WO2012039679A2

WO2012039679A2 - System for near-infrared autofluorescence measurement of a subject, and method thereof

Info

Publication number: WO2012039679A2
Application number: PCT/SG2011/000322
Authority: WO
Inventors: Zhiwei Huang
Original assignee: National University Of Singapore
Priority date: 2010-09-22
Filing date: 2011-09-20
Publication date: 2012-03-29
Also published as: WO2012039679A3; SG179311A1

Abstract

A system for near-infrared autofluorescence measurement of a subject, the system comprising: a light source configured to emit an excitation light at near-infrared; a medium configured for delivering the excitation light to the subject; and a processing section for processing an autofluorescence signal from the subject at near-infrared resulting from the excitation light impinging upon the subject so as to analyze the subject. For example, the processing section comprises a spectrometer for detecting the autofluorescence signal at near-infrared to analyze the subject. As another example, the processing section comprises an image sensor for generating an image based on the autofluorescence signal.

Description

SYSTEM FOR NEAR-INFRARED AUTOFLUORESCENCE MEASUREMENT OF A SUBJECT, AND METHOD THEREOF

FIELD OF INVENTION

The invention relates to a system for near-infrared (NIR) autofluorescence (AF) measurement of a subject and a method thereof. In particular, the invention relates to NIR AF spectroscopy system and NIR AF imaging system and methods thereof for tissue measurements for detection and/or diagnosis of cancer such as colon cancer.

BACKGROUND Gastrointestinal malignancies are the second leading cause of cancer-related death, and also the third most common cancer in the world. In certain countries, colonic cancer has become the most common malignancy for males while the second most common for females. In general, the diagnosis of colonic cancer is based on conventional white light colonoscopic inspections followed by the histopathological examinations of biopsied tissues. However, for example, flat and depressed neoplastic lesions are difficult to identify due to the lack of obvious morphological changes under white light colonoscopy which heavily relies on the observation of gross morphological changes of tissues associated with neoplastic transformation. The accurate diagnosis and localization of early neoplastic tissue represents an important measure for planning effective treatment to improve the survival rates of patients with colonic malignancies.

In the past, autofluorescence (AF) spectroscopy and imaging techniques, which are capable of probing the changes of tissue morphological structures and concentrations of intrinsic fluorophores such as collagen, nicotinamide adnine dinucleotide (NADH), and flavin adenine dinuccleotide (FAD) in tissue associated with neoplastic transformation, have been investigated for improving the diagnostic sensitivity of malignant lesions in various organs, including the colon. For example, there exist tissue AF studies which are focused on the use of ultraviolet (UV) or short visible (VIS) wavelengths of excitation light. However, UV or short VIS excitation light has a limited penetration depth and cannot detect lesions in deeper areas.

It is against this background that the present invention has been developed. SUMMARY

The present invention seeks to overcome, or at least ameliorate, one or more of the deficiencies of the prior art mentioned above, or to provide the consumer with a useful or commercial choice.

According to a first aspect of the present invention, there is provided a system for near-infrared autofluorescence measurement of a subject, the system comprising: a light source configured to emit an excitation light at near-infrared;

a medium configured for delivering the excitation light to the subject; and a processing section for processing an autofluorescence signal from the subject at near-infrared resulting from the excitation light impinging upon the subject so as to analyze the subject.

Preferably, the processing section comprises a spectrometer for detecting the autofluorescence signal at near-infrared to analyze the subject.

Preferably, the medium is a fiber probe comprising a first fiber for delivering the excitation light and a plurality of second fibers for collecting the autofluorescence signal of the subject.

Preferably, the plurality of second fibers is arranged so as to surround the first fiber. Preferably, the fibre probe has a first end portion located towards the spectrometer and a second end portion located towards the subject in use, and wherein the fiber probe comprises a first filter section at the first end portion and a second filter section at the second end portion.

Preferably, the first filter section comprises a first filter coupled to the first fiber for suppressing laser noise and a second filter coupled to the plurality of second fibers for suppressing reflected excitation light. Preferably, the second filter section comprises a first filter coupled to the first fiber for suppressing noise generated in the first fiber and a second filter coupled to the plurality of second fibers for suppressing reflected excitation light while allowing the autofluorescence signal to be collected by the plurality of second fibers. Preferably, the fiber probe is configured to be insertable through an instrument channel of an endoscope for autofluorescence measurement of the subject under the guidance of the endoscope.

Preferably, the endoscope is part of a wide-field endoscopic system for providing white-light reflecting imaging.

Preferably, the system combined with the wide-field endoscopic system constitutes an integrated near-infrared autofluorescence spectroscopy and imaging system. Preferably, the subject is a tissue, and the system is operable to analyze the tissue based on the autofluorescence signal and to detect the presence of cancer.

Preferably, the system is further operable to classify the tissue into a category of tissue including normal, hyperplastic polyp, adenomatous polyp and cancer. Preferably, the system further comprises a processor for analysing an autofluorescence spectrum received from the spectrometer based on one or more multivariate statistical techniques. Preferably, the processor is operable to process the autofluorescence spectrum under the control of a program comprising a set of executable instructions for performing the one or more multivariate statistical techniques.

Preferably, the one or more multivariate statistical techniques comprise at least one of principal component analysis and linear discriminant analysis.

Preferably, the classification of the type of tissue is determined based on the principal component analysis and linear discriminant analysis. Alternatively, the processing section comprises an image sensor for generating an image based on the autofluorescence signal.

Preferably, the medium comprises a series of lens. Preferably, the series of lens comprises a first lens system for transmitting the excitation light and a dichroic mirror for reflecting the excitation light to the subject.

Preferably, the first lens system comprises a collimator and a narrow band-pass filter.

Preferably, the processing section further comprises a second lens system for receiving the autofluorescence signal from the subject at near-infrared resulting from the excitation light impinging upon the subject. Preferably, the image sensor is operable to receive the autofluorescence signal and generate a near-infrared autofluorescence image based on the received autofluorscence signal. Preferably, the subject is a tissue and the near-infrared autofluorescence image generated by the image sensor enables the visual detection of a cancer portion based on differences in intensity level between normal and cancer portions.

Preferably, the first lens system further comprises a polarizer for polarizing the excitation light so as to enhance the differences in intensity level between normal and cancer portions in the near-infrared autofluorescence image generated by the image sensor.

Preferably, the excitation light is perpendicularly polarized. Alternatively, the excitation light is horizontally polarized.

Preferably, the system further comprises a second light source and a third lens system for diffuse reflectance imaging.

Preferably, the second light source emits a light to illuminate the subject through the third lens system, and near-infrared diffuse reflectance signal from the subject resulting from the light impinging upon the subject are collected by the imagine sensor after passing through the second lens system.

Preferably, the processing section further comprises a processor for receiving a near-infrared diffuse reflectance signal and a corresponding near-infrared autofluorescence signal from the subject, and ratio imaging the near-infrared diffuse reflectance signal to the corresponding near-infrared autofluorescence signal.

Preferably, the subject is a tissue and said ratio imaging produces an image whereby the differences in intensity level between normal and cancer portions of the tissue in the image is enhanced.

Preferably, the third lens system further comprises a polarizer for polarising the light from the second light source. Preferably, the second lens system further comprises an analyzer such that polarized autofluorescence and diffuse reflectance images can be acquired in tandem through rotation of the analyzer.

Preferably, the system is coupled to an endoscope for in vivo autofluorescence measurement of the subject.

According to a second aspect of the present invention, there is provided a method for near-infrared autofluorescence measurement of a subject, the method comprising:

emitting an excitation light at near-infrared from a light source;

transmitting the excitation light to the subject; and

processing an autofluorescence signal from the subject at near-infrared resulting from the excitation light impinging upon the subject so as to analyze the subject.

Preferably, said processing comprises detecting the autofluorescence signal at near-infrared using a spectrometer to analyze the subject.

Preferably, said transmitting comprises transmitting the excitation light to the subject through a fiber probe, and wherein the fiber probe comprises a first fiber for transmitting the excitation light and a plurality of second fibers for collecting the autofluorescence signal of the subject.

Preferably, said transmitting further comprising filtering the excitation light for suppressing laser noise

Preferably, said collecting further comprises filtering the autofluorescence signal for suppressing reflected excitation light. Preferably, the method further comprises inserting the fiber probe through an instrument channel of an endoscope for autofluorescence measurement of the subject under the guidance of the endoscope.

5 Preferably, the endoscope is part of a wide-field endoscopic system for providing white-light reflecting imaging.

Preferably, the method combined with said white-light reflecting imaging constitutes an integrated near-infrared autofluorescence spectroscopy and imaging method.0

Preferably, the subject is a tissue, and said processing comprises analyzing the tissue based on the autofluorescence signal for detecting the presence of cancer.

Preferably, said processing further comprises classifying the tissue into a category5 of tissue including normal, hyperplastic polyp, adenomatous polyp and cancer.

Preferably, said processing comprises analysing an autofluorescence spectrum received from the spectrometer based on one or more multivariate statistical techniques.

0

Preferably, said processing is under the control of a program comprising a set of executable instructions for performing the one or more multivariate statistical techniques.

:5 Preferably, the one or more multivariate statistical techniques comprise at least one of principal component analysis and linear discriminant analysis.

Preferably, wherein said classifying is determined based on the principal component analysis and linear discriminant analysis.

D

Alternatively, said processing comprises generating an image based on the autofluorescence signal using an image sensor. Preferably, said transmitting comprises transmitting the excitation light through a series of lens. Preferably, the series of lens comprises a first lens system and a dichoric mirror, and said transmitting comprises transmitting the excitation light through the first lens system and reflecting the excitation light to the subject using the dichroic mirror.

Preferably, said processing further comprises the processing section further comprises a second iens system for receiving the autofluorescence signal from the subject at near-infrared resulting from the excitation light impinging upon the subject.

Preferably, said processing further comprises said image sensor receiving the autofluorescence signal and generating a near-infrared autofluorescence image based on the received autofluorscence signal. Preferably, the subject is a tissue and said method further comprising visual detection of a cancer portion based on differences in intensity level between normal and cancer portions in the near-infrared autofluorescence image generated by the image sensor. Preferably, the method further comprises polarizing the excitation light so as to enhance the differences in intensity level between normal and cancer portions in the near-infrared autofluorescence image generated by the image sensor.

Preferably, said polarizing comprises perpendicular polarization. Alternatively, said polarizing comprises horizontal polarization. Preferably, the method further comprises diffuse reflectance imaging using a second light source and a third lens system.

