Diagnosing clean margins through Raman spectroscopy in human and animal mammary tumour surgery: a short review

Abstract

Breast cancer frequency in human and other mammal female populations has worryingly increased lately. The acute necessity for taxonomy of the aetiological factors along with seeking for new diagnostic tools and therapy procedures aimed at reducing mortality have yielded in an intense research effort worldwide. Surgery is a regular method to counteract extensive development of breast cancer and prevent metastases provided that negative surgical margins are achieved. This highly technical challenge requires fast, extremely sensitive and selective discrimination between malignant and benign tissues even down to molecular level. The particular advantages of Raman spectroscopy, such as high chemical specificity, and the ability to measure raw samples and optical responses in the visible or near-infrared spectral range, have recently recommended it as a means with elevated potential in precise diagnostic in oncology surgery. This review spans mainly the latter 10 years of exceptional efforts of scientists implementing Raman spectroscopy as a nearly real-time diagnostic tool for clean margins assessment in mastectomy and lumpectomy. Although greatly contributing to medical discoveries for the wealth of humanity, animals as patients have benefitted less from advances in surgery diagnostic using Raman spectroscopy. This work also dedicates a few lines to applications of surface enhanced Raman spectroscopy in veterinary oncological surgery.

1. Background

Estimates of the cancer incidence and mortality in Europe in 2006 showed breast cancer as the second most common cause of death among women [1]. From a diagnostic approach, the yearly rate of new patients would amount to approximately 450 000.

A remarkable improvement in survival rates and more conservative treatments result when early detection of tumours occurs owing to mammography screening. X-ray mammography has a high rate of false positives [2] requiring confirmation from histopathology that makes the definitive diagnoses of breast cancer [3].

Mammary tumours are found also in other species. Cows, mares, goats, ewes and sows rarely develop mammary neoplasia, whereas dogs are the most frequently affected, with an incidence of about three times higher than in women [4]. Dogs develop naturally occurring cancers whose biology exhibits strong similarities with cancer in humans including patterns of response/resistance to conventional therapy, metastasis and recurrence [5–9]. Histology and molecular studies have revealed that many specific cancers are functionally identical in dogs and humans, e.g. osteosarcoma, mucosal melanoma, non-Hodgkin lymphoma, bladder cancer, breast cancer and others. [6,7,10,11]. Also, similarities in gene dosage between corresponding cancers in dog and human have been found through genome-wide studies [10–13].

Metastases [14,15] occur mainly through the lymphatic system, as in human mammary neoplasm. The similarity in cancer tumours developed by humans and dogs has led to the concept of ‘comparative oncology’ that would allow cross progress in diagnostic, monitoring and therapies in human and veterinary medicine [16]. Breast tumours are often subject to surgery in either species. The influence of surgical procedure on the survival time, disease-free interval and local recurrence rely on the surgical margin status [3,17,18].

2. The clean margin debate

The definition of a ‘clean’ or ‘negative’ margin has been under a continuous debate. For more than a decade various descriptions have been reported: 45.9% of radiation oncologists defined negative margins as no cancer cells at inked margins, 7.4% as no cells within 1 mm and 21.8% accepted no cells within 2 mm [19]. The increase in the number of studies correlating the width of uninvolved margins to local recurrence [20,21] has allowed surgeons to use 2 mm, or even 1 mm to define a clean margin in their approach towards breast conservation. However, no matter the margin width accepted, a general agreement exists among surgeons that the key predictor of local recurrence is the margin status [22–29].

This margin issue has been less discussed in veterinary surgery and options are more limited in comparison with human medicine [30]. To consider the closest to the human example, the dog, it is to underline that there are on the average five pairs of mammary glands in female dogs positioned on two symmetrical chains extending from axillary to inguinal zones [4,15,31]. The appropriate decision on which surgical technique to employ (lumpectomy, mammectomy, regional mastectomy, unilateral mastectomy, bilateral mastectomy or radical mastectomy) is made for each case accounting for the species on one hand and the number, size and location of the tumours on the other hand [18]. In the aggressive alternatives, e.g. radical mastectomy, the surgeon has to account thoroughly on both the status of the margins and the available surface of healthy skin to close the wound.

Since the key predictor of local recurrence in breast cancer is the margin status, an intraoperative diagnosis option should be made available whether human or animal patient were involved.

3. What option for rapid and reliable diagnostic of margins?

Various techniques have been interrogated for tumour margin assessment in both human and veterinary oncological breast surgery [32–35]. The following paragraph briefly addresses their capabilities and limitations.

