- Email: [email protected]
Fluorescence and Raman spectroscopy
Gastrointest Endoscopy Clin 13 (2003) 279 – 296
Fluorescence and Raman spectroscopy Louis-Michel Wong Kee Song, MDa, Norman E. Marcon, MDb,* a
Division of Gastroenterology and Hepatology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, USA b Division of Gastroenterology, The Centre for Therapeutic Endoscopy and Endoscopic Oncology, St. Michael’s Hospital, University of Toronto, 16-062 Cardinal Carter Wing, 30 Bond Street, Toronto, ON M5B 1W8, Canada
Patients with Barrett’s esophagus (BE) are at risk for the development of esophageal adenocarcinoma. Despite technological improvements in standard white-light endoscopy (WLE), the ability to visually detect dysplasia or early cancer within the Barrett’s segment remains limited. Current practice guidelines recommend periodic endoscopic surveillance with random biopsies in an attempt to identify these lesions at an early and potentially curative stage [1]. This approach, however, is hindered by random sampling error, low sampling yield, biopsy-associated risks, pathology-related costs, interobserver variability in pathology readings, and delay in diagnosis. In part, these limitations have generated increasing interest in the development of optical techniques, such as fluorescence spectroscopy and imaging, elastic scattering spectroscopy, Raman (inelastic scattering) spectroscopy, and optical coherence tomography, as complementary means of enhancing the detection of Barrett’s dysplasia or early adenocarcinoma at endoscopy [2,3]. The common thread in these techniques is the analysis of particular light-tissue interactions (eg, fluorescence or light scattering signals), which relate diagnostic information about the underlying microstructure or biochemical content of tissue. Because these interactions can be promptly measured by an optical probe or imaging device, these techniques offer the potential to rapidly survey a larger surface area of the Barrett’s mucosa than that afforded by conventional biopsy and to identify areas more likely to harbor dysplasia or cancer for targeted biopsies. This article focuses on the in vivo potential of fluorescence and Raman scattering techniques during endoscopic evaluation of BE.
* Corresponding author. E-mail address: [email protected] (N.E. Marcon). 1052-5157/03/$ – see front matter D 2003, Elsevier Inc. All rights reserved. doi:10.1016/S1052-5157(03)00013-8
280
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Fluorescence spectroscopy and imaging Autofluorescence versus drug-induced fluorescence Fluorescence occurs when tissue absorbs light of a shorter wavelength (eg, blue) and partially re-emits it at a longer wavelength (eg, red). Autofluorescence refers to fluorescence emitted by naturally occurring fluorescent molecules (fluorophores) comprising the tissue. Different excitation wavelengths activate different fluorophores, each of which emits over a particular range of wavelengths (Table 1). For example, tissue porphyrins emit red light (fluorescence intensity peaks/bands in the range of 630– 690 nm) when excited with violet-blue light (400 –450 nm). Because tissue contains a variety of fluorophores, its overall autofluorescence line-shape is typically broad and relatively featureless, caused by the composite contributions and overlapping intensity peaks and bands of these fluorophores. In BE, the histologic sequence of events leading to adenocarcinoma is likely accompanied by particular changes in fluorophore content or distribution, which can be detected by differences in spectral shape. At the excitation wavelength or wavelengths typically used (violet-blue range), Barrett’s epithelium generally emits a fluorescence spectrum that demonstrates a relative increase in the red-togreen fluorescence intensity ratio when compared with normal squamous mucosa. Although spectral variability exists, this ratio is seemingly accentuated as Barrett’s tissue evolves from intestinal metaplasia (IM) to dysplasia and eventually adenocarcinoma (Fig. 1). The mechanisms responsible for these observed spectral differences are manifold, including variable concentrations or spatial distributions of fluorophores, altered microvasculature, and changes in metabolic status or microstructure between tissue types [2,4]. The measurement of tissue fluorescence following administration of an exogenous fluorophore (drug-induced fluorescence) is another diagnostic option in BE. These drugs include photosensitizers used in photodynamic therapy, such as hematoporphyrin derivative, or they can be a precursor molecule, such as 5-aminoTable 1 Examples of endogenous fluorophores Fluorophore
Origin
Tryptophan Tyrosine Phenylalanine Collagen Elastin NADH Flavins Porphyrins
Protein — — Connective tissue — Respiratory chain — Heme biosynthesis by-products; bacterial fauna
Abbreviation: NADH, nicotinamide adenine dinucleotide.
Optimal excitation wavelength, nm
Peak fluorescence emission, nm
280 275 260 330 360 340 450 400 – 450
350 300 280 390 410 450 520 635, 690
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
281
Fig. 1. Average normalized fluorescence emission spectra of normal squamous esophagus, nondysplastic Barrett’s esophagus (BE), and dysplastic BE following 437-nm excitation. The spectral dip at around 580 nm is attributed to hemoglobin absorption. (From Wong Kee Song LM, Marcon NE. Novel optical diagnostic techniques for the recognition of metaplasia and dysplasia. In: Sharma P, Sampliner RE, editors. Barrett’s esophagus and esophageal adenocarcinoma. Boston: Blackwell Science; 2001. p. 123 – 36; with permission.)
levulinic acid (5-ALA), which induces the production of protoporphyrin IX, an endogenous fluorophore. The advantage of drug-induced fluorescence over autofluorescence is that the fluorescent signal generated by these drugs is typically strong and can be detected by simpler and cheaper instruments. Furthermore, wavelength selection is facilitated because the optimum excitation and emission wavelengths of these agents are known a priori, unlike in autofluorescence. Their success in BE, however, is evidently dependent on selective drug uptake and retention in dysplastic or early malignant tissues. In this regard, agents such as 5-ALA appear promising, although the optimal dose, mode of administration, time interval between drug intake and spectral measurement, and degree of selectivity in various Barrett’s tissues remain to be clearly defined. Moreover, issues involving adverse effects, extra costs, and regulatory approval present important obstacles for the application of drug-induced fluorescence in endoscopic practice. Fluorescence point spectroscopy versus fluorescence imaging There are primarily two methods for measuring tissue fluorescence during endoscopy. Fluorescence point spectroscopy involves the use of a fiberoptic probe that is introduced into the working channel of a standard endoscope and placed in
282
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Fig. 2. Schematic of a simplified fluorescence point spectroscopic system.