Preferably, said diffuse reflectance imaging comprises emitting a light from the second light source to illuminate the subject through the third lens system, and collecting near-infrared diffuse reflectance signal from the subject resulting from the light impinging upon the subject by the imagine sensor after passing through the second lens system. Preferably, said processing further comprises receiving a near-infrared diffuse reflectance signal and a corresponding near-infrared autofluorescence signal from the subject, and ratio imaging the near-infrared diffuse reflectance signal to the corresponding near-infrared autofluorescence signal. Preferably, the subject is a tissue and said ratio imaging produces an image whereby the differences in intensity level between normal and cancer portions of the tissue in the image are enhanced.

Preferably, said diffuse reflectance imaging further comprises polarising the light from the second light source.

Preferably, said processing further comprises acquiring polarized autofluorescence and diffuse reflectance images in tandem through rotation of an analyzer in the second lens system.

Preferably, the method is coupled to an endoscope for in vivo autofluorescence measurement of the subject.

According to a third aspect of the present invention, there is provided a system according to the first aspect of the present invention for non-invasive in vivo diagnosis or detection of cancer in an organ. Preferably, the organ is a colon. According to a fourth aspect of the present invention, there is provided a system according to the first aspect of the present invention for ex vivo diagnosis or detection of cancer in an organ. Preferably, a tissue specimen of the organ is resected for diagnosis or detection of cancer. Still preferably, the organ is a colon.

According to a fifth aspect of the present invention, there is provided a method according to the second aspect of the present invention for non-invasive in vivo diagnosis or detection of cancer in an organ. Preferably, the organ is a colon. According to a sixth aspect of the present invention, there is provided a method according to the second aspect of the present invention for ex vivo diagnosis or detection of cancer in a tissue of the subject an organ. Preferably, a tissue specimen of the organ is resected for diagnosis or detection of cancer. Still preferably, the organ is a colon.

According to a seventh aspect of the present invention, there is provided a data storage medium having stored therein a set of instructions executable by a computer processor for processing said autofluorescence signal from the subject at near-infrared according to any one of the second, fifth and sixth aspects of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the invention will be better understood and readily apparent to one of ordinary skill in the art from the following written description, by way of example only, and in conjunction with the drawings, in which:

Figure 1 depicts a system for near-infrared autofluorescence measurement of a subject according to an embodiment of the present invention;

Figures 2(a) to (c) depict exemplary colonic tissues of normal type, polyp type and cancer type, respectively; Figure 3 depicts the in vivo mean autofluorescence spectra ±1 standard error from normal (n=116), hyperplastic polyp (n=48), and adenomatous polyp (n=34) colonic tissues;

Figure 4 depicts the eight significant principal components (PCs);

Figure 5 depicts box charts of the eight significant principal component (PC) scores for three colonic types (normal, hyperplastic polyp and adenomatous polyp);

Figure 6 depicts a two-dimensional ternary plot of the posterior probability belonging to normal tissue, hyperplastic and adenomatous polyp achieved by PCA-LDA algorithms, together with the leave-one tissue site-out, cross validation method;

Figure 7 depicts the in vivo mean NIR AF spectra ±1 standard error of benign

(normal and hyperplastic polyps, n=164), precancer (adenomatous polyps, n=34) and cancer (n=65) colonic tissue;

Figure 8 depicts the relationship between the number of PLS factors-latent variables (LVs) and the cross validation classification error for correct classification of benign, precancer and cancer colonic tissues;

Figure 9 depicts the first three diagnostically significant LVs accounting for about 80% of the total variation in the AF spectral dataset;

Figure 10 depicts a two-dimensional ternary plot of the posterior probability belonging to benign tissue, precancer and cancer achieved by PLS-DA algorithms, together with the leave-one tissue site-out, cross validation method;

Figure 11 depicts a mean diffuse reflectance (DR) spectra ±1 standard error (SE) of normal (n=58) and cancer (n=48) colonic tissue;

Figure 12 depicts a system for near-infrared autofluorescence measurement of a subject according to another embodiment of the present invention;

Figure 13 depicts a system for near-infrared autofluorescence measurement of a subject according to a further embodiment of the present invention;

Figure 14 depicts polar diagrams for a full sample rotation of every 20 degrees for six paired colonic tissues;

Figures 15(a) to (c) depict representative NIR DR images of colonic tissues acquired using tungsten halogen light illumination under different polarization conditions;

Figures 15(d) to (f) depict representative NIR AF images of colonic tissues acquired using 785 nm laser excitation under different polarization conditions; Figure 16(a) and (b) depict the average AF intensity for the normal and cancer colonic tissues based on the selected region on NIR DR image and NIR AF images, respectively;

Figures 17(a) to (c) depict representative pseudocolor NIR AF images of colonic tissues acquired using 785 nm excitation under different polarization conditions;

Figure 17(d) depicts the intensity profiles along the lines as indicated on the NIR AF images in Figures 17(a) to (c);

Figures 18(a) to (c) depict pair-wise comparison of NIR AF intensities of all 48 paired (normal vs. cancer) colonic tissues under the three different polarization conditions;

Figure 19(a) depicts a processed polarization ratio image of normal and cancer tissues;

Figure 19(b) depicts depolarized ratio values along the line across the normal and cancer tissues as indicated on the polarization ratio image in Figure 19(a);

Figures 20(a) to (c) depict NIR DR images of colonic tissues acquired using a broadband light source under different polarization illumination;

Figure 20(d) depicts intensity profile along the lines as indicated in the NIR DR images in Figures 20(a) to (c);

Figures 21 (a) to (c) depict the ratio imaging of the NIR DR image to the NIR AF image under different polarization conditions; and

Figure 21 (d) depicts a comparison of ratio intensity profiles along the lines as indicated in the ratio images in Figures 21 (a) to (c).

Figure 22 depicts a schematic flowchart illustrating a method for near-infrared autofluorescence measurement of a subject according to an embodiment of the present invention.

DETAILED DESCRIPTION

Figure 1 depicts the schematic diagram of an exemplary system 10 for near-infrared (NIR) autofluorescence (AF) measurement of a subject according to an embodiment of the present invention. For example, the subject can be one of various types of tissue such as a tissue of an organ (e.g., colon). In the embodiment, there is provided an exemplary integrated NIR AF spectroscopy and imaging system 10 for the NIR AF measurements of tissue for the detection and/or diagnosis of cancer in various organs (e.g., colon) under the guidance of wide-field endoscopic imaging. For the sake of clarity, the system 10 will hereinafter be described with respect to the in vivo detection and/or diagnosis of colon cancer. However, it will be apparent to persons skilled in the art that the present invention is not limited to only the in vivo detection/diagnosis of colonic cancer, and other types of detection/diagnosis technique, including ex vivo, and other types of cancers regardless of in vivo or ex vivo detection/diagnosis fall within the scope of the present invention.

As described hereinbefore in the background, ultraviolet (UV) or short visible (VIS) excitation light has a limited penetration depth and cannot detect lesions in deeper areas. Unlike UV excitation light, NIR light can penetrate deeper into the tissue of up to about 1 mm. Further, NIR light is non-carcinogenic and thus is safe for tissue diagnosis.

The system 10 comprises an NIR AF spectroscopy system 20 for in vivo NIR AF measurements of tissue for the detection and/or diagnosis of cancer. The spectroscopy system 20 comprises a light source 22 for generating an excitation light, a processing section 23 comprising a spectrometer 24 for receiving and processing collected tissue AF spectrum resulting from the excitation light hitting the tissue, and a medium 25 comprising an endoscopic fiber probe 26 for laser light delivery. In the embodiment, the fiber probe 26 is also configured for AF spectrum collection. For example, the light source 12 can be a spectrum-stabilized 785 nm laser diode with a maximum output of 300 mW. For example, the spectrometer 24 may be any scientific- grade spectrometer.

The fiber probe 26 according to the embodiment comprises a light delivery fiber (or central excitation fiber) 28 and a plurality of collection fibers 30 surrounding the light delivery fiber 28. As a non-limiting example, the fibre probe 18 may comprise 32 collection fibers 30 surrounding the central light delivery fiber 28 as illustrated in Figure 1. In an embodiment, the light delivery fiber 28 has a core diameter of about 200 pm and a numerical aperture ("NA") of about 0.22. The fibre probe 26 comprises two stages of optical filtering, a first filter section and a second filter section, respectively incorporated at a first end (or a proximal end) 32 and a second end (or a distal end) 34 of the fibre probe 26. The distal end 34 of the fibre probe 26 incorporates two different types of filters. In particular, the distal end 34 of the light delivery fiber 28 is coated with a narrow bandpass filter (centered at 785 nm, FWH = ±2.5nm) and the distal end 34 of the plurality of collection fibers 30 are coated with edge long-pass filters (cut off at 800 nm). The narrow bandpass filter is operable to reduce most of the fused-silica noise generated in the light delivery fiber 28 of the fibre probe 26 before the excitation beam transmitted from the light source 22 hits the tissue under inspection. The edge long-pass filters are operable to reinforce the blocking of the reflected excitation light but yet allow the scattered tissue NIR AF signal to be collected by the collection fibers 30 and transmitted to the spectrometer 24 for examination. At the proximal end 32 of the fiber probe 26, the light delivery fiber 28 is coupled to a first filter 36 and the collection fibers 30 are coupled to a second filter 38 as illustrated in Figure 1. The first filter 36 is an in- line (or laser-line) filter module integrated with a narrow bandpass filter for suppressing laser noise, fluorescence, and Raman emissions from the fiber connecting the light source 22 to the first filter 36 for tissue excitation. The second filter 38 is an in-line (or laser-line) filter module integrated with an edge long-pass filter for suppressing the reflected excitation light (if not already blocked by the edge long-pass filter at the distal end 34) while permitting the scattered-tissue AF signals to pass through toward the spectrometer 24.

The first filter 36 further comprises a lens system 40 for effectively coupling the excitation light generated from the light source 22 to the light delivery fiber 28 of the fibre probe 26. The second filter 38 also further comprises a lens system 40 for effectively coupling the collected AF signals from the collection fibers 30 to the spectrometer 24.

The spectroscopy system 20 further comprises a computer 42 for performing various operations such as triggering data acquisition and background spectrum subtraction (primaryily dark current in the detector). For example, the computer 42 may comprise a processor 43 for processing a set of executable instructions and a storage medium 44 having stored therein a computer program comprising a set of executable instructions for controlling the operations of the computer 42 when executed. The computer 42 may further comprise a display 46 for displaying various information received and processed by the computer 42, such as AF spectra images. The system 10 further comprises an imaging system 60 for providing wide-field endoscopic imaging (or white-light reflectance (WLR) imaging). The imaging system 60 comprises an endoscope (or a video colonoscopy) 62 operable to be inserted into a hollow organ or a cavity of the body for examination, a light source 64 for illuminating the tissue under inspection, and a video system 66 comprising a video processor for WLR imaging. The system 10 may also comprise a display 68 for displaying the wide-field endoscopic images. For example, the light source 64 can be a 300W dedicated short-arc xenon light source, the endoscope 62 can be a commercially available video colonoscopy. In the embodiment, the endoscope 62 comprises an instrument (or biospy) channel 70. The fibre probe 26 is configured to be inserted through the instrument channel 70 of the endoscope 62 for excitation light delivery and in vivo AF spectrum collection.

