When Sir Venkata Raman discovered the effect which was to be named after him in 1928, he had visions of shedding light on problems relating to radiation, optics and thermodynamics. Little did he know that less than a century later his discovery could be used as an innovative method to detect the disease which will affect one in three of the population – cancer.
##The Raman Effect When a photon is incident on a molecule there are three scenarios which can occur (shown in Figure 1). The case where there is no net energy transfer within the molecule is known as Rayleigh radiation, which was observed by Lord Rayleigh in the late 19th Century(4,5,6,7). In this instance, frequency of the incident photon is equal to that of the emitted photon, which is by far the most common scenario.
Figure 1: An incident photon absorbed by a molecule can undergo either Rayleigh (a), Stokes (b) or Anti-Stokes (c) radiation.
The effects which were observed by Raman are far less common. If the incident photon is of a higher frequency than the emitted photon (i.e. the photon has lost energy to the molecule) this is known as Stokes radiation. Only around 1 in 107 photons illustrate this effect.Anti-Stokes radiation occurs when the molecule is already in an excited state before absorbing the incoming photon. The incoming photon absorbs the energy from the molecule, resulting in the emitted photon being of higher frequency than that of the incident. From the Boltzmann Distribution, fewer molecules begin in an excited state therefore Anti-Stokes radiation is observed even less frequently than Stokes. It is for this reason that Raman Spectroscopy concerns itself only with Stokes radiation.</p>
What is Raman Spectroscopy?
Raman Spectroscopy is a form of emission spectroscopy – an excited molecule falls from a high energy state E2 to a lower energy state E1, emitting a photon in the process.A Raman spectrum shows a plot of scattered intensity as a function of the difference in energy between the incident photons and the scattered photons. When only considering the Raman case (i.e. Stokes and Anti-Stokes scattering) the spectrometer shows that the difference in frequency solely corresponds to the vibrational modes of the material being analysed. Every substance has a different Raman spectrum, making one compound differentiable from another.
Fluorescence spectroscopy was one of the first optical spectroscopic techniques to be used for the detection of cancer. Although it has proved successful in distinguishing between normal and abnormal tissue, there is a lack of ability to differentiate between these abnormalities. Since the early 1990s, Raman Spectroscopy has become of increasing interest to those researching medical diagnostics as a solution to this problem. Many biological molecules are Raman active –each having their own unique fingerprint. The subtle differences indicating cancer include an increased nucleus-to-cytoplasm ratio, changes in lipid and protein levels, a high metabolic activity and disordered chromatin.xi Raman spectroscopy has the capability to detect each of these subtleties independently. For this reason, Raman is being seen as the modern approach to the detection of cancers and pre-cancers.
Detection by Raman spectroscopy of cancer of the skin, breast, colon and cervix have already been illustratedxi, however the in vivo applications of this technique are by no means limited to cancer diagnosis. Raman spectroscopy has also been proven to diagnose other diseases such asatherosclerosis and vulnerable plaque.
The most commonly found setup for creating a Raman spectrum of a tissue sample is shown in Figure 2. A non-ionising laser illuminates the sample via a fibre-optic probe. This probe then filters out
Figure 2: A typical diagram of apparatus for creating Raman spectra in vivo (Adapted from“Raman Spectroscopy for Cancer Diagnosis”)xi
both the transmitted laser light and the Rayleigh scattered light, transferring only the information regarding the Raman scattering to the spectrograph. This is the preferred method when performing experiments in vivo: its non-invasive manner is particularly appealing to both researchers and (in future) patients alike. It also allows the diagnostic testing to be done in real time and with no biopsy necessary.
When using Raman spectroscopy in vivo there are several filters found within the probe. The first of these is the band-pass filter which is required to prohibit the illumination of the sample by the Raman scattered photons. The long-pass filter prevents both the Rayleigh scattering and the non-absorbed laser light from creating additional Raman scattering. Since the optical fibres are fused silica-based, they will produce their own Raman scattering. A third filter must therefore be used in order to minimize the signal produced by them.xxiii
The Signal-to-Noise Ratio
The higher the signal-to-noise ratio, the more reliable the results will be. In order to reduce the significance of the noise, the length of time over which the data is acquired (integration time) should be maximised. When considering clinical applications however, the practicalities of this must be considered. Any movement of the probe will increase the error in measurement. The longer the integration time, the more likely it is that the probe will be moved – inducing further error.
An Example – Cervical Cancer
Cervical cancer is the third most common cancer found in women. Of these cases, the human papilloma virus (HPV) is found in 99.7%.xixEarly detection of this virus is therefore crucial to the diagnosis of HPV-associated neoplasia.
The current method for detecting the virus is the Papanicolaou (Pap) smear. This involves the exfoliation of cells from the cervix and for them to then be examined under a microscope for abnormalities. Although this method is effective, a significant number of false positives/negatives have been recorded. This process is also rather labour-intensive; probing researchers to look for a new, innovative method to detect HPV.
