WO2006079143A1

WO2006079143A1 - Detection of cancer

Info

Publication number: WO2006079143A1
Application number: PCT/AU2005/001615
Authority: WO
Inventors: Sarah Pearson
Original assignee: University of New England
Current assignee: University of New England
Priority date: 2004-11-15
Filing date: 2005-10-19
Publication date: 2006-08-03
Anticipated expiration: 2007-05-15

Abstract

The method for determining whether a tissue includes a tumour includes: -utilising second harmonic generation (SHG) signals obtained by application of an optical wave to a region of a tissue to from an image of a collagen matrix at the region; -assessing whether the image of the matrix is characteristic of a tumour. The method for determining whether a tumour is malignant includes: -utilising SHG signals obtained by application of an optical wave to a region of a tissue to from an image of a collage matrix at the region; -assessing whether the image of the matrix is characteristic of a malignant tumour. The method for determining whether a tissue includes a tumour includes: -applying an optical wave to a region of a tissue to obtain a SHG signal that corresponds with a collagen matrix at the region; -assessing whether the anisotropy of the signal is characteristic of a tumour.

Description

Detection of cancer

Technical field

The invention relates to using second harmonic generation signals for detecting cancer, especially breast cancer.

Background

Current diagnostic techniques used for detecting malignant breast tumours frequently misdiagnose, are subject to sampling errors, are ineffective for young women and are highly subjective¹'².

Optical second-order harmonic generation (herein "SHG") is a second-order nonlinear optical process, the signal being produced by constructive interference in non- symmetric environments. Because of this constraint it is ideally suited to imaging surfaces, interfaces and other mediums without inversion symmetry, such as chiral molecules.

US patent no. 6,208,886 relates to forming signal intensity maps from the application of SHG signals to tissue to resolve symmetry and content properties of layers in biological tissues.

There is a need for improvements in the detection of cancer, especially breast cancer.

Description

In certain embodiments there is provided a method for determining whether a tissue includes a tumour including:

-utilising SHG signals obtained by application of an optical wave to a region of a tissue to form an image of a collagen matrix at the region;

-assessing whether the image of the matrix is characteristic of a tumour. As discussed herein, the inventor has found that a collagen matrix that is characteristic of a tumour tends to be associated with collagen that has a straight conformation. The median curvature of this straight collagen is 0, with a range of 0 to 0.1, as opposed to that of the typically curly collagen in normal tissue which has a median value of 0.32, range of 0.2 to 0.56.

As discussed herein, the "curvature" of a collagen fibre is expressed as the height of a curve of a collagen fibre divided by width of the curve.

Thus in certain embodiments, the method includes a step of assessing from the image whether the matrix at the tissue region is composed predominantly of collagen that has a straight conformation, with curvature generally less than 0.2. Typically the collagen has a curvature of about 0.1, although it may be less than 0.1, for example, between about.0 and 0.1.

The collagen characteristic of a tumour has also been found to be typically organised into thinner fibres than those in normal tissue. Fibre diameters in malignant tissue range from 0.18 μm to 0.55 μm, with a median of 0.320 μm, whereas fibres from normal tissue have diameters ranging from 0.24 μm to 1.03 μm, with a median value of 0.445 μm. The larger diameter fibres in normal tissue suggests that these fibres constitute a greater range in the number of fibrils in a fibre (up to 11) than for those studied in malignant tissue (up to 8).

Thus, in other embodiments, the method includes a step of assessing from the image whether the matrix at the tissue region is composed predominantly of collagen that is organised into thin fibres. These fibres typically have a diameter of at least about 0.18 μm and generally less than 0.55 μm.

In certain embodiments where the method is performed in vivo, the diameter of collagen fibres may be up to 30% more than the values discussed above. For example, in these embodiments, the method may include the step of assessing from the image whether the collagen fibres have a diameter of at least about 0.234 μm and generally less than 0.715 μm. The inventor has also found that fibrils from malignant tissue have a smaller diameter than those present in normal tissue. Fibrils in normal tissue show a median of 0.05 μm, and range of 0.03 μm to 1.0 μm, whereas fibrils in malignant tissue have a median of 0.04 μm, range 0.04 μm to 0.05 μm.

Thus, in other embodiments, the method includes the step of assessing from the image whether the matrix at the tissue region is composed predominantly of collagen that is organised into fibrils that have a diameter of at least about 0.04 μm and generally less than 0.05 μm.

In certain embodiments where the method is performed in vivo, the diameter of collagen fibrils may be up to 30% more than the values discussed above. For example, in these embodiments, the method may include the step of assessing from the image whether the collagen fibrils have a diameter of at least about 0.052 μm and generally less than

0.065 μm.The experiments discussed herein suggest that the collagen matrix that is characteristic of a tumour tends to be one that is comprised of newly laid down collagen, or otherwise immature forms of collagen.

The above described method is particularly useful for determining whether breast tissue, including inter- and intra-lobular connective tissue, glandular and ductal tissue, includes a tumour.

A tumour is a cancer, neoplasm or other form of tissue that is characterised by a loss of normal control of tissue proliferation and/ or differentiation. A tumour may be a single cell or multi- cellular mass.

Where the tissue is breast tissue, the method is particularly useful for- the detection of carcinoma, especially invasive and non invasive carcinoma.

Examples of non malignant carcinoma include intraductal carcinoma, lobular carcinoma in situ and intraductal papillary carcinoma, benign Phyllodes Tumour. Examples of invasive carcinoma include invasive ductal carcinoma, Paget's disease, invasive lobular carcinoma, medullary carcinoma, mucinous carcinoma, tubular (well- differentiated) carcinoma, invasive papillary carcinoma, adenoid cystic carcinoma, secretory carcinoma, apocrine carcinoma, carcinoma with metaplasia, malignant Phyllodes carcinoma, and mixed type carcinoma.

