device 10 is adapted to capture a time- resolved signal 500 of the SMI signal 520 as a function of time 510 for each SMI sensor 200 in a signal capturing step 410 of the measuring method 400 . An example of such a time- resolved signal 500 is schematically depicted in figure 3 . The time-resolved signal 500 is indicative of the position 130 of the respective point 120 on the surface of the eyeball 110 .

[0040] I f the respective SMI sensor 200 is operated in constant mode with a constant wavelength of the laser beam 220 , the time- resolved signal 500 may be indicative of a displacement 132 of the respective point 120 due to eye motion . The displacement 132 is a change of the position 130 of the respective point 120 in a direction towards or away from the respective laser diode 110 .

[0041] I f the respective SMI sensor 200 is operated in FMCW ( frequency modulated continuous wave ) mode with a wavelength of the laser beam 220 that changes as a function of time , the time-resolved signal 500 may be indicative of an absolute distance 131 between a respective point 120 and the respective laser diode 210 , as well as a local displacement 132 of the point 120 .

[0042] The eyeball measuring device 10 comprises a processing unit 300 . The processing unit 300 may comprise a microcontroller or a microprocessor, for example . The processing unit 300 is adapted for calculating a model 600 of the eyeball 110 in a model calculation step 440 of the measuring method 400 . The model 600 of the eyeball 110 is calculated on the basis of the time-resolved signals 500 obtained with the SMI sensors 200 of the eyeball measuring device 10 .

[0043] The model 600 may be a parameterised 3D model 610 of the eyeball 10 , as schematically shown in figure 4 . The parameterised 3D model 610 may model the eyeball 110 using a set of simple geometric shapes such as ellipsoids and a number of parameters 611 that describe the si ze and orientation of these shapes .

[0044] The model 600 may alternatively be a parameterised topographical model 620 of the surface of the eyeball 110 , as schematically shown in figure 5 . The parameterised topographical model 620 may use a set of functions and parameters 611 to describe the topography of the surface of the eyeball 110 .

[0045] The model 600 may alternatively be a numerical topographical map 630 of the surface of the eyeball 110 , as schematically shown in figure 6 . The numerical topographical map 630 may be a height map of the topography of the surface of the eyeball 110 .

[0046] In some variants of the eyeball measuring device 10 , more than one model 600 of the eyeball 110 is calculated . The eyeball measuring device 10 may be adapted for calculating any combination of the parameterised 3D model 610 , the parameterised topographical model 620 and the numerical topographical map 630 , for example .

[0047] The measuring method 400 may comprise an optional pre-processing step 420 that is carried out before the model calculation step 440 . In the pre-processing step 420 , each time- resolved signal 500 may be pre-processed using standard timeseries signal processing routines such as a detrending routine , a bandpass filter, a scaling routine , or a temporal smoothing routine . The pre-processing step 420 may help to increase the signal-to-noise ratio of the respective time-re- solved signal 500 .

[0048] The measuring method 400 may include a feature extraction step 430 that is carried out before the model calculation step 440 . In the feature extraction step 430 , at least one input feature 550 is extracted from each time-resolved signal 500 . Each input feature 550 may be a scalar value or a multidimensional vector . The model 600 of the eyeball 110 is then calculated on the basis of the input features 550 in the model calculation step 440 .

[0049] In some variants of the measuring method 400 , each time-re- solved signal 500 is trans formed into the frequency domain for extracting the respective input feature 550 . This trans formation may be carried out using an FFT ( fast Fourier trans formation) or a DCT ( discrete cosine trans formation) , for example .

[0050] In some variants of the measuring method, the extracted input feature 550 may include the displacement 132 of the respective point 120 or the absolute distance 131 of the respective point 120 . The displacement 132 may be extracted from the time-resolved signal 500 using a classical fringe counting approach, for example . The absolute distance 131 may be extracted from the time-resolved signal 500 as the input feature 500 in a known manner by identi fying a beat frequency i f the respective SMI sensor 200 is operated in FMCW mode . In some variants , the beat frequency may be identi fied by a peak finding routine that uses an argmax approach and looks for the highest frequency peak, a peak finding routine that looks for the highest peak that corresponds to a speci fic width and height , or a peak finding routine that uses a machine learning based approach .

