Eye-imaging device

Abstract

The invention provides a device (30) for imaging and / or measuring an eye (42) having a cornea (40), the device comprising a camera (32) which has a field of view and which is for obtaining a camera image of the eye viewed along a first direction, a mirror (44, 56) which is mounted in a predetermined position and orientation relative to the camera and in the camera's field of view to provide, in the camera image, a reflection showing the cornea viewed along a second direction which is not parallel to the first direction, and a processing device configured to determine, by analysis of the reflection, a distance from the camera to the cornea.

Description

[0001] EYE-IMAGING DEVICE

[0002] The present invention is concerned with imaging and / or measurement of the eye.

[0003] One application of the present invention is in relation to corneal topography, a technique used to measure the shape of the cornea's front surface. Corneal topography is used, for example, to identify corneal irregularities and distortions, to help plan refractive surgeries and to improve fit of contact lenses.

[0004] The principles of corneal topography are familiar to the skilled person. A known pattern of light is projected onto the cornea and is imaged using a camera of some form. Information about the profile of the cornea is determined from the reflected light pattern in the resultant image, typically by computer analysis.

[0005] The reflected light pattern depends not only on the profile of the cornea but also on the alignment of the eye with respect to optics of the corneal topography machine, and especially on distance of the cornea from the optics, so that variation of that distance is a potential cause of error. To provide for control of the alignment of the cornea, commercial corneal topography machines often have a chin rest and a forehead rest to assist the patient in keeping their head in a fixed position, with provision for adjustment of the position of these rests relative to the machine's optics to achieve a desired alignment.

[0006] Provision may be made to measure the distance of the cornea from some reference point on the corneal topography machine, enabling the distance from the cornea to the machine's optics to be adjusted to a desired value by movement of the rests. Specifically, it is known to make this measurement using the arrangement schematically depicted in Figure 1, in which a corneal topography machine 10 comprises a camera 12 directed toward cornea 14. A sensor unit 16 in the corneal topography device emits ultrasound waves and receives their reflection from the cornea, the time from transmission to reflection being used to estimate the distance from the sensor unit 16 to the cornea 14. The required ultrasound sensor 16 adds cost and weight to the machine, and is not well suited to miniaturisation.

[0007] Another arrangement for determining misalignment of the cornea is depicted in Figure 2 and uses a light source 18 to emit a light beam 20 along an inclined direction onto the cornea, and a light sensor 22 to detect the beam's reflection from the cornea, the arrangement being such that the detected intensity of the reflected beam is maximal when the cornea is correctly positioned. Actual distance information is not provided by this technique.

[0008] Corneal topography machines are typically substantial and expensive desktop devices.

[0009] An improved apparatus and method are thus sought for determining the distance of the cornea from a device being used to image it.

[0010] According to the present invention, there is a device for imaging and / or measuring an eye having a cornea, the device comprising a camera which has a field of view and which is for obtaining a camera image of the eye viewed along a first direction, a mirror which is mounted in a predetermined position and orientation relative to the camera and in the camera's field of view to provide, in the camera image, a reflection showing the cornea viewed along a second direction which is not parallel to the first direction, and a processing device configured to determine, by analysis of the reflection, a distance from the camera to the cornea.

[0011] According to a second aspect of the present invention, there is a method of determining distance of a cornea of an eye from a camera of a device, the camera having a field of view and the method comprising:

[0012] - mounting a mirror in a predetermined position and orientation relative to the camera and in the field of view,

[0013] -imaging the eye with the camera to provide a camera image showing the eye viewed along a first direction and containing a reflection from the mirror which shows the cornea viewed along a second direction which is not parallel to the first direction, and

[0014] - analysing the reflection to determine the distance of the cornea from the camera.

[0015] The applications of the invention are not limited to corneal topography, nor to ophthalmology in general. Rather, the invention can be applied in any situation where the eye is imaged by a device, and where it is desired to know the distance of the eye from the device.