Examples of forms of breast disease include benign neoplasia of the breast, such as fibroadenoma, other neoplasms, including Phyllodes tumour, lipomas, hemangiomas, leiomyomas, fibromas and angiosarcomas, sclerosing adenosis, hyperplasia, and non neoplastic diseases including fibrocystic disease.

Examples of other tissue include those having a connective tissue component that includes collagen, such as bowel, prostate, skin, cervical, ovarian tissue and the like.

Thus in certain embodiments there is provided a method for determining whether a tissue includes a tumour including:

-assessing whether the image of the matrix is characteristic of a tumour.

Typically the tissue is selected from the group consisting of bowel, prostate, skin, cervix, ovarian tissue and the like. These and other examples of tissue may include epithelial cells. Other tissue such as cartilage and bone may also be determined according to the method.

The image may be assessed by assessing from the image whether the matrix at the tissue region is composed of collagen that predominantly has a straight conformation, as discussed above.

Alternatively, the image may be assessed by assessing from the image whether the matrix at the tissue region is composed of collagen that is organised into thin fibres, as discussed above. Also, the image may be assessed by assessing whether the matrix at the tissue region is composed of collagen that is organised into fibrils, the fibrils having a particular diameter, as discussed above.

The optical wave is typically a laser light. In one embodiment, the laser light has a wavelength in the range of about 700 to 900nm. The second harmonic is typically in the range of about 350 to 450nm. In another embodiment, the laser light has a wavelength of about 830 nm and the second harmonic has a wavelength of about 410 to 415 nm. It will be understood that a laser light and second harmonic may have a wavelength that is more or less than those wavelengths described herein.

Any device for production of an optical wave having a wavelength as described herein may be used. Particularly preferred are those devices that are compatible with an apparatus for application of the optical wave to a tissue and/or for forming an image from the second harmonic (i.e the SHG signal) formed by application of the optical wave.

In a particularly useful form, the tumour is detected in situ for example by using an optical fibre to apply an optical wave to a tissue. One advantage of this form is that in certain embodiments, a diagnosis can be made without biopsy of subject tissue.

However, in certain embodiments, the method is performed in vitro. Thus there is provided a method for determining whether a tissue includes a tumour including:

-providing a sample of tissue;

-utilising SHG signals obtained by application of an optical wave to the sample of tissue to form an image of a collagen matrix at the region;

-assessing whether the image of the matrix is characteristic of a tumour.

Typically the image is assessed by assessing from the image whether the matrix at the tissue region is composed of collagen that predominantly has a straight conformation; or collage that is organised into thin fibres; or collagen that is organised into fibrils having a particular diameter, as discussed above.

As described herein, collagen is typically expressed in a tissue in the form of a matrix, or in other words, an association, collection or network of like fibres, and sometimes other fibres such as elastin and keratin.

The inventor has found that images of normal tissue formed from SHG signals tend to represent a collagen matrix that has a particular degree of organisation. For example in an image of normal breast tissue, a matrix tends to have a curly or crimped appearance, and it is distributed more or less evenly throughout the tissue. Further, acini and alveoli are clearly defined.

The inventor has further found that images of abnormal tissue formed from SHG signals tend to represent a collagen matrix in which organisation is relatively diminished as tissue tends toward neoplasia. For example, in a breast tumour, collagen tends to have a predominantly stringy or straight structure and it tends to be distributed unevenly through the tissue. Further, alveoli and acinus structures are poorly represented in these images.

Accordingly, the inventor has found that an image of a collagen matrix, formed according to the method, that is characteristic of a tumour tends to be one in which a collagen matrix is represented as a stringy and uneven distribution of collagen. Further, where the tissue is breast tissue, the image may further have a poorly defined acini and alveoli morphology.

The inventor has further found that the collagen fibril diameters, measured using the SHG images, may be used to distinguish between normal and malignant tissue.

Accordingly, the inventor has found that the collagen fibrils in malignant tissue are smaller in diameter than those in normal tissue, and that these fibrils correspond to immature collagen fibrils, whereas those in normal tissue correspond to a mixture of immature and mature fibrils. Further, as discussed herein, the inventor has found that an image of a collagen matrix of a malignant tumour can be distinguished from the image of a collagen matrix of a benign tissue. Specifically, it has been found that collagen of a matrix associated with a malignant tumour tends to be composed predominantly of collagen having a straight conformation, as defined above. Collagen of a matrix associated with a benign tumour tends to be composed of collagen having both straight and curly or crimped conformation, as defined above.

Thus in further embodiments there is provided a method for determining whether a tumour is malignant including:

-assessing whether the image of the matrix is characteristic of a malignant tumour.

Alternatively, the image may be assessed by assessing from the image whether the matrix at the tissue region is composed of collagen that is organised into thin fibres, as discussed above.

Also, the image may be assessed by assessing whether the matrix at the tissue region is composed of collagen that is organised into fibrils, the fibrils having a particular diameter, as discussed above.

It will be understood that the image of the matrix formed from the SHG signal can be assessed by any method or instrument in which the information in the image can be presented so as to be visible. An example of a suitable method and instrument is described further herein. Some components of this instrument are: a pulsed laser source linked to a microscope; a sample stage; appropriate filters to isolate the SHG signal; a detector of suitable resolution; imaging software to collect the signal and form an image; software for measuring fibre/fibril diameters, and morphological parameters.