proximity or in contact with tissue (Fig. 2). A typical probe may consist of a central optical fiber delivering excitation light to tissue and a surrounding circular array of optical fibers collecting the emitted fluorescence. The light source can be a laser or a wavelength-filtered lamp. The collected fluorescence is separated into its component wavelengths and corresponding intensities by a spectrograph coupled to a detector and computer, and the resultant fluorescence spectrum of tissue is displayed as an intensity-versus-wavelength curve on the computer monitor. Fluorescence spectra (and other types of spectra) can be analyzed by a variety of statistical/computational means to extract relevant spectral features or regions that discriminate between tissue types. These methods may use simple empiric procedures, such as taking the intensity ratio of two fluorescence emission wavelength bands to determine a threshold that optimally differentiate between tissue categories or use sophisticated multivariate techniques of spectral analysis, such as principal component analysis, linear discriminant analysis, and artificial neural networks. Diagnostic algorithms derived from these methods can be automated, so that the collection of an unknown tissue spectrum at the time of endoscopy and input of this spectrum into the computerized algorithm provides a rapid diagnosis of tissue based on its spectral characteristics. Spectral diagnosis is then compared with histologic diagnosis to determine sensitivity and specificity of the algorithm. The entire fluorescence point spectroscopic system can be integrated into a compact, mobile, and user-friendly device. An example of this is the WavSTAT system (SpectraScience, Inc, Minneapolis, MN). Moreover, a unique and helpful feature of this spectroscopic device is the incorporation of the optical fiber within a standard biopsy forceps. The obvious advantage here is the ability to accurately perform a biopsy at the site immediately following fluorescence measurement. The WavSTAT system uses 337-nm excitation and a proprietary algorithm to analyze the spectra, with the result displayed in a few seconds on an liquid crystal display (LCD) panel. At the excitation wavelength or wavelengths typically used, the sampling depth of the light into tissue is approximately 500 mm, which effectively probes
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
283
the mucosa, the critical layer of interest. The obvious limitation with the ‘‘pointand-shoot’’ spectroscopic technique, however, is that the optical probe samples only a small volume of tissue ( 1mm3) and is directed to the Barrett’s mucosa in a somewhat random fashion, akin to conventional biopsy. Multiple fluorescence measurements (‘‘optical biopsies’’), however, can be collected in the time it takes to obtain a single pinch biopsy, thus increasing the sampling yield. To circumvent the sampling limitations imposed by point spectroscopy, much effort has been devoted toward designing fluorescence imaging systems, which are simply an extension of the spectroscopic technique whereby a two-dimensional surface map of tissue fluorescence is generated. Clearly, the advantage is the visualization of a much larger surface area of the mucosa, with fields of view approaching those obtained with a standard endoscope. The goal here is to detect dysplastic or early cancerous lesions otherwise occult to WLE. Although no new physical concepts are involved, there is a considerable increase in technological complexity and cost relative to spectroscopy, especially if multiwavelength imaging is considered. Several fluorescence imaging systems are currently undergoing evaluation in BE and other sites [4,5]. For example, the second-generation light-induced fluorescence endoscopy system (LIFE-II; Xillix Technologies Corporation, Richmond, BC, Canada) uses blue-light excitation (400 – 450 nm) and two intensified charged-coupled device (CCD) cameras to detect selected fluorescence emission bands in the green (490 –560 nm) and in the red (630 – 750 nm; Fig. 3). LIFE-II has the two CCD cameras incorporated in a detachable lightweight module, which connects to the eyepiece of a conventional fiberoptic endoscope. The two fluorescence images are digitized and combined to produce a single composite pseudocolor image in real time, where normal tissue typically appears green/cyan and abnormal tissue appears red. The system can alternate rapidly between standard WLE and endoscopic fluorescence imaging by means of a simple switch ( 4 s). Sensitivity and specificity values are determined by correlating positive (red fluorescence) and negative (green/cyan fluorescence) images with histologic diagnosis of corresponding biopsy samples. The Storz system is another fluorescence imaging device, which consists of an incoherent light source (D-Light, Storz, Tuttlingen, Germany) and a special camera (Endovision Telecam SL, Storz, Tuttlingen, Germany) specifically tailored for the detection of protoporphyrin IX (PpIX) fluorescence induced by 5-ALA [6,7]. Violet-blue light (375 –440 nm) serves as the excitation source, with a red channel displaying the PpIX fluorescence, a green channel displaying tissue autofluorescence, and an optional blue channel displaying a portion of the excitation light reflected from the tissue surface. Lesion detection can be accomplished through observation of increased red-to-green or red-to-blue contrast. Endoscopic applications Autofluorescence spectroscopy In 1995, Panjehpour et al [8] assessed an in vivo point spectroscopic device with 410-nm excitation and a spectral classification algorithm based on linear
284
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Fig. 3. Schematic of an investigational fluorescence imaging system for in vivo endoscopic applications (Xillix LIFE-II). A combination of green and red fluorescence channels produces the composite false-colored fluorescence image. Note the difference in the red-to-green fluorescence spectral ratio of normal and tumor sites, and how these differences are depicted in the composite image (See also Color Plate 5). (From Wong Kee Song LM, Marcon NE. Novel optical diagnostic techniques for the recognition of metaplasia and dysplasia. In: Sharma P, Sampliner RE, editors. Barrett’s esophagus and esophageal adenocarcinoma. Boston: Blackwell Science; 2001. p. 123 – 36; with permission.)
discriminant analysis to separate esophageal cancer from normal squamous mucosa. Using this technique, malignant tissues (squamous cell cancer and adenocarcinoma) were differentiated from normal mucosa with a sensitivity of 100% and a specificity of 98%. Although these malignant lesions were endoscopically obvious, this study demonstrated for the first time the diagnostic potential of laser-induced autofluorescence spectroscopy for the differentiation of esophageal tissues in vivo. This group then described the use of a simpler algorithm for spectral analysis, the differential normalized fluorescence (DNF) index, which could differentiate between normal esophagus and esophageal cancer with greater than 98% accuracy [9,10]. In this technique, each fluorescence spectrum was first normalized by dividing the intensity at each wavelength by the integrated area under the spectrum. This normalization procedure allowed comparisons to be made between spectra, independent of the variable intensity factor, and amplified spectral differences between normal and malignant tissues. A baseline curve was then established as the mean average of normalized fluorescence spectra from a selected number of normal tissue samples. Finally, the DNF index for a given sample was determined by the relative intensity difference between its normalized fluorescence spectrum and the baseline curve.