[0016] Specific embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which:-

[0017] Figure 1 is a highly simplified sectional view of some components of a corneal topography device belonging to the prior art;

[0018] Figure 2 is a similarly simplified sectional view of another corneal topography device belonging to the prior art;

[0019] Figure 3 is a highly simplified sectional view of some components of a corneal topography device embodying the present invention;

[0020] Figure 4 is an image taken using a device of the type depicted in Figure 3;

[0021] Figure 5 shows part of the Figure 4 image to an enlarged scale;

[0022] Figure 6 depicts an experimental apparatus used in calibration of the corneal topography device of Figure 3;

[0023] Figure 7 is a graph showing variation of corneal-image location (in pixels, on the horizontal axis) with corneal separation from an experimental apparatus (in millimetres, on the vertical axis); and

[0024] Figure 8 is a section in a longitudinal plane through a front end of housing of a corneal topography device embodying the present invention. The arrangement depicted in Figure 3 comprises major components of a corneal topography device 30 embodying the present invention, having a camera 32 which is provided with a lens schematically represented at 36 and is mounted in a housing 38 to image the cornea 40 of an eye 42 of a patient, which will typically be a human patient although there is no reason why the invention cannot be applied to the imaging of animal corneas.

[0025] A light-emiting arrangement, e.g. in the form of a ring projection system or so-called Placido disc system, is provided to project a known patern of light onto the cornea. This patern may for example comprise concentric rings, or dots arranged in concentric rings. LEDs may be used. This aspect of corneal topography is well known to the skilled person and need not be described further here.

[0026] Figure 3 shows a measurement axis Z which passes through the camera 32 and extends along its field of view, in the present example, the axis Z extends along the centre of the camera's field of view, although this may differ in other embodiments. The eye 42 is seen in Figure 3 to be aligned (in X and Y directions which are orthogonal to the Z direction and to each other) such that the eye's centre lies on the Z axis. The eye's pupil, at the cornea's centre, also lies on the axis Z, although this is not seen. That is to say that the patient is looking along the axis Z.

[0027] The present invention provides for the position of the cornea 40 along the axis Z to be determined. To put this another way, the invention provides for measurement of a corneal distance D from a chosen point on the device 30 to the cornea 40. For this purpose, and in accordance with the present invention, at least one mirror 44 is disposed in the field of view of the camera 34. The position and orientation of the mirror relative to the camera are known, and in this embodiment are fixed. The mirror 44 is in this embodiment carried by the housing 38.

[0028] The mirror 44 is positioned and oriented to provide the camera 32 with a reflected view of the cornea 40 along a lateral direction. The direction of view of the cornea 40 provided to the camera by the mirror 44 may be tangential or close to tangential to the corneal anterior surface. Figure 3 shows the path of a notional light ray travelling at 46 from the cornea to the mirror 44, and at 48 from the mirror 44 to the lens 36 of the camera 32. Hence the mirror provides a view of the cornea along the direction of the light ray at 46, which can be seen to be lateral with respect to the axis Z. The direction of view may be perpendicular, or close to perpendicular, to the axis Z. It may be within plus or minus 30 degrees of a line perpendicular to the axis Z. It may be within plus or minus fifteen degrees of a line perpendicular to the axis Z.

[0029] To provide the desired view of the cornea, the mirror 44 is laterally offset from the axis Z. The offset of the mirror 44 from the axis Z is large enough that the mirror 44 does not obscure the camera's view of at least the iris of the eye 42. The offset may be large enough that the mirror does not substantially obscure the camera's view of the exposed anterior surface of the eye 42. The offset may be large enough that the mirror 44 lies in a part of the image provided by the camera 32 that is not needed for corneal topography.

[0030] The mirror 44 is disposed at or close to an end of the housing 38 which lies toward the eye 42, in use. The mirror 44 of the present embodiment has a reflecting face which is a flat plane, although use of a curved mirror is possible, if its shape is suitably allowed for in calculations to be explained below. A normal 54 to the reflecting face of the mirror 44 (that is, a line perpendicular to the plane of the reflecting face) forms an angle A to the axis Z which, measured on the side facing toward the camera 34, is obtuse. The angle A may be between 100 and 170 degrees. It may be between 115 and 155 degrees.

[0031] The camera 34 / lens 36 have a focal plane which at least substantially coincides with the anterior surface of the cornea 40, to provide a focussed image of it, and the mirror 44 lies in or close to the focal plane.

[0032] A single mirror 44 is sufficient to determine the distance D, but in the present embodiment there is a second mirror 56 disposed on the opposite side of the axis Z from the mirror 44. The arrangement of the two mirrors is symmetrical about the axis Z. In use, one of the mirrors 44, 56 lies on a nasal side of the axis Z and the other lies on a temporal side of the axis Z. The mirror 44, 56 on the nasal side may receive weaker light intensity due to the presence of the patient's nose. By providing a pair of mirrors, it is ensured that the corneal imaging device 30 performs equally well in imaging the patient's left and right eyes.

