WO2014096835A1

WO2014096835A1 - Improvements in and relating to ophthalmoscopes

Info

Publication number: WO2014096835A1
Application number: PCT/GB2013/053366
Authority: WO
Inventors: Derek Swan; Jano VAN HEMERT; Michael VERHOEK; Gonzalo Muyo
Original assignee: Optos Plc
Priority date: 2012-12-21
Filing date: 2013-12-19
Publication date: 2014-06-26
Also published as: GB201223184D0

Abstract

A method of determining a correction for image distortion of an ophthalmology system is provided, comprising creating a target, computing an image of the target, using the ophthalmology system to scan the target to create a distorted image of the target, and comparing the computed image of the target with the distorted image of the target to determine the correction for image distortion of the ophthalmology system. The invention further provides a method of correcting distortion in a retinal image acquired using an ophthalmology system, a corrected retinal image and an image distortion correction system.

Description

Improvements in and relating to Ophthalmoscopes

The invention relates to improvements in and relating to ophthalmoscopes particularly determining a correction for distortion in eye representations of ophthalmoscopes.

Ophthalmoscopes generally comprise a system for directing rays from a source onto a portion of a subject's eye and for collecting rays reflected from the portion of the subject's eye in a ray detector. A number of optical elements and scan elements are commonly used to direct and collect the rays, and the collected rays are used to form an image of the portion of the subject's eye, often a portion of the retina of the subject's eye. Due to the three dimensional nature of the eye, the inherent characteristics of the optical elements and scan elements and customary operation modes used for such ophthalmoscopes, distortion can be introduced into the eye images produced by the ophthalmoscopes. The distortion in the images makes it difficult to determine accurately the size, structure and position of features of the portion of the subject's eye and to compare ophthalmoscope eye images with eye images taken using other devices. Retinal feature measurement is therefore highly inaccurate and meaningful comparison to other imaging modalities in order to provide a consistent measurement or coordinate system not possible. This can lead to difficulties in diagnosing and treating conditions of the eye.

According to a first aspect of the invention there is provided a method of determining a correction for image distortion of an ophthalmology system comprising

creating a target,

computing an image of the target,

using the ophthalmology system to scan the target to create a distorted image of the target, and

comparing the computed image of the target with the distorted image of the target to determine the correction for image distortion of the ophthalmology system. Computing the image of the target may comprise calculating coordinates of a plurality of points of the image of the target.

Scanning the target may comprise passing a plurality of rays through the ophthalmology system to determine coordinates of a plurality of points of the distorted image of the target.

Comparing the computed image of the target with the distorted image of the target may comprise comparing points of the computed image of the target with corresponding points of the distorted image of the target.

Determining the correction for image distortion of the ophthalmology system may comprise deriving an analytical transformation that maps coordinates of points of the distorted image of the target onto coordinates of corresponding points of the computed image of the target.

Determining the correction for image distortion of the ophthalmology system may comprise constructing a correction look up table comprising, for each of a plurality of possible points of the distorted image of the target, coordinates of the possible point of the distorted image against coordinates of the corresponding point of the computed image of the target.

The computed image of the target may be a 2-dimensional image and the distorted image of the target may be a 2-dimensional image. The computed 2-dimensional image may comprise a projection of the target. The projection of the target may comprise a stereographic projection. Certain projections are more suitable for a particular task than others. For example, several images taken from different overlapping parts of the retina can be montaged (also known as stitched) if a stereographic projection is used as it is conformal. Conformal projections preserve angle, which is a necessary property for montaging images. The computed image of the target may be a 3-dimensional image and the distorted image of the target may be a 3-dimensional image. For instance, optical coherence tomography is a technology that can be used to create 3-dimensional images of the target and to produce 3-dimensional images of the back of the eye which can then be corrected.

The target may comprise a plurality of contrasting shapes. The target may have a predefined pattern on it. According to a second aspect of the invention there is provided a method of correcting distortion in a retinal image acquired using an ophthalmology system comprising

obtaining a correction for image distortion of the ophthalmology system using the method of the first aspect of the invention, and

using the correction to map points of the acquired retinal image to corresponding points of a corrected retinal image.

The method may further comprise measuring a gaze angle using a fovial location of the acquired retinal image, determining an image distortion correction for the gaze angle and using the selected correction to map points of the acquired retinal image to corresponding points of a corrected retinal image.

The corrected retinal image may be used to calculate measurements of features of the image. According to a third aspect of the invention there is provided a corrected retinal image obtained using the method of the second aspect of the invention.