Therefore, as depicted in Figure 1 , an integrated NIR AF spectroscopy and imaging system 10 is provided for in vivo NIR AF measurements of tissue for the detection and/or diagnosis of cancer in various organs (e.g., colon) of a subject under the guidance of wide-field endoscopic imaging. Further, with the system 10, wide-field endoscopic images and the corresponding real-time in vivo tissue AF spectra images can be simultaneously acquired, displayed and recorded in the video system 66 and the computer 42, respectively.

Experiments performed using the system 10 for in vivo tissue AF measurement during colonoscopy will now be described. Figure 2a to 2c show the WLR images of different types of colonic tissues (namely, normal, polyp and cancer colonic tissues, respectively) obtained during experimental clinical colonoscopy. Under the guidance of wide-field endoscopic WLR imaging, when a suspicious lesion was discovered during routine colonoscopy screening, the fiber probe 26 was placed through the instrument channel 70 to collect AF spectra at 785 nm laser excitation. Following the acquisition of AF spectra from colonic polyps or cancer, an endoscopically normal-appearing area of colonic mucosa, approximately 1 cm distant from the polyp, was selected as a normal site for AF spectra collection. Due to increases proliferation of mucosal cell, the polyp appears as a lump that protrudes into the inside of the colon and can be endoscopically differentiated from healthy colon that has a smooth surface with a visible patter of fine blood vessels. Immediately after NIR AF spectra acquisition, the polyps were resected and fixed in formalin for routine histopathologic examination and the cancer tissues were done biopsy for histopathological confirmation.

In this experiment, in vivo AF spectra of 263 colonic tissue sites were acquired from 100 patients (57 male and 43 female, with a median age of 51 years) who underwent colonoscopy screening, in which 116 spectra were from normal, 48 spectra were from hyperplastic polyp (confirmed by histopathology), 34 spectra were from precancer (adenomatous polyps (confirmed by histopathology)), and 65 spectra were from colon cancer tissues (confirmed by histopathology).

AF spectroscopy can probe great wealth information of intrinsic fluorophores, such as collagen, elastin, and porphyrin. But the AF spectrum usually contains many overlapping bands and the data interpretation can not be easily based on simple visual inspection for subtle change in tissues. Hence, according to embodiments of the present invention, different statistical techniques are provided to analysis AF spectrum for tissue diagnosis and classification. The AF spectrum data usually consists of many different variables (e.g., intensity, spectral shape and wavelength) for different cases (e.g., normal tissue, hyperplastic polyp, adenomatous polyp, and cancer). Each of theses variables can be considered to represent a different dimension. Given n variables (where 'n' represents the number of variables), each of the different cases may be regarded to be located in a respective position in an n-dimensional hyperspace. However, this can be difficult to visualize due to the number of variables. Therefore, various statistical algorithms according to embodiments of the present invention are provided to address such a problem by reducing the high-order dimensional space to a lower-order and interpretable dimensional space. The statistical algorithms according to embodiments of the present invention will be described hereinafter.

Principal component analysis (PCA)

5

PCA is a mathematical procedure that uses an orthogonal transformation to convert a set of observations variables, which are possibly correlated with each other, into a set of values of uncorrelated variables called principal components (PCs). PCA decomposes the spectroscopic data matrix S into scores T and loading P, according to the relation,0

S = T - P (4.1 )

With this equation, PCA transforms a number of correlated variables into a number of uncorrelated variables (PCs) which describe the greatest variance of the spectral data. The number of PCs is less than or equal to the number of original variables. It can be 5 employed as a method for variable or data reduction by retaining the first few PCs. The first PC accounts for as much of the variability in the data, and each succeeding PC has the highest variance possible in turn. Each PC is orthogonal to the preceding PCs. In addition, inspection of the plots generated by scores provides a mean to assess the relationship between samples, since it helps to identify some clusters related to a certain0 feature and detect potential outliners.

Linear discriminant analysis (LDA)

LDA is used for pattern recognition and machine learning to find linear combination of5 the measurements variables that separate the objects from different classes as much as possible. The distance (e.g. Euclidean distances) between groups and the compactness of each group are used to determine the separability of classes. Then LDA follows the rule that the ratio of the between-to-within variability of the transformed training data vectors (i.e. spectra) should be maximized.

\0

^SMAx = {Variance _helween I Variance _withill } (4.2) In other word, the aim of LDA is to find a discriminant function line that maximizes the variance in the data between groups and minimizes the variance between members of the same group. LDA is closely related to PCA and both methods look for linear combinations of variables that best explain the data. LDA explicitly attempts to model the difference between the classes of data, while PCA does not take into account any difference in class.

In this experiment, PCA-LDA was applied for multivariate analysis. PCA was used to reduce the dimensions of the data and also retain significant details to perform classification. Before performing PCA, the data set is standardized to eliminate the intra and inter-patients spectral variability. As a result of the standardization, the mean and variance of the spectra become zero and one respectively. It transforms a number of possibly correct variables into a smaller number of uncorrelated variables called PCs. The combinations of variables produce the PCs when they are projected on the unit vector space. The variance can be probed by perturbing the factor loadings. The first PC corresponds to maximum variability in the data set, and the variability goes on decreasing for successive PCs. These low variance PCs are largely accounting for noise that can be removed for further analysis without losing the significant details. These PCs are orthogonal to each other. The PCA is a method which decomposes the eigenvector of the covariance matrix to project the data onto new set of orthogonal axes, so the PCs are formed by maximizing the variance within the dataset. After PCA, unpaired student's T test is done on the reduced data to find the most significant PCs (p<0.05). The significant PCs are further classified by performing LDA (supervised classification algorithm) that maximizes the variances between the groups and minimizes the variances within a group. Therefore the group clustering can be achieved. If there are more than two groups, the analysis of variance (ANOVA) is performed on PCs to identify the significant PCs. These significant PCs are given to LDA model for further classification. If the number of groups is n, then n-1 discriminant function are required for classification.

The high dimension of florescence spectral space (each fluorescence spectrum ranging from 810-1050 nm with a set of 324 intensities) will result in computational complexity and inefficiency in optimization and implementation of the LDA algorithms. As such, PCA was first performed on tissue fluorescence data set to reduce the dimension of fluorescence spectral space while retaining the most diagnostically significant information for tissue classification. To eliminate the influence of intersubject and/or intrasubject spectral variability on PCA, all spectra were standardized so that the mean of the spectra was zero. Mean centering ensures that the PCs form an orthogonal basis. Standardized AF data sets were assembled into data matrices with wavelength columns and individual case rows. The PCA was performed on the standardized spectral data matrices to generate PCs comprising a reduced number of orthogonal variables that accounted for most of the total variance in original spectra. Each loading vector is related to the original spectrum by a variable called PC score, which represents the weight of that particular component against the basis spectrum. PC scores reflect the differences between different classes. In the embodiment, one-way ANOVA was then used to identify the most diagnostically significant PCs (P<0.05) for separation of three different tissue classes. These significant PC scores were selected as input for the development of LDA algorithms for multiclass classification. LDA determines the discriminate function that maximizes the variances in the data between groups while minimizing the variances between members of the same group. The performance of the diagnostic algorithms rendered by the LDA models for correctly predicting the tissue groups was estimated in an unbiased manner using the leave-one-sample-out, cross-validation method on all model spectra. In this method, one sample (one spectrum) was held out from the data set, and the entire algorithm including PCA and LDA was redeveloped using the remaining tissue spectra. The algorithm was then used to classify the withheld spectrum. This process was repeated until all withheld spectra were classified. For the assessment of diagnostic sensitivity and specificity of NIR AF spectroscopy technique, histopathologic examination results serve as the gold standard (i.e., assumed to be 100% accurate). In this experiment, the multivariate statistical analysis was performed in the STATISTICA version 7 (Statsoft Inc., Tulsa, Oklahoma, USA). Partial least square - discriminant analysis (PLS-DA)

PLS-DA is a classification method based on the PLS algorithm. PLS is a statistical method that bears some relation to PCs regression. It finds a linear regression model by projecting the predicted variables and the observable variables to a new space. In other words, PLS is a two block regression method, the independent block is predicted from the dependent block. Both X and Y blocks are given to PLS method, where X block is the dataset with each row representing each ample and Y block represents the class labels. The X and Y blocks are mean centered to bring the intercept term zero. PLS is an iterative algorithm used to produce the desired response from the dependent variables. The latent scores, loadings, and weights are extracted for the corresponding number of the selected components. The complexity of the model is controlled by the number of the selected components. By selecting the optimum number of components, it is possible to avoid over fitting and under fitting. It extracts the latent variables in a decreasing order of their respective singular values from the input dataset. The number of optimal components is determined from leave one out cross validation error values. For two classification (e.g., normal and cancer), one class can be 1 and other may be 0 or -1. The predicted values are calculated for test sample by building the model with remaining samples.

According to embodiments of the present invention, PLS-DA can advantageously be applied for multi-class classification problems by encoding the class membership of zeros and ones, representing group affinities. PLS-DA employs the fundamental principle of PCA to explain the diagnostically relevant variations but rotates the components (Latent variables (LVs)) by maximizing the covariance between the spectral variation and group affinity. Therefore, the LVs explain the diagnostic relevant variations rather than the most prominent variations in the spectral dataset. In most cases, this ensures that the diagnostic significant spectral variations are retained in the first few LVs. In this experiment, the performance of the PLS-DA diagnostic algorithm was validated in an unbiased manner using the leave-one-tissue spectra-out, cross validation methodology. In the validation procedure, one tissue AF spectrum was left out and the PLS-DA modeling was redeveloped using the remaining AF spectra. The redeveloped PLS-DA diagnostic algorithm was then used to classify the withheld AF spectra. This process was repeated iteratively until all withheld AF spectra were classified. The number of retained LVs was determined based on the cross validation error classification curves. In this experiment, multivariate statistical analysis was performed using the PLS toolbox (Eigenvector Research, Wenatchee, WA) in the Matlab (Mathworks Inc., Natick, MA) programming environment.

