DE102007055922A1 - Method for determining properties and / or the position of characteristic ocular components - Google Patents

Abstract

The invention is based on a method for determining the position of the center of characteristic ocular components, in particular of the limbus or the pupil, during an eye examination or treatment. According to the invention, a digital image of at least a section of an eye is taken with the camera. This image is then correlated with an annular comparison object of defined size. From the values of the correlation function, the center of the best matching position is determined and chosen as the limbus / pupil center.

Description

The The invention relates to a method for detection or tracking the position of the center of characteristic ocular components the preamble of claim 1.

such Methods are, for example, in the field of eye surgery known.

For example in corneal surgery for the removal of defective vision of the human eye (LASIK), in which a part of the cornea by means of a Laser is removed, it is of interest to the surgeon at which point the visual axis of the patient pierces the cornea. On the basis of the exact determination of this point during the Surgery can make the laser ablation more precise from this point take place when choosing a theoretically accepted or estimated Center of the cornea.

One Another example is cataract surgery which is a natural lens of the human eye which has clouded through an artificial lens is replaced. Such a procedure takes the surgeon under one Operating microscope in front. After a circular opening of the anterior capsule sheet usually becomes the lens smashed and sucked off. Subsequently, in the empty capsular bag used an artificial lens.

From the DE 10 2004 055683 A1 For example, a surgical microscope for eye surgery is known which superimposes a pattern on the eye to be operated on. The pattern may aid in setting the cutting position, but it may also serve as a guide in the placement of toric intraocular lenses, or may assist in the insertion of a suture in a corneal transplant. To position the pattern in the right place, it is necessary to determine the position of the pupil or iris on the eye to be treated. Ideally, the position is also constantly redefined or tracked during the operation, as it may come during the procedure to movements of the entire eye or the pupil.

It is also of fundamental importance for other applications in the field of ophthalmic surgery to determine the position or diameter of the iris of the eye to be treated. For example, the diameter of the iris is necessary to calculate the strength of an intraocular lens to be implanted after cataract surgery. This and other possible applications, as well as a method for determining positions and magnitudes within an ocular segment, are described in the DE 101 08 797 A1 discussed in more detail.

Some methods are known in which the position of the pupil is determined on the basis of the current image of the eye section to be operated, which is obtained with the camera on the surgical microscope. Both in the DE 10 2004 055683 A1 as well as in the DE 101 08 797 A1 Methods are proposed in which a binary image is first generated by thresholding to determine the dark areas in the image. Thereafter, the largest contiguous region in the dark regions is identified, which is identified as a pupil. In order to determine the edge of the pupil or iris in more detail, an edge detection is usually performed in this method. These methods have some disadvantages. On the one hand, the pupil is not always the largest contiguous dark area; on the contrary, the pupil may be disturbed by a reflex and have a completely different appearance. On the other hand, the edge detection during the introduction of microsurgical instruments during surgery can be greatly disturbed. Basically, it is difficult to define a meaningful threshold value for all methods which work with thresholding, which on the one hand does not leave too much information in the picture, but on the other hand does not remove important details from the picture.

Especially then, if the localization over the whole investigation or treatment as an eye-tracking, so to speak, It is also essential that the detection process works extremely fast, so the result of the localization always immediately back into the recording can be displayed it has been derived, and if possible without any time offset.

Of the Invention is based on the object, a method for determining or tracking the position of characteristic ocular components to develop, which is robust against disturbing influences is independent of the individual design of the Eye reliable and fast working.

Solved the object is achieved according to the invention by a Method for identifying or tracking the center more characteristic Eye components with the features of claim 1.