As described herein, one particular advantage of the method is that it has a much higher sensitivity for the detection of abnormal or neoplastic tissue than conventional techniques, such as histological techniques. In particular, as described herein, by assessment of an image formed from SHG signals, the inventor has detected a tumour in the form of a stringy, unevenly dispersed collagen matrix in tissue that was otherwise considered by conventional histological technique to be normal. This is a particularly important advantage in the detection of breast cancer as these cancers tend to metastasise by forming infiltrating fenestrations that are often difficult to detect by conventional techniques.

In one embodiment, the method comprises the steps of:

-utilising SHG signals obtained by application of an optical wave to a selection of regions of a tissue to form an image of a collagen matrix at each region of tissue;

-determining whether an image of a matrix formed at one region is different to an image of a collagen matrix formed at another region, to determine whether the tissue includes a tumour.

The inventor has further found that images of abnormal tissue formed from SHG signals tend to represent a collagen matrix in which the directional distribution of collagen fibres is anisotropic and highly directional. In contrast, images of normal tissue formed from SHG signals tend to represent a collagen matrix in which the directional distribution of collagen fibres is highly isotropic.

Thus, also provided is a method for determining whether a tissue includes a tumour including:

-applying an optical wave to a region of a tissue to obtain a SHG signal that corresponds with a collagen matrix at the region; -assessing whether the anisotropy of the signal is characteristic of a tumour.

The invention is described below by reference to certain non-limiting examples. It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the invention as described in the examples without departing from the spirit or scope of the invention as broadly described. The following examples are, therefore, to be considered in all respects as illustrative and not restrictive.

Brief description of the figures.

Figure 1. SHG image of tumour site breast tissue.

Figure 2. SHG image of breast tissue at 6 mm from the tumour site.

Figure 3. SHG image of breast tissue at the tumour site.

Figure 4. High resolution SHG image of breast tissue 6 mm from the tumour site.

Figure 5. SHG image of adipose tissue.

Figure 6. SHG image of ductal carcinoma in situ.

Figure 7. SHG image of cyst.

Figure 8. Short profile plot through a long (200μm) straight section of fibres in a densely packed region of malignant tissue.

Figure 9. Short profile plot through a long (130μm) straight section of sparsely distributed fibres in malignant tissue.

Figure 10. Profile plot through a short straight section of fibre groups in normal tissue.

Figure 11. 110 μm profile plot through an area where straight fibre groups are sparsely distributed in malignant tissue. Figure 12. 110 μm profile plot through an area where straight fibre groups are densely distributed in malignant tissue.

Figure 13. 300 μm profile plot through an SHG image from malignant tissue showing a sparse, low intensity distribution.

Figure 14. 300 μm profile plot through an SHG image from normal tissue showing a more even, higher intensity distribution.

Figure 15. Surface plots of a) malignant tissue SHG signal, b) normal tissue SHG signal, demonstrating the sparse distribution of collagen in malignant tissue versus the more even distribution in normal tissue.

Figure 16. Histogram of fibril diameters in malignant tissue.

Figure 17. Histogram of fibril diameters in normal tissue.

Figure 18. SHG images of a) normal tissue, b) malignant tissue at the same PMT voltage settings, c) malignant tissue at different PMT settings, showing the presence of straight, fine fibres.

Figure 19. SHG images of a) normal tissue at 20⁰C, b) normal tissue at 60⁰C at the same PMT voltage settings.

Figure 20. SHG maximum surface plots of a) normal tissue, at 20⁰C, and b) normal tissue at 60⁰C at the same PMT settings.

Figure 21. SHG average surface plots of a) normal tissue at 20⁰C, and b) normal tissue at 60⁰C at the same PMT settings.

Figure 22. SHG minimum surface plots of a) normal tissue at 20⁰C, and b) normal tissue at 60⁰C at the same PMT settings. Figure 23. SHG images of a) malignant tissue with the same PMT settings as for figure 19, b) the same sample, the image having been collected at different PMT voltage settings.

Figure 24. High resolution image of fine fibres at the end of a thick fibre group (image is 58 μm by 30 μm).

Figure 25 a) and b). High resolution images of fine fibres peeling away or adhering to a thicker fibre group (image a) is 55 μm by 23 μm, b) is 100 μm by 50 μm).

Figure 26. a) H&E stained histopathology slide image of tissue in a benign area of tissue showing fibroblasts in a highly activated state , b) the SHG^' image of the same sample in this region.

Figure 27. a) H&E stained histopathology slide image of tissue in a malignant area of the same tissue sample shown in figure 19 showing fibroblasts in a highly activated state and b) the SHG image of the same sample in this region.

Figure 28. H&E stained histopathology slide image showing tumour cells forming along collagen fibres in Indian file in malignant tissue.

Example

SHG imaging was conducted as discussed below using malignant tissue from 12 patients, benign tissue from 6 patients and normal tissue from 10 patients. In the order of 800 SHG images have been produced from these tissue samples. The lesion types include: invasive ductal, ductal hyperplasia with atypia, fibroadenoma, ductal hyperplasia without atypia, sclerosis adenosis, intraduct papilloma, fibrofatty, and malignant tissue of grades 1 , 2 and 3. Classification rates of 100% were achieved when classifying the tissue as malignant, benign or normal using collagen fibre morphology and macroscopic tissue morphology, as discussed below, as measures of disease. In one case a 20 μm piece of malignant tissue was observed in a benign sample. Conventional histopathology was conducted on the tissue that had been imaged using SHG. The SHG imaging process discussed herein was found to be more sensitive than H&E staining, although there was agreement between these processes where a result could be obtained by H&E staining.

