WO2003044522A1

WO2003044522A1 - Method of registration of visible light image to fluroescent light image of protein spots

Info

Publication number: WO2003044522A1
Application number: PCT/AU2002/001557
Authority: WO
Inventors: Edmond Joseph Breen
Original assignee: Proteome Systems Intellectual Property Pty Ltd
Priority date: 2001-11-16
Filing date: 2002-11-15
Publication date: 2003-05-30

Abstract

A method of registration of a white light image of a gel to a fluorescence image for gel is disclosed in which the boundary of the gel itself is used to determine the mapping of the protein spot locations imaged in fluorescent light to locate the proteins within the white light imaging system. Typically, a radial decomposition function of the boundaries of the white light image and fluorescent image of the gels is used to map any point on the fluorescent image onto the white light image.

Description

METHOD OF REGISTRATION OF VISIBLE LIGHT IMAGE TO FLUORESCENT LIGHT

IMAGE OF PROTEIN SPOTS

Field of the Invention

This invention relates to a tool for cutting spots from a gel under white light. In particular, the present invention relates to the cutting of spots of macromolecules from a drab of gel in which a 2-D electrophoresis process has been carried out in order to separate macromolecules as proteins in the sample into an array of spots on the gel.

Background of the Invention

2-D electrophoresis is a very common method for separating samples of macromolecules such as proteins but which may be any biomolecule particularly large biomolecules produced by biological processes. The protein spots in the gel are not easy to see with the naked eye and can be visualised either by staining the protein spots with a dye that is visible under white light or by fluorescence of a protein itself. Some proteins naturally fluoresce: In some cases it is common to apply a fluorescent marker to the proteins.

Whilst identification under fluorescent imaging system provides a clear means of imaging different protein spots on a gel chromatogram, the subsequent excision of the spots from the gel for further analysis is expensive as it needs to be conducted under fluorescence imaging conditions on expensive custom built equipment. If imaging were possible on an ordinary desktop scanner such as is proposed in the applicant's co-pending Australian provisional application PR5226 entitled Imaging Means for Excision Apparatus, the entire contents of which are incorporated herein by reference, the process would be considerably cheaper and easier to carry out.

The present invention relates to a method of excising fluorescent dyed protein spots under a white light without the need of fluorescent conditions. In the method, the protein spots on a 2-D electrophoresis gel are initially imaged under fluorescence imaging. Subsequently, the image data on the gel is then imported to the white light system and the spots excised.

In order to carry out this process successfully, it is necessary to achieve accurate registration of the white light image of the gel to the fluorescence image. There are a number of existing methods of undertaking this registration.

It is known to insert fiducial markers into the gel which are visible both under fluorescent and white light. However, this approach is not without problems since the fiducial markers invade the gel area and may obscure key proteins in the gel. Further, the placing of fiducial markers into the gel may damage and weaken the gel. Further, a large number of fiducial markers are required to position the gel to micrometer precision necessary for the exact location of spots.

Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is solely for the purpose of providing a context for the present invention. It is not to be taken as an admission that any or all of these matters form part of the prior art base or were common general knowledge in the field relevant to the present invention as it existed in Australia or elsewhere before the priority date of each claim of this application.

Summary of the Invention

In a first broad aspect, the present invention provides a method of registration of a white light image of a gel to a fluorescence image for gel in which the boundary of the gel itself is used to determine the mapping of the protein spot locations imaged in fluorescent light to locate the proteins within the white light imaging system.

Typically, a radial decomposition function of the boundaries of the white light image and fluorescent image of the gels is used to map any point on the fluorescent image onto the white light image.

The accuracy of this method is determined by the accuracy with which the edge of the gel can be determined. The boundary of the gel may be extremely irregular. The boundary of the gel may also be difficult to determine due to digitisation and aliasing effects. Although any suitable method for determining the edge of the gel could be used, in one preferred embodiment the edge detection method used is that described in the applicant's co-pending International patent application filed on even date and entitled "Method for locating the edge of an object". In that method the boundaries of an object of the gel are determined by a) creating a binary image of the object, the binary image having a central region, a boundary region, and an outside region with the boundary region encompassing the boundary of the object; and b) performing a homotopic thinning of the binary image, iteratively until no more pixels can be removed.

For optimum results, where the object is a gel, the homotopic thinning iteration is based on a grey scale image of the gel edge and surrounding area, and most preferably based on an edge threshold image. Thus, in a preferred aspect of the present invention, step (b) comprises the steps of: c) producing a grey scale edge image of the gel having high response values or intensity where local intensity changes are highest within the image; and d) performing a grey scale controlled homotopic thinning of the binary image in which the pixels which can be removed homotopicaly from the binary image are ordered and the pixel which is removed is that which corresponds in location to the pixel which has the smallest edge/intensity value in the grey scale edge image of the gel. This process is continued iteratively until no more pixels can be removed.

Various methods can be used to produce an edge image of the gel, including Sobel, dilation erosion or laplacian.

In one preferred embodiment step a) involves the following steps i) creating a crude binary image of the gel and cleaning the image to remove noise to provide an image that depicts the gel segmented into one group with the background as a second group, with the one group including the gel and gel boundary areas; and ii) subtracting an erosion of the binary image from a dilation of the binary image.