According to the invention, an annular light-dark transition of corresponding size in the recording is determined on the basis of the correlation of the digital image acquisition to be analyzed with an annular comparison object with a fixed radius. The invention is based on the finding that, in the case of the images of the eye recorded in the course of an eye treatment, ocular components such as the limbus or the pupil rim are ring-shaped Represent gangsobjekte and that they are particularly simple and reliable to find on the basis of the comparison with a corresponding comparison object or on the convolution with a corresponding annular filter in particular with difference. The search for an annular transition object is extremely robust to impairments that image-dominating features may corrupt the image during the operation. For example, when instruments partially obscure the eye, further strong edges emerge which can cause great difficulty for any edge detection method, as these are detected instead of the limbus or the pupil edges. However, all these disturbances do not change the fact that the limbus or the pupil edge continues to exist, albeit as an exposed or slightly deformed, dominant, at least in the middle, annular element in the image and is still reliably found in the search by means of a corresponding annular comparison object , It should also be stressed at this point that an absolute thresholding, which is necessary in the conventional edge detection method, and thus the problem of choosing a suitable threshold in this method is not necessary and should also be avoided. Any absolute thresholding or binarization falsifies the image and can destroy the ring-shaped character of limbus or pupil edge, so that it may no longer be found in a subsequent search for an annular object. The fact that the search for the annular object, so the limbus or the pupil edge is made directly on the grayscale image, the annular character is certainly given, even if it affects disturbances.

In In a preferred embodiment, the radius of the comparison object in an initialization step determined by the image to be analyzed each with ring-shaped comparison objects of different Radii is correlated. In the correlation of the image section with comparison objects of different sizes in each case the best match for the comparison object determined and based on the comparison of the values of best match the absolutely best match among all and thus the comparison object with the best to the object to be examined adjusted size. This is done advantageously in that the respective maximum response of the correlation function, which results in the correlation with the image section, for the comparison object is plotted against its radius, wherein gives a function that always maximizes when the radius is good to the radius of a corresponding object in the fits the image to be examined. The maximum that occurs during largest ones belonging to a maximum value Radius is equal to the radius of the largest annular object in the image detail and thus the limbus radius, what was detected in the course of this procedure. This determined so Radius becomes a fixed radius for a comparison object chosen, which is suitable, the limbus and thus also the limbus center during eye examination or too Treatment, so in the evaluation of the following shots the same Object to track with. This initialization step requires a much larger computational effort and thus Time factor as the subsequent tracking or Tracking the already known object with a comparison object defined size. Since he only once to Beginning of the examination or treatment is done this is acceptable. During the examination or treatment can and must then however favored by the retention of the radius of the comparison object a detection of the limbus center so to speak in real time respectively. Since then the hit probability when using the predetermined exact matching radius for the comparison object is much higher than an arbitrary one chosen radius, it is definitely justified, this greater time for the investigation of the radius. However, it is important that she does not with each admission takes place again but that after the once made determination the appropriate radius of the comparison object and set again and again can be used.

In a further preferred embodiment, the annular object is constructed from two concentrically arranged annular components. The fact that the comparison object has at least two components, the possibility arises, in each case a component to the eye area outside the density transition, z. B. the Sclera and the second component of the lying within the density transition eye area, z. As the iris to adapt. By means of these two components, the density transition can thus be strengthened to a certain extent via a correlation with the comparison object. The optimal match with the comparison object results when the inner ring of the comparison object z. B. on the iris, the outer z. B. is located on the sclera, and thus the transition region is thus comprised in this case, the limbus edge of the two annular components. At the same time, the limbus center comes into coincidence with the center of the comparison object. In this design of the comparison object, not only the shape feature, that is, the ring-shaped or circular appearance of the limbus / pupil, but also the surface feature, the density transition, is used to search for the transitional object. Preferably, the two components of the reference object are two narrow annular components which are spaced so far that none of the Be constituents in the transition region of the density increase, but one in each case in the region of lower, the other in the region of high density comes to rest. This allows a clear identification of the object. If the limbus radius is not determined in advance, the distance between the two annular components is preferably to be chosen so that the outer ring is safe outside, the inner safely within the edge of the transition object, ie the limbus or the pupil edge. A suitable distance can be found empirically in this embodiment by examining a wider selection of typical images taken in such eye examinations or treatments. At a resolution of the capture of at least 100x100 pixels, which is necessary to determine the center of the limbus with an accuracy of at least 1 mm required for an operation, a distance between one and five pixels, preferably two pixels, has been determined appropriate ,

In a further advantageous embodiment of the invention takes place within the framework of the correlation of the comparison object with the Recording a difference of the annular components of the comparison object or of the areas correlated with these within of the recorded eye detail. So preferably in the Correlation of an annular component of the comparison object with positive, the other with negative sign. The comparison object is preferably designed, or the correlation function so chosen that when correlated with a gray area, So a surface without trained density transition a neutral result such as the value reaches zero becomes. Only at a trained density transition, on the comparison object comes to lie, there is an increased Value of the correlation function.