Frozen section analysis has a sensitivity of 73.08% and a specificity of 98.32% in breast cancer in comparison with paraffin section analysis. Its main drawback refers to the inability to be done over the entire surface area of the tissue specimen. Like paraffin section analysis it can sample only 10–15% of the surface area [32].

Touch prep cytology can rapidly assess the entire surface area, while preserving the integrity of the specimen with a sensitivity of 75% and a specificity of 82.8% [23]. It can be employed when tumour cells are at the surface and detached. Therefore, malignant cells beneath the surface cannot be detected and that poses a limit in distinguishing close and negative margins.

Intraoperative radiography permits visualization of the margin in depth through two-dimensional X-ray projections. Its inability to identify diffuse microscopic processes results in 49% sensitivity and 73% specificity [33], especially on weakly defined tumour boundaries [34].

Radiofrequency spectroscopy performs in bulk (100 µm depth) over a circular area of 0.7 cm diameter [35]. Its reported sensitivity and specificity are 71% and 68% respectively, and the resolution is quite low. All these features make this technique rather inappropriate for margin assessment within 1–2 mm.

Optical coherence tomography (OCT) is a high-resolution microscopic optical imaging technique that yields real-time multidimensional images of subsurface tissue structure [36]. A penetration depth of 1–2 mm in breast tissue and images of the same resolution scaling as histopathology owing to the use of near-infrared (NIR) light makes it a suitable technique for intraoperative tumour margin assessment.

Hyperspectral imaging (HIS) [37–39] has been recently applied to characterize canine mammary tumours of unknown histopathology (pre-surgery) and to correlate the results with the post-surgical histopathology outcomes. HIS allows for acquiring hundreds of images in many adjacent narrow spectral bands and compressing them into a single file—the hypercube. Every pixel of the hyperspectral image corresponds to a reflectance spectrum. The combination of spectral and spatial information provides detailed insight on the differences between normal and malignant tissues.

This emerging imaging technology reports 71% sensitivity and 76% specificity. A combination of tactile and HIS sensors [39], where the tactile and spectral data were appropriately fused, has been achieved with increased sensitivity and specificity to 86% and 97%, respectively. Tactile imaging measures the elastic modulus of a tumour, and HIS detects biochemical markers. Although not dedicated to clean margins assessment in canines, these non-invasive imaging sensors may eliminate unnecessary surgeries in either dog or human patients.

It is now obvious that real-time/in vivo/intraoperative margin diagnostic are still developing topics in breast cancer surgery and new alternative techniques should be established to replace or complement existing technology.

Optics provides a wide and gentle way of interaction with biological specimens when it comes to be used in analyses. ‘Non-invasive’ and ‘non-destructive’ are attributes of most optical spectral/imaging measurement techniques. They offer the potential to provide rapid and reliable procedures for in vivo diagnoses.

4. Optics and nanotechnology in surgical margin detection

The state of the art of optical methods in biomedical applications was reviewed in 2011 [40] stressing the ‘highs’ and ‘lows’ of each method. Interestingly, those providing better specificity are spectroscopy based with typically weak signals. That drawback has been counteracted at least in part when combining nanotechnology with optics and spectroscopy.

During the last decade, much research effort has focused on developing nanoparticles that enhance scattering or absorption in the NIR, relying on the high transmittance of tissue in this spectral range. Particles such as fluorescence dyes, gold nanoshells and gold nanorods have been exploited either as diagnostic tools [40–42] or as photothermal therapy-mediated agents [41–45]. Gold nanorods have been highly considered as a diagnostic tool in view of their biocompatibility and bioconjugation with biomolecules for targeting, on their unique optical properties, on their surface plasmon resonance wavelength and, last but not least, their uncomplicated fabri­cation [46]. The potential use of gold nanoparticles as a novel contrast agent for photothermal molecular imaging of cancer has been demonstrated [47]. These features made them widely attractive as imaging contrast agents for NIR fluorescence, Raman scattering and diffusion reflection.

Although not discussed here in relation to breast cancer surgery, Fourier transform infrared (FTIR) imaging spectroscopy [48] has been recently explored as a means to identify histologic features in the normal epithelium and cervical intraepithelial neoplasia (CIN) stages I and III. Focusing primarily on human cervical cancer and then as a model for venereal cancers in dogs, this technique demonstrates its potential to provide molecular and spatial information at the single-cell level. As an in vitro diagnostic tool, FTIR spectroscopy will obviously have a great future in the clinical laboratory once the actual barriers are overcome. Those are in part related to following the outcome of the patient. In humans, it can extend over several years whereas in dogs, where lifespans are shorter, the FTIR capability as a cancer diagnostic technique could be proved much sooner.