Materials and methods.

Laser light from a Coherent Mira tuneable pulsed titanium sapphire laser, of wavelength 830 nm and pulsewidth approximately 100 fs, was utilised in order to produce SHG signals in human breast tissue. The wavelength was checked with a Rees spectrum analyser and pulsewidth with an APE autocorrelator.

The tissue samples imaged comprised human breast tissue sourced from biopsy samples as follows;

« 1 mm² surface area tissue blocks at the tumour site and at distances of up to 6mm from this site (at intervals of 2mm from this, up to 6mm from the tumour site). Pathology diagnosis indicated that the tissue at the tumour site was malignant and that all other tissue was 'normal' (healthy).

• Tissue blocks continuous from the tumour site, along the line of invasion and out to the edge of invasion.

• Tissue blocks from benign samples.

• Tissue blocks from normal samples.

These samples were fixed in formalin, embedded in paraffin (for ease of transport), sliced at 15-20 μm and mounted on conventional microscope slides.

The equipment used to produce the SHG signals included a Leica DMIRBE inverted stand equipped with a Leica TCS2MP confocal system. The microscope is equipped with dual photomultiplier transmitted light detectors. A 415/10-nm narrow bandpass filter (with the laser tuned to 830 nm) was used to exclude fluorescent signals in the transmission detector. The main objective used was a 25x NA 0.75 oil-immersion planapochromat and an oil-immersion condenser was also used. The second objective used was a 40. OX/1.25 oil-immersion planapochromat. The laser light was passed through the tissue sample, then filtered using the 415/10 nm narrow bandpass filter to select the SHG signals which were registered using photomultiplier detectors. SHG images were collected automatically as frame-by-frame sequential series, the average of four images at the same position being used in order to reduce the effects of background signals. Leica software was then utilised to form images using the SHG signals. Formation of each image took roughly 1-2 seconds.

Results and discussion.

The inventor has found that the collagen matrix of tumour tissue can be distinguished from that of normal tissue, and hence that tumour tissue can be detected, according to one or more of an assessment of: (i) an image formed from SHG signals; (ii) SHG signal anisotropy. Of particular interest in the assessment of an image formed from SHG signals is the structure of collagen fibres, tissue morphology highlighted by collagen distribution, fibre curvature, fibre distribution, fibre/fibril diameters, and presence of immature collagen fibres.

We have found that each of the above parameters are markers for disease, in particular breast cancer, and hence represent a technique for the detection and diagnosis of cancer.

Collagen structure and tissue morphology

The parameters, collagen structure and tissue morphology, have been studied with the aid of whole images produced using the SHG signals. An example is given in Figure 1. This image was formed using the SHG signals from a 600μm x 600μm area of breast tissue taken from the tumour site.

Figure 1 demonstrates the unique nature of the collagen structure at the tumour site. Note the 'stringy' nature of the collagen fibres in some areas of the image. Also note the uneven distribution of the collagen matrix, some areas of the image indicating a very sparse collagen presence.

Figure 2 shows a representative image generated by collagen at 6 mm from the tumour site. This tissue had been diagnosed as 'normal' tissue by the pathologist. The collagen structure here differs greatly from that observed at the tumour site; the collagen matrix has a 'curly' or 'crimped' arrangement and its distribution, in the areas where acinus or alveoli are not present, is more uniform.

Figure 3 gives another example of the 'stringy' nature of the collagen matrix at the tumour site. This effect was observed in all tumour site images, irrespective of tissue morphology.

The 'stringy' collagen was observed to become increasingly 'curly' as tissue further away from the tumour site was imaged, traces of the 'stringy' collagen still being observed 6 mm away from the tumour site. This is indicated in Figure 4; an image of the collagen matrix at 6 mm from the tumour site at a higher resolution than that of previous images (2048x2048 vs 512x512).

Figure 4 also indicates the presence of traces of 'stringy' collagen, normally associated with tumour site tissue, close to relatively well-defined alveoli. This is a significant observation, indicating the possibility that the method can image the production of 'stringy' collagen, created by the tumour to enable metastasis. The remainder of the image shows evenly distributed 'curly' collagen, associated with normal tissue, and the presence of well-defined alveoli.

Given that this method of analysing SHG images highlights the changing nature of the collagen structure with distance from the tumour site, evidence of tumour associated structure being present at least 6 mm away ^"from this site, the method will make a valuable contribution to the detection of breast cancer. It should be noted that whilst the standard method used by the pathologist for diagnosing the tissue at a distance of 6 mm indicated that the tissue was normal, the SHG imaging method would have been able to correctly diagnose the patient even with tissue at this distance from the tumour.

The second parameter studied, tissue morphology, has also been observed to differ greatly between normal and tumour tissue. This can be seen when contrasting Figure 1 with Figure 2. In Figure 2, for the case where normal tissue was imaged, the alveoli are well defined and circular. Since they are acini they do not contain collagen and hence are observed as black saclike structures. By contrast, Figure 1 indicates that the alveoli at the tumour site are poorly defined, showing infiltration by 'stringy' collagen and not possessing clear edges. The distribution of the collagen matrix also differs between tissue states; the 'normal' tissue indicates an even distribution of collagen matrix throughout the tissue, where alveoli are not present, whereas the images from tumour tissue indicate sparse representation of collagen in some areas. This could be explained by tissue necrosis causing destruction of tissue morphology, reduction in the collagen matrix, and the formation of new 'stringy' collagen as a path for metastasis. This change in tissue morphology was observed to change gradually from the tumour site to 6 mm where the tissue morphology was very well defined. .