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
285
The DNF indices at 480 nm and 660 nm were selected because normal and malignant tissues exhibited significant spectral differences around these wavelengths. Using 410-nm excitation and the DNF algorithm, Panjehpour et al [11] subsequently reported the first in vivo study of autofluorescence spectroscopy in BE. Fluorescence spectra were collected at various spots within the Barrett’s segment (N = 36 patients), followed by conventional biopsy for corresponding histologic diagnosis. Analysis of the fluorescence spectra using the DNF index at 480 nm showed that 96% of nondysplastic Barrett’s (NDB) samples and all lowgrade dysplasia (LGD) samples were classified as benign, 90% of high-grade dysplasia (HGD) samples as premalignant, but only 28% of LGD samples with focal HGD were classified as premalignant. When both DNF indices (480 nm and 660 nm) were used concurrently, all patients with any HGD were classified as premalignant, but 7 of 23 NDB patients were classified as having premalignant lesions, accounting for a 30% false-positive rate. Hence, this technique was sensitive at detecting HGD but not LGD, based on two DNF indices. Moreover, the inability of fluorescence to reliably differentiate between reactive/inflammatory and dysplastic changes is a well-known phenomenon [4], which likely partially contributed to the significant false-positive rate observed in this report. Despite these shortcomings, this important study demonstrated the potential of autofluorescence point spectroscopy to identify endoscopically occult but clinically important lesions, such as HGD, in BE, and to enable targeted biopsies of spectroscopically suspicious sites. In addition, six to seven ‘‘optical biopsies’’ could be collected during the amount of time taken to obtain a standard pinch biopsy specimen. Preliminary findings using the WavSTAT system in BE were recently reported [12]. A total of 87 patients were studied and 326 independent optical biopsy samples were analyzed. Similar to the findings by Panjehpour et al [11], this particular spectroscopic modality demonstrated a sensitivity of 95% and a specificity of 80% in differentiating HGD from LGD and NDB samples. As an alternative to a laser, Mayinger et al [13 – 15] used a filtered light source (Storz GmbH, Tuttlingen, Germany) capable of delivering either white light or violetblue light directly by means of the endoscope, followed by collection of the emitted autofluorescence by an optical probe placed in contact with tissue. A total of 129 spectra were acquired from normal mucosa and malignant lesions in nine patients with squamous cell cancer and four with esophageal adenocarcinoma. Using a simple algorithm based on intensity ratios, a sensitivity of 97% and a specificity of 95% were obtained for the diagnosis of esophageal carcinoma [15]. Georgakoudi et al [16] recently described the use of a multimodal in vivo optical spectroscopic system in BE, combining the techniques of fluorescence, reflectance, and elastic scattering. Fluorescence spectra were acquired at 11 different excitation wavelengths (between 337 –620 nm) using a fast excitationemission matrix (EEM) instrument. Reflectance spectra were also collected from the same site using white light (350 –700 nm) coupled into the same probe. The reflectance spectra were used to correct the measured fluorescence spectra for distortions induced by changes in tissue absorption or scattering, rather than by
286
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
tissue biochemistry, because reflectance is affected by the same absorption and scattering processes. The intrinsic (undistorted) fluorescence spectra were then analyzed using multivariate techniques of principal component analysis and logistic regression. In this study, the use of intrinsic autofluorescence of tissue differentiated HGD from LGD and NDB with a sensitivity and a specificity of 100% and 97%, respectively. The differentiation of LGD and HGD from NDB, however, was less accurate, with a sensitivity of 79% and a specificity of 88%. Moreover, the combination of all three spectroscopic modalities significantly improved the diagnostic accuracy, achieving a sensitivity of 100% and a specificity of 100% for identifying HGD from NDB and LGD, and a sensitivity and specificity of 93% and 100%, respectively, for differentiating dysplastic (LGD and HGD) from nondysplastic Barrett’s samples. These remarkable results, however, were obtained in the context of a limited number of patients and samples (26 NDB, 7 LGD, 7 HGD) and require confirmation on a prospective and much larger sample size. Nevertheless, this study hinted at a multimodal optical technique as the ideal approach for optimizing diagnostic accuracy in the differentiation of Barrett’s tissue. The use of ultraviolet (UV) light for fluorescence excitation in the esophagus has also been reported [17]. Using fluorescence intensity ratio thresholds (I390nm/ I550nm) for tissue classification, the technique differentiated dysplastic/neoplastic tissue (HGD, intramucosal squamous cell cancer, and adenocarcinoma) from normal esophageal and nondysplastic Barrett’s mucosa with a sensitivity and a specificity of 86% and 95%, respectively. The major caveat with the use of UV light, however, is its mutagenic potential, which likely will hinder its application in endoscopic practice. In summary, autofluorescence point spectroscopy appears promising in the identification of Barrett’s lesions that are HGD or frankly neoplastic in nature. The technique seems limited in its ability to identify LGD lesions, or to differentiate between dysplastic and reactive changes. Despite the encouraging results discussed previously, much work remains to be done to determine the true potential of autofluorescence point spectroscopy in BE. For instance, the optimal excitation wavelength or wavelengths for autofluorescence detection of dysplasia or early cancer in BE remain to be determined; wavelength selection is currently done in a somewhat empiric fashion. Mechanistic studies using techniques such as confocal microscopy, microspectrofluorimetry, and fast EEM instruments may shed some light on the origins of the spectral differences observed between nondysplastic and dysplastic Barrett’s tissues and optimize the selection of wavelengths for fluorescence diagnostics in BE [18]. Caution is warranted, however, because the translation of findings obtained ex vivo may not apply in vivo because of changes in the biochemical state of tissue. Autofluorescence imaging Prospective trials are currently underway in assessing the utility of autofluorescence imaging in BE. A preliminary report by Haringsma and Tytgat [19] demonstrated the potential of the Xillix LIFE-II imaging system in 111 patients
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
287
with biopsy-proven BE. Lesions seen within the Barrett’s segment at either WLE or LIFE-II were judged to be dysplastic or nondysplastic, and targeted biopsy samples were taken of these abnormalities, in addition to standard four-quadrant surveillance biopsy samples at each 2-cm interval. LIFE-II detected many more HGD (20 of 24) and early cancerous (17 of 17) lesions than WLE (11 of 24 HGD and 16 of 17 cancers). Similar to autofluorescence spectroscopy, the technique was insensitive at detecting LGD. In the current authors’ experience [20], the LIFE-II system occasionally diagnosed lesions containing HGD that were not recognized by WLE, but the columnar-lined esophagus (with or without dysplasia) routinely produced a falsepositive (red) fluorescence image (Fig. 4), making identification of dysplastic lesions difficult. Moreover, the present system is configured to be used with fiberoptic endoscopes, instead of video endoscopes, which limits its practical utility. Another limitation is the contamination of fluorescence images by the presence of blood in the field of view (following biopsy), because blood is highly fluorescent. The system is currently undergoing technical modifications specifically addressing these issues. Drug-induced fluorescence spectroscopy A limited number of studies are available regarding the use of exogenous fluorophores with fluorescence spectroscopy in BE. These studies to date have mainly exploited photosensitizers used in photodynamic therapy, many of which
Fig. 4. Endoscopic fluorescence imaging of Barrett’s esophagus using the Xillix LIFE-II system. Areas of high-grade dysplasia appear brick red (A), which may be difficult to discriminate from an already red fluorescent metaplastic background (B). Areas were confirmed by histology. This figure appears in color at the front of this issue (See also Color Plate 6). (From Wong Kee Song LM, Marcon NE. Novel optical diagnostic techniques for the recognition of metaplasia and dysplasia. In: Sharma P, Sampliner RE, editors. Barrett’s esophagus and esophageal adenocarcinoma. Boston: Blackwell Science; 2001. p. 123 – 36; with permission.)