In the experiment, the NIR AF spectroscopy system 20 was successfully employed to acquire 263 in vivo NIR AF spectra under the guidance of the imaging system 60 from different types of colonic tissues, including normal (n=116), hyperplastic polyp (n=48), adenomatous polyp (n=34) and cancer (n=65) colonic tissues. Based on spectral characters among these different types of colonic tissues, PCA-LDA was applied to run three group classifications among normal, hyperplastic polyp, and adenomatous polyp for distinguishing the subtypes of colonic polyps. Subsequently, PLS-DA was applied to identify the precancer (adenomatous polyps) from benign (normal and hyperplastic polyp) and cancer tissues. Good classification results demonstrate the diagnostic capability of integrated NIR AF spectroscopy to be a clinical complement to conventional WLR endoscopy for the rapid, non-invasive, in vivo differentiation of precancer (adenomatous polyps) and cancer in clinical colonoscopy procedures.

As illustrative examples according to embodiments of the present invention, the classification of subtype of colonic polyps using PCA-LDA, detection of precancer using PLS-DA and classification of colon cancer based on diffuse reflectance spectra will be described hereinafter.

Classification of subtype of colonic polyps using PCA-LDA

Figure 3 shows the in vivo mean autofluorescence spectra ±1 standard error (SE) from normal (n=116), hyperplastic polyp (n=48), and adenomatous polyp (n=34) colonic tissues. The adenomatous polyps are characterized by notably lower AF intensity than that of the normal tissue and hyperplastic polyps in the whole region (810-1050 nm). The comparison of AF spectra between the two subtypes of colonic polyp tissue (hyperplastic polyp vs. adenomatous polyp) shows that the hyperplastic polyp has a significantly higher NIR AF intensity than the normal tissue by 1.2-fold with the p-value of 4E-4 (paired 2-sided Student's t-test). The spectral differences in both the intensity and line shape confirm the capability of NIR AF spectroscopy according to embodiments of the present invention for in vivo differentiation of adenomatous polyps from hyperplastic polyps in the colon. In addition, Figure 3 also shows the Raman features, which are indicated with dashed lines, are distinguishable within each classified AF spectra of colonic tissues. In Figure 3, the in vivo NIR AF spectra of normal colonic tissue are vertically shifted for better visualization. The shaded areas in tissue AF spectra represent the respective standard error. The lines indicate the prominent Raman peaks in the AF spectra. For instance, the peaks in the AF spectra at 874, 885, and 1015 nm correspond to Raman peaks 1302 cm^"1 [CH2CH3 twisting of proteins and nucleic acids], 1450 cm^"1 [δ (CH2) of proteins and lipids], and 2885 cm ^"1 [CH2 stretching of lipids] respectively. According to embodiments of the present invention, these integrated NIR AF and Raman spectral features are utilized to further improve the identification of hyperplastic and adenomatous colonic polyps.

As can been seen in Figure 3, the experimental results demonstrate that the AF intensity decreases progressively from normal tissue to hyperplastic polyps and adenomatous polyps. The decrease in broadband AF peak intensities can be attributed to the changes of tissue optical properties associated with adenoma-carcinoma progression. Due to cellular hyperproliferation, the thickening of the mucosa layer could significantly attenuate the excitation light penetration and also obscure the tissue AF emission from submucosa layer in the polyp tissue compared to normal colonic tissue. Moreover, the lower intensity of adenomatous polyp compared to hyperplastic may be caused by the proliferation of neoplastic crypt cells. In the adenomatous polyp, the growth of these crypt cells displace the lamina propria and further reduce the collagen AF emission from the lamina propria. Thus, a combination of factors related to morphological architecture resulted in an overall decrease of NIR AF intensity in adenomatous polyp as compared to benign colonic tissue (normal and hyperplastic tissue). In addition, the changes in concentrations of endogenous biochemicals in colonic tissue, such as NADH, collagen, flavins, porphyrin, etc., associated with neoplastic transformation may also attribute to the differences in the NIR AF emission among normal, hyperplastic and adenomatous polyp colonic tissue.

In general, AF spectroscopy is very sensitive to tissue biochemical and morphological changes but is inaccurate in determining the types of specific changes during the progression of colon polyps to colon cancer due to its very broad spectra line shapes. According to an embodiment of the present invention, to investigate the endogenous fluorophores that are responsible for the NIR spectral differences between different types of polyps, an integrated NIR AF and Raman spectroscopy system (i.e., the NIR AF spectroscopy system 20 combined with Raman technique) for in vivo measurement of normal, hyperplastic, and adenomatous polyp tissues during clinical colonoscopic examination is provided. With the integrated NIF AF and Raman spectroscopy system, the Raman peaks which represent biological molecules, such as proteins, lipids, and DNA, can be observed from in vivo colonic NIR raw spectra. The prominent Raman signals at 1302 cm ^"1, 1450 cm^"1, and 2885 cm^"1 can be attributed to proteins, lipids, and DNA that are involved in the metabolic activities. These biochemical alterations can be attributed to the increased cellular metabolism in the dysplastic tissues. For instance, during development and progression of colonic polyp to cancer, epithelial cells undergo transformations that result in increased metabolic activity (e.g., increased mitotic activities that include enzymes, hormones, etc.) and the increased hyperchromatism and the nucleic acids-to-cytoplasm ratio of dysplastic cells. These spectral differences show that, although the NIR AF technique alone can be used to identify adenomatous and hyperplastic polyps, the raw spectra, which contains both NIR AF and Raman signatures, can provide more diagnostic information of intrinsic biomolecules which are the Raman-active scatterers in tissue. Hence, it is found that the integrated NIR AF and Raman spectroscopy system has the ability to probe the changes of morphology and endogenous fluorophores as well as the biomolecular structure and composition for improving diagnosis of adenomatous polyps in the colon.

In an attempt to determine the most diagnostically significant AF features for identifying adenomatous polyp tissue from normal tissue and hyperplastic polyp colonic tissue, the multivariate statistical technique (e.g., PCA-LDA) was employed by utilizing the entire NIR AF spectra (each fluorescence spectrum ranging from 810-1050 nm with a set of 324 intensities). One-way ANOVA test on the obtained PC scores shows that eight PCs (PC1- PC8) (see Figure 4) accounting for -99% of the total variance contain the most diagnostically significant AF features (p<0.05) for classification of normal, hyperplastic and adenomatous. The first PC accounted for the largest variance (~95.5% of the total variance), and generally represents the raw AF spectra line shape (see Figure 3). The successive PCs contribute progressively smaller variances (PC2~3.08 %, PC3~0.99%, PC4~0.1%, PC5~0.09%, PC6~0.02%, PC7~0.01%, and PC8~0.01%) (see Figure 4). The PC features, such as the peaks and troughs are correlated with the representative biochemicals associated with structural or cellular metabolic progression in colonic precancer and cancer.

Figure 5 shows the relationship between the diagnostically significant difference colonic tissue types. In particular, Figure 5 shows box charts of the eight significant PC scores for the three colonic types (normal, hyperplastic polyp and adenomatous polyp): PC1 , PC2, PC3, PC4, PC5, PC6, PC7, and PC8. The line within each notch box represents the median, and the lower and upper boundaries of the box indicate first (25 percent percentile) and third (75 percent percentile) quartiles respectively. Error bars (whiskers) represent the 1.5-fold interquartile range, where the asterisk in Figure 5 represents P<0.05 (pairwise comparison of tissue types with post boc multiple comparison tests (Fisher's Least Significant Differences (LSD)))

Fisher's LSD tests show that different PC scores were largely associated with different degrees of diagnostic utility for classification of different colonic tissue types (normal tissue, hyperplastic polyp and adenomatous polyp). For instance, PC1 can be used for differentiating hyperplastic poly from normal tissue and adenomatous polyp, PC2 is optimal in discriminating normal tissue from hyperplastic and adenomatous polyp, PC3 and PC7 can be used to distinguish adenomatous polyp from hyperplastic polyp and normal tissue, PC4 and PC6 can be used to separate adenomatous polyp from normal tissue, PC5 show efficacy in classification of the three different colonic tissue types, PC8 can be used to separate hyperplastic polyp from normal tissue. Figure 6 is a two-dimensional ternary plot of the posterior probability belonging to normal tissue, hyperplastic and adenomatous polyp, and illustrates the good clusterings of the three different colonic tissue types achieved by PCA-LDA algorithms, together with the leave-one tissue site-out, cross validation method. In particular, the two-dimensional ternary plot is derived when all eight diagnostically significant PCs were loaded into the LDA model to generate effective diagnostic algorithms for colonic polyp tissue identification.

The plot depicts probabilistic outcome in association with data for normal, hyperplastic and adenomatous polyp colonic tissues, providing a three-class diagnostic model for classification. The final diagnostic category of each data point was determined by the nearest proximity of data to the diagnostic category related to the vertex of the ternary plot. The vertices in Figure 6 represent the 100 per cent posterior probability belonging to normal tissues, hyperplastic, and adenomatous polyps, respectively.

Table 1 shown hereinafter summarizes the diagnostic indices for in vivo NIR AF spectra using PCA-LDA together with leave-one tissue site-out, cross validation method in classifying the three different types of colonic tissue. Diagnostic sensitivities of 83.6%, 77.1 % and 88.2%, specificities of 96.3%, 88%, and 92.1%, and accuracies of 88.9%, 85.4% and 91.4%, respectively, were achieved for differentiation between normal tissue, hyperplastic and adenomatous polyps.

Autofluorescence prediction

Normal Hyperplastic polyp Adenomatous Total polyp

Normal 97 14 5 116

Hyperplastic 3 37 8 48 polyp

Adenomatous 0 4 30 34 polyp

Sensitivity (%) 83.6 77.1 88.2

Specificity (%) 96.3 88 92.1 Accuracy (%) 88.9 85.4

Table 1 - Classification results of in vivo NIR AF spectra prediction for the three colonic tissue groups using PCA-LDA algorithms, together with the leave-one tissue site-out, cross validation method.

Detection of precancer using PLS-DA

Based on the good classification of subtype of colonic polyps as described hereinbefore, using the NIR AF spectroscopy system 20 for differentiating precancer (adenomatous polyp) from benign and cancer colonic tissues is now described as an illustrative example.

Figure 7 shows the in vivo mean NIR AF spectra ±1 standard error (SE) of benign (normal (n=116) and hyperplastic polyp (n=48)), precancer (adenomatous polyp (n=34)) and cancer (n=65) colonic tissues. The shaded areas in tissue AF spectra represent the respective standard error. In Figure 7, it should be noted that the in vivo NIR AF spectra of benign colonic tissue has been vertically shifted for better visualization. All spectra are intensity calibrated using the calibrated tungsten-halogen light source. This intensity calibration will retrieve the real tissue NIR AF signals without depending on the instrument response function.