Raman spectroscopy is based on inelastic scattering of incident monochromatic photons by chemical bonds in a sample. Since the process develops with energy and momentum conservation, a photon of lower energy is emitted (Raman photon) accompanied by a transition from one energy level to another that results in a frequency shift of the emitted photon. This method offers particular advantages in biology and medicine, very attractive to precise diagnostic purposes. A brief description of Raman spectroscopy as a diagnostic means should undoubtedly mention: (i) sensitivity to many different functional groups, with access to C=C, S–S and CS bonds that are weak in the infrared (IR); (ii) high selectivity in discriminating similar compounds and providing their fingerprints; (iii) non-invasive and non-destructive when appropriate excitation photon energy and spectra acquisition parameters are chosen; (iv) no sample preparation is required, which enhances its amenability to ex vivo and in vivo applications; (v) compatibility with aqueous solutions that enables analyses of blood, lymph and other body fluids; (vi) high spatial resolution permits single-cell and sub-cell level analysis; (vii) sensitivity to molecular orientation through polarization measurements. In medical research, Raman spectroscopy was reported to have a sensitivity of 100%, a specificity of 100%, and overall accuracy of 93% in identifying carcinomas [49].

A few optical meth­ods have been developed lately for real-time/in vivo margin detection. Some of them require contrast agents and chromophores, some others are strictly imaging techniques. However, there remains an urgent need to develop a highly specific and sensitive real-time method for margin status assessment that will reduce the risk of cancer recurrence and consequently the need for reoperation. Raman spectroscopy shows exceedingly attractive promise to this end.

5. Raman spectroscopy, a golden opportunity for surgical margin assessment?

When reviewing the progress of a method employed in a particular field one should account for both knowledge and technological advances. Instrumentation evolves along with specific challenges and with cross-disciplinary steps forward. Raman spectroscopy has been widely applied to various areas from physics to materials science, organic and inorganic chemistry, molecular biology, atmosphere research and medicine. Contributions to theoretical developments have arisen from every domain according to the needs and requirements of the moment. Information technology has become a true ‘must’ to speed up data acquisition and processing for series or personalized Raman spectrometers and made good partnership with optical component and fibre-optics technologies. The portability of Raman instruments and their deployment in medicine (especially for intraoperative diagnostic) raise new challenges related to spectrum collection time, resolution and processing. In addition, issues connected with excitation wavelength and choice of the most appropriate version of Raman scattering technique are shared with molecular biology, analytical chemistry and, last but not least, physics.

5.1. Diagnosing breast cancer through Raman spectroscopy

Although this review is focused on the last decade, it is worth mentioning pioneer work in the 1990s ([50] and references herein).

The main goal was to learn whether the molecular information accessible through Raman spectroscopy could provide a marker associated with abnormality and enable the development of a rapid and minimally invasive method of medical diagnosis of breast disease and potentially other disorders. Raman spectra were collected on human breast biopsy samples using a Ti-sapphire laser operating at 784 nm (NIR) and delivering 20–200 mW at the sample. The wavelength was chosen owing to minimal fluorescence interference and acceptable incident power level. A 640 mm single spectrograph and 1152 × 296 pixel charge-coupled device (CCD) detector were preceded by a holographic band reject filter. The spectral range was 400–1800 cm⁻¹ and the acquisition times varied from 10 to 500 s in the 20–200 mW incident power. Spectra of diseased, normal and benign tissues were compared showing smaller differences between benign and diseased specimen than those between normal and malignant ones. Fibre optics and remote probing approaches have been also demonstrated as a 6 × 1 fibre bundle (one for the laser, six for collection) inserted in a 1 mm diameter hypodermic needle penetrating in the tissue samples. The fibre was 15 m long. The authors' remark was that the possibility of rapid diagnosis with Raman spectroscopy was considered. Also, in terms of spectra interpretation with a view to diagnostic, they have underlined that if the differences in the 850–950 and 1200–1400 cm⁻¹ ranges were larger than patient to patient variations, those might form the basis of a clinical diagnosis [50].