[0033] Figure 4 depicts an image 58 obtained by the camera 32 of the corneal topography device 30, in which can be seen the eye 42 including its pupil 60 and iris 62, as well as the eyelids 64, 66. The reflection from the mirror 44 is seen at 68 and the reflection from the second mirror 56 is seen at 70. It can be appreciated that the mirrors 44, 56 are sufficiently laterally offset that they do not obscure the desired view of the eye 42, A pattern of dots 72 is also visible: this is the light pattern projected onto the cornea 40 by the device 30 to make corneal topography possible. The means used to project this light pattern are not depicted in Figure 3.

[0034] Figure 5 is an enlargement of the reflection 68 from the mirror 44, showing the approximately tangential view of the cornea 40 reflected by the mirror 44 to the camera 32. It can be seen that the reflection 68 includes both the anterior and posterior surfaces 74, 76 of the cornea 40, and that the cornea, being curved, appears as an approximately part-circular arc.

[0035] The image 58, and in particular the portion of it which shows the reflection 68, is analysed to determine the corneal distance D. The principle is that the position of the cornea in the image, along a direction J which is indicated in Figure 5, is a function of the corneal distance D. Knowing the relationship between the position of the cornea in the image and the distance D, one can thus use the former to determine the latter. This relationship could be established geometrically, but was in the present embodiment determined experimentally, using the arrangement depicted in Figure 6, in which the corneal topography device 30 was immovably mounted on a plinth 76 and a model cornea (not itself visible in Figure 6) was carried on a stem 78 movable along the Z axis (toward and away from the corneal topography device 30) by means of a screw adjuster 80, allowing the model cornea to be positioned at known - and adjustable - positions along the Z axis of the corneal topography device 30. With this arrangement, a number of test images were obtained with the model cornea at different positions along the Z axis. The position of the cornea in each of the test images was established. This can be done simply by counting image pixels along the direction J in the test image. Of course the image of the cornea is an arc, but to get a consistent measurement the largest distance along the J direction of Figure 5 was determined in each case.

[0036] Figure 7 shows the relationship determined in this way between corneal distance in the real world and pixel count in the image, which is substantially a straight line, easily parameterizable using the well-known function y=mx+c.

[0037] In the working corneal topography device 30, the determination of corneal position along the Z axis thus involves (a) determination of the corneal position in the image 58 (and more specifically, determination of the corneal position in the reflection 68) and (b) application of the above function to the determined corneal position in the image to obtain the corneal distance D.

[0038] The determination of the corneal position in the image may in principle be carried out manually, or by a computer-based algorithmic method, but in the present embodiment this part of the image analysis is made using a machine learning system, and more specifically using a convolutional neural network ("CNN"). Training data was created by imaging a training population of human eyes to obtain a set of training images. Each training image was studied by a human operative who made a pixel count along the direction J from a fixed image point to the image of the cornea 40. The set of training images, each with an associated pixel count, constituted the training data supplied to the CNN. Trained in this manner, the CNN is able to determine the corneal position - in terms of pixels - for the images acquired in operation of the corneal topography device 30.

[0039] In Figure 3, the eye is aligned so that the (notional) Z axis of the corneal topography device 30 passes through the centre of the eye. If the eye is displaced from that ideal position either along the X axis (which is orthogonal to the Z axis and horizontal) or the Y axis (which is orthogonal to the Z axis and vertical) then the result is a change in the position of the cornea in the reflected image 72. This is a potential source of error, but can be corrected for by analysing the camera image to determine the X-axis / Y-axis displacement of the eye in it, providing a basis for correction. An embodiment of the invention that operates in this manner will now be described.

[0040] To develop a system to make the correction for X and Y axis displacement, use was again made of an experimental apparatus having a model cornea to train a machine learning system, and more specifically a convolutional neural network ("CNN"), to determine the X and Y displacement of the eye by analysis of the camera images. In this experimental apparatus the model cornea was provided with a movable mounting enabling it to be positioned at known positions which can be characterised by cartesian coordinates X, Y and Z, where as before the Z axis lies along the depth direction, the X axis is horizontal and orthogonal to the Z axis, and the Y axis is vertical and orthogonal to the Z axis. These coordinates represent the position of the peak of the cornea. Using this experimental arrangement a training data set was obtained by positioning the model cornea at a large number of different positions and obtaining images from the camera 32, with the known coordinates of the corneal peak being associated with the corresponding image in the training data set. The training data set was used to train the CNN to analyse, in use, the image 58 of a real eye obtained by the camera 32 to determine its displacements x_shift and y_sh ift along the X and Y axes from the ideal position (i.e. from the Z axis). This analysis does not rely only on the reflection 68. Rather, the pattern of dots 72 used in corneal topography is utilised. The analysis may involve consideration of the position of the centre of the pattern of dots 72 in the image 58.