Figure 3: Overlay of High Grade Dysplasia Raman spectrum on normal cervical tissue Raman spectrum (Top) and results of a t-test performed at each wavenumber (Bottom). Adapted fromRobichaux-Viehoeveretal.xvii
In 1998, Mahadevan-Jansen et al. developed the first fibre-optic probe to be used to detect cervical cancer in vivo.Four years later, Robichaux-Viehoever completed her PhD thesis under Mahadevan-Jansen, conducting the largest in vivo study using Raman spectroscopy for the detection of cancer up to that point in time. Her study entitled “Characterization of Raman Spectra Measured InVivo for the Detection of Cervical Dysplasia” analysed data from 66 patients (33 normal and 33 dysplasia) using the statistical method known as logistic regression. It was found that the difference in Raman spectra between high-grade dysplasia and normal cervical tissue was small yet distinct.As shown in Figure 3, several spectral regions including 1006, 1055, 1305-1330 and 1450 cm-1show statistical significance at p<0.001 and other peaks (i.e. at 1550 and 1655 cm-1) show significance at p<0.01. This significance at low p-values suggests that these results are reliable and are not simply the result of pure chance.
A problem encountered by this study was that variance from patient to patient altered the results more than had been hypothesized. Robichaux-Viehoever et al. showed that the woman’s menopausal state varied the results significantly and that the spectra would need to be modified accordingly. Kanter et al. took this into account and proposed methods in which both the menopausal state and the stage of the menstrual cycle can be corrected for.
The studies above have all focussed on the difference in intensities within the fingerprint region – the case where the Raman shift is between 800 and 1800 cm-1. This trend is not unique to the works mentioned here – most in vivo Raman spectroscopy data features only the results from this region(,,). In most forms of optical spectroscopy, the fingerprint region is the only area considered when testing organic molecules. Any peaks observed out with this region are generally ignored as they are often the result of noise.
In 2005 however, Santos et al presented a setup for the characterisation of tissue in vivo by the analysis of the high wavenumber region (2400-3800 cm-1). The fused-silica core only produces significant Raman scattering within the fingerprint region therefore the simple setup of a single, unfiltered optical fibre could be used in this case. This study also tested a range of different materials from which the optical fibre could be made, cladded and coated. The combination which gave rise to the least noise was the fused-silica core and cladding, coated with acrylate. When tested ex vivo, this gave positive results leading to the conclusion that the next step would be to repeat the experiment in vivo. Continuing on this theory, Mo etalxxiii performed a study which did just that.
Using a specially designed fibre-optic probe based on the one sampled by Santos et al, data from 92 Raman spectra (46 normal and 46 dysplasia) was collected from a total of 46 patients to be analysed in the high-wavenumber region. The addition of a ball lens allows the light to be coupled both in to and out of the fibre as shown in figure 4. Using this arrangement, several spectral differences can be seen between the normal and dysplastic tissue (Figure 5).In the region of 2800-3000 cm-1, the intensity of the Raman shift is lower for the dysplasia case whereas for higher wavenumbers (3100-3700 cm-1) the intensity is greater. Each of these regions show statistical significance at p<0.001. Multivariate Statistical Analysis was then performed to confirm this result. The study concluded there was enough evidence to be able to differentiate between normal cervical tissue and cervical tissue at different grades of dysplasia.
Figure 4: Optical layout of Raman probe.
Picture from Mo et al.
Figure 5: Comparison of mean in vivo high wavenumber Raman spectra ± one standard deviation. (Intensity measured in arbitrary units) Picture from Mo etal.xxiii
By considering the full spectrum (both the fingerprint region and high wavenumber region) even more accurate predictions can be made. The two techniques are complementary and the simultaneous Raman spectroscopy of the two could become a very promising diagnostic tool in future.
After authoring several papers on the subject, Dr Fiona Lyng and her colleagues at the Dublin Institute of Technology have recently patented a method for analysing a biological sample by Raman spectroscopy. Having won Enterprise Ireland’s “One to Watch” Award in 2011, the team are now planning to commercialise this technology, specifically in the form of a cervical cancer analyser.
Named a “High Sensitivity, High Specificity Cervical Cancer Analyser”, this product has a lot to live up to. It is advertised as being low cost, fast and easy to use – not to mention the significant decrease in the risk of human error. This is an in vitro application of Raman spectroscopy – patients will continue to attend their regular Pap smear and the exfoliated cells will be analysed with this new technology. For those studying the diagnostic applications of Raman spectroscopy this is a huge step in the right direction. It is hoped that in the near future, Raman spectra will be a common sight in many cancer diagnostics departments and humankind will be one step closer to winning the battle against cancer.