In other tissue samples, these differences detailed above between malignant and normal tissue were observed all along the line of invasion and up to the invasion front.

Other tissue morphology recognised using this technique includes adipose (figure 5), ductal carcinoma in situ (figure 6), and cysts (figure 7). The presence of a thick layer of collagen around the cyst should be noted, this differentiating it from ductal carcinoma in situ.

Curvature

Curvature measurements of collagen were conducted using the SHG images from malignant and normal tissue.

Malignant Tissue

All collagen fibre groups follow relatively straight lines. When a measure of curvature is conducted (height divided by width of curve), the range of values found for fibres in malignant tissue is from 0 to 0.1 , the majority of fibres having a median curvature of 0. The maximum curvature for the very few non-straight collagen fibres is 0.1. When a straight line fit is conducted on the majority of fibres the R² value ranges from 0.93 to 0.98, with a median of 0.97, demonstrating the straightness of these fibres.

Normal Tissue Most fibre groups follow curves in random directions. When a straight line fit is conducted on the curved fibres, the R² value ranges from 0.007 to 0.4, indicating a very poor fit to a straight line. When fitting a polynomial (second order), the R² value ranges from 0.73 to 1.0, representing good fits to curves. When a measure of curvature is conducted (height divided by width of curve), the range of values found for the curved fibres in normal tissue is 0.2 to 0.56, the median being 0.32. Straight sections are observed in places but they are significantly different from those observed in malignant tissue: their length is up to 60 μm, versus up to 200 μm in malignant tissue; they are more densely packed and exist in thicker fibre groupings (see section on fibre group distribution in straight sections).

The mean curvature for malignant tissue is significantly different from that for normal tissue, a t-test giving a p value of p<= 0.001 , when curved malignant and straight normal fibres are included, and 0.0003 when they are excluded (ie only representative fibres are compared).

Fibre distribution in straight fibre sections

Collagen fibre distribution was studied using the SHG images, and compared in straight fibre sections within both malignant and normal tissue.

Figures 8 and 9 show intensity profiles over a short region (15 μm) of the image for malignant tissue, figure 8 being for fibres in a densely packed region, figure 9 for sparsely distributed fibres. These figures indicate that densely packed straight fibres in malignant tissue exist in slightly thicker groupings than those in sparsely packed regions. When both figures 8 and 9 are compared with that for normal tissue (figure 10) it can be seen that the straight fibres in normal tissue are more densely packed and exist in thicker fibre groupings. Gray value comparisons cannot be made between the plots for malignant and normal tissue in this case as they were obtained using different voltage settings on the PMT.

Overall fibre distribution Most of the fibres in malignant tissue are distributed very sparsely, others exist in a more evenly spread distribution. This can be seen in Figures 11 and 12.

The densely populated regions of the image yield a higher intensity reading than in areas where the fibres are more sparsely distributed, as can be seen in figures 1 and 2 on the gray value scale. This is due to coherent addition of the SHG signal, hence in regions where the distribution of fibres is dense, the signals from these fibres add up coherently to produce a large signal.

The overall distribution of the collagen fibres in malignant tissue is sparse when compared with that of normal tissue. The intensity of the SHG signal is also much lower for malignant tissue than normal tissue. This can be seen by comparing figures 13 and 14. Note that these figures have a different gray scale range and intensity comparisons can be made with regards intensity of the signal in these images as they were obtained using the same voltage settings on the PMT.

This difference in overall distribution of the collagen fibres between normal and malignant tissue can be elegantly observed using surface plots of the SHG images. No regions of sparsely distributed collagen are observed in the images from normal tissue and the SHG signal from collagen is evenly distributed (except where acini are present, containing no collagen and hence giving no SHG signal). Figures 15 a) and b) are representative surface plots from malignant and normal tissue respectively. Note that intensity comparisons cannot be made with regards intensity of the signal in these images as they were obtained using different voltage settings on the PMT. No acini are included in either figure.

Fibre and fibril diameters

Plot profiles were produced for SHG images from both malignant and normal tissue. The diameters of fibres and individual fibrils were measured and comparisons made. It was found that the fibres from normal tissue had diameters ranging from 0.24 μm to 1.03 μm, with a median value of 0.445 μm and variance 0.034 μm. For malignant tissue the fibre diameters ranged from 0.18 μm to 0.55 μm, with a median of 0.320 μm, variance 0.006 μm. From this it can be seen that the mean diameter for malignant fibres differs significantly from that for normal tissue, a t-test giving a p value of p <= 1x10^~6. As expected, fibres from normal tissue are found to be larger than those observed in malignant tissue, this being due either to collagen degradation or fibrillogenesis in the malignant tissue. It is suggested that the cause might be fibrillogenesis, this being discussed in the section "fibrillogenesis and degradation". The larger number of measured diameters (representing 11 fibre diameter groupings, 11 peaks in a histogram of fibre diameters) in normal tissue suggests that these fibres constituted a greater range in the number of fibrils in a fibre than for those studied in malignant tissue (representing 8 fibre diameter groupings, 8 peaks in a histogram of fibre diameters).