In In another preferred embodiment, the comparison object realized by a filter, with which the image is folded. Of the Annular filter is designed to be always then gives a maximum filter response when the annular filter on an annular density transition like that Limbus or pupil edge comes to rest. This maximum of the filter response The better the agreement, the clearer it is of the filter radius with the radius of the searched object. Therefore is preferred according to the previously explained method determines the radius of the filter in an initialization step, before the eye tracking accompanying the examination or treatment with fixed filter radius. As a preferred annular Filter is a filter chosen, which two concentric arranged components, so that a maximum filter response can then be achieved if the one component is complete in the area lower, the other completely in the area higher density and thereby the transition region to be identified, such as As the limbus, included by the two components becomes. The distance between the two annular components is chosen so big that no part in the transitional area, but both are clearly in one Dense region, but at the same time so small that one possible exact radius determination is possible.

Around this initialization step is not a change the object radius due to changes in the recording conditions, like the microscope magnification factor, perform again and again to be in a preferred embodiment any change to the device settings that based on the size of the captured eye-opening affects the size of the annular Included in the comparison object or the filter radius accordingly customized. This ensures that the radius of the Comparison object automatically adapted to the shooting conditions and indeed only has to be determined once in advance. For this it is necessary to provide an interface between the device parameter changing Device, for example the microscope and the device, at which the correlation with the comparison object takes place.

Even though the annular configuration of the comparison object important is, it would not change anything essential to the process if a polygon or something similar would be used. It is also not necessary to use a closed ring becomes. The comparison object may as well be made of ring-shaped Be composed segments. Essential for the procedure is only that total of the ring-shaped character of the comparison object preserved. Especially in the border area of the picture it is even more reliable to use only ring segments. In these Ring segments is preferably exposed to the area that is at the edge lies, in which the comparison object in the correlation and so that the limbus in the picture approaches. This corresponds the comparison object in the correlation better to find that Object, which as soon as it gets to the edge of the picture, partially is cut off.

In a further preferred embodiment, the red-extraction of the image is always used for the correlation with the comparison object. Surprisingly, it has been shown that this is the least affected by disturbances during eye treatment, since in this color separation, the red of the hemorrhages and veins with the white of the sclera forms a homogeneous surface. This makes it possible to achieve a more reliable result in this color channel than in other color separations. Thereby, Thus, always using the red channel as a grayscale image can provide high accuracy and speed the process over using a multicolor image or everlasting selection of the currently most contrastive color channel.

Around To further optimize the method, it is advantageous to use this grayscale image as far as the required accuracy permits. With correspondingly high computing capacity, this step also omitted. Furthermore, it is advantageous to homogenize the image, to small, unimportant contrast transitions that the Falsify result would eliminate.

In In a preferred embodiment, both the, from the position of pupil or limbus derived assistance for the surgeon, as well as an indication of the Reliability of the determined size in the view of the eye-opening considered by the surgeon appears. This will help the surgeon on the one hand for the operation, on the other hand, he can also directly assess the extent to which this assistance is reliable is and thereby decide for himself whether he uses it or himself but rather rely on his experience.

Further Details and advantages of the invention will become apparent from the dependent claims in connection with the description of an embodiment, which will be explained in detail with reference to the drawings.

It demonstrate:

1 schematically an apparatus for carrying out the method according to the invention,

2 an example of an advantageous ring filter superimposed on a picture of an eye-opening,

3 an example of a filter response,

4 an example of a threshold image of a filter response and

5 an eye opening and associated with this example of ring filter, which are exposed in the edge region.

The 1 shows a schematic representation of the basic structure, as is typical in an eye treatment in which the inventive method can be used particularly advantageous. The eye to be treated 1 the patient, which is illuminated by a light source, not shown, is on the one hand by means of an eyepiece 2 , on the other hand by means of a video camera 3 observed, wherein the observation beam path through a beam splitter 4 is split into two observation beam paths for the observing instruments. The on the video camera 3 recorded data is sent to a computer 5 pass, where the data is stored and analyzed. Based on the data, an auxiliary pattern is calculated by means of a pattern generation unit 6 formed and in the eyepiece 2 superimposed on the visible image, leaving the surgeon 7 the eye to be treated 1 together with the superimposed pattern attached to the pattern generation unit 6 was formed, can consider. The pattern generation unit 6 can be used, for example, as a projector with a ring-shaped LED display that is above the beam splitter 4 a pattern fades into the eye, be executed.