Within about a decade, the ability of Raman spectroscopy to diagnose benign and malignant breast lesions was demonstrated ex vivo at a laboratory level [51]. ‘Ex vivo’ has kept its meaning of ‘frozen for a short while’ as it was in the 1990s.

The results, based on chemical–morphological changes that accompany breast disease, have been a step towards data acquisition in hospital settings. Both ex vivo directly (no freezing) and in vivo experiments were performed. The first in vivo Raman diagnostic has been attempted during partial mastectomy in a human patient in 2006 [52]. The authors emphasize the progress in development of a Raman optical fibre probe designed for medical applications optimized to collect high signal-to-noise ratio data from tissue in clinically relevant times (1 s). The gained features of the remote probing element within a decade (1995–2006) are remarkable: 4 m long instead of 15 m long fibre; the probe is less than 2 mm in diameter and consists of a single central excitation fibre surrounded by 15 collection fibres. Other outstanding advances are registered through the connection with a dual-filter module that rejects the intense interfering signals generated in the fibres and the sapphire ball lens employed to collimate the excitation light. In terms of excitation wavelength, the NIR light at 830 nm replaces the 784 nm [50], but this will be a variation up until the present according to the research groups' purposes and facilities. The spectral range extends from 600 to 1800 cm⁻¹ as it is expected for the NIR Raman excitation, the spectral resolution was 8 cm⁻¹ and power values of 82–125 mW at the sample have been used. To date, progress in instrumentation development has been reported for intra-cavity operation purposes, e.g. brain operation. A handheld contact fibre-optic Raman probe was developed recently [53]. The fibre-optic cables were connected to an NIR (785 nm) spectrum-stabilized laser and also to a high-speed and high-resolution CCD detector. A spectral resolution between 1.6 and 2.1 cm⁻¹ was achieved over the spectral range 381–1653 cm⁻¹. Brain cancer (glioma) could be distinguished from normal brain owing to single-point submillimetre Raman signal detection. The total acquisition time was 0.2 s. This high-performance set-up might be adapted to clean margin assessment in breast cancer and other oncological surgeries.

Getting back to the first in vivo experiment for margin assessment during partial mastectomy (2006), it is worth noticing that it occurred with all room and surgical lights turned off during the measurements [52] to prevent deleterious effects of all lights on the Raman spectra. For several samples of margins (at most six as the local protocol had required), the diagnosis duration had amounted to less than 1 min, i.e. short enough to avoid any stressful impact of the darkness on both patient and surgeons. However, the influence of potential sources of ambient light in operating theatres has been an issue of concern and several groups dealt with searching for optimal measuring conditions according to the specific surgery performed [54,55]. It was found that the measurements are mainly influenced by the surgical microscope light sources (white or blue), ambient fluorescent lights and standard operating room (OR) lights when they are pointed directly at the measured sample [54]. A simple and effective solution to overcoming the deleterious effects of theatre lighting on the Raman spectra of human lymph node tissue was reported in 2016 [55]. The Raman probe of a MiniRam II Raman spectroscopy device that incorporated a 785-nm laser was mounted inside a light eliminator and the light sources in the theatre were turned on. The resulted Raman spectra were very similar to those obtained under standard laboratory conditions.

Although successful in its yes/no result, the debut in vivo Raman diagnosis of surgical margins had not been adopted in the clinical practice. However, the Raman spectrometer deployed for that performance has become a clinically validated instrument for in vitro and ex vivo diagnosis of breast cancer and further served for an ample and accurate prospective study [56]. In that work, a Raman spectroscopic algorithm developed in vitro was validated for the diagnosis of breast cancer on a large prospective ex vivo dataset with strong similarities of the target patient population foreseen for in vivo clinical applications. In contrast with the first in vivo experiment [52], where collection and processing each occurred in 1 s, Raman spectra in the prospective study were taken with 10–30 s integration time, depending on signal intensity, and a spectral resolution of 8 cm⁻¹ was recorded. The average laser excitation power varied between 100 and 150 mW. This integration time is not above acceptable limits even for in vivo applications as long as the spectrum processing performs for 20 s at longest. The authors concluded that the negative predictive value of their diagnostic algorithm in that first prospective study was excellent. They were also aware that developments of optical techniques for medical applications would require testing of algorithms retrospectively in order to enable correct use of diagnostic information contained in any bias of the measurements. Thus, to obtain quantitative and objective diagnosis, multivariate calibration and classification models obtained from Raman spectra developed on large ‘training’ datasets will be then used for new patients.