According to a fourth aspect of the invention there is provided an image distortio correction system comprising

an ophthalmology system, a target,

a first processor which computes an image of the target,

storage means which stores a distorted image of the target created using the ophthalmology system to scan the target, and

a second processor which compares the computed image of the target with the distorted image of the target to determine a correction for image distortion of the ophthalmology system.

Thus this invention relates to a method to generate an ultrawide field retinal image that corrects for system distortion and non-linearities such that required image geometry is maintained in order to allow retinal measurements and multi-modal retinal coordinates to be derived.

The ability to transform images of the back of the eye into a target projection has several benefits.

Devices that create images from the back of the eye can be generalised to any ophthalmology system that measures one or more properties on the back of the eye and places this information in a coordinate system that represents the back of the eye. Examples include field analysers (measure functional response), optical coherence tomography (measure changes in coherence of light), fundus and scanning laser ophthalmoscopes (measure the intensity of reflected light). The same eye imaged on different devices, i.e. with different distortions resulting in different device projections, will produce images that cannot directly be compared nor be overlaid. Once the differently-distorted images have been corrected to the same computed image of the target, i.e. transformed into the same target projection, it is possible to compare features of the images directly. Moreover, if the computed image of the target is a projection which is conformal, i.e. it is an angle-preserving projection, it is possible to overlay images using registration methods. Choosing one projection for the computed image of the target and transforming images from different devices into that one projection allows direct comparison and registration of the images. Choosing one projection for the computed image of the target and many distortion corrections for different device projections to that one target projection, results in a uniform coordinate system in which images from different devices and models can be integrated. This allows image data from the same eye obtained from different devices with potentially different image modalities (e.g., angiography, fundus photography, functional measurements and 3-dimensional data) to be correlated.

If the computed image of the target is calibrated against measuring geometry, it is possible to accurately measure features on images of the back of the eye once the images have been corrected for distortion, i.e. transformed from a device projection into the target projection. The method can be used to calibrate geometry measurements (area, distance and angle) by ensuring these measurements are accurately known in the computed image of the target. By imaging a physical phantom or by using a reference measurement, it is possible to make accurate measurements in physical units (mm², mm and degrees). Also, by reversing the distortion correction, it is possible to measure directly on the retinal image rather than on the computed image of the target. By correcting images taken at different gaze angles and having different distortions to the same computed image of the target, i.e. creating transformations for different gaze angles of the eye all into the same target projection, it is possible to integrate images from these gaze angles into one large image using registration and montaging techniques if the computed image of the target is a projection which is conformal, i.e. angle-preserving.

By choosing several or many different device distortions it is possible to explore various parameters that influence technical aspects and may affect imaging quality such as, variance in patients' eye alignment, biological variance, different pathologies and changes in system design, by analysing images produced by the devices after distortion correction to the same computed image of the target.

An embodiment of the invention will now be described by way of example only with reference to the accompanying drawings, in which:

Figure 1 is a schematic representation of an image distortion correction system according to the fourth aspect of the invention;

Figure 2 is a schematic representation of part of an ophthalmology system of Figure 1; Figure 3 is a flowchart representing the method of the first aspect of the invention; Figure 4 is a schematic representation of a computer image of a target used in the method of the first aspect of the invention;

Figure 5 is a schematic representation of a distorted image of a target used in the method of the first aspect of the invention, and

Figure 6 is a flowchart representing the method of the second aspect of the invention.

Referring to Figure 1, an image distortion correction system 1 is shown, comprising an ophthalmology system 2, a target 3, a first processor 4, storage means 5 and a second processor 6. The first processor 4, storage means 5 and second processor 6 are shown as comprised in the ophthalmology system 2, but it will be appreciated that these could be provided separate to the ophthalmology system.

Referring to Figure 2, part of the ophthalmology system 2 of Figure 1 is shown. This comprises an ophthalmoscope 10 which, when used to acquire images of a patient's eye, such as images of a portion of the retina of the patient's eye, will introduce distortion into the acquired images. The ophthalmoscope 10 comprises a source of collimated light 12, a first scanning element 14, a second scanning element 16, scan compensation means 18 and scan transfer means 20. The source of collimated light 12 directs a light beam 13 onto the first scanning element 14. This produces a scan of the beam (depicted by rays A, B and C) in a first, vertical, direction. The beam is incident on the scan compensation means 18 and is reflected from there onto the second scanning element 16. This produces a scan of the beam in a second, horizontal, direction. The beam is incident on the scan transfer means 20, which has two foci, the second scanning element 16 is provided at a first focus and an eye 22 of a patient is provided at the second focus. The beam from the second scanning element 16 which is incident on the scan transfer means 20 will be directed to the eye 22 and will impinge on a portion of the retina of the eye. The components of the ophthalmoscope 10 combine to provide a two dimensional scan of an incident beam from an apparent point source onto the retina portion. As the incident beam is scanned over the retina portion, it will be reflected therefrom to produce a reflected beam which is transmitted back through the components of the ophthalmoscope 10 and received by the one or more detectors (not shown).