288
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
fluoresce and demonstrate some selective localization in neoplastic tissues [18]. Fluorescence generated by these drugs is less ambiguous because of distinguishable intensity peaks or bands. The photosensitizer Photofrin (QLT Therapeutics, Vancouver, Canada), for instance, contains porphyrins that emit characteristic fluorescence peaks in the red (630 and 690 nm) and that are generally retained in neoplastic tissue to a higher degree than in normal tissue. Von Holstein et al [21] initially assessed tissue fluorescence induced by low-dose Photofrin (0.35 mg/kg) in a small number of patients with BE. Following 405-nm excitation, fluorescence spectra were collected and intensity ratios calculated as the quotient of Photofrin fluorescence (at 630 nm) divided by the autofluorescence background of tissue (at 500 nm). From initial ex vivo results, a ratio of 0.55 was used as the cut-off level to distinguish malignant from normal mucosa or Barrett’s epithelium in vivo. In this study, there was a high probability that fluorescence ratios greater than 0.55 represented malignancy; ratios less than 0.55 were more difficult to categorize because of considerable overlap between normal mucosa, nondysplastic or dysplastic Barrett’s epithelium, and malignant tissue. This small-scale study was a first step in the assessment of drug-induced fluorescence in the esophagus in vivo, although Photofrin seemed limited in its ability to differentiate between grades of dysplasia in BE. There also has been interest in the use of 5-ALA in combination with fluorescence spectroscopy in BE. Although not a photosensitizer in itself, the drug is intracellularly converted to PpIX, which has a characteristic fluorescence emission in the red between 625 and 725 nm. Because of low ferrochelatase activity in tumor cells, PpIX preferentially accumulates in malignant tissues and, unlike Photofrin, undergoes rapid metabolic breakdown resulting in a low incidence of side effects and short duration of skin photosensitivity [22]. In a preliminary study, 16 patients with BE received 5-ALA (DUSA Pharmaceuticals, Inc., Wilmington, MA) 10 mg/kg orally 3 hours before endoscopy and spectroscopy. The excitation source was 400-nm light, and the normalized PpIX fluorescence intensity peak at 635 nm was quantitatively analyzed and used for tissue differentiation. A total of 98 spectra were collected. Based on whether the normalized PpIX fluorescence intensity of each spectrum was above or below 0.15, HGD was distinguished from NDB samples with sensitivity and specificity of 86% and 89%, respectively. There was significant overlap between spectra from squamous epithelium, NDB, and indefinite/LGD samples, however, suggesting that fluorescence induced by 5-ALA is more appropriate for detection of HGD and early cancerous lesions in BE [23]. Drug-induced fluorescence imaging In the upper gastrointestinal (GI) tract, initial studies of fluorescence imaging aided by exogenous fluorophores have produced encouraging results [6,7, 24 – 26]. Endlicher et al [24] assessed the feasibility of detecting Barrett’s dysplasia or early adenocarcinoma using fluorescence imaging after sensitization with 5-ALA. The study was performed on 47 patients with biopsy-proven BE, 10
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
289
with known LGD or HGD. The 5-ALA was given either orally at different dosages (5, 10, 20, or 30 mg/kg) or sprayed onto the Barrett’s mucosa (500 – 1000 mg). Fluorescence imaging using the Storz system was performed 4 to 6 hours after systemic sensitization or 1 to 2 hours after topical spray. Interim analysis revealed that 5 mg/kg of 5-ALA orally was too low for the detection of dysplasia, whereas 30 mg/kg orally achieved high sensitivity (100%) at the expense of poor specificity (27%). Subsequent patients were sensitized with 10 or 20 mg/kg of 5-ALA orally or 500 mg topically. A total of 243 biopsy specimens from PpIX red fluorescent (n = 113) and nonfluorescent (n = 130) areas were taken; 33 of these were LGD or HGD by histology. Fluorescence imaging at 4 to 6 hours following drug administration (10 – 30 mg/kg orally) showed relatively high sensitivity (80% –100%) but poor to fair specificity (27% – 56%). Fluorescence imaging 1 to 2 hours following 5-ALA topical spray (500 mg) demonstrated moderate sensitivity (60%) and specificity (69%) for the detection of Barrett’s dysplasia. In addition to the 10 patients with known dysplasia, 2 patients with early cancers and 1 patient with dysplasia were solely detected by fluorescence imaging. Adverse drug effects were limited to nausea and vomiting in 4 patients and increased liver enzymes in 2 patients. Mild skin photosensitivity was observed only in patients receiving 20 to 30 mg/kg of 5-ALA. False-positive fluorescence was induced mainly by inflammation, bile, and metaplasia. Another preliminary study using 7.5 to 10 mg/kg of 5-ALA orally and fluorescence evaluation at 3 hours following drug administration reported a sensitivity of 85% and a specificity of 70% for the detection of dysplastic and early cancerous lesions involving the upper GI tract, including BE [25]. Ortner et al [27] evaluated both fluorescence spectroscopy and imaging following topical application of 5-ALA (500 mg) in 42 patients with BE. Of 36 patients with negative WLE, 4 (11%) had dysplastic or malignant lesions identified by fluorescence spectroscopy and imaging. Fluorescence imaging achieved a sensitivity of 92% but a specificity of 32% in detecting these lesions, whereas fluorescence spectroscopy demonstrated a sensitivity of 76% and a specificity of 74%. The complementary use of both techniques appeared to improve the detection rate of these lesions and diminished the number of biopsy specimens required for optimal diagnosis. The initial findings to date suggest that fluorescence imaging aided by 5-ALA enhances the detection of dysplastic lesions, particularly HGD and early cancer, albeit with a decrease in specificity. A relatively high number of false-positive results (red fluorescence) can be attributed to confounding factors such as inflammation and even metaplasia itself, similar to autofluorescence. Regardless of the technique used (spectroscopy or imaging), it remains to be seen whether drug-induced fluorescence will outperform autofluorescence in the detection and grading of dysplasia in BE. Perhaps the development of agents by means of conjugation of a fluorophore dye to a monoclonal antibody or other dysplastic/ tumor-targeting moiety may tip the balance in favor of drug-induced fluorescence, but this principle will require much work demonstrating efficacy and safety for in vivo use.