As can be seen from Figure 7, significantly lower AF intensity was observed in cancer tissues compared to the benign and precancer colonic tissue. The differences of fluorescence intensity among benign, precancer, and cancer tissue may be attributed to the changes of tissue optical properties in the colon, such as the thickening of mucosa layer due to hyperproliferation, which could significantly attenuate the excitation light penetration and also obscure the tissue AF emission from the precancer and cancer tissue compared to benign colonic tissue. In addition, the changes in concentrations of endogenous fluorophores such as NADH, collagen, flavins, porphyrin, etc., associated with malignant transformation may also be attributed to the differences in the NIR AF emission among benign, precancer, and cancer colonic tissue. According to an embodiment of the present invention, to develop effective algorithms for identifying precancer (adenomatous polyps) from benign (normal tissue and hyperplastic polyps) and cancer colonic tissue, the standardized AF spectra were assembled into a dataset with wavelength columns and individual case rows (each fluorescence spectrum ranging from 810-1050 nm with a set of 324 intensities). Prior to data analysis, the constructed dataset was mean centered to remove common variances. The PLS-DA is then employed with leave-one-spectrum-out, cross validation method to generate diagnostic algorithms, such that 7 LVs were found to be the optimal numbers of retained components as defined by the cross validation (CV) classification error indicated in Figure 8, accounting for 80% of the total AF spectral variances. Figure 8 illustrates the relationship between the number of PLS factors-latent variables (LVs) and the cross validation classification error for correct classification of benign, precancer and cancer colonic tissues. Figure 9 illustrates the first three diagnostically significant LVs (i.e., LV1 , LV2 and LV3) accounting for 80% of the total variation in the AF spectral dataset revealing the diagnostically significant AF spectral features for tissue classification. In particular, Figure 9 shows the first three diagnostic significant LV loadings accounting for the largest AF spectral variance (30.5%, 13.8%, 16.7%) and generally represent line shape of AF spectra and the peaks in the AF spectra at 874, 885, and 1015 nm correspond to Raman peaks 1302 cm^"1, 1450 cm^"1 and 2885 cm ^"1. Successive components accounted for distinctive amount of spectral variations {i.e., LV4, 5.56%; LV5, 2.22%; LV6, 3.41 %; LV7, 2.76%). Figure 10 shows the ternary plot of the NIR AF spectra cross validated prediction results. In particular, Figure 10 shows the two-dimensional ternary plot of the posterior probability belonging to benign tissue, precancer and cancer. This depicts probabilistic outcome in association with data for each tissue type, providing a three-class diagnostic model for classification. The final diagnostic category of each data point was determined by the nearest proximity of data to the diagnostic category related to the vertex of the ternary plot. The vertices in Figure 0 represent the 100 per cent posterior probability belonging to benign, precancer, or cancer colonic tissue. Table 2 as shown hereinafter summarizes the diagnostic indices for in vivo NIR AF spectra using PLS-DA together with leave-one tissue site-out, cross validation method in classifying the three different types of colonic tissue. Diagnostic sensitivities of 85.4%, 76.5% and 84.6%, specificities of 89.9%, 93.4%, and 91.4%, and accuracies of 87.1 %, 91.3% and 89.7% respectively, were achieved for differentiation between benign, precancer, and cancer colonic tissues. Figure 10 illustrates the good clusterings of the three different colonic tissue types achieved by PLS-DA algorithms, together with the leave-one tissue site-out, cross validation method. The high predictive accuracy reinforces the robustness of AF endoscopic technique according to embodiments of the present invention for in vivo diagnosis of colonic precancer at the molecular level.

Tissue type NIR AF prediction

Benign Precancer Cancer Total

(Normal + Hyperplastic)

Benign 140 10 14 164

(Normal + Hyperplastic)

Precancer 5 26 3 34

Cancer 5 5 55 65

Sensitivity (%) 85.4 76.5 84.6

Specificity (%) 89.9 93.4 91 .4

Accuracy (%) 87.1 91 .3 89.7

Table 2 - Classification results of in vivo NIR AF spectra prediction for the three colonic tissue groups using PLS-DA algorithms, together with the leave-one tissue site-out, cross validation method.

Classification of colon cancer based on diffuse reflectance spectra

To compare NIR AF spectroscopy techniques according to embodiments of the present invention with the conventional diffuse reflectance techniques for colon cancer detection and diagnosis, the classification results using broadband diffuse reflectance (DR) spectra are provided in Table 3 shown hereinafter. In an experiment according to an embodiment of the present invention, 106 DR spectra of colonic tissue were acquired from 7 paired ex vivo colonic tissues, in which 48 were from cancer tissues and 58 were from normal tissues. Figure 1 1 shows the mean DR spectra ± standard error (SE) from normal (n=58) and cancer (n=48) colonic tissues within the spectral bandwidth of 375-1000 nm. The shaded areas in tissue AF spectra represent the respective standard error. In Figure 1 1 , it should be noted that that DR spectra of normal colonic tissue are vertically shifted for better visualization. The colonic cancer tissues are characterized by higher absorption from hemoglobin and water than that of the normal tissues near 540 nm, 580 nm, and 970 nm. Hemoglobin is present in vascularized tissues and has a strong Soret band absorption near 420 nm, 540 nm, and 580 nm. Water, which is one of the main components in the human body, have a prominent absorption band at 970 nm. These absorption peaks can frequently produce valleys in tissue fluorescence spectra through reabsorption of emitted fluorescence. The higher absorption in the colonic cancer tissues compared to the normal ones could be attributed to the proliferation of cancerous cells. The cancer progression result in increased metabolic activities and higher concentration for hemoglobin and water.

In addition, PCA-LDA was employed by utilizing different region of DR spectra, including 400-700 nm, 700-1000 nm, and 400-1 OOOnm. Table 3 shown hereinafter summarizes the sensitivity and specificity for identification of cancer tissues from normal tissues within different wavelength regions, respectively. The wavelength within 400-1000 nm, which also covers the NIR region, yields the best diagnostic sensitivities of 87.5% and specificity of 94.8%. This is because the whole region includes spectral features of both water and hemoglobin for diagnosis. Thus, the broadband DR spectra has been demonstrated to be capable of detection and diagnosis of colon cancer based on alteration of tissue absorption. According to an embodiment of the present invention, the fiber probe 26 comprises changeable filters for enabling the combination of DF spectroscopy and AF spectroscopy for improving the diagnosis of colon cancer during endoscopic examination.

Wavelength region Sensitivity Specificity

400-700 nm 85.4% (41 /48) 81 .3% (47/58) 700-1000 nm 87.5% (42/48) 93.1 % (54/58)

400-1000 nm 87.5% (42/48) 94.8% (55/58)

Table 3 - Comparison of diagnostic results of DR spectra prediction within different wavelength regions using PLS-DA algorithms. As demonstrated in the experiments described hereinbefore, the NIR AF spectroscopy system 20 is capable of acquiring for high quality in vivo NIR AF spectra from normal, hyperplastic and adenomatous polyp colonic tissue under WLR imaging guidance during clinical colonoscopy. The in vivo NIR AF spectra can also be obtained quickly and can be as quick as within 2 seconds. As tissue AF spectroscopy technology possesses the ability of providing wealthy information about tissue morphology structure and biochemical conformation associated with diseases transformation, the NIR AF spectroscopy according to embodiments of the present invention can be an effective diagnostic approach to complement conventional white-light colonoscopy for improving classification of subtype of polyp without obvious macroscopic difference.

Significant differences in in vivo NIR AF spectra were observed among normal, hyperplastic polyp, adenomatous polyp and cancer colonic tissue in both spectral intensity and shape. As biological tissue is very complex, there are many factors related to morphological and biomolecular changes taking part in a myriad of biochemical processes to influence disease concurrently. Multivariate statistical techniques (e.g., PCA-LDA and PLS-DA) according to embodiments of the present invention can fully use the entire tissue spectra and elucidate diagnostic information for classification of multiple pathologies for cancer diagnosis and detection. As demonstrated in the experiments described hereinbefore, good classification among different types of colonic tissue can be achieved using PCA-LDA and PLS-DA diagnostic algorithms, indicating the capability of NIR AF to be a clinically complement to conventional WLR endoscopy for the rapid, non-invasive, in vivo identification of precancer (adenomatous polyp) during clinical colonoscopic examination for example. According to another embodiment of the present invention, there is provided an exemplary system 100 for NIR AF measurement of a subject as depicted in Figure 12. For example, the subject can be one of various types of tissue such as a tissue of an organ (e.g., colon). In the embodiment, there is provided an exemplary NIR AF imaging system 100 for NIR AF measurements of tissue for the detection and/or diagnosis of cancer in various organs (e.g., colon). For the sake of clarity, the system 100 will be described with respect to the ex vivo detection and/or diagnosis of colon cancer. However, it will be apparent to persons skilled in the art that the present invention is not limited to only the ex vivo detection/diagnosis of colonic cancer and other types of detection/diagnosis technique, including in vivo, and other types of cancers regardless of in vivo or ex vivo detection/diagnosis fall within the scope of the present invention.

Figure 12 depicts a schematic diagram of the exemplary NIR AF imaging system 100 according to the embodiment. The imaging system 100 comprises a light source 102 for generating excitation light, a fibre 104 as a medium for transmitting the excitation light, a first lens system 106 for transmitting the excitation light, a dichroic mirror 108 for reflecting the excitation light to the tissue 1 10, and a processing section 111 comprising a second lens system 112 for transmitting the induced AF emission from the tissue 0 and an image sensor 13 for generating an image based on the received AF emission. In the embodiment, the light source 102 is a laser diode operable to emit a NIR excitation light, preferably at 785 nm. For example, the light source 102 has a maximum output of 300 mW. The fibre 104 may be a commercially available 200μιτι fibre or any other suitable fibre apparent to persons skilled in the art for transmitting the NIR excitation light. In the embodiment, the first lens system 106 comprises a collimator 114 coupled with a narrow bandpass filter 116 for removing interference of fiber background fluorescence and laser noise. Preferably, the first lens system 106 comprises a polarizer 118 for polarising the excitation light such that polarization technique is integrated into the system 100. The first lens system 106 may also further comprises a neutral density filter 120. The excitation light transmitted from the first lens system 106 is then reflected by the dichroic mirror 108 and shined onto the tissue 1 10. The induced AF emission from the tissue 110 passes through the dichroic mirror 108 and the second lens system

112 before being collected by the image sensor 1 13. For example, the image sensor

113 can be a charge-coupled device (CCD) camera such as a NIR-optimized back- illuminated, deep-depletion CCD camera. In the embodiment, the second lens system 112 comprises an analyzer 122 for adjusting the polarization angle, a long- pass filter 124 and a lens 126 for converging the light onto the image sensor 113. For example, the long-pass filter is a commercially available 850nm long-pass filter. It will be apparent to persons skilled in the art that the above described arrangement of the lens is provided only as an example and the arrangement can be modified as appropriate to achieve similar or equivalent functions.