A particularly well-documented review [2] on use of Raman spectroscopy in medical diagnosis was published in 2015 with a special focus on diagnosing breast cancer. In addition to the complex information, the authors provide the readers with a very comprehensive and enjoyable piece of work.

Setting a diagnostic on even a yes/no basis requires at least one criterion to relate effects to their causes. That could be called a marker. To draw Raman spectroscopy into the clinic for assessment of surgical margins ex vivo and in vivo would need one marker or several. This would make a good relative qualitative approach for clinical validation of the method and a step forward towards absolute quantitative diagnostic.

5.2. In quest of a marker

The marker should be a fingerprint of a particular molecule, or a combination of fingerprints. A ‘set’ sounds more appropriate because any tissue is composed of various molecules, i.e. chemical bonds whose vibrations are not always straightforwardly identified.

Most of the marker research for surgical margin assay through Raman spectroscopy has been run on ex vivo samples usually prepared as in histopathology.

A rigorous study in vitro on the potential of Raman spectroscopy as a diagnostic tool to detect biochemical changes associated with cervical cancer evolution was published in 2007 [57]. Raman spectra of histological samples of normal, CIN and invasive carcinoma tissue (prepared the usual way) were compared with each other and also evaluated against spectra acquired from unprepared polycrystalline proteins (albumin (bovine), β-galactosidase, and collagen (calf skin)), purines (adenine and guanine), pyrimidines (cytosine and thymine), nucleic acid (salmon DNA), carbohydrates (glucose, glycogen), lipids (phosphatidylcholine and phosphatidylinositol), amino acids (all 20 amino acids) and a dipeptide (arginine–lysine). The instrument used to acquire the spectra was a Labram Raman spectroscopic confocal microscope. The excitation radiation of an Ar ion laser operating at 514.5 nm gave a power of 6.5 ± 0.05 mW at the sample through a 50× objective lens focused to a spot size of about 2 µm. Spectra were collected within 30 s from the polycrystalline samples and within 150 s from tissues, over the 200 to approx. 3000 cm⁻¹ range (or similar) as the reader could learn from the peaks assignment. A classification model was developed using multivariate analysis of the spectra to discriminate normal from abnormal tissue. The authors concluded that Raman spectroscopy would be exceptionally suitable for a clinical rapid and non-invasive diagnostic tool for cervical and other cancers.

Much effort has been devoted to studying the vibrational properties of water confined in non-cancerous and cancerous human breast tissue in comparison with the properties of water confined at interfaces for DNA (single-stranded DNA and a double-stranded DNA), and phospholipids (DPPC—D-α-Phosphatidylcholine, dipalmitoyl) [58–61]. The normal tissue from a negative margin contains a significantly higher number of hydrophobic adipose cells than the cancerous tissue. The latter is mainly composed of hydrophilic proteins [62]. The first Raman ‘optical biopsy’ images of the non-cancerous and cancerous (infiltrating ductal cancer) human breast tissues are provided. The results show that water confined in the cancerous tissue exhibits a single band at 3311 cm^–1. A sensitivity of 83% for lipid markers (2800–3000 cm⁻¹) and 69% for carotenoids (1158, 1520 cm⁻¹) have been found in this series of studies extending over 7 years. The comparison between the micro-Raman spectra of normal and malignant tissues demonstrates that the normal tissue provides strong signals from carotenoids (1158 and 1520 cm⁻¹ and at 1161 and 1527 cm⁻¹). These bands are absent in the spectrum of cancerous tissue where peaks assigned to proteins are dominant [63].

A remarkable step forward on diagnosing negative margins through Raman spectroscopy has been registered with the exclusive assessment of carotenoid-like bands in spectra of healthy tissue and respectively water (OH 3311 cm⁻¹) in spectra of the cancerous tissue. To date, adoption of the markers and of the method into the clinical practice still requires enhancements and trials. The in vivo applications would need yet faster acquisition and processing of the data should those be images or spectra. This is the conclusion of the year 2015.

5.3. Plasmons act for intraoperative assay of surgical margins

A particularly interesting and also feasible way of reducing the acquisition times of spectra without diminishing the accuracy of the diagnostic would be surface enhanced Raman scattering (SERS). Metal nanostructures provide up to 10¹² signal intensity amplification due to the resonant interaction of light with the surface plasmons excited at the surface of the structure. The spectral resolution increases and the time required to collect the spectra is significantly reduced.