To acquire a retinal image, the incident beam is scanned over the retina portion of the eye 22 in a raster scan pattern, produced by the first and second scanning elements 14, 16 operating perpendicularly to each other. The first and second scanning elements are operated continuously to generate a continuous incident beam on, and a continuous reflected beam from, the retina portion from a scan start time to a scan stop time. The detectors are operated to record, or sample, the reflected beam over fixed time intervals of, for example, 30ns throughout the scan. During each time interval, the incident beam is scanned over a part of the retina portion and the reflected beam from that part of the retina portion is recorded and assigned to a pixel associated with that part of the retina portion. The result is a pixelated image of the retina portion.

In the ophthalmoscope 10 due to the inherent characteristics of the optical elements and scan elements 14, 16, 18 20 (and the 3-dimensional nature of the eye), the scan angle at the pupil plane is not uniform. In the customary operation mode of the ophthalmoscope 10, the angle covered in each, equal, time interval of the scan, and the part of the retina portion scanned in each time interval, is therefore non-linear with retinal arc length, resulting in distortion in acquired images of the retina portion, which are comprised of non-uniform pixels. The pixels are not consistent in either angular or spatial extent introducing non-linearity or warping in the acquired images of the retina portion when compared to the real retina portion. The geometry of the acquired images is partly determined by the time interval of sampling of the reflected beam and the angular scan pattern of the ophthalmoscope.

Each image produced by the ophthalmoscope 10 is a 2-dimensional image of a 3- dimensional retina portion. Every 2-dimensional image that represents the back of the eye is a projection of a 3-dimensional object - the retina - into a 2-dimensional space - the image. Therefore, every image of the back of the eye, whether acquired with a fundus camera or scanning laser ophthalmoscope, etc., comprises a particular 3- dimensional to 2-dimensional projection. In ophthalmoscopy, a portion of the spherical retina - the back of the eye can be approximated by a sphere - is projected to a plane, i.e., a 2-dimensional image. An infinite number of projections of a sphere to a plane exist and each projection has its own unique features. Every ophthalmoscope and its position relative to the back of the eye is one unique projection. Each 3-dimensional to 2-dimensional projection results in compromising several specific properties- properties that exist in the spherical representation, i.e. the back of the eye, will be lost in the planar representation, i.e. the image. Well-known and important properties that may be lost are area-preservation, distance-preservation and angle-preservation. This produces distortion in the resultant 2-dimensional images of the retina portion. Note, every system that captures images from the back of the eye has some kind of distortion as it creates a 2-dimensional image from a 3-dimensional object. The distortion it introduces depends on the method it employs to generate the image. Therefore in operation of the ophthalmoscope 10, the non-linearity of the scan and the projection of the 3-dimensional retina portion into a 2-dimensional image have the effect of introducing distortion into the 2-dimensional image of the retina portion. Correction of this distortion has a number of advantages, as referred to above. A correction for image distortion in retinal images of the ophthalmology system 3 is determined as follows, Figure 3. A target is created (step 30), an image of the target is computed (step 32), the ophthalmology system 3 is used to scan the target to create a distorted image of the target (step 34), and the computed image of the target is compared with the distorted image of the target to determine the correction for image distortion of the ophthalmology system 3 (step 36).

The first processor 4 of the image distortion correction system 1 (Figure 1) is used to compute the image of the target 3 of the image distortion correction system 1 (Figure 1). The target 3 comprises a plurality of contrasting shapes in the form of series of concentric circles with straight lines across the vertical, horizontal and 45 degree angles. The computed image of the target 3 (Figure 4) comprises a representation of the concentric circles with straight lines. The image of the target is computed by calculating coordinates of a plurality of points of the image of the target. The ophthalmology system 2 is used to scan the target 3 to produce a distorted image of the target as shown in Figure 5. Scanning the target comprises passing a plurality of rays through the ophthalmology system to determine coordinates of a plurality of points of the distorted image of the target. It is clear from the distorted image that neither angle, distance or area is preserved, with each image pixel representing a different retinal geometry. The storage means 5 of the image distortion correction system 1 stores the distorted image of the target.