[0051] In alternative variants of the measuring method 400 , the input feature 550 is extracted from each time-resolved signal 500 using a trained neural network . In this case , the respective input feature 550 may be a multi-dimensional vector with e . g . between 50 and 200 dimensions . The trained neural network may extract the input feature 550 from the time-resolved signal 500 in the time domain or after trans forming the time- resolved signal 500 into the frequency domain . Another alternative is to first calculate a power spectral density from the time-resolved signal 500 and extract the input feature 550 from the power spectral density .

[0052] The trained neural network may comprise convolutional layers with residual blocks and squeeze-and-excitation blocks , for example .

[0053] Calculating the model 600 of the eyeball 110 in the model calculation step 440 of the measuring method 400 may use further information in addition to the input features 550 extracted from the time-resolved signals 500 . For example , the model calculation step 440 may use an eye-gaze direction that describes a rotation of the eyeball 110 of the person 100 . The eye-gaze direction may be determined using the SMI sensors 200 of the eyeball measuring device 10 or using additional sensors of the eyeball measuring device 10 such as additional SMI sensors or an optical camera .

[0054] In the model calculation step 440 of the measuring method 400 , the model 600 of the eyeball 110 may be calculated using a numerical optimi zation routine . This approach is convenient i f the input features 550 are scalar values such as the distance 131 and the displacement 132 of the respective point 120 on the surface of the eyeball 110 and i f the model 600 is a parameterised 3D model 610 or a parameterised topographical model 620 .

[0055] In other variants of the measuring method 400 , the model 600 of the eyeball 110 is calculated using a trained machine learning model in the model calculation step 440 . Also in this variant , it is convenient i f the input features 550 extracted from the time-resolved signals 500 are scalar values such as the distance 131 and the displacement 132 of the respective point 120 . The trained machine learning model may include a multi-linear and multi-label regression model with 11 or 12 regularisation such as Lasso , ElasticNet , or Ridge regression, for example . Other regression methods such as RandomForest Regression or Support Vector Regression can alternatively be used .

[0056] The machine learning model may be trained using data that has been obtained using an arti ficial eyeball , for example .

[0057] In other variants of the measuring method 400 , the trained machine learning model that is used for calculating the model 600 of the eyeball 110 in the model calculation step 440 , is a deep learning model . In this case , the extracted input features 550 may be scalar values such as the distance 131 and the displacement 132 of the respective point 120 , or multi-dimensional vectors to allow for calculating a detailed model 600 . The deep learning model may conform to the U-NET architecture and may comprise convolutional layers and skip layers , for example . Other neural network architectures such as attention layers , or trans former architectures to encode the relevant information may be used as well .

[0058] The deep learning model may be trained using data that has been collected beforehand using arti ficial eyeballs . In case that the input features 550 are extracted using a trained neural network as well , both neural networks may be trained together .

[0059] In some variants of the measuring methods 400 , a time series of several time-resolved signals 500 is captured successively with each of the SMI sensors 200 of the eyeball measuring device 10 . This data is collected over a period of time such as several minutes , several days or several weeks . Only after enough data has been collected, the model 600 of the eyeball 110 is calculated on the basis of the time series of time-re- solved signals 500 of all SMI sensors 200 .

[0060] It is possible to re-calculate the model 600 of the eyeball 110 after a set amount of time to identi fy changes of the eyeball 110 which might have medical importance .

[0061] The invention has been illustrated and described in more detail with the aid of exemplary variants . The invention is not , however, restricted to the examples disclosed . Rather, other variants may be derived therefrom by the person skilled in the art . REFERENCE SYMBOLS eyeball measuring device pair of glasses person eyeball point position distance displacement SMI sensor laser diode laser beam optical element processing unit measuring method signal capturing step pre-processing step feature extraction step model calculation step time-resolved signal time SMI signal input feature model parameterised 3D model parameter parameterised topographical model numerical topographical map

Claims

CLAIMS1. A method for measuring an eyeball, the method comprising- irradiating a laser beam (220) on a point (120) on a surface of an eyeball (110) with a laser diode (210) of an SMI sensor (200) ;- capturing a time-resolved signal (500) indicative of a position (130) of the point (120) with the SMI sensor (200) ;- calculating a model (600) of the eyeball (110) on the basis of the time-resolved signal (500) .