[0041] In the present embodiment, a separate machine-learning model is used to determine the position of the corneal peak in the reflection 68. This machine-learning model is trained as described above with reference to the first embodiment, but in this instance the position is represented by not one but two variables u and v, representing the corneal peak's position along respective orthogonal axes.

[0042] So, the two machine learning models of the present embodiment provide (a) coordinates x_shift and y_shift representing the physical displacement of the corneal peak from the ideal position (i.e. from the Z axis) and (b) coordinates u and v representing the position of the corneal peak in the reflection 68 portion of the image 58. From these coordinates, a determination is made of the displacement of the cornea along the Z axis. While various mathematical approaches may be taken to this determination, in the present embodiment the following formula is applied:

[0043] Z=(Ci * x_shift) + (C2 * y_shift) + (C3 * u) + (C4 * v) + C5

[0044] ..where constants Ci .. C5 are determined by optimisation against experimental data. The training data set may be employed for this purpose. In this way, a determination is made of the displacement of the corneal peak from the camera 32 which is compensated for X and Y misalignment.

[0045] A further refinement of the technique according to a third embodiment of the invention makes use of reflections from both the first and the second mirrors 44, 56. Any misalignment between the camera 32 and the front portion of the device 30, carrying the mirrors, can be a source of systematic error in the values of X, Y and Z. Sufficiently accurate alignment of the camera 32 to remove such error is challenging. Calibrating individual devices to allow for it would add undesirably to costs, and disturbance of camera orientation (e.g. if the device 30 suffers an impact) would reintroduce error. These problems are obviated by analysing reflections from both mirrors 44, 56. If u_right and v_right represent the position of the corneal peak in the right-hand reflection from the first mirror 44 while ujeft and vjeft represent the position of the corneal peak in the left-hand reflection in the second mirror 56 then the displacement of the corneal peak along the Z axis can be determined using:-

[0046] Z= (Di * x shift) + (D2 * y_shift) + (D3 * (u right + u left)) + (D4 * (v right + vjeft)) + D5

[0047] ...where as before constants Di ... D5 are determined by optimisation against experimental data. Misalignment of the camera (which may for example increase u_right whilst correspondingly decreasing ujeft) is thus compensated for.

[0048] In this approach, machine-learning is used to separately determine (a) the coordinates u_right and v_right of the corneal peak in the right-hand reflection and (b) the coordinates ujeft and vjeft of the corneal peak in the left-hand reflection.

[0049] It will be appreciated that the invention provides for determination of corneal position along the Z axis simply by the use of two mirrors, which can be small, compact and economical, along with the required software. The present invention lends itself especially well to implementation of the functionality of a corneal topography device in a unit which, in contrast to the bulky desk-bound devices which are commercially available at the time of writing, takes the form of a compact device able to be held in one hand and in that way presented directly to the eye.

[0050] The arrangement of Figure 3 also makes possible straightforward determination of corneal thickness profile, simply by analysis of the mirror's reflection 68, since the anterior and posterior surfaces of the cornea 40 are both visible in it.

[0051] Figure 8 shows in longitudinal section a housing 38a of such a device which is substantially circular in cross section and has a circular rim 90 which is disposed around the eye, close to the face, in use. The camera 32 is omitted from Figure 8 but is to be mounted in a cavity 92. Frusto-conical housing wall 94 leading from the cavity 92 to the rim 90 has a plurality of openings 96 through which light emitters (which may be LEDs, and which are omitted from the drawing) project respective rays onto the cornea, creating the dot pattern 72 on which corneal topography is based. The mirrors 44, 56 are coupled through L-shaped brackets 98 to the wall 94. Provision is not made for controlled adjustment of the device along the Z axis. Instead, using the measured distance D from the device to the cornea, compensation for this dimension is applied in software when interpreting the reflected light pattern 72 to establish the contours of the cornea.

[0052] The device 30 is configured to analyse the camera image 58 to determine the corneal topography, taking account of the determined distance of the cornea from the camera 32.