Fibril diameters were also measured for both malignant and normal tissue. It was found that the fibrils from malignant tissue had a smaller diameter than those from normal tissue. Fibrils in normal tissue showed a median of 0.05 μm, variance 0.02 μm, and range of 0.03 μm to 1.0 μm, whereas fibrils in malignant tissue gave a median of 0.04 μm, variance 0.005 μm, range 0.04 μm to 0.05 μm. From this it can be seen that the mean diameter for malignant fibrils differs significantly from that for normal tissue, a t- test giving a p value of p <= 0.05. It was also observed that the fibrils in malignant tissue do not have diameters greater than 0.05 μm, whereas fibrils in normal tissue can have diameters of up to 0.1 μm. In addition, the distribution of fibril diameters was found to be unimodal for malignant tissue (median 0.04 μm with a standard deviation of 0.005 μm, indicating that a unimodal description is appropriate), and multimodal for normal tissue (see figures 16 and 17). Fibrils in both malignant and normal tissue have diameters that are multiples of 8 - 10 nm. This agrees with research conducted using small-angle x-ray scattering (SAXS) and electron microscopy¹'²'³. It should be noted that these values may be reduced with respect to those for live tissue since the tissue preparation (fixation and embedding) can cause shrinkage or expansion of the structures under investigation².

Fibrillogenesis & Degradation Degradation of collagen is expected to be observed in malignant tissue as the tumour cells cause necrosis and remodelling of the extracellular matrix (ECM) in order to invade the surrounding tissue. This is observed in our images by a reduction in SHG signal intensity from malignant tissue versus normal tissue, and also by a reduction in the number and density of collagen fibres observed in the image (as previously mentioned). This is demonstrated in figures 18 a) and b), taken at the same voltage across the PMT and hence directly comparable in terms of intensity. In addition to this, when the SHG image from malignant tissue is collected at PMT settings such that the collagen fibres present are highlighted, straight, fine fibres are observed that are not present in the images of normal tissue. This is shown in figure 18 c).

The SHG images from malignant tissue demonstrate that the majority of the collagen has been degraded (hence leading to a low SHG signal), but straight fibres of higher intensity signal are observed. The presence of high intensity straight collagen in malignant tissue (high relative to the low signal from the majority of the image), suggests that these fibres are not a product of degradation. In order to obtain a high intensity SHG signal the collagen must exhibit chirality and a degree of crystallinity, which would not be expected for degraded collagen. It is suggested that these strands may be immature fibres produced by fibrillogenesis. This has been substantiated by a number of observations made using the following quantities:

a) SHG signal from heat degraded tissue b) Measurement of fibre diameters c) Measurement of fibril diameters d) Fine fibre images e) Fibroblast activity

a) SHG signal from heat degraded tissue

Normal breast tissue was subjected to degradation by heating the tissue to 60⁰C, this degradation being similar to proteolysis caused by enzymatic catalysis present in malignancy. Small-angle x-ray scattering results also indicate that the collagen fibres unravel in the presence of malignancy, producing a larger scattering surface area for amorphous scatter⁴. Hence by heating the normal tissue, degradation caused by enzymatic activity in malignant tissue was mimicked. The SHG images obtained from this degraded tissue were analysed, the results indicating that the straight collagen observed in malignant tissue is not a product of degradation.

Evidence for this conclusion includes: a lower intensity SHG signal being observed for the degraded versus non-degraded tissue (as expected for collagen degradation); tissue morphology is relatively unaffected by heat degradation; and the absence of straight collagen fibres in the degraded tissue, as presented below.

Figures 19 a) and b) show representative SHG images obtained for normal tissue at 20⁰C and 60⁰C respectively, taken with the same voltage settings on the PMT. From this it can be seen that heating the tissue to 60⁰C causes a small but noticeable degradation of collagen, which is translated into a decrease in the SHG signal. The images are from different slices of the tissue sample and hence morphology is not directly matched, however, the morphology of the tissue remains largely unaffected by heating to 60⁰C. The outline structure of the ducts is only slightly blurred on heating. It should also be noted that there are no straight, fine collagen fibres present in either case.

The surface plots given in figure 20 a) and b) show the maximum intensity distributions observed in a) normal tissue at 20⁰C, b) normal tissue at 60⁰C at the same PMT voltage as used to produce the images in figure 19. The surface plots given in figure 21 a) and b) show the average intensity distributions observed in a) normal tissue at 20⁰C, b) normal tissue at 60⁰C at the same PMT voltage as used to produce the images in figure 19. The surface plots given in figure 22 a) and b) show the minimum intensity distributions observed in a) normal tissue at 20⁰C, b) normal tissue at 60⁰C at the same PMT voltage as used to produce the images in figure 19, In all these comparisons it can be seen that the intensity distribution is greater for the tissue kept at 20⁰C than for that heated to 60⁰C. When the mean SHG intensity is calculated for images from tissue at both temperatures and a comparison made, it is found that the median of the mean SHG signal intensities for the normal tissue at 20⁰C is 36.5, and 21.8 for the tissue heated to 60⁰C. A t-test conducted on these mean intensity values gave a p value of p <=9x10^"7, indicating that the difference between the intensity means in these two groups is statistically significant.

Figure 23 shows the SHG image from a malignant tissue sample, taken at a) the same PMT settings as those used to produce figure 19, and b) different settings used to highlight the straight fibres.