290
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Raman spectroscopy Principles and techniques Among several optical techniques currently under investigation for in vivo endoscopic applications, Raman spectroscopy provides the most detailed information about the molecular composition of tissue. Light can be absorbed or scattered in interacting with tissue molecules. Whereas almost all of the scattered light is of the same wavelength as the incident light (elastic scattering), a small fraction undergoes so-called Raman (inelastic) scattering, in which slight shifts in energy (or wavelength), relative to the incident light, occur from an exchange of energy between the light photons and the molecular structure of tissue. These wavelength shifts correspond to specific vibrations or rotations of particular molecular bonds. A Raman spectrum is then a plot of the intensity of the Raman scattered light as a function of the wavelength shift,1 with each peak corresponding to a specific molecular state (Fig. 5). Compared with fluorescence or elastic scattering, Raman spectra of tissue are spectrally highly detailed, providing the potential for better tissue discrimination. Molecular vibrations originating from proteins, lipids, or nucleic acids exhibit distinct Raman signatures. For instance, the peak at 1002 cm 1 in Fig. 5 corresponds to the ‘‘in and out’’ vibrations of phenyl rings of aromatic amino acids. Raman scattering within tissue is more difficult to measure because the signal is much weaker relative to fluorescence or elastic scattering and is masked by a broad tissue autofluorescence background. In addition, the endoscopic application of Raman spectroscopy using fiberoptic probes has been technically challenging because of spectral contamination from background fluorescence and Raman signals generated in the fiberoptic materials. Recent technological progress in high-throughput spectrographs, high-sensitivity near-infrared detectors, and filtered fiberoptic probes, however, has enabled Raman spectroscopy instruments to be developed for clinical use [28,29]. Shim et al [30] originally demonstrated the endoscopic feasibility of obtaining near-infrared Raman spectra with reasonable signal-to-noise ratio throughout the GI tract, using a system developed in-house. The instrumentation design is similar to that shown in Fig. 2, except that the technology is more expensive because of the spectroscopic components required to extract the weak Raman signal of tissue. Although Raman scattering can be induced by UV, visible, or near-infrared light, the light source used is usually a near-infrared laser (eg, 785 nm), which minimizes excitation of tissue autofluorescence while generating Raman signals sensitive enough to be detected by existing CCD cameras.
1
The shift is usually expressed in terms of the difference in wave numbers (1/wavelength) in units of cm 1 between the incident and scattered light. Here, only the Raman signals corresponding to transfer of energy from the light to the molecules, resulting in the scattered light having less energy and so longer wavelength, are considered.
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
291
Fig. 5. Raman versus fluorescence spectrum of tissue. Note the spectrally detailed features associated with the Raman signal, which may translate into more accurate tissue diagnosis.
The use of UV light to excite Raman scattering in tissue is attractive because it results in fluorescence that occurs outside the spectrum where the Raman bands are observed, which can then be easily filtered out. Furthermore, UV excitation that corresponds to an absorption band of a molecule produces Raman bands that are significantly amplified, a process known as UV-resonance Raman. The mutagenic potential of UV light is a serious concern, however, which is likely to deter the development of UV Raman for clinical applications.
Fig. 6. Average in vivo Raman spectra of nondysplastic (n = 80), low-grade dysplastic (LGD; n = 26), and high-grade dysplastic (HGD; n = 30) Barrett’s samples obtained from 13 patients with Barrett’s esophagus. Typical tissue Raman peaks are noted (labeled). Subtle or visually unapparent (but potentially discriminatory) spectral differences are present throughout, requiring multivariate techniques to sort them out.
