Method and apparatus for the morphometric analysis of cells of a corneal endothelium

Abstract

The present invention relates to a method and apparatus for the morphometric analysis of endothelial cells, which is based on the use of an image taken from a camera connected to a biomicroscope which is digitally reprocessed and subsequently analyzed.

Claims

1. A method for morphometric analysis of corneal endothelium cells comprising the following steps: A. acquiring at least one real digital image of the corneal endothelium cells with a bio-microscope equipped with a digital camera, said real digital image being composed of a plurality of single pixels; B. selecting at least one area of interest of said real digital image; C. detecting a luminance value for each pixel of said area of interest; D. generating a first matrix having elements that contain a luminance value of the single pixels; E. modelling said area by reconstructing one or more model cells of the endothelium, effected at least by assignment, to each model cell, of pixels substantially having the same luminance value; F. calculating, for each model cell, a barycenter of the pixels of which said model cell is composed and a radius thereof; G. generating a second matrix 3×N wherein N is a number of identified cells; H. scanning said second matrix 3×N and identifying, for each model cell, a nearby model cell which satisfies a relation (d*K)≦(r1+r2), r1 and r2 being radiuses of the two model cells, d a distance between barycenters of the two model cells, and K a form factor; I. for each pair of model cells for which the relation of step (H) has been verified, considering the two model cells of the pair of model cells in contact; and J. calculating one or more of a number of model cells per mm.sup.2, an area of each model cell, an average area of the model cells, or a standard deviation on the area of each model cell with respect to one or both of the average area or a number of cells touched by a specific cell.

2. The method according to claim 1, wherein step (A) comprises acquiring a plurality of images and selecting a preferred image among the plurality of images acquired by calculation of a relative Modulation Transfer Function (MTF) obtained by measuring contrast of two consecutive images acquired in subsequent times.

3. The method according to claim 2, wherein step (C) comprises a step of applying image analysis filters to the pixels of said area of interest, adapted to eliminate false information and homogenizing a luminance value over an entire cell.

4. The method according to claim 3, wherein said image analysis filters are at least Richardson-Lucy, Contrast Enhancement filter, Morphology filter (erosion, dilation), and Segmentation Watershed filter.

5. The method according to claim 3, wherein the applied filters applied are: i) Richardson-Lucy, ii) Contrast Enhancement filter, iii) Morphology filter (erosion, dilation), and iv) Segmentation Watershed filter, said applied filters being applied consecutively a i), ii), iii), iv) order.

6. The method according to claim 1, wherein said form factor K in step (H) ranges from 0.7 to 1.

7. An apparatus for morphometric analysis of a cells of a corneal endothelium, wherein said apparatus is configured to: A. acquire at least one real digital image of the corneal endothelium cells with a bio-microscope equipped with a digital camera, said real digital image being composed of a plurality of single pixels; B. select at least one area of interest of said real digital image; C. detect a luminance value for each pixel of said area of interest D. generate a first matrix having elements that contain a luminance value of the single pixels; E. model said area by reconstructing one or more model cells of the endothelium, effected at least by assignment, to each model cell, of pixels substantially having the same luminance value; F. calculate, for each model cell, a barycenter of the pixels of which said model cell is composed and a radius thereof; G. generate a second matrix 3×N wherein N is a number of identified cells; H. scan said second matrix 3×N and identify, for each model cell, a nearby model cell which satisfies a relation (d*K)≦(r1+r2), r1 and r2 being radiuses of the two model cells, d a distance between barycenters of the two model cells, and K a form factor; I. for each pair of model cells for which the relation of step (H) has been verified, consider the two model cells of the pair of model cells in contact; and J. calculate one or more of a number of model cells per mm.sup.2, an area of each model cell, an average area of the model cells, or a standard deviation on the area of each model cell with respect to one or both of the average area or a number of cells touched by a specific cell.

8. The apparatus according to claim 7, further comprising at least an electronic processor and a screen.

Description

[0038] The method is first described hereunder in its general embodiment and then in greater detail.

[0039] A real image of the corneal endothelium cells is acquired by means of a biomicroscope and digital camera connected to the same and to a personal computer, with a resolution sufficient for the subsequent processing.

[0040] For this purpose, it should be noted that a cell is preferably represented by about 60/70 pixels, but theoretically it should also be possible to work with a lower number of pixels.

[0041] The technique preferably used for selecting the most suitable image consists in acquiring a variable frame number starting from two frames up to any number “n” and calculating the relative MTF (Modulation Transfer Function) for each of these and, on the basis of the necessity of the software, selecting that with the most suitable value for the subsequent processings.

[0042] It should be remembered that the MTF function—in general—is given by the ratio between two contrasts, C1/C0, wherein C1 is the contrast of the image acquired by the sensor of the digital camera (e.g. CCD) and C0 is the contrast of the image before the acquisition process.

[0043] In the present case, the relative MTF is calculated by measuring the contrast C0 at time t0 and the contrast C1 at time t1, again after the acquisition process.

[0044] In this way, there is a relative MTF measurement between two frames acquired at two different moments, one at the moment t0 and the other at the moment t1.