The method is particularly suited to edge detection where the object is a gel of the type used in the chromatography process referred to as gel electrophoresis, however it could be used in other applications where edge detection is required for image segmentation or locating objects. Step b) may utilise any suitable edge detection method such as a sobel method.

Once the gel edge has been determined, smoothing can be used to remove local perturbations and warping effects in the gel. Brief Description of the Drawings

A specific example of the present invention will now be described by way of example only and with reference to the accompanying drawings in which:- Figure 1 shows a first image of a gel; Figure 2 shows a second image of the gel of Figure 1 ; and

Figure 3 is a graph comparing the distance of a gel edge of two gel images from their centroids with angle;

Figures 4a to 4f are schematic diagrams illustrating various steps in an edge determining process; and Figures 5a and 5b illustrate a homotopic edge thinning process.

Detailed Description of a Preferred Embodiment

In the present invention the boundary of the gel itself is used to determine the mapping of the protein spot locations imaged in fluorescent light to locate the proteins within the white light imaging system. The accuracy of the method is determined in part by the accuracy with which the edge of the gel can be determined. Although various methods exist to predict the location of the boundary of a gel, the preferred method is that disclosed in the applicant's co-pending PCT patent application filed on even date and entitled "Method for locating the edge of an object" which is described in detail below.

A radial decomposition function of the boundaries of the white light image and fluorescent image of the gels is used to map any point on the fluorescent image onto the white light image. Radial decomposition is defined as the length, as a function of angle, for any position in the boundary of the gel to the centroid of the gel. Thus, scale and rotational differences between the two gel images are determined from the cross correlation of the radial decomposition of the fluorescent and white light images. Figure 1 illustrates a first gel image 10 which might say be an image made under fluorescent conditions. Figure 2 illustrates a second image 12 of the same gel which might say be an image made under white light which is an affine of the first image but which may have been translated, rotated and/or scaled up or down.

Figure 3 shows a graphical representation of the distance of the edge or boundary 16, 18 of the two images 10 and 12 from their centroids 14. Line 20 represents the boundary of the first image 10 and line 22 the boundary of the second image 14. The two images can be compared an the transformation (scale, translation and rotation) required to map one onto the other determined. In particular, the relative locations of the centroid gives the translation between the two images, the cross-correlation the rotational difference and the relative amplitude of the lines in Figure 3 gives the scale.

Once the scale, rotation and translation factors are known, arbitrary X, Y locations, such as those of protein spots can be mapped from the fluorescent image to their corresponding positions in the white light image. These mapped protein spots can then be accurately excised from the gel.

The accuracy of the mapping can be quickly determined by remapping the white light and image locations back to the fluorescent image and then measuring the difference and positions of corresponding points on the original and remapped fluorescent.

In the edge detection process thresholding of the gel image is first undertaken to provide an approximation of the image boundary as a crude segmentation. In Figure 4a, a scanned image 110 of a gel 112 against a background 114, which might be a scanner is shown. All parts of the image which are darker than a threshold intensity T are ascribed binary value 1 , all other parts of the image are set at binary value 0. This crude binary image includes the gel, the separated background and noise. Next, noise cleaning is first undertaken on the binary image to provide a binary image 116 shown in Figure 4b that essentially depicts a gel segmented into one group 118 with the background as a second group 120. The group 118 which includes the gel will include the gel and also boundary areas around the gel where shadowing and diffraction makes the background adjacent the edge of the gel look darker than the rest of the background. The threshold T is set to include this shadowed boundary area so that the group 118 is guaranteed to include the edge of the gel.

The cleaned binary image 116 is then put through a morphological edge detection. This is defined as the dilation δr(B) of the cleaned binary image minus the erosion εr(B) of the image using a disk or square structuring element of a specified radius r. The value of r is chosen so that it is comfortably large enough to include the edge and any shadowing but not so large as to include many spots in the gel. Typically, r is 11 to 20 pixels.

Figure 4c shows the dilation (expansion) of B at 122 and Figure 4d the erosion of B at 124. Both are binary images having the same shape as the binary gel image 118. The subtraction of the erosion from the expansion produces a thick boundary 128 of width 2r, within which the real boundary is located. This thickened boundary estimate 128 shown at 126 can now be thinned to produce a medial or skeletal boundary as an estimate of the boundary of the gel.

At the same time as the boundary estimate is generated, a traditional local edge detection method (edge(f) = δr1(f) - εr1(f))is used to produce an image with high response values where local intensity changes are highest within the image which occurs on the edges and where the spots in the gel are present. This image is shown in Figure 4e at 130. In the image there are regions of high intensity changes at 32 where there are spots and at 134 and 136 for example at boundary of the gel. Various known edge detection methods such as Sobel, dilation, erosion or Laplacian can be used. For example, in a Laplacian method, to produce the edge value for a particular pixel, (the "central" pixel) the adjoining pixels intensity values are summed and averaged and the average value subtracted from the intensity value of the central pixel. Essentially, high values are produced where changes in intensity are greatest - which is indicative of a boundary or edge.