292
First author
Technique
lexc, nm
lem, nm
No. patients No. Spectra/tissue
Diagnostic algorithm Findings/comments
Panjehpour (1995) [8] Vo-Dinh (1995) [9]
AFS
410
430 – 716
32
LDA
AFS
410
430 – 720
48
Panjehpour (1996) [11]
AFS
410
430 – 716
36
Bourg-Heckly (2000) [17]
AFS
330
350 – 650
24
Wang (2001) [12] AFS
337
350 – 800
87
Georgakoudi (2001) [16]
AFS
11 lexc (between 337 – 620)
350 – 750
18
Mayinger (2001) [15]
AFS
375 – 478
478 – 700
13
Von Holstein (1996) [21]
DFS (Photofrin)
405
450 – 750
5
123 normal 36 SCC and ACA > 200 normal and SCC/ACA spectra 216 NDB 36 LGD 46 focal HGD 10 diffuse HGD 66 normal 116 NDB 11 dysplastic 25 SCC and ACA 266 NDB 46 LGD 14 HGD 26 NDB 7 LGD 7 HGD 57 55 17 48 17
normal SCC ACA normal/NDB ACA
DNF
DNF480nm and DNF660nm
I390nm/I550nm
Sn 100% and Sp 98% for SCC/ACA vs normal > 98% accuracy
Sn 100% and Sp 70% for HGD vs rest. Technique insensitive at detecting LGD Sn 86% and Sp 95% for HGD/ACA/SCC vs normal/NDB
Proprietary algorithm
Sn 95% and Sp 80% for HGD vs NDB/LGD
PCA and LRA
Sn 100% and Sp 97% for HGD vs NDB/LGD Sn 79% and Sp 88% for LGD/HGD vs NDB Sn 97% and Sp 95% for SCC/ACA vs normal
I500 – 549nm/ I657 – 700nm I630nm/I500nm
Sn 88% and Sp 94% for ACA vs normal/NDB
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Table 2 Summary of in vivo fluorescence and Raman studies in Barrett’s esophagus
DFS (5-ALA)
400
16
Haringsma (2001) [19]
AFI Blue light (Xillix system) (400 – 450)
490 – 560 (green) 630 – 750 (red)
Mayinger (2001) [26]
DFI (Storz system)
Violet-blue light (375 – 440)
5-ALA induced 22 (6 BE) PpIX fluorescence
Quantitative normalized PpIX fluorescence intensity peak at 635 nm Red fluorescence abnormal; green/ cyan fluorescence normal PpIX red fluorescence
Endlicher (2001) [24]
DFI (Storz system)
Violet-blue light (380 – 440)
5-ALA induced 47 PpIX fluorescence
PpIX red fluorescence
Ortner (2002) [27]
DFS and DFI
505
5-ALA induced 42 PpIX fluorescence
PpIX red fluorescence
Wong Kee Song (2002) [31]
RS
785
900 – 1800 cm
1
16 normal 23 NDB 23 indefinite/LGD 7 HGD
111
13
80 NDB 26 LGD 30 HGD
PCA and LDA
Sn 86% and Sp 89% for HGD vs NDB
Sn 90% and Sp 89% for detecting HGD/ACA Technique insensitive at detecting LGD DFI: Sn 85% and Sp 53%; WLE: Sn 25% and Sp 94% for detecting premalignant/ malignant lesions Dose ranging study Sn 80% – 100% and Sp 27% – 56% (5-ALA 10 – 30 mg/kg p.o.) and Sn 60% and Sp 69% (5-ALA 500 mg topical spray) for detecting dysplasia DFS: Sn 76% and Sp 74% DFI: Sn 92% and Sp 32% for detecting LGD/HGD/ACA Sn 79% and Sp 79% for LGD/HGD vs NDB
293
Abbreviations: ACA, adenocarcinoma in Barrett’s esophagus; AFI, autofluorescence imaging; AFS, autofluorescence spectroscopy; 5-ALA, 5-aminolevulinic acid; BE, Barrett’s esophagus; DFI, drug-induced fluorescence imaging; DFS, drug-induced fluorescence spectroscopy; DNF, differential normalized fluorescence; HGD, highgrade dysplasia; I, fluorescence intensity; LDA, linear discriminant analysis; LGD, low-grade dysplasia; LRA, logistic regression analysis; NDB, nondysplastic Barrett’s; PCA, principal component analysis; PpIX, protoporphyrin IX; RS, Raman spectroscopy; SCC, squamous cell cancer; Sn, sensitivity; Sp, specificity; WLE, white-light endoscopy; lexc, excitation wavelengths; lem, emission wavelengths.
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Brand (2000) [23]
294
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Endoscopic applications Wong Kee Song et al [31] recently reported on the in vivo potential of nearinfrared Raman spectroscopy for the identification of dysplasia in BE during endoscopy. A custom-made filtered fiberoptic probe was placed in gentle contact with the Barrett’s mucosa and a single Raman spectrum was obtained in 5 seconds, followed by biopsy of the site for corresponding histologic diagnosis. Prior studies have demonstrated the effects on the spectra of varying the pressure of the probe tip on the tissue and of the probe-tissue angle to be insignificant [30]. Thirteen patients were studied, and 136 spectra and corresponding biopsy specimens (80 IM, 26 LGD, and 30 HGD) were obtained. Raman spectra of BE demonstrated typical tissue Raman peaks, including the protein amide I band, CH2 bending mode, CH2 twisting mode, protein amide III band, and the phenyl ring breathing mode. As in other tissues, the spectral differences among the categories of Barrett’s tissue were subtle (Fig. 6), yet contained diagnostically important information. In this setting, multivariate analytic techniques are well suited, indeed may be essential, for spectral analysis rather than simply using algorithms based on empirically determined peak intensities, widths, or peak ratios. Using the statistical methods of principal component analysis and linear discriminant analysis to extract and analyze the full diagnostic content of Raman spectra, it was possible to differentiate dysplastic (LGD and HGD) from nondysplastic (IM) Barrett’s samples with a sensitivity of 79% and a specificity of 79% [31]. These initial findings were based on a custom-built Raman spectroscopic system, and it is anticipated that further improvements in Raman instrumentation or methods of spectral analysis will increase the diagnostic performance of this technique in BE. Furthermore, the technique is being refined and tested to assess its discriminating power in classifying dysplasia and early cancer (LGD versus HGD versus early adenocarcinoma). Unlike fluorescence, the prospect for a real-time endoscopic Raman imaging system remains doubtful because of the weak Raman signals and technological limitations. Endoscopic Raman diagnosis can thus be accomplished only by means of an optical probe at the moment, with sampling limitations inherent to all point spectroscopic techniques.
Summary Table 2 provides a summary of selected in vivo fluorescence and Raman studies performed in BE. Although the findings from these studies appear promising, these techniques are still under development, and it is anticipated that technological refinements will further enhance their diagnostic accuracy. Ultimately, however, large-scale prospective clinical trials are required to determine their true diagnostic potential in BE and other sites. Ideally, the instrumentation of choice would be a real-time endoscopic system that combines excellent diagnostic accuracy with wide-area sampling. In this regard, fluo-
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
295
rescence imaging is most appealing, although a variety of issues remain to be resolved, including the choice between autofluorescence versus drug-induced fluorescence and the problematic distinction between dysplastic (true positive) and confounding background metaplastic fluorescence (false positive), among others. It is also not clear whether exogenous fluorophores are necessary to achieve clinically useful sensitivity and specificity for lesion detection in BE. Point spectroscopic techniques, either fluorescence or Raman scattering, are inherently limited by the small volume of tissue (biopsy specimen size) they sample, but more detailed information can be extracted from the spectra, which may increase diagnostic accuracy. Moreover, it may be that the optimal system will be a combination of multiple optical spectroscopic or imaging techniques (multimodality approach), as suggested by Georgakoudi et al [16]. For instance, a lesion could be detected by fluorescence imaging and its dysplastic nature characterized (graded) by Raman spectroscopy. In this era of cost containment, however, the critical challenge is to demonstrate whether an increase in diagnostic accuracy merits investment in costly technology, regardless of the technique used.