In an embodiment, the processing section 111 further comprises a computer 128 for performing various operations as described herein. For example, the computer 128 may comprise a processor 130 for processing a set of executable instructions and a storage medium 132 having stored therein a computer program comprising a set of executable instructions for controlling the operations of the computer 128 when executed. The computer 128 may further comprise a display 134 for displaying various information received and processed by the computer 128, such as AF images. According to a further embodiment of the present invention, there is provided a system 150 for NIR AF measurement of a subject as depicted in Figure 13. Similarly as described above, for example, the subject can be one of various types of tissue such as a tissue of an organ (e.g., colon). In the embodiment, there is provided an integrated NIR AF and diffuse reflectance (DR) imaging system 150 for tissue measurement. The NIR AF and DR imaging system 150 comprises the NIR AF imaging system 100 as described hereinabove and further comprises a second light source 152 and a third lens system 154. The third lens system 154 comprises a collimator 156, a beam expander 158 and a polarizer 160. For example, the light source can be a tungsten halogen light source and the light is coupled into a fiber (e.g., a 200 μηι fiber) and passes through the third lens system 154 to illuminate the tissue directly. The NIR diffuse reflectance photons from the tissue 110 are collected by the imagine sensor 113 after passing through the dichroic mirror 108 and the second lens system 1 12.

As illustrated Figure 13, to acquire AF and DR images under different polarization conditions, two linear polarizers 1 18, 160 are placed along the AF and DR illumination light paths respectively. The parallel and perpendicular polarized AF/DR images can be acquired in tandem by rotating the analyzer 122 positioned in front of the imagine sensor 1 13. In an experiment according to the present invention, to explore the fluorescence polarization technique for colon cancer detection and diagnosis, a series of polarized AF and DR images under different polarization angle by rotating the analyzer 122 in front of the CCD camera 1 13. For each image, a homogenous area (e.g., about 2 x 2 mm²) on colonic tissue is selected to estimate the average intensity for both normal and cancer tissue respectively. Figure 14 shows polar diagrams to illustrate the NIR AF and DR signals as a function of polarization for a total rotation of 360°. In particular, the Polar diagrams represent a full sample rotation of every 20° for six paired colonic tissues, (a) NIR AF imaging for normal tissues, (b) NIR AF imaging for paired cancer tissues, (c) NIR DR imaging for normal tissues, and ^'(d) NIR DR imaging for cancer tissues. The results of six pairs of colonic tissues are drawn in different colors in Figure 14.

From Figure 14, it can be observed that the trend in the intensity changes of AF and DR images are consistent in all six paired colonic tissues (normal vs. cancer). The maximum intensity is under the parallel-polarized condition (0° and 180°) and the minimum intensity is under the perpendicular-polarized condition (90° and 270°). The intensity differences between parallel polarization and perpendicular polarization condition are consistent. The oriented polarized light components of parallel-polarization image (l_pai) are roughly 3% higher than perpendicular-polarization image (l_per), and the randomly polarized components of l_pai and l_per are roughly the same. Thus, it is found that parallel- polarized images have the higher intensity than the perpendicular-polarized images for NIR AF and DR imaging. Moreover, it is found that the parallel-polarized images mainly contain the information of structures on the surface or subsurface of the tissue while the perpendicular-polarized images predominantly contain information from deep areas of the tissue. Thus, parallel polarization and perpendicular polarization conditions can be selected accordingly for tissue diagnosis. Furthermore, polarization AF imaging has been found to suppress the superficial information, which includes artifacts because of uneven or irregular surface of tissues. In addition, the intensity change of colonic tissues with the function of polarization angles might also reflect the role of collagen in the colon. Collagen is the structural protein in the extracellular matrix of the colonic wall and the dominant fluorophore in the submucosa of colon tissues. Because of its hierarchical structure and fibril alignment, the optical anisotropy properties of collagen can be used to reflect the structure and alignment of the fibrils. Thus, according to embodiments of the present invention, the polarized NIR AF is utilised to provide additional diagnostic information associated with morphological changes due to the cancer progression.

In an exemplary experiment according to an embodiment of the present invention, the integrated NIR AF and DR imaging system 150 combined with polarization technique was used to obtain a set of six images for colonic tissues in tandem, i.e., NIR AF image and the corresponding NIR DR image under three different excitation light polarization conditions (i.e., non-polarization; parallel and perpendicular polarization). The system 150 acquired NIR AF images and DR images within the spectral bandwidth of 850-1100 nm, and each NIR AF image was acquired within 1 second with the 785 nm laser light irradiance of 0.15 W/cm² (which is less than the American National Standards Institute (ANSI) maximum permissible skin exposure limit set out for a 785-nm laser beam), while each NIR DR image was acquired within 0.01 second with the tungsten light and the incident optical power on sample surface is 2 mW. in the experiment, a total of 48 paired (i.e., normal vs cancer) colonic tissue specimens (average size of about 4 x 2 mm³) were collected from 48 patients (20 men and 28 women with a mean age of 62) who underwent partial colectomy or surgical resections with clinically suspicious lesions or histopathologically proven malignancies in the colon.

Immediately after surgical resections, the tissue specimens are immersed in physiological saline solution and sent for NIR AF/DR imaging measurements using the system 150. The paired tissue specimens from each patient were placed on a standard glass slide (e.g., about 26x76x1.2 mm³). The cancer tissue was placed at the upper part of the slide while the normal one was placed at bottom part of the slide for NIR imaging measurements. After the NIR imaging acquisitions, the tissue specimens were fixed in 10% formalin solution and then submitted back to the hospital for histopathological examinations. The histopathogical examinations confirmed that 48 tissue specimens were normal, and 48 tissue specimens were cancer (moderately differentiated adenocarcinoma). NIR AF and PR Imaging

Using the integrated NIR AF/DR imaging system 150 combined with the polarization technique, NIR AF/DR images of 48 paired colonic tissues were successfully obtained under the three light excitation polarization conditions (i.e., non-polarization, parallel and perpendicular polarization). Figure 15 shows the representative NIR DR images and AF images of one pair of colonic tissue (normal (about 9 x 6 mm²) vs. cancer (about 6 ^χ 3.5 mm²) confirmed by histological examinations): (a) NIR DR image without polarization; (b) NIR DR image with parallel polarization, and (c) NIR DR image with perpendicular polarization; (d) NIR AF image without polarization; (e) NIR AF image with parallel polarization, and (f) NIR AF image with perpendicular polarization. Cancer tissue was located at the upper frame of images while the normal tissue was located at bottom frame of images.

It can be observed that NIR DR images under three polarization conditions as shown in Figures 15a to 15c cannot distinguish the difference between cancer and normal tissues by naked eyes. However, in contrast, the NIR AF images under three different polarization conditions as shown in Figures 15d to 15f can easily and noticeably distinguish the cancer tissue (having lower intensity) from the normal tissue (having higher intensity). In order to further quantitative analyze the intensity difference between colonic normal and cancer tissue, the intensity ratio between the normal and cancer colonic tissue was calculated. W 201

36

As shown in Figure 16, a homogenous area (about 2 x 2 mm²) is selected to estimate the average intensity of both normal and cancer tissue respectively. Due to the size variation of colonic tissues from different patients, the size and the location of the homogenous area for calculating the average intensity are different individually. Figure 16 shows the average intensity for both normal and cancer tissue in the NIR DR and NIR AF images. The intensity ratio of cancer to normal tissue is calculated to be 1.19 for the NIR DR image in Figure 16(a) and 2.7 for the NIR AF image in Figure 16(b). The experimental results demonstrate that 37 colon cancer tissues have lower AF intensity than the corresponding normal tissues. In other word, for 77.1% of paired colonic tissue, the intensity ratio of normal to cancer tissues is higher than 1. The mean value of the intensity ratios from all NIR AF imaging results is 1.2 (p=3.5E-4). Whereas, the mean intensity ratio for all NIR DR images is 0.98 and not statistical significant for classification the cancer from normal colonic tissues. Accordingly, the integrated NIR AF and DR imaging system 150 (or equivalentiy the NIR AF imaging system 100) has been demonstrated to successfully acquire NIR AF images of colonic normal and cancer tissues. The experimental results demonstrate that the colon cancer tissue can be identified from normal tissues with significant lower AF intensity in NIR AF images under different polarization conditions. It is found that this is because AF imaging takes advantage of the intrinsic fluorophores that are associated with the structural matrix of tissues or are involved in cellular metabolic processes to assess both the structural and the biochemical progression of colon cancer. The differences of fluorescence intensity between normal and cancer tissue could be attributed to the changes of tissue optical properties of cancer tissue in the colon. For example, the proliferation of neoplastic cells caused the thickening of mucosal tissue in cancer tissue, which could significantly attenuate the excitation light penetration and also obscure the tissue AF emission from the tissue, resulting in an overall decrease of NIR AF intensity from cancer tissue as compared to normal colonic tissue. In addition, the changes in concentrations of endogenous fluorophores such as NADH, collagen, flavins, porphyrin, etc., in tissue associated with malignant transformation may also attribute to the differences in the NIR AF emission between normal and cancer colonic tissue. Accordingly, the experimental results demonstrate that the capability of NIR AF imaging according to embodiments of the present invention to complement the conventional white light diffuse reflectance imaging for improving the diagnosis of cancer such as colon cancer.

Polarization Fluorescence Imaging

As described hereinbefore, polarization technique was integrated into the NIR AF imaging system 100 to examine the diagnostic ability of polarization fluorescence imaging. Figure 17 shows the representative NIR AF images of the paired colonic specimens under the different polarization conditions in pseudo-color: (a) NIR AF image without polarization; (b) NIR AF image with parallel polarization, and (c) NIR AF image with perpendicular polarization. Figure 17(d) shows the intensity profiles along the vertical line as indicated on the NIR AF images in Figures 17(a) to (c) respectively. In Figure 1 (d), the AF intensity profiles under the parallel and perpendicular polarizations have been magnified by 4 times in Fig. 3.4 d for better visualization.

It can be observed from Figure 17 that the cancer tissue shows a relatively lower NIR AF intensity than the normal tissue. Since the heterogeneous structure of tissue, the line was drawn to cross the homogenous area which located at the center of field of view. Then the ratio of average intensity of last 20 points to the first 20 points in intensity profile was calculated. For instance, NIR AF emission arising from cancer tissue reduces by 2.0-, 2.2-, and 2.4- fold, respectively, in intensity as compared to the normal tissue under the non-polarization, parallel and perpendicular polarization conditions. From Figure 17(d), it can be observed that the contrast of NIR AF emission between normal and cancer colonic tissue is enhanced in the parallel- and perpendicular polarization condition compared to the non-polarization condition.