Figure 23 a) indicates that collagen degradation in malignant tissue is much greater than that produced when heating tissue to 60⁰C. A comparison of figure 23 b) with figure 19 b) also indicates that the straight collagen observed in malignant tissue is not observed in normal tissue degraded by heating. The morphology of the malignant tissue also shows significant distortion, to a much greater extent than for the heated normal tissue. This is as expected since the presence of tumour cells can lead to both necrosis, regions where tumour cells have infiltrated normal tissue and mimicked ductal tissue, all of which will yield unevenly outlined areas where no collagen is present.

b) Measurement of fibre diameters

As has been mentioned, collagen fibres in normal tissue were found to have diameters ranging from 0.24 μm to 1.03 μm, whilst those in malignant tissue ranged from 0.18 μm to 0.55 μm. Fibres from normal tissue were therefore found on average to have larger diameters than those observed in malignant tissue. The low diameter values for fibres in normal tissue may be a result of collagen turnover present in healthy tissue. It was also noted that the diameter distribution suggested that fibres in normal tissue comprise a larger number of fibrils than those in malignant tissue. No thick fibres (>0.55 μm) were observed in the malignant tissue. These results could be attributed to both collagen degradation and fibrillogenesis in the malignant tissue; degradation leading to a stripping of collagen fibrils from the fibres and hence a decrease in the diameter of the fibres; fibrillogenesis leading to the presence of thinner fibres due to the smaller fibril diameters of immature collagen, and the presence of fewer fibrils in the immature fibres.

c) Measurement of fibril diameters A number of observations on fibril diameters indicate that malignant tissue contains immature fibrils. These include:

1. Fibril diameters in malignant tissue are significantly smaller than those in normal tissue.

2. The distribution of fibril diameters is unimodal in malignant tissue and multimodal in normal tissue.

Research conducted using electron microscopy^{2 3} indicates that immature collagen fibrils exhibit a unimodal distribution of small diameters (mean fibril diameters ranging from 16-49 nm), which are quantised, diameters being multiples of about 8 nm³. These results were obtained using a variety of tissue (such as skin, tendon, cornea from both human, rat, sheep and other animals), and were found to be tissue invariant. This distribution has been shown to evolve to a broad distribution of fibril diameters as growth proceeds, with larger fibril diameters. It is more difficult to ascertain whether the mean fibril diameters are multiples of 8 nm for mature collagen due to the broad distribution of diameters and the fact that fibrilar cross-sections become less regular as the tissue ages. Results from SAXS experiments¹ suggest that fibril diameters are slightly larger (46 nm at 13 days fetal to 58 nm at 19 days fetal for chick metatarsal tendons using SAXS, versus 42 nm at 13 days fetal to 52 nm at 18 days fetal), but the trend is the same. Diameters measured using electron microscopy are reduced due to shrinkage occurring during tissue preparation. This would also be expected in our SHG images since the tissue is fixed and embedded in paraffin, these processes leading to a slight shrinkage of the structures under investigation. It can be seen that our results showing that fibril diameters in malignant tissue are significantly smaller than those in normal tissue, and that the distribution of fibril diameters is unimodal in malignant tissue and multimodal in normal tissue, are consistent with the claim that malignant tissue contains immature collagen colonies.

d) Fine fibre images

It was also observed that in some sparsely populated regions of the image, the straight collagen fibres appeared to split into a number of finer pointed ends (see figure 24 below), and in places appeared to be either peeling away or adhering to thicker fibre groups (figures 25 a) and b)). This effect is not observed in the normal tissue, and it is suggested the process responsible for this is fibrillogenesis since these fibres are positioned in regions of sparse, degraded collagen and yet these fibres have a high SHG intensity.

e) Fibroblast activity

20 μm slices of breast tissue samples were imaged using SHG and 6 μm slices adjacent to these were then stained with H&E for conventional histopathology imaging. The two images were then compared. The conventional histopathology imaging highlights cell nuclei and tissue morphology, whereas the SHG highlights the collagen fibres. It was found that in all cases the diagnosis using SHG agreed with that concluded using conventional methods. In addition to this, fibroblasts were observed in the histopathology images. Fibroblasts that are in an active, collagen synthesis state posses a dispersed chromatin and well-developed cytoplasmic organelles⁵. It was found that fibroblasts in malignant tissue were much more active than in benign areas of the same tissue sample. Figure 26 a) shows fibroblasts in a benign area of tissue, these fibroblasts being in a normal active state. Figure 26 b) shows the SHG image of the same sample in this region. The image contains no straight collagen associated with malignancy, and would be classed as benign when observing the state of the collagen and tissue morphology.

Figure 27 a) shows fibroblasts in a malignant area of tissue, these fibroblasts posses a dispersed chromatin, being in a highly active state. Figure 27 b) shows the SHG image of the same sample in this region. The image contains a large amount of straight collagen associated with malignancy, and would be classed as malignant when observing the state of the collagen and tissue morphology. This suggests that the collagen observed in malignant tissue using SHG is newly synthesised collagen.

Studies have suggested that fibril assembly occurs in close association with deep recesses of the fibroblast cell surface and that assembly of the fibrils may occur by the addition of collagen aggregates to the end of the fibril⁵. The collagen molecules add to each other initially to form long linear subunits and later such aggregates add both linearly and laterally to form a subfibril. These studies showed that recesses in the fibroblast cell contained recently synthesised collagen fibrils which were in the process of deposition in the extracellular space. The fibrils were observed to be striated. These results indicate that it would be highly likely that the immature collagen would take a striated form initially, and could therefore account for the straightness of the fibres observed in malignant tissue. Again, this could indicate that the collagen observed in malignant tissue using SHG is newly synthesised collagen. Results also suggest that the tumour cells may move along these collagen fibres in Indian file. This is shown in figure 28).

Bringing all these results together appears to indicate that the major difference between normal and malignant tissue observed using the SHG method is the maturity of the collagen fibres. Normal tissue has a heterogeneous distribution of mature and immature collagen fibres, whereas the mature collagen fibres in malignant tissue have been degraded and immature collagen is formed by highly active fibroblasts.

In addition to these studies with normal and malignant tissue, benign tissue was studied. It was found that the collagen fibres in benign tissue exhibited a structure between the curliness found in normal tissue and the straightness observed in malignant tissue. The tissue morphology was also degraded. This is shown in figure 26 b.