References [1] Sampliner RE. Updated guidelines for the diagnosis, surveillance, and therapy of Barrett’s esophagus. Am J Gastroenterol 2002;97:1888 – 95. [2] DaCosta RS, Wilson BC, Marcon NE. New optical technologies for earlier endoscopic diagnosis of premalignant gastrointestinal lesions. J Gastroenterol Hepatol 2002;17:S85 – S104. [3] Pfau PR, Sivak Jr MV. Endoscopic diagnostics. Gastroenterology 2001;120:763 – 81. [4] Wagnieres GA, Star WM, Wilson BC. In vivo fluorescence spectroscopy and imaging for oncological applications. Photochem Photobiol 1998;68:603 – 32. [5] Haringsma J, Tytgat GN, Yano H, et al. Autofluorescence endoscopy: feasibility of detection of GI neoplasms unapparent to white light endoscopy with an evolving technology. Gastrointest Endosc 2001;53:642 – 50. [6] Mayinger B, Reh H, Hochberger J, et al. Endoscopic photodynamic diagnosis: oral aminolevulinic acid is a marker of GI cancer and dysplastic lesions. Gastrointest Endosc 1999;50:242 – 6. [7] Messmann H, Knuchel R, Baumler W, et al. Endoscopic fluorescence detection of dysplasia in patients with Barrett’s esophagus, ulcerative colitis, or adenomatous polyps after 5-aminolevulinic acid-induced protoporphyrin IX sensitization. Gastrointest Endosc 1999;49:97 – 101. [8] Panjehpour M, Overholt BF, Schmidhammer JL, et al. Spectroscopic diagnosis of esophageal cancer: new classification model, improved measurement system. Gastrointest Endosc 1995;41: 577 – 81. [9] Vo-Dinh T, Panjehpour M, Overholt BF, et al. In vivo cancer diagnosis of the esophagus using differential normalized fluorescence (DNF) indices. Lasers Surg Med 1995;16:41 – 7. [10] Vo-Dinh T, Panjehpour M, Overholt BF. Laser-induced fluorescence for esophageal cancer and dysplasia diagnosis. Ann N Y Acad Sci 1998;838:116 – 22. [11] Panjehpour M, Overholt BF, Vo-Dinh T, et al. Endoscopic fluorescence detection of high-grade dysplasia in Barrett’s esophagus. Gastroenterology 1996;111:93 – 101. [12] Wang KK, Buttar NS, Wong Kee Song LM, et al. The use of an optical biopsy system in Barrett’s esophagus [abstract]. Gastroenterology 2001;120:A2112. [13] Mayinger B, Horner P, Jordan M, et al. Light-induced autofluorescence spectroscopy for tissue diagnosis of GI lesions. Gastrointest Endosc 2000;52:395 – 400.
296
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
[14] Mayinger B, Horner P, Jordan M, et al. Endoscopic fluorescence spectroscopy in the upper GI tract for the detection of GI cancer: initial experience. Am J Gastroenterol 2001;96:2616 – 21. [15] Mayinger B, Horner P, Jordan M, et al. Light-induced autofluorescence spectroscopy for the endoscopic detection of esophageal cancer. Gastrointest Endosc 2001;54:195 – 201. [16] Georgakoudi I, Jacobson BC, Van Dam J, et al. Fluorescence, reflectance, and light-scattering spectroscopy for evaluating dysplasia in patients with Barrett’s esophagus. Gastroenterology 2001; 120:1620 – 9. [17] Bourg-Heckly G, Blais J, Padilla JJ, et al. Endoscopic ultraviolet-induced autofluorescence spectroscopy of the esophagus: tissue characterization and potential for early cancer diagnosis. Endoscopy 2000;32:756 – 65. [18] DaCosta RS, Wilson BC, Marcon NE. Light-induced fluorescence endoscopy of the gastrointestinal tract. Gastrointest Endosc Clin North Am 2000;10:37 – 69. [19] Haringsma J, Tytgat GNJ. Fluorescence imaging of high-grade dysplasia in Barrett’s esophagus [abstract]. Gastrointest Endosc 2001;53:A4132. [20] Wong Kee Song LM, Marcon NE. Novel optical diagnostic techniques for the recognition of metaplasia and dysplasia. In: Sharma P, Sampliner RE, editors. Barrett’s esophagus and esophageal adenocarcinoma. Boston: Blackwell Science; 2001. p. 123 – 36. [21] Von Holstein CS, Nilsson AMK, Andersson-Engels S, et al. Detection of adenocarcinoma in Barrett’s oesophagus by means of laser induced fluorescence. Gut 1996;39:711 – 6. [22] Messmann H. 5-aminolevulinic acid-induced protoporphyrin IX for the detection of gastrointestinal dysplasia. Gastrointest Endosc Clin North Am 2000;10:497 – 512. [23] Brand SA, Wang TD, Schomacker KT, et al. Detection of high-grade dysplasia in Barrett’s esophagus by 5-aminolevulinic acid (ALA) induced protoporphyrin IX (PPIX) fluorescence spectroscopy [abstract]. Gastroenterology 2000;118:A193. [24] Endlicher E, Knuchel R, Hauser T, et al. Endoscopic fluorescence detection of low and high grade dysplasia in Barrett’s oesophagus using systemic or local 5-aminolaevulinic acid sensitisation. Gut 2001;48:314 – 9. [25] Gossner L, Stepp H, Sroka R, et al. Photodynamic diagnosis (PDD) of high-grade dysplasia and early cancer in the upper GI-tract using 5-aminolaevulinic acid (ALA) [abstract]. Gastrointest Endosc 1998;47:AB29. [26] Mayinger B, Neidhardt S, Reh H, et al. Fluorescence induced with 5-aminolevulinic acid for the endoscopic detection and follow-up of esophageal lesions. Gastrointest Endosc 2001;54:572 – 8. [27] Ortner ME, Voderholzer W, Ebert B, et al. Fluorescence imaging and white light endoscopy system with fluorescence spectroscopy are complementary methods in detecting malignant lesions in Barrett’s esophagus [abstract]. Gastroenterology 2002;122:A31. [28] Choo-Smith LP, Edwards HG, Endtz HP, et al. Medical applications of Raman spectroscopy: from proof of principle to clinical implementation. Biopolymers 2002;67:1 – 9. [29] Hanlon EB, Manoharan R, Koo TW, et al. Prospects for in vivo Raman spectroscopy. Phys Med Biol 2000;45:1 – 59. [30] Shim MG, Wong Kee Song LM, Marcon NE, et al. In vivo near-infrared Raman spectroscopy: demonstration of feasibility during clinical gastrointestinal endoscopy. Photochem Photobiol 2000;72:146 – 50. [31] Wong Kee Song LM, Molckovsky A, Wang K, et al. Raman spectroscopy for in vivo classification of Barrett’s tissue [abstract]. Gastroenterology 2002;122:A288.