[0053] This compact, hand held and relatively economical device has the potential to make corneal topography more widely available, rather than being the preserve of specialist institutions. The hand-held device is preferably self contained, in that the processing capacity needed to implement its functions is onboard. Nonetheless, the hand-held device may network with additional processing capacity through a local or wide area network to provide some or all of the necessary processing capacity. For example, the trained machine-learning models may be implemented in one or more remote servers to which data from the handheld device is transmitted. In such embodiments, the device and its processing device must be understood to include such remote servers. For example, the handheld device may comprise a smartphone, with analysis software residing on a cloud system accessed through an application running on the smartphone.

Claims

CLAIMS1. A corneal topography device comprising: a light emitting arrangement for emitting a predetermined light pattern onto a cornea of an eye, a camera which has a field of view and which is configured to obtain a camera image of the eye viewed along a first direction to obtain an image for use in corneal topography, a mirror which is mounted in a predetermined position and orientation relative to the camera and in the camera's field of view to provide, in the camera image, a reflection showing the cornea viewed along a second direction which is not parallel to the first direction, and a processing device configured to determine, by analysis of the reflection, a value characterising a position of the cornea's image in the reflection, and using the said value to determine a distance from the camera to the cornea.

2. A device as claimed in claim 1 in which the processing device comprises a machine-learning system trained to determine the position of the cornea's image in the reflection.

3. A device as claimed in claim 1 or claim 2 in which the processing device is configured to apply a predetermined mathematical function to the value characterising the cornea's image in the reflection to determine the distance of the cornea from the camera.

4. A device as claimed in which the field of view has a central axis Z and the mirror is offset from the axis Z.

5. A device as claimed in claim 4 in which the mirror is positioned and orientated such that the second direction is within 30 degree of being perpendicular to the axis Z.

6. A device as claimed in claim 4 in which the mirror is positioned and oriented such that the second direction is within 30 degree of a tangent to the cornea's centre.

7. A device as claimed in any preceding claim which is configured to be held in the hand and in that way presented to the eye of a subject.

8. A device as claimed in any preceding claim in which the field of view has a central axis Z, the processing device being configured to:- analyse the camera image to determine a displacement of the cornea from the central axis Z, and make allowance for the determined displacement in determining the distance from the camera to the cornea.

99. A device as claimed in claim 8 in which the processing device is configured to implement a machine learning model trained to analyse the camera image to determine the displacement of the cornea from the central axis Z.

10. A device as claimed in claim 8 or claim 9 in which the processing device is configured to determine two coordinates representing the position of the cornea's image in the reflection.

11. A device as claimed in any of claims 8 to 10 in which the processing device is configured to determine two coordinates representing the displacement of the cornea from the central axis Z.

12. A device as claimed in claim 11 when dependent on claim 10 in which the processing device is configured to apply a predetermined mathematical function to the two coordinates representing the position of the cornea's image in the reflection and the two coordinates representing the displacement of the cornea from the central axis Z to determine the distance from the camera to the cornea.

13. A device as claimed in claim 4 or in any subsequent claim when dependent on claim 4, comprising a further mirror which is mounted in a predetermined position and orientation relative to the camera on an opposite side of the Z axis from the mirror and in the camera's field of view to provide, in the camera image, a further reflection showing the cornea viewed along a third direction which is not parallel to the first direction or to the second direction.

14. A device as claimed in claim 13 in which the processing device is configured to determine, by analysis of the further reflection, a further value characterising a position of the cornea's image in the further reflection, and to use the value and the further value to determine the distance from the camera to the cornea.

15. A method of determining corneal topography comprising:- emitting a predetermined light pattern onto a cornea of an eye, providing a camera having a field of view, and using the camera to obtain a camera image of the eye along a first direction for use in corneal topography, providing a mirror at a predetermined position and orientation relative to the camera and in the camera's field of view to provide, in the camera image, a reflection showing the cornea viewed along a second direction which is not parallel to the first direction, and analysing the reflection to determine the distance of the cornea from the camera.

16. A method as claimed in claim 15 in which analysing the reflection comprises determining a position of the cornea's image in the reflection, and from that position determining the distance from the camera to the cornea.

17. A method as claimed in claim 16 in which determining the position of the cornea's image in the reflection is carried out by a trained machine-learning system.

18. A method as claimed in claim 16 or claim 17 which further comprises determining a parameterised function relating the position of the cornea's image in the reflection to the distance of the camera from the cornea, and applying the said function to the position of the cornea's image in the reflection to determine the distance from the camera to the cornea.

19. A method as claimed in any preceding claim, further comprising determining thickness of the cornea by analysis of the reflection.

Images