To compare the diagnostic performance of NIR AF imaging under different polarization conditions, NIR AF intensities are as calculated from the homogenous area on the normal and cancer NIR AF images, respectively (see Figure 16). Figures 18 (a) to (c) show the pair-wise comparison of NIR AF intensities of all 48 paired (normal vs cancer) colonic tissues under the three polarization conditions (i.e., (a) non-polarization, (b) parallel and (c) perpendicular polarization). It can be seen that NIR AF intensities of cancer tissue are significantly lower than those of normal tissue with the p-values of 3.5E-4, 3.2E-8 and 5.8E-9, respectively, under the non-polarization, parallel and perpendicular polarization light excitation conditions (paired 2-sided Student's Mest, n=48). Based on the intensity ratio of normal to cancer tissue (Inorma/lcancer), the diagnostic accuracies of 79.2% (38/48), 91.7% (44/48) and 93.8% (45/48), respectively, can be achieved by using the NIR AF imaging under the non-polarization, parallel and perpendicular polarization light excitation. Hence, the polarized NIR AF imaging was able to enhance the contrast between normal and cancer colonic tissue with a higher diagnostic accuracy (of ~ 92-94%) compared to the non-polarized AF imaging (accuracy of -79%).

To examine the possible reason that the polarized NIR AF imaging performs better than the non-polarized NIR AF imaging technique for detection of colon cancer, NIR AF polarization properties of normal and cancer colonic tissue were studied by calculating the polarization ratio values (Ratio=(l_Par-lper)/(lpar⁺lper)) in NIR AF images, where Ipar and Iper are the NIR AF intensities under the parallel and perpendicular polarization conditions (see Figure 19(a)). It was observed that the polarization ratio values of cancer colonic tissue are in the range of 0.0001 to 0.01 , while the polarization ratio values of normal tissue are much higher, ranging from 0.012 to 0.075 as shown in Figure 19(b). Figure 19(b) shows the polarized ratio values along the line across normal and cancer colonic tissue as indicated on the polarization ratio image in Figure 19(a). Similar to the polarized DR imaging, the parallel-polarized NIR AF imaging contains the information mainly from the surface or subsurface of the tissue, whereas the perpendicular-polarized NIR AF imaging reveals the information predominantly from deep areas of the tissue. A strong linear polarization of cancer tissue reflects that much more multiple light scatterings may occur in deeper regions of tissue due to the disorganized structures of tissue in colonic adenocarcinoma, resulting in a larger perpendicular polarized light component as compared to the normal tissue. Accordingly, it has been demonstrated that the polarized NIR AF imaging technique according to embodiments of the present invention has the ability to selectively probe the AF light photons that arise from the subsurface or deep areas of tissue for improving cancer diagnosis and characterization.

Ratio imaging of NIR DR/NIR AF

Tissue NIR AF image acquired depends on not only the tissue status (e.g., tissue surface structures, physiology or histopathology status, etc.), but also the measurement conditions (e.g., light excitation-tissue-collection configurations with respect to the tissue surface, illumination light power variation, etc.). According to an embodiment of the present invention, to eliminate the geometrical effects on NIR AF measurements such as the variations of the light source-tissue distance, the varying angles for the incident light and tissue fluorescence collections, and the irregularities of the tissue surface which are naturally encountered in practical tissue fluorescence imaging, the NIR DR images from normal and cancer tissue serving as background image were measure to normalize the NIR AF image for correcting the artifacts of NIR AF image non-uniformity.

Figures 20(a) to 20(c) show the NIR DR images of normal and cancer colonic tissue acquired using a broadband light source under the three polarization conditions: (a) non-polarization, (b) parallel polarization, (c) perpendicular polarization. Figure 20(d) illustrates the intensity profiles along the line as indicated on the NIR DR images in Figures 20(a) to 20(c). In Figure 20(d), it should be noted that the AF intensity profiles under the parallel and perpendicular polarizations have been magnified by 12 times in Fig. 3.7 d for better visualization. From Figure 20, it can be observed that the NIR DR images shown gave no significance differences in NIR DR intensities between normal and cancer tissue (p-values of 0.20, 0.28 and 0.17, respectively for the non-polarization, parallel and perpendicular polarization conditions, paired 2-sided Student's t-test, n=48). When normalize the NIR DR images (Figures 20(a) to 20(c)) to the corresponding NIR AF images (Figures 17(a) to 17 (c)), much enhanced differences in NIR ratio imaging between normal and cancer tissue can be observed clearly in Figures 21 (a) to (c) (with W

40

the p-values of 5.0E-5, 2.5E-9 and 7.8E-10, respectively under the non-polarization, parallel and perpendicular polarization conditions (paired 2-sided Student's f-test, n=48)). Figures 21 (a) to (c) depicts the ratio imaging of the NIR DR image to the NIR AF image under different polarization conditions: (a) noh-polarization, (b) parallel polarization, (c) perpendicular polarization. Figure 21 (d) depicts a comparison of ratio intensity profiles along the lines as indicated on the ratio images in Figures 21 (a) to (c). In Figure 21 (d), it should be noted that the ratio intensity profiles under parallel and perpendicular polarization have been magnified by 3 times for better visualization. From Figure 21(d), it can be observed that the NIR DR/NIR AF ratio values of cancer tissue can be ~2.8-fold larger than those of normal tissue. The diagnostic accuracies of 83.3% (40/48), 93.8% (45/48) and 95.8% (46/48), respectively, can also be achieved by using the NIR DR/NIR AF ratio imaging under the non-polarization, parallel and perpendicular polarization conditions. Therefore, with the ability of correcting the geometrical effects on NIR AF measurements, the NIR DR/NIR AF ratio imaging technique according to embodiments of the present invention can further improve the diagnostic accuracy (of ~94 to 96%) for colon cancer detection. Since the white light source was shined on the tissue with an angle, the intensity gradient can be observed from left to right (see Figures 20 and 21). In an embodiment, the non-uniform illumination in DR image is corrected to optimise the system 150 by coupling the white light from the second light source152 into the same light path as laser excitation from the first light source 102. In a further embodiment, the system 150 is integrated into conventional colonoscopy system in order to enable the NIR DR/NIR AF ratio imaging technique to be used for assisting in delineating the margins of tumors for surgical operation.

The diagnostic accuracies and p-values for discrimination between normal and cancer colonic tissues using different NIR imaging methods have been summarized in Table 4 shown hereinafter. The diagnostic ability of the integrated NIR AF imaging system is evaluated - in total nine imaging methods: NIR AF image and the NIR DR image under non-polarization, parallel-polarization and perpendicular polarization conditions and corresponding ratio image NIR DR/NIR AF. W

41

Non- Parallel Perpendicular

Imaging modalities

polarization polarization polarization

35.4%(17/48) 43.8%(21/48) 31.2%(15/48)

NIR DR

p=0.20 p=0.28 p=0.07

77.1 %(37/48) 91.6%(44/48) 93.8%(45/48)

NIR AFI

p=3.5E-4 p=5.8E-09 p=3.2E-08

Ratio imaging of 83.3% (40/48) 93.8% (45/48) 95.8% (46/48)

NIR DR to NIR AF

p=5.0E-4 p=1.3E-08 p=6.2E-08

Table 4 - Comparison of diagnostic accuracy and p-value (paired 2-sided Student's t-test) of different NIR imaging modalities (i.e., NIR AF imaging and NIR DR image under non-, parallel- and perpendicular polarization, and the ratio imaging of NIR DR to NIR AF for detection of colon cancer.

First, the NIR AF imaging takes advantage of changes of endogenous fluorophores and structure of tissue layer to investigate the differences of NIR AF emission between normal and cancer colon tissue without using chemical dyes. The distinctive difference of intensity in NIR AF images between normal and cancer tissue was employed to identify the cancer from normal colonic tissues. Second, the polarization fluorescence technique with the ability to select the fluorescence light that backscatters from the superficial tissues or deeper region of the tissue was developed to improve the diagnostic ability of NIR AF imaging. When the lesions invaded into different layer of the tissue, the polarization NIR AF imaging can yield images whose contrast is in the region of interest. As a result, the NIR AF imaging under polarized conditions gives a higher diagnostic accuracy (~ 92-94%) than the non-polarized AF imaging (~79%), while the NIR DR imaging did not reach significance on statistical testing (P=0.20, 0.28 and 0.07). Finally, the ratio imaging of NIR DR to NIR AF further improves the diagnostic accuracy and achieved the best diagnostic accuracy (~96%) under perpendicular-polarization condition due to correction of the geometrical effects.

In summary, it is discovered that under the 785 nm laser excitation, NIR AF emission from colonic tissue can be detected and imaged by the sensitive NIR imaging system. Significant differences in AF intensity are observed between normal and cancer colonic tissue, indicating the feasibility of NIR AF imaging technique for detection of colon cancer. The ratio imaging of NIR DR to NIR AF under polarization condition further improves the colon cancer diagnosis and characterization. In addition, it will be apparent to persons skilled in the art that the integrated NIR AF and NIR DR imaging with polarization excitation technique according to embodiments of the present invention has the capability to be a clinically useful tool for in vivo diagnosis and detection of cancer such as colon cancer during colonoscopic examination. Figure 22 is a schematic flowchart 200 illustrating an exemplary method for near- infrared autofluorescence measurement of a subject according to an embodiment of the present invention. Step 202 of the method includes emitting an excitation light at near-infrared from a light source. Step 204 includes transmitting the excitation light to the subject. Step 206 includes processing an autofluorescence signal from the subject at near-infrared resulting from the excitation light impinging upon the subject so as to analyze the subject.

It will be apparent to persons skilled in the art that the above-described method and various processing of data performed to produce various results/outcomes as described herein can be implemented in one or more computer programs. For example, just to name a few, the various processing of data includes multivariate analysis (such as PCA- LDA, PLS-DA and/or ANOVA algorithms), classification of types of tissue, and ratio imaging of NIR DR/NIR AF as described hereinabove. The computer program comprises a set of executable instructions, which when executed by a computer processor 43, 130, controls a computer 42, 128 to perform the above method and/or the various processing of data. For example, the computer program product may be embodied or stored in a date storage or computer readable medium 44, 132 of a computer 42, 128 such as an internal memory device or a data storage or computer readable medium which can be interfaced with the computer 42, 128 such as an optical disk or a portable memory device. It will be appreciated by a person skilled in the art that numerous variations and/or modifications may be made to the present 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 to be illustrative and not restrictive.

Claims

1. A system for near-infrared autofluorescence measurement of a subject, the system comprising:

a light source configured to emit an excitation light at near-infrared;

2. The system according to claim 1 , wherein the processing section comprises a spectrometer for detecting the autofluorescence signal at near-infrared to analyze the subject.