The final parameter studied was the anisotropy of the SHG signal. The SHG polarization anisotropy can be used to determine the absolute orientation and degree of organisation of the collagen in tissues. The anisotropy parameter, β, describing the

molecular orientation is defined as , where l_par = intensity of signal

whose polarisation is parallel to the polarisation of the incident laser beam, l_perp = intensity of signal whose polarisation is perpendicular to the polarisation of the incident laser beam. It ranges in value from -0.5 to 1. If β= 0 then the collagen strands are arranged isotropically, when l_par = Iperp- A value of 1 refers to complete ordering relative to the incident laser, and so indicating well aligned, well structured helices. Polarisation of the laser light was achieved using a linear polariser, and a rotatable polariser placed between the condenser and the PMT detector allowed the SHG signal to be separated into parallel and perpendicular components.

Measurement of the anisotropy parameter, β, was carried out for tissue at the tumour site and 6mm away. A mean value of 0.32 (SD 0.05) for tumour tissue and 0.21 (SD 0.05) for 'normal' tissue at 6mm was observed, indicating that the 'normal' tissue is more isotropic than the tumour tissue. This can be explained by the observation that the collagen strands in 'normal' tissue are 'curly', uniformly distributed and pointing in all directions; whereas the collagen in the tumour tissue is very straight and 'stringy' and appears highly directional. This indicates that β is another discriminator between tissue states.

Other measurements of directionality (such as the use of Gabor transforms) will also yield methods for differentiation using the strong directional nature of the collagen in tumour tissue.

References:

1. Bjurstam N, et al., The Gothenburg breast screening trial: first results on mortality, incidence, and mode of detection for women ages 39-49 years at randomization, Cancer, 1997, 80(11): p. 2091-2099

2. Rosenberg, R. D., et. al., Effects of age, breast density, ethnicity, and estrogen replacement therapy on screening mammographic sensitivity and cancer stage at diagnosis: review of 183, 134 screening mammograms in

^■ Albuquerque, New Mexico, Radiology, 1998, 209: p. 511-518.

3. Raymond, W. and A.-Y. Leong, Assessment of invasion of in breast lesions using antibodies to basement membrane components and myoepithelial cells, Pathology, 1991 , 32: p. 291-297 4. Kauppila, S., et al., Aberrant type I and type III collagen gene expression in human breast cancer in vivo, J Pathol, 1998, 186(3): p. 262-8 5. Wang, W, et al., Single cell behaviour in metastatic primary mammary tumours correlated with gene expression patterns revealed by molecular profiling, Cancer Research, 2002 (November 1), 62: p. 6278-6288 6. Eikenberry EF, Brodsky BB and Parry DAD, Collagen fibril morphology in developing chick metatarsal tendons: 1. X-ray diffraction studies, lnt J Biol Macromol, 1982 (October), 4: p 322-328.

7. Eikenberry EF, Brodsky BB, Craig AS and Parry DAD, Collagen fibril morphology in developing chick metatarsal tendon: 2. Electron microscope studies, lnt J Biol Macromol, 1982 (December), 4: p 393-398 . 8. Parry, DAD and Craig AS, Electron microscope evidence for an 8OA unit in collagen fibrils, Nature, 1979, 282: p 213-215

9. Fernandez, M., et al., Small angle X-ray Scattering studies of human breast tissue samples, Physics in Medicine and Biology, 2002, 47: p. 577-592

10-Trelstad, R. L., and K. Hayashi, Tendon collagen fibrillogenesis: intracellular subassemblies and cell surface changes associated with fibril growth, Dev. Biol., 71 : p. 228-242.

Claims

1. A method for determining whether a tissue includes a tumour including:

-assessing whether the image of the matrix is characteristic of a tumour.

2. A method according to claim 1 wherein the method includes a step of assessing from the image whether the matrix at the tissue region is composed predominantly of collagen that has a curvature (as defined herein) of less than about 0.2, to assess whether the image of the matrix is characteristic of a tumour.

3. A method according to claim 2 wherein the curvature is about 0.1.

4. A method according to claim 3 wherein the curvature is between about 0 and 0.1.

5. A method according to claim 1 wherein the method includes the step of assessing from the image whether the matrix at the tissue region is composed predominantly of collagen fibres having a diameter of less than about 0.55 μm, to assess whether the image of the matrix is characteristic of a tumour.

6. A method according to claim 5 wherein the diameter is about 0.32 μm.

7. A method according to claim 6 wherein the diameter is between about 0.18 and 0.32 μm.

8. A method according to claim 1 wherein the method includes the step of assessing from the image whether the matrix at the tissue region is composed predominantly of collagen that is organised into fibrils that have a diameter of less than about 0.05 μm, to assess whether the image of the matrix is characteristic of a tumour.

9. A method according to claim 8 wherein the diameter is about 0.04 μm.

10. A method according to claim 9 wherein the diameter is between about 0.04 and 0.05 μm

11. A method according to claim 1 wherein the tissue is selected from the group , consisting of breast, bowel, prostate, skin, cervical, ovarian tissue, bone and cartilage.

12. A method according to claim 1 wherein the optical wave has a wavelength in the range of about 700 to 900nm.

13. A method according to claim 12 wherein the optical wave has a wavelength of about 830 nm

14. A method according to claim 13 wherein the second harmonic has a wavelength of about 41 O to 415 nm.

15. A method for determining whether a tumour is malignant including:

16. A method for determining whether a tissue includes a tumour including:

-applying an optical wave to a region of a tissue to obtain a SHG signal that corresponds with a collagen matrix at the region;

-assessing whether the anisotropy of the signal is characteristic of a tumour.