An enormous enhancement of the Raman scattering signal from molecules at or close to the surface can be achieved through the amplification of electromagnetic fields generated by the excitation of localized surface plasmons at metallic surfaces [64–66]. It actually relies on both electromagnetic and chemical mechanisms [67,68]. Raman spectra of a single molecule can be measured in the enhanced local fields of metal nanostructures and giant SERS amplification has been reported when aggregates of silver and gold nanoparticles were used as plasmonic structures [69]. SERS imaging using biocompatible nanoprobes has been recently reported for in vivo cancer diagnosis [70], where gold nanoprobes are injected into the body and accumulate in the tumour. This recently published work gives accurate details on the new generation of gold nanoparticles used with SERS or surface enhanced resonant Raman scattering (SER(R)S) that provide high sensitivity and specificity, and unique multiplexing capabilities. The authors present their advancements in connection with an orthotopic glioblastoma mouse model. Gold nanoparticles were injected into the tail vein of the mice and sequential tumour resections were performed on live animals. SERS imaging allowed for detection of microscopic tumours in resection beds. The actual limits of SER(R)S with injected gold nanoprobes, e.g. the penetration depth of light into the tissue, toxicity-related issues and side effects (e.g. light inflammatory response in the liver), are highlighted and mitigating solutions suggested. Although ‘nearly into the clinic’, SER(R)S imaging on injected nanoprobes needs a few steps forward from advances in Raman instrumentation and Food and Drug Administration approval of the nanoparticles themselves.

Among many other imaging candidates for cancer diagnosis in vivo with subsequent particular interest in tumour margin assay, SERS shows prominence in terms of sensitivity, specificity and speed of analysis. Injecting gold nanoprobes into the body could be claimed as ‘minimally invasive’ as long as the side effects are mild and non-toxicity is confirmed. Live mice have been used to prove the efficacy of this cancer diagnosis technique.

5.4. SERS on a blade for animal patients

An analysis of human and dog gene expression data derived from tumour and normal mammary glands indicates a significant overlap of genes [71]. The ongoing development of comparative oncology [5–8] may also benefit of the outstanding progress towards in vivo assay of surgical margins using Raman spectroscopy.

A part of our group's original results on SERS applied to diagnosis of surgical margins in dog mammary surgery is included here. The aim of our work has been to implement Raman spectroscopy in veterinary oncological surgery and check if the marker-like spectra (see §5.2) found in human research would match at least in part the findings in the dog research.

Providing a Raman solution for intraoperative margins assay still remains a challenge for both extra- and intra-cavity surgery. There are also issues related to the non-invasive character of Raman spectroscopy. As long as surgery is involved, any associated in vivo technique becomes ‘invasive’, but efforts should be put to making it ‘minimally invasive’. ‘Non-destructive’ is a must because laser radiation could damage the sample and easily change its composition if the incident power was too high, or inappropriate wavelengths were employed, or a combination of the two.

The necessity of in vivo margin assessment in dog mammectomy (radical in most cases) arises from the need to minimize the probability of re-excisions and also of conserving as large as possible skin surface to close the wound. With regard to the time scale of completing the analysis and making a decision 1 min seems acceptable.

With those considerations in mind, we have designed a SERS substrate directly attached to the surgical instrument and intend to collect the spectra with a fibre-optics portable Raman spectrometer [31] that as yet needs some enhancements in terms of resolution (aimed at better than the actual 4 cm⁻¹) and software adjustments. We have already experienced the surgical SERS substrate in a series of direct ex vivo (5 min between excision and Raman measurement) assays using a LABRAM HR 800 micro-Raman spectrometer (Horiba Jobin-Yvon). In the development of Raman protocols, the choice of wavelength is intimately linked with the choice of substrate [72]. All Raman spectra were acquired in the backscattering geometry with a 632 nm HeNe laser for excitation source. This wavelength allows for: (i) a reasonable signal intensity, because I_Raman ∼ (1/λ)⁴; (ii) smaller energetic impact with the samples than 514 and 488 nm thus avoiding damage and subsequent modifications of samples; (iii) penetration in the biological tissue beyond the subcutaneous adipose layer; (iv) laser power at the sample surfaces of about one order of magnitude smaller than for IR (e.g. 785 and 830 nm) excitation; (v) a spectral range extending between 100 and 4000 cm⁻¹ to include all lines above 2200 cm⁻¹ that cover all vibrations of biological interest such as, for instance, the 3311 cm⁻¹ line [61] marking only cancerous cells. The first experiments were performed directly on the commercially available scalpels [31] whose surfaces are not optically flat. The encouraging result was the absence of fluorescence and the disappointment arose from the long accumulation times. Silver coatings on the blades provided a spectacular enhancement of the signal [31,73].