In a cataract operation is constantly in very short time series with the camera 3 the eye 1 digitally recorded or analog recorded data converted to digital and the digital data of the recording of the eye detail, as he, for example, in 2 (to explain the comparison object with superimposed ring filter) can be seen, to the arithmetic unit 5 transmitted. There, the eye center is determined according to the method according to the invention, so that the optimal cutting position for the cut, for removing the clouded and for insertion of the artificial lens, can be determined. Once the center of the limbus and at the same time, with known radius of the limbus 11 , which is derived cutting position, the image that is the eye of the surgeon 7 over the eyepiece 2 sees, at, the pattern generation unit 6 generated pattern indicating this cutting position superimposed. This will see the eye of the surgeon 7 during the treatment always the optimal cutting position for the preparation of the cut.

In order to ensure reliable operation of the method, the radius and the center of the limbus are determined on the basis of the start image in a very precise but relatively time-consuming initialization process. Such a method is described in detail in co-pending patent application "Characteristic Properties and / or Characteristic Ocular Position Detection" filed by the same Applicant, the entire contents of which are hereby incorporated by reference. In it, as in the method claimed here, a ring filter 8th as he in 2 is schematically represented, an eye patch is superimposed, folded with the inclusion of the eye opening. The ring filter 8th is made of two annular components, an outer ring 9 and an inner ring 10 composed of which adapted to the limbus radius of the ring filter 8th symmetrical around the examined one Lirabus 11 are laid. The ring filter 8th is normalized so that the outer ring 9 delivers positive contributions to the filter response, while the inner ring 10 negative contributions. In addition, the ring filter 8th normalized such that the filter response at convolution with a gray area is zero. That means the two rings 9 and 10 weighted according to their filter components in the image. To explain the principle of the method, it is assumed below that the radius of the limbus and thus the size of the ring filter 8th derived in an initialization step to be described later and therefore known. For all subsequently taken pictures of sections of the same eye, the limbus becomes 11 keep, it can always, at least as long as the recording mode on the microscope is not changed, the same ring filter 8th be used. This ring filter 8th will now be folded with the image section. That is, the filter response is determined at each point of the image. The result of a convolution with the ring filter 8th is in 3 exemplified. At the place where filter radius and limbus 11 lie one above the other, results in the maximum filter response, which is seen here as a bright area. The center of this bright area or the position of the absolute maximum of the filter response corresponds to the center of the limbus which is connected to the pattern generation unit 6 is handed over.

To exactly determine this center, the image of the filter response, which is in 3 by thresholding into a binary image as shown in FIG 4 shown is converted. The threshold used for this purpose is not determined in advance but determined from the image itself. A threshold of at least 90% of the maximum value of the filter response has been found to be useful. The value 90% was used as the threshold for the in 4 used representation shown. From this, in the 4 visible, small white area, the focus is determined. This center of gravity corresponds to the sought-after limbus center which is connected to the pattern generation unit 6 is handed over. This method can be used with the pre-determined, suitable ring filter 8th for each additional shot of the eye-view extremely fast and reliable, so to speak in real time, after this procedure the current Limbuszentrum be determined.

However, it is, as already mentioned, necessary in advance in a somewhat more complex process, the radius of the limbus 11 and thus the appropriate radius for the ring filter 8th to determine. For this purpose, as described in detail in the aforementioned parallel application, the image of an eye patch with ring filters 8th folded different size and thus different radius. For every fold with a ring filter 8th the maximum filter response is determined. This maximum filter response is against the radius of the ring filter used 8th plotted so that a curve results which always shows maximum values when a radius of a selected ring filter 8th coincides with the radius of a round object in the image section. It was recognized that the radius of the largest of these round objects in the image section is the radius of the limbus 11 matches, giving the largest radius of the correlated ring filter 8th associated maximum in the curve or its filter radius to the radius of the limbus 11 equivalent. This will be referred to below as the radius of the ring filter 8th to be used in all further images taken in the same examination or treatment of the same eye. For all the following pictures it is only necessary to track the limbus center, this with the ring filter 8th Folded radius to binarize the filter response and to determine the center of gravity. This has created a highly efficient and reliable method, the radius of the limbus 11 and to derive the limbus or pupil center extremely fast in all further images after a one-time, detailed and slightly longer analysis step and thus to determine this center in a real-time manner for each image over the entire eye diagnosis or treatment and to derive the position therefrom. to display a pattern or the like intended as an aid to the surgeon.