2. The method as claimed in claim 1, wherein the SMI sensor (200) is operated in FMCW mode.

3. The method as claimed in any one of the preceding claims, wherein the time-resolved signal (500) is pre-processed using a detrending routine, a bandpass filter, a scaling routine, or a temporal smoothing routine.

4. The method as claimed in any one of the preceding claims, wherein an input feature (550) is extracted from the time-resolved signal (500) , wherein the model (600) of the eyeball (110) is calculated on the basis of the input feature (550) .

5. The method as claimed in claim 4, wherein the time-resolved signal (500) is transformed into the frequency domain before extracting the input feature ( 550 ) .

6. The method as claimed in any one of claims 4 and 5, wherein the input feature (550) is a distance (131) between the point (120) and the laser diode (210) or a displacement (132) of the point (120) relative to the laser diode (210) .

7. The method as claimed in any one of claims 4 and 5, wherein a power spectral density is calculated from the time-resolved signal (500) , wherein the input feature (550) is extracted from the power spectral density.

8. The method as claimed in any one of claims 4, 5, and 7, wherein the input feature (550) is a multi-dimensional vector, wherein the input feature (550) is extracted using a trained neural network.

9. The method as claimed in claim 8, wherein the trained neural network comprises convolutional layers with residual blocks and squeeze-and-exci- tation blocks.

10. The method as claimed in any one of claims 1 to 6, wherein the model (600) of the eyeball (110) is calculated using a numerical optimization routine.

11. The method as claimed in any one of claims 1 to 9, wherein the model (600) of the eyeball (110) is calculated using a trained machine learning model.

12. The method as claimed in claim 11, wherein the machine learning model is a deep learning model .

13. The method as claimed in any one of the preceding claims, wherein the model (600) of the eyeball (110) includes a parameterised 3D model (610) .

14. The method as claimed in any one of claims 1 to 12, wherein the model (600) of the eyeball (110) includes a parameterised topographical model (620) of the surface of the eyeball (110) .

15. The method as claimed in any one of claims 11 and 12, wherein the model (600) of the eyeball (110) includes a numerical topographical map (630) of the surface of the eyeball (110) .

16. The method as claimed in any one of the preceding claims, wherein a time series of time-resolved signals (500) is captured successively with the SMI sensor (200) , wherein the model (600) of the eyeball (110) is calculated on the basis of the time series of time-resolved signals ( 500 ) .

17. The method as claimed in any one of the preceding claims, wherein several time-resolved signals (500) indicative of positions (130) of several points (120) on the surface of the eyeball (110) are captured with several SMI sensors (200) , wherein the model (600) of the eyeball (110) is calculated on the basis of the several time-resolved signals (500) .

18. An eyeball measuring device comprising an SMI sensor (200) and a processing unit (300) , wherein the SMI sensor (200) comprises a laser diode (210) adapted for irradiating a laser beam (220) on a point (120) on a surface of an eyeball (110) , wherein the SMI sensor (200) is capable of capturing a time-resolved signal (500) indicative of a position (130) of the point (120) , wherein the processing unit (300) is adapted for calculating a model (600) of the eyeball (110) on the basis of the time-resolved signal (500) .

19. The eyeball measuring device as claimed in claim 18, comprising a plurality of SMI sensors (200) , wherein each SMI sensor (200) comprises a respective laser diode (210) adapted for irradiating a laser beam(220) on a respective point (120) on the surface of the eyeball (110) , wherein each SMI sensor (200) is capable of capturing a time-resolved signal (500) indicative of a position (130) of the respective point (120) , wherein the processing unit (300) is adapted for calculating the model (600) of the eyeball (110) on the basis of the plurality of time-resolved signals (500) .

20. The eyeball measuring device as claimed in any one of claims 18 and 19, wherein the eyeball (110) measuring device is integrated into a pair of glasses (11) .

Images