In figure 1, the SERS surgical blade is shown together with the corresponding atomic force microscopy (AFM) image of its surface coated with cyanide free gold. The root mean square (RMS) roughness is about 3.3 nm. The SERS enhancement factor was 10⁵ and the accumulation time of the spectra was 1 min.

Figure 1.

An example of SERS surgical blade coated with cyanide free gold and the corresponding AFM image of its surface. The RMS roughness is about 3.3 nm.

Direct ex vivo (i.e. no freezing, no staining, no conservation after excision) samples of normal (healthy) tissue, breast tumours, skin and pure fat from female dog patients were investigated following regional mastectomy (removal of one mammary gland chain, the skin covering the breasts and the corresponding one side lymph nodes). The tumours were immediately divided into two parts, from which one followed the common way of cytological and histopathology analyses and the second part was sampled for direct Raman exploration. Pure adipose tissue was collected also from incisions on the median line in healthy subjects operated for spaying.

The sequences from tumour excision to fluorescence confocal microscopy imaging are displayed in figure 2. The microscopy image corresponds to the twin of one sample investigated through micro-Raman spectroscopy on a SERS scalpel. An LSM 710 NLO Zeiss confocal microscope has been employed to explore in vitro samples.

Figure 2.

The sequences from tumour excision to fluorescence confocal microscopy imaging: (a) tumour at clinical examination, (b) tumour before excision, (c) excised tumour, (d) cross section of tumour, (e) confocal fluorescence microscopy image of a sample taken from the centre of the section.

Tumour cells circulating in body fluids (e.g. lymph, blood) during surgery could also provide high input assay in negative margins, especially when excision of large surfaces of tissue is required, as for example in the dog regional or radical mastectomy. Therefore, we have analysed smears on the SERS scalpel blades used to section the samples of all kinds.

Screening of body fluids in general [72] and with SERS in particular may represent a strategic target for ex-vivo diagnostics in mammary surgery and other applications [73].

A correspondence between the microscopic image of the sample and its spectrum on an SERS scalpel is shown in figure 3. The healthy sample comes from the margin (a), the malignant sample (b) has been collected from the middle of the tumour (figure 2d) and the pure fat spectrum (c) was taken from a healthy dog that had undergone spaying. One can notice that carotenoids are found only in the margin and not in the malignant or the adipose tissue. On the contrary, the 3311 cm⁻¹ line attributed to water (OH) is seen solely in the tumour spectrum. Figure 3 is just one typical example of the many scalpel-SERS measurements (about 60/patient on 20 dogs) during the 18 months span of our work. The spectrum of pure adipose tissue (c) is quite neat and intense especially in the region between 2500 and 3020 cm⁻¹. It easily fits in the spectra of the other samples because all of them contain fat whose Raman scattering cross section is particularly large in comparison with those of other components. Judging from the spectra, we would say that the Raman spectrum of fat does not signal the malignancy/non-malignancy of a tissue at least in the dog case. Therefore, we would propose for our next step in the processing algorithm to subtract first the healthy fat spectrum from any spectrum acquired from the margins and afterwards do the analysis on DNA, carotenoids and water as the case would be.

Figure 3.

The scalpel-SERS spectra of (a) healthy tissue (margin), (b) malignant tissue (centre of the tumour) and (c) subcutaneous adipose tissue from healthy spayed female dog. The confocal microscopy images correspond to the Raman spectra in the graph: (a) margin, (b) tumour and (c) pure fat.

To round up the discussion on the last decade of progress in diagnosing clean margins through Raman spectroscopy we have presented original results on applications of SERS using surgical instruments for substrates. It is rather a 1 min ex vivo assessment but it would match quite well the needs of a veterinarian surgeon once a sufficient number of cases could be involved. The spectral markers for normal and for malignant tissues found in vitro for human subjects were also found in direct ex vivo spectra for dog patients.