3. The system according to claim 2, wherein the medium is a fiber probe comprising a first fiber for delivering the excitation light and a plurality of second fibers for collecting the autofluorescence signal of the subject.

4. The system according to claim 3, wherein the plurality of second fibers is arranged so as to surround the first fiber.

5. The system according to claim 3 or 4, wherein the fibre probe has a first end portion located towards the spectrometer and a second end portion located towards the subject in use, and wherein the fiber probe comprises a first filter section at the first end portion and a second filter section at the second end portion.

6. The system according to claim 5, wherein the first filter section comprises a first filter coupled to the first fiber for suppressing laser noise and a second filter coupled to the plurality of second fibers for suppressing reflected excitation light.

7. The system according to claim 5 or 6, wherein the second filter section comprises a first filter coupled to the first fiber for suppressing noise generated in the first fiber and a second filter coupled to the plurality of second fibers for suppressing reflected excitation light while allowing the autofluorescence signal to be collected by the plurality of second fibers.

8. The system according to any one of claims 3 to 7, wherein the fiber probe is configured to be insertable through an instrument channel of an endoscope for autofluorescence measurement of the subject under the guidance of the endoscope.

9. The system according to claim 8, wherein the endoscope is part of a wide- field endoscopic system for providing white-light reflecting imaging.

10. The system according to claim 9, wherein the system combined with the wide-field endoscopic system constitutes an integrated near-infrared autofluorescence spectroscopy and imaging system.

11. The system according to any one of claims 2 to 10, wherein the subject is a tissue, and the system is operable to analyze the tissue based on the autofluorescence signal and to detect the presence of cancer.

12. The system according to claim 11 , wherein the system is further operable to classify the tissue into a category of tissue including normal, hyperplastic polyp, adenomatous polyp and cancer.

13. The system according to any one of claims 2 to 12, wherein the system further comprises a processor for analysing an autofluorescence spectrum received from the spectrometer based on one or more multivariate statistical techniques.

14. The system according to claim 13, wherein the processor is operable to process the autofluorescence spectrum under the control of a program comprising a set of executable instructions for performing the one or more multivariate statistical techniques.

15. The system according to claim 13 or 14, wherein the one or more multivariate statistical techniques comprise at least one of principal component analysis and linear discriminant analysis.

16. The system according to claim 15, when dependent from claim 12, wherein the classification of the type of tissue is determined based on the principal component analysis and linear discriminant analysis.

17. The system according to claim 1 , wherein the processing section comprises an image sensor for generating an image based on the autofluorescence signal.

18. The system according to claim 17, wherein the medium comprises a series of lens.

19. The system according to claim 18, wherein the series of lens comprises a first lens system for transmitting the excitation light and a dichroic mirror for reflecting the excitation light to the subject.

20. The system according to claim 19, wherein the first lens system comprises a collimator and a narrow band-pass filter.

21. The system according to claim 20, wherein the processing section further comprises a second lens system for receiving the autofluorescence signal from the subject at near-infrared resulting from the excitation light impinging upon the subject.

22. The system according to claim 21 , wherein the image sensor is operable to receive the autofluorescence signal and generate a near-infrared autofluorescence image based on the received autofluorscence signal.

23. The system according to claim 22, wherein the subject is a tissue and the near-infrared autofluorescence image generated by the image sensor enables the 47

visual detection of a cancer portion based on differences in intensity level between normal and cancer portions.

24. The system according to claim 23, wherein the first lens system further comprises a polarizer for polarizing the excitation light so as to enhance the differences in intensity level between normal and cancer portions in the near- infrared autofluorescence image generated by the image sensor.

25. The system according to claim 24, wherein the excitation light is perpendicularly polarized.

26. The system according to claim 24, wherein the excitation light is horizontally polarized.

27. The system according to any one of claims 17 to 26, wherein the system further comprises a second light source and a third lens system for diffuse reflectance imaging.

28. The system according to claim 27, wherein the second light source emits a light to illuminate the subject through the third lens system, and near-infrared diffuse reflectance signal from the subject resulting from the light impinging upon the subject are collected by the imagine sensor after passing through the second lens system.

29. The system according to claim 28, wherein the processing section further comprises a processor for receiving a near-infrared diffuse reflectance signal and a corresponding near-infrared autofluorescence signal from the subject, and ratio imaging the near-infrared diffuse reflectance signal to the corresponding near- infrared autofluorescence signal.

30. The system according to claim 29, wherein the subject is a tissue and said ratio imaging produces an image whereby the differences in intensity level between normal and cancer portions of the tissue in the image is enhanced.

31. The system according to any one of claims 28 to 30, wherein the third lens system further comprises a polarizer for polarising the light from the second light source.

32. The system according to any one of claims 31 , wherein the second lens system further comprises an analyzer such that polarized autofluorescence and diffuse reflectance images can be acquired in tandem through rotation of the analyzer.

32. The system according to any one of claims 7 to 32, wherein the system is coupled to an endoscope for in vivo autofluorescence measurement of the subject.

33. A method for near-infrared autofluorescence measurement of a subject, the method comprising:

emitting an excitation light at near-infrared from a light source;

transmitting the excitation light to the subject; and

34. The method according to claim 33, said processing comprises detecting the autofluorescence signal at near-infrared using a spectrometer to analyze the subject.

35. The method according to claim 34, wherein said transmitting comprises transmitting the excitation light to the subject through a fiber probe, and wherein the fiber probe comprises a first fiber for transmitting the excitation light and a plurality of second fibers for collecting the autofluorescence signal of the subject.

36. The method according to claim 35, wherein said transmitting further comprising filtering the excitation light for suppressing laser noise

37. The method according to claim 35 or 36, wherein said collecting further comprises filtering the autofluorescence signal for suppressing reflected excitation light.

38. The method according to any one of claims 35 to 37, further comprising inserting the fiber probe through an instrument channel of an endoscope for autofluorescence measurement of the subject under the guidance of the endoscope.

39. The method according to claim 38, wherein the endoscope is part of a wide- field endoscopic system for providing white-light reflecting imaging.

40. The method according to claim 39, wherein the method combined with said white-light reflecting imaging constitutes an integrated near-infrared autofluorescence spectroscopy and imaging method.

41. The method according to any one of claims 34 to _.40, wherein the subject is a tissue, and said processing comprises analyzing the tissue based on the autofluorescence signal for detecting the presence of cancer.

42. The method according to claim 41 , wherein said processing further comprises classifying the tissue into a category of tissue including normal, hyperplastic polyp, adenomatous polyp and cancer.

43. The method according to any one of claims 34 to 42, wherein said processing comprises analysing an autofluorescence spectrum received from the spectrometer based on one or more multivariate statistical techniques.

44. The method according to claim 43, wherein said processing is under the control of a program comprising a set of executable instructions for performing the one or more multivariate statistical techniques.

45. The method according to claim 43 or 44, wherein the one or more multivariate statistical techniques comprise at least one of principal component analysis and linear discriminant analysis.

46. The method according to claim 45, when dependent from claim 42, wherein said classifying is determined based on the principal component analysis and linear discriminant analysis.

47. The method according to claim 33, wherein said processing comprises generating an image based on the autofluorescence signal using an image sensor.

48. The method according to claim 47, wherein said transmitting comprises transmitting the excitation light through a series of lens.

49. The method according to claim 48, wherein the series of lens comprises a first lens system and a dichoric mirror, and said transmitting comprises transmitting the excitation light through the first lens system and reflecting the excitation light to the subject using the dichroic mirror.

50. The method according to claim 49, wherein the first lens system comprises a collimator and a narrow band-pass filter.

51. The method according to claim 50, wherein said processing further comprises the processing section further comprises a second lens system for receiving the autofluorescence signal from the subject at near-infrared resulting from the excitation light impinging upon the subject.

52. The method according to claim 51 , wherein said processing further comprises said image sensor receiving the autofluorescence signal and generating a near-infrared autofluorescence image based on the received autofluorscence signal.

53. The method according to claim 52, wherein the subject is a tissue and said method further comprising visual detection of a cancer portion based on differences in intensity level between normal and cancer portions in the near-infrared autofluorescence image generated by the image sensor.

54. The method according to claim 53, further comprising polarizing the excitation light so as to enhance the differences in intensity level between normal and cancer portions in the near-infrared autofluorescence image generated by the image sensor.

55. The method according to claim 54, wherein said polarizing comprises perpendicular polarization.

56. The method according to claim 54, wherein said polarizing comprises horizontal polarization.

57. The method according to any one of claims 47 to 56, further comprising diffuse reflectance imaging using a second light source and a third lens system.

58. The method according to claim 57, wherein said diffuse reflectance imaging comprises emitting a light from the second light source to illuminate the subject through the third lens system, and collecting near-infrared diffuse reflectance signal from the subject resulting from the light impinging upon the subject by the imagine sensor after passing through the second lens system.

59. The method according to claim 58, wherein said processing further comprises receiving a near-infrared diffuse reflectance signal and a corresponding near-infrared autofluorescence signal from the subject, and ratio imaging the near- infrared diffuse reflectance signal to the corresponding near-infrared autofluorescence signal.

60. The method according to claim 59, wherein the subject is a tissue and said ratio imaging produces an image whereby the differences in intensity level between normal and cancer portions of the tissue in the image are enhanced.

61. The method according to any one of claims 58 to 60, wherein said diffuse reflectance imaging further comprises polarising the light from the second light source.

62. The method according to claim 61 , wherein said processing further comprises acquiring polarized autofluorescence and diffuse reflectance images in tandem through rotation of an analyzer in the second lens system.

63. The method according to any one of claims 47 to 62, wherein the method is coupled to an endoscope for in vivo autofluorescence measurement of the subject.

64. A system according to any one of claims 1 to 32 for non-invasive in vivo diagnosis or detection of cancer in an organ.

65. The system according to claim 64, wherein the organ is a colon.

66. A system according to any one of claims 1 to 31 for ex vivo diagnosis or detection of cancer in an organ.

67. The system according to claim 66, wherein a tissue specimen of the organ resected for diagnosis or detection of cancer.

68. The system according to claim 66 or 67, wherein the organ is a colon.

69. A method according to any one of claims 33 to 63 for non-invasive in vivo diagnosis or detection of cancer in an organ.

70. The method according to claim 69, wherein the organ is a colon.

71. A method according to any one of claims 33 to 62 for ex vivo diagnosis or detection of cancer in an organ.

72. The method according to claim 71 , wherein a tissue specimen of the organ is resected for diagnosis or detection of cancer.

73. The method according to claim 71 or 72, wherein the organ is a colon.

74. A data storage medium having stored therein a set of instructions executable by a computer processor for processing said autofiuorescence signal from the subject at near-infrared according to any one of claims 33 to 63 and 69 to 71.