Homotopic thinning is then carried out on the image 128. Specifically, this homotopic thinning approach does not change the number of regions in the image during boundary thinning and therefore is guaranteed to keep closed boundaries. For example, in the Figure 126 there are three regions the inside 140 the thickened boundary region 128 and the outside 142. The algorithm used in homotopic thinning examines each pixel within the thickened boundary region to see if can be removed without altering the homotopy of the image. If it can be removed as described below, it is removed. This process is repeated until no further pixels can be removed without breaking the homotopy.

Figures 5a and 5b illustrate homotopic thinning. In Figure 5a which shows one section of the inner edge of the boundary of the thickened boundary region in which in theory any of the pixels marked with an x or pixel 50, could be removed from the boundary region without creating a new region and destroying the homotopy. If pixel 50 is removed (ie changed from binary 1 to binary 0 ), pixel 52 is now eligible for removal -see Figure 5b. If pixel 52 is then removed any of the seven pixels surrounding it could also be removed without creating a new region.

In order to determine the edge location most accurately all those pixels that can be removed homotopically at each iteration are ordered, according to their edge element value from an edge thresholding image, as determined from the corresponding location in the edge thresholding image 130 which is grey scale. In Figure 4f the thickened boundary region 128' is shown superposed over the edge thresholding image 130. Only the pixel that has the smallest edge value is removed at each iteration. This is the darkest pixel. After each pixel is removed from image 128 any new pixels which can now be homotopically removed are also ordered

The boundary detected by this process will be one that has maximised the minimum value along the contour and has produced a contour that lines up along the strongest edge elements pertaining to the boundary of the gel. It is guaranteed to be continuous as the homotopic thinning process guarantees a connected path. Thus, an optimal contour location is produced. This determination of the gel edged boundary lines up naturally on the most responsive parts of the edge as determined from the edge element image.

Although this method may appear to be computationally intensive, it can be performed extremely quickly using priority queues and neighbourhood analysis. Thus, the method of the present invention offers a fast and repeatable approach to define closed edge boundaries of gels. For further refinement, control smoothing is applied. As well as being used with gels the method of the present invention may also be used to image other objects particularly membranes to which an array of biomolecules have been transferred by electroblotting or the like.

It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the invention as shown in the specific embodiments without departing from the spirit or scope of the invention as broadly described. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.

Claims

1. A method of registration of a visible light image of an object to a fluorescence image of an object characterised by using the boundary of the object itself to determine the mapping of the protein spot locations imaged in fluorescent light to locate the proteins under a white light imaging system.

2. A method as claimed in claim 1 wherein the object contains or supports an array of biomolecules.

3. A method as claimed in claim 2 wherein the object is a gel.

4. A method as claimed in claim 2 wherein the object is a membrane.

5. A method as claimed in any preceding claim 1 wherein a radial decomposition function of the boundaries of the visible light image and fluorescent image of the objects is used to map any point on the fluorescent image onto the visible light image.

6. A method as claimed in claim 5 including the step determining scale and rotational differences between the visible light image of the object and the fluorescent image of the object from the cross correlation of the radial decomposition of the fluorescent and visible light images

7. A method as claimed in any preceding claim wherein the visible light image is a white light image.

8. A method as claimed in any one of claims 2 to 7 further comprising the step of mapping X, Y locations of biomolecule spots in the array from the fluorescent image to their corresponding positions in the white light image.

9. A method as claimed in claim 8 further including the step of directing a robotic cutting head or the like to the mapped biomolecule spots and excising them from the gel.

10. A method as claimed in any preceding claim wherein the boundary of the object is determined by the steps of:- a) creating a binary image of the object, the binary image having a central region, a boundary region, and an outside region with the boundary region encompassing the boundary of the object; and b) performing a homotopic thinning of the binary image, iteratively until no more pixels can be removed.

11. A method as claimed in claim 10 wherein the step of performing a homotopic thinning iteration is performed on a grey scale image of the object and an area surrounding the object.

12. A method as claimed in claim 11 wherein the step of performing homotopic thinning iteration is performed on an edge threshold image of the object.

13. A method as claimed in any one of claims 10 to 12 wherein step (b) comprises the steps of: bi) producing a grey scale edge image of the object having high response values or intensity where local intensity changes are highest within the image; and bii) performing a grey scale controlled homotopic thinning of the binary image in which pixels which can be removed homotopicaly from the binary image are ordered and the pixel which is removed is that which corresponds in location to the pixel which has the smallest edge/intensity value in the grey scale edge image of the gel.

14. A method as claimed in claim 12 or claim 13 wherein the edge threshold image of the object is produced by a dilation of the binary image minus an erosion of the binary image.

15. A method as claimed in claim 13 wherein the grey scale edge image of the object is produced by a Sobel, dilation/erosion or laplacian method.

16. A method as claimed in any one of claims 10 to 15 wherein step a) includes the following steps ai) creating a crude binary image of the object and cleaning the image to remove noise to provide an image that depicts the object segmented into one group with the background as a second group, with the one group including the object and object boundary areas; and aii) subtracting an erosion of the binary image from a dilation of the binary image.