6. Summary and outlook

We have reviewed the research dedicated within the last decade to diagnosis of margins in breast cancer surgery with a focus on their intraoperative assessment. It is now obvious that the status of margins is the key predictor of local recurrence in both human and animal patients. An in vivo, or at least close to it, diagnosis approach should be made clinically available, and this is a topic urgently demanding research. The non-invasive and non-destructive requirements for an in vivo assay method immediately suggest optical spectral/imaging techniques. High specificity, high sensitivity and high speed are the three ‘S’—features to define an in vivo diagnostic tool, and these must be supported by accuracy and reproducibility. Raman spectroscopy has proved exceedingly attractive from the first two S factors but it is known for low efficacy with biological samples. However, combinations among optics, nanotechnology and software developments have demonstrated large steps forward, including a first in vivo experiment in woman at the beginning of the decade and several other attempts since. More steps are needed but the end of the road is coming into sight. Using IR light (wavelengths of 785 and 830 nm) to excite the Raman effect has permitted various comparisons among the spectra taken from normal and malignant tissues in the range 400–2000 cm⁻¹. The search for one or several markers to clearly differentiate between a normal and malignant status of a margin has required an extension of the spectral window up to 4000 cm⁻¹, thus accessing contributions from fatty molecules (above 2200 cm⁻¹) and water (e.g. 3311 cm⁻¹). That extension has been possible in very recent in vitro studies using visible Raman excitation wavelengths. To attenuate fluorescence of tissues in the visible and enhance the intensity and resolution of the Raman signal, gold nanoparticle technology and plasmonic effects have been tested mainly with SERS imaging that shows prominence in terms of sensitivity, specificity and speed of analysis. Although we should think of ‘minimally invasive’ instead of ‘non-invasive’ methods, this is indeed a huge step forward to intraoperative margin assessment and from 2016 on there will be just a matter of agreement between instrumentation, software and legal entities' approvals to get SERS and its relatives translated into clinics.

The similarity in cancer tumours developed by humans and dogs has led to the concept of ‘comparative oncology’ that would allow cross progress in diagnostic, monitoring and therapies in human and veterinary medicine. Despite those findings, veterinary oncological surgery has not benefitted from the powerful Raman tool, although an analysis of human and dog gene expression data derived from tumour and normal mammary glands indicated a significant overlap of genes. We have brought into discussion some recent results of our young research work on SERS applied to margin assay in dog mammectomy showing that exploration with 632 nm Raman excitation can provide fluorescence-free spectra of fresh ex vivo samples if appropriate substrates are employed. Those spectra contain lines of carotenoids, lipids and water also found in spectra of human tissue.

Ethics

All dogs involved in the original research described in §5.4 have been current patients of the FVM-UASVM, Bucharest and ROXY VETERINARY, Magurele clinics. They have undergone therapeutic surgery with the consent of their human companions and following the usual pre-surgery protocols. The consent had been also obtained to measuring the ex vivo samples, according to the guidelines of the ethics committees of the clinics and of INOE 2000.

Authors' contributions

I.A.B. coordinated the team of veterinary surgeons at FVM-UASVM, Bucharest and supervised completion of pre-surgery and surgery protocols as well as post-surgery care and follow-ups, and revised the paper. D.T., R.M.M. and C.M. performed surgery, pre-surgery protocols, post-surgery care and follow-ups at FVM-UASVM, Bucharest, made the link between histopathology and Raman laboratories, took pictures and contributed to §§1–4 of this review. C.R. performed surgery, pre-surgery protocols, post-surgery care and follow-ups at Roxy Veterinary SRL in Magurele, provided adipose tissue from spaying, prepared the in vitro samples for confocal microscopy, contributed to §§1–4 of this review and revised the paper. M.I.R and M.T. performed Raman experiments and depositions of nanostructured films on scalpels, contributed to §§4 and 5 of this review, and revised the references. L.B. and A.P. performed profilometry and AFM measurements on the blades, contributed to §§4 and 5 of this paper, revised the paper and the references. L.O.S. contributed to software and algorithm issues, provided spectroscopic analysis issues, contributed to §§4 and 5 of this review and revised the paper. M.V.U and N.D.B.-B. revised issues related to Raman instruments, laser sources and wavelengths, spectral windows, signal intensity, signal/noise ratio and resolution, contributed to §§4 and 5 of this review and revised the paper. A.T. and R.N. performed confocal fluorescence microscopy, contributed to §§3–5 of this paper and revised the paper. C.E.A.G. conceived the SERS scalpels, coordinated the work of the groups, contributed analyses of spectra, conceived and wrote the paper.

Competing interests

We have no competing interests.

Funding

This work has been supported from PCCA 2013-UEFISCDI Romania, contract no. 20/2014.