[0045] By repeating the relative MTF measurement for each acquisition (t0, t1, t2 . . . tn), it is possible to evaluate the variation in the MTF moment by moment and therefore position the biomicroscope in the highest-quality area of the image.

[0046] For this purpose, the generation of a warning or feedback signal towards the operator is envisaged, who controls the apparatus (for example a sound feedback) determined by the most suitable MTF (Modulation transfer function) value.

[0047] In order to favour the acquisition, after the area of interest of the endothelium has been identified by the operator, a centering step can be envisaged, which comprises the phase of maintaining the portion of the endothelium of interest in the centre of the screen or monitor in which it is displayed.

[0048] After the acquisition, the area to be analyzed can be selected, for example by selection on the screen, with a frame, and this information can be memorized.

[0049] The processing comprises a preliminary modelling of the real image by means of suitable treatment procedures of the image in a fixed sequence.

[0050] The image is represented by a matrix of Cartesian coordinate points (x, y) to which a sequence of functions f (x, y) is applied, from which the parameters of interest i.e. the area of each single cell and the number of first nearby cells, are deduced.

[0051] The data are finally processed to provide a statistical description of the image, i.e. at least some of the parameters of the morphometric analysis of the corneal endothelium cells.

[0052] More specifically, the method for the morphometric analysis of the corneal endothelium cells, comprises the following steps:

[0053] A. acquiring a real digital image of corneal endothelium cells in a specific area by means of a bio-microscope equipped with a digital camera, said real digital image being composed of a plurality single pixels,

[0054] B. selecting at least one area of interest of said image, preferably an area containing endothelial cells only,

[0055] C. detecting a luminance value for each pixel of said area of interest,

[0056] D. generating a first matrix whose elements contain a luminance value of the single pixels,

[0057] E. modelling said area by reconstructing one or more model cells of the endothelium, effected at least by assignment, to each model cell, of pixels substantially having the same luminance value,

[0058] F. calculating, for each model cell, the barycentre of the -pixels of which said cell is composed” and the radius,

[0059] G. generating a second matrix 3×N wherein N is the number of cells identified,

[0060] H. scanning said matrix 3×N identifying, for each model cell, a nearby model cell which satisfies the relation (d*K)≦(r1+r2), r1 and r2 being the radiuses of the two model cells, d the distance between the barycentres of the two model cells and K a form factor,

[0061] (d*K)e, with each model cell, of pixels substantially having the same distance between the barycentres of the two model cells and K a form factor.

[0062] I. for each pair of model cells for which the relation of step H has been verified, considering the two model cells of the pair in contact,

[0063] calculating the number of model cells per mm.sup.2 and/or the area of each model cell and/or the average area of the model cells and/or the standard deviation on the area of each model cell with respect to the average area and/or the number of cells touched by a certain cell.

[0064] In this way, in short, the morphometric analysis can be effected on the basis of data acquired by the biomicroscope and camera, thus overcoming the limitation linked to the state of the art.

[0065] In this way, it is also possible to select the corneal portion whose morphometric analysis is to be effected with precision.

[0066] According to an optional and advantageous improvement, step C) comprises the step of applying image analysis filters to the pixels of said area, suitable for eliminating false information and homogenizing the luminance value over a whole cell.

[0067] Filters of this type are known in literature; in particular, at least one, preferably all, of the following filters are applied:

[0068] Richardson-Lucy:

[0069] This is an algorithm based on Bayes' theorem which allows the deconvolution of an image to be effected by means of an iterative process. In general, if Y(i) is a latent “non-confused” image, X(i) the acquired image and P(i|j) the point spread function, it can be said that X(i)=ΣP(i|j)Y(j), wherein j is the actual pixel and i is the pixel. It can be demonstrated that Y(j) can be obtained by means of an iterative process assuming that P(i|j) PSF is known and assuming that the distribution of the photons is a Poisson distribution.

[0070] Contrast Enhancement Filter:

[0071] The algorithm for obtaining an improvement in the contrast of the image is based on the equalization of the histogram of the luminance values.

[0072] Morphology Filter (Erosion, Dilation):

[0073] This is an algorithm for image processing based on form analysis. It is based on the use of a structuring element convoluted with the image to be analyzed. The pixel resulting from the convolution has a value depending on the nearby pixel values. Two basic operations are generally used called EROSION and DILATION. In the DILATION operation, pixels on the contours of the image object are summed up. In the EROSION operation pixels are eliminated from the contours of the object in the image.

[0074] Segmentation Watershed Filter:

[0075] This algorithm was introduced by Luc Vincent and Pierre Soille and is based on the immersion concept. Each local minimum is considered as a hole of a surface. The filling of the basin limited by this surface is simulated until only the crest is visible.

[0076] The above filters are known in literature and consequently no further specifications are required.

[0077] All of these filters are preferably applied and in a precise order (time, consecutive) defined above.

[0078] With respect to the form factor K according to step H). this takes into consideration the fact that the form of each single cell is not perfectly spherical; it preferably ranges from 0.7 to 1, so as to accept, in short, a deviation of up to 30%.

[0079] The objectives proposed above have therefore been achieved.

[0080] The protection scope of the invention is defined by the following claims.