Only when the center of the eye moves into the edge areas of the image section, which occurs more often during an eye operation, part of the limbus is cut off by the edge of the image and the iris no longer corresponds to an annular object, which with a ring filter 8th easy and reliable to find. In this case, the process becomes significantly more unreliable. To counter this, as in 5 shown proposed, exposed in the edge regions of the image ring filter 8th to use where the lying on the edge of the ring segment is cut off.

The filters to be preferably used for the ranges are in the 5 represented and provided to simplify the assignment, each with the same Roman numerals, which were also placed on the edge regions of the eye portion shown above. Thus, for example, for the convolution with an image of the eye detail in the area of the upper left corner, which is marked with I, the filter also marked with I instead of the fully continuous ring filter which is marked V in the image and in the middle region of the image used is used. Thus, the filter used in each case also corresponds significantly better in the edge region of the eye cutout to that cut off in this region, Limbuskreis. This measure significantly increases the safety of the process.

Nonetheless, it is not possible to achieve a consistent 100% hit probability, especially in the margins. Ultimately, the surgeon still has to decide for himself whether he will follow the help he has received or rely on his feeling or experience and continue to work without the help. To make this decision, it is extremely helpful for the surgeon to be provided with information about the reliability of the displayed assistance in addition to the assistance. The reliability of the method can be particularly easily derived in the method described here. For example, the value of the maximum in the image of the filter response can be used as an absolute measure of the safety of the determination of the limbus center. The higher this value, the better the correspondence of radius of the limbus 11 and corresponding ring filter 8th and the more reliable the result. This security or reliability of the determined center can be represented for example in the form of a bar chart or in the form of an identification of the displayed assistance, for example as a solid, dashed or dotted object.

1

eye

2

eyepiece

3

video camera

4

beamsplitter

5

computer unit

6

Pattern generating unit

7

eye of the surgeon

8th

ring filter

9

Outer filter ring

10

inner filter ring

11

limbo

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• - DE 102004055683 A1 [0005, 0007]
• - DE 10108797 A1 [0006, 0007]

Claims (11)

Method for determining or tracking the position of the center of characteristic eye components, in particular of the limbus or the pupil, during an eye examination or treatment, characterized in that - a digital image of at least a section of an eye taken with the camera, - this image with a correlated annular comparison object of defined size, - and / or the center of the best matching position is selected as the limbus / pupil center. Method for identifying or tracking the position the center of characteristic ocular components according to claim 1, characterized in that the size of the annular comparison object via the correlation an image of the cutout with ring-shaped comparison objects different radii is determined in advance. Method for identifying or tracking the position the center of characteristic ocular components according to claim 1 or 2, characterized in that the annular Comparison object two concentric annular components having. Method for identifying or tracking the position the center of characteristic ocular components according to claim 3, characterized in that the size of the ring-shaped comparison object is chosen so that the outer annular component safely outside and inside safely within the limbo / Pupil lies. Method for identifying or tracking the position the center of characteristic ocular components according to claim 3 or 4, characterized in that in the correlation a Difference formation of the annular components takes place. Method for identifying or tracking the position the center of characteristic ocular components according to claim 1, characterized in that - the digital picture folded radius defined with an annular filter, - in the filter response determines the maximum and - of the Location of the maximum filter response as the position of the center of the limbus Pupil is determined. Method for identifying or tracking the position the center of characteristic ocular components according to claim 1, characterized in that the defined radius at a Changing the shooting mode of the camera automatically adjusted becomes. Method for identifying or tracking the position the center of characteristic ocular components according to claim 1, characterized in that as an annular comparison object a suspended ring is used. Method for identifying or tracking the position the center of characteristic ocular components according to claim 8, characterized in that the exposed ring in the edge region of the digital image is used. Method for identifying or tracking the position the center of characteristic ocular components according to claim 1 or 6, characterized in that one derived from the position Help for the surgeon is displayed. Method for identifying or tracking the position the center of characteristic ocular components according to claim 10, characterized in that an indication of the reliability of the method.