The second processor 6 of the image distortion correction system 1 then compares the computed image of the target (Figure 4) with the distorted image of the target (Figure 5) to determine a correction for image distortion of the ophthalmology system 2. Comparing the computed image of the target with the distorted image of the target comprises comparing points of the computed image of the target with corresponding points of the distorted image of the target. Determining the correction for image distortion of the ophthalmology system 2 comprises deriving an analytical transformation that maps coordinates of points of the distorted image of the target onto coordinates of corresponding points of the computed image of the target. The analytical transformation takes as its input the distorted image of the target and produces as output an approximation of the computed image of the target. The parameters of the transformation are optimised by ensuring this output image matches as closely as possible the computed image of the target. The optimised parameters are then saved. Determining the correction for image distortion of the ophthalmology system further comprises constructing a correction look up table comprising, for each of a plurality of possible points of the distorted image of the target, coordinates of the possible point of the distorted image against coordinates of the corresponding point of the computed image of the target.

In this embodiment, the computed image of the target is a 2-dimensional image and the distorted image of the target is a 2-dimensional image. The computed 2-dimensional image of the target may comprise a chosen projection of the target, for example a stereographic projection. Stereographic projections are conformal, i.e. they preserve angle in the image. It will be appreciated that the computed image of the target may be a 3-dimensional image and the distorted image of the target may be a 3-dimensional image. Distortion in a retinal image acquired using the ophthalmology system 2 is corrected as follows, Figure 6. A correction for image distortion of the ophthalmology system 3 is obtained (step 60) and the correction is used to map points of the acquired retinal image to corresponding points of a corrected retinal image (step 62). The distortion correction is obtained using the method described above. Retinal images acquired with any ophthalmology device with the same characteristics, i.e. model, can be directly transformed into the chosen target projection by running the transformation with the saved parameters. For example, retinal images can be transformed from a scanning laser ophthalmoscope projection to a stereographic target projection. The correction method may further comprise measuring a gaze angle using a fovial location of the acquired retinal image, determining an image distortion correction for the gaze angle and using the selected correction to map points of the acquired retinal image to corresponding points of a corrected retinal image. The corrected retinal image may then be used to calculate measurements of features of the image, i.e. measurements of features of the retina.

Claims

1. A method of determining a correction for image distortion of an ophthalmology system comprising

creating a target,

computing an image of the target,

comparing the computed image of the target with the distorted image of the target to determine the correction for image distortion of the ophthalmology system.

2. A method according to claim 1 in which computing the image of the target comprises calculating coordinates of a plurality of points of the image of the target.

3. A method according to claim 1 or claim 2 in which scanning the target comprises passing a plurality of rays through the ophthalmology system to determine coordinates of a plurality of points of the distorted image of the target.

4. A method according to claim 2 and claim 3 in which comparing the computed image of the target with the distorted image of the target comprises comparing points of the computed image of the target with corresponding points of the distorted image of the target.

5. A method according to any preceding claim in which determining the correction for image distortion of the ophthalmology system comprises deriving an analytical transformation that maps coordinates of points of the distorted image of the target onto coordinates of corresponding points of the computed image of the target.

6. A method according to any preceding claim in which determining the correction for image distortion of the ophthalmology system comprises constructing a correction look up table comprising, for each of a plurality of possible points of the distorted image of the target, coordinates of the possible point of the distorted image against coordinates of the corresponding point of the computed image of the target.

7. A method according to any preceding claim in which the computed image of the target is a 2-dimensional image and the distorted image of the target is a 2-dimensional image.

8. A method according to claim 7 in which the computed 2-dimensional image comprises a projection of the target.

9. A method according to any of claims 1 to 6 in which the computed image of the target is a 3-dimensional image and the distorted image of the target is a 3-dimensional image.

10. A method according to any preceding claim in which the target comprises a plurality of contrasting shapes.

11. A method of correcting distortion in a retinal image acquired using an ophthalmology system comprising

obtaining a correction for image distortion of the ophthalmology system using the method of any of claim 1 to 10, and

12. A method according to claim 11 further comprising measuring a gaze angle using a fovial location of the acquired retinal image, determining an image distortion correction for the gaze angle and using the selected correction to map points of the acquired retinal image to corresponding points of a corrected retinal image.

13. A method according to claim 11 or claim 12, in which the corrected retinal image is used to calculate measurements of features of the image.

14. A corrected retinal image obtained using the method of any of claims 11 to 13.

15. An image distortion correction system comprising

an ophthalmology system,

a target,

a first processor which computes an image of the target,