Fluorescence and Raman spectroscopy Louis-Michel Wong Kee Song, MDa, Norman E. Marcon, MDb,* a
Division of Gastroenterology and Hepatology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, USA b Division of Gastroenterology, The Centre for Therapeutic Endoscopy and Endoscopic Oncology, St. Michael’s Hospital, University of Toronto, 16-062 Cardinal Carter Wing, 30 Bond Street, Toronto, ON M5B 1W8, Canada
Patients with Barrett’s esophagus (BE) are at risk for the development of esophageal adenocarcinoma. Despite technological improvements in standard white-light endoscopy (WLE), the ability to visually detect dysplasia or early cancer within the Barrett’s segment remains limited. Current practice guidelines recommend periodic endoscopic surveillance with random biopsies in an attempt to identify these lesions at an early and potentially curative stage [1]. This approach, however, is hindered by random sampling error, low sampling yield, biopsy-associated risks, pathology-related costs, interobserver variability in pathology readings, and delay in diagnosis. In part, these limitations have generated increasing interest in the development of optical techniques, such as fluorescence spectroscopy and imaging, elastic scattering spectroscopy, Raman (inelastic scattering) spectroscopy, and optical coherence tomography, as complementary means of enhancing the detection of Barrett’s dysplasia or early adenocarcinoma at endoscopy [2,3]. The common thread in these techniques is the analysis of particular light-tissue interactions (eg, fluorescence or light scattering signals), which relate diagnostic information about the underlying microstructure or biochemical content of tissue. Because these interactions can be promptly measured by an optical probe or imaging device, these techniques offer the potential to rapidly survey a larger surface area of the Barrett’s mucosa than that afforded by conventional biopsy and to identify areas more likely to harbor dysplasia or cancer for targeted biopsies. This article focuses on the in vivo potential of fluorescence and Raman scattering techniques during endoscopic evaluation of BE.
* Corresponding author. E-mail address: [email protected] (N.E. Marcon). 1052-5157/03/$ – see front matter D 2003, Elsevier Inc. All rights reserved. doi:10.1016/S1052-5157(03)00013-8
280
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Fluorescence spectroscopy and imaging Autofluorescence versus drug-induced fluorescence Fluorescence occurs when tissue absorbs light of a shorter wavelength (eg, blue) and partially re-emits it at a longer wavelength (eg, red). Autofluorescence refers to fluorescence emitted by naturally occurring fluorescent molecules (fluorophores) comprising the tissue. Different excitation wavelengths activate different fluorophores, each of which emits over a particular range of wavelengths (Table 1). For example, tissue porphyrins emit red light (fluorescence intensity peaks/bands in the range of 630– 690 nm) when excited with violet-blue light (400 –450 nm). Because tissue contains a variety of fluorophores, its overall autofluorescence line-shape is typically broad and relatively featureless, caused by the composite contributions and overlapping intensity peaks and bands of these fluorophores. In BE, the histologic sequence of events leading to adenocarcinoma is likely accompanied by particular changes in fluorophore content or distribution, which can be detected by differences in spectral shape. At the excitation wavelength or wavelengths typically used (violet-blue range), Barrett’s epithelium generally emits a fluorescence spectrum that demonstrates a relative increase in the red-togreen fluorescence intensity ratio when compared with normal squamous mucosa. Although spectral variability exists, this ratio is seemingly accentuated as Barrett’s tissue evolves from intestinal metaplasia (IM) to dysplasia and eventually adenocarcinoma (Fig. 1). The mechanisms responsible for these observed spectral differences are manifold, including variable concentrations or spatial distributions of fluorophores, altered microvasculature, and changes in metabolic status or microstructure between tissue types [2,4]. The measurement of tissue fluorescence following administration of an exogenous fluorophore (drug-induced fluorescence) is another diagnostic option in BE. These drugs include photosensitizers used in photodynamic therapy, such as hematoporphyrin derivative, or they can be a precursor molecule, such as 5-aminoTable 1 Examples of endogenous fluorophores Fluorophore
Origin
Tryptophan Tyrosine Phenylalanine Collagen Elastin NADH Flavins Porphyrins
Protein — — Connective tissue — Respiratory chain — Heme biosynthesis by-products; bacterial fauna
Abbreviation: NADH, nicotinamide adenine dinucleotide.
Optimal excitation wavelength, nm
Peak fluorescence emission, nm
280 275 260 330 360 340 450 400 – 450
350 300 280 390 410 450 520 635, 690
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
281
Fig. 1. Average normalized fluorescence emission spectra of normal squamous esophagus, nondysplastic Barrett’s esophagus (BE), and dysplastic BE following 437-nm excitation. The spectral dip at around 580 nm is attributed to hemoglobin absorption. (From Wong Kee Song LM, Marcon NE. Novel optical diagnostic techniques for the recognition of metaplasia and dysplasia. In: Sharma P, Sampliner RE, editors. Barrett’s esophagus and esophageal adenocarcinoma. Boston: Blackwell Science; 2001. p. 123 – 36; with permission.)
levulinic acid (5-ALA), which induces the production of protoporphyrin IX, an endogenous fluorophore. The advantage of drug-induced fluorescence over autofluorescence is that the fluorescent signal generated by these drugs is typically strong and can be detected by simpler and cheaper instruments. Furthermore, wavelength selection is facilitated because the optimum excitation and emission wavelengths of these agents are known a priori, unlike in autofluorescence. Their success in BE, however, is evidently dependent on selective drug uptake and retention in dysplastic or early malignant tissues. In this regard, agents such as 5-ALA appear promising, although the optimal dose, mode of administration, time interval between drug intake and spectral measurement, and degree of selectivity in various Barrett’s tissues remain to be clearly defined. Moreover, issues involving adverse effects, extra costs, and regulatory approval present important obstacles for the application of drug-induced fluorescence in endoscopic practice. Fluorescence point spectroscopy versus fluorescence imaging There are primarily two methods for measuring tissue fluorescence during endoscopy. Fluorescence point spectroscopy involves the use of a fiberoptic probe that is introduced into the working channel of a standard endoscope and placed in
282
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Fig. 2. Schematic of a simplified fluorescence point spectroscopic system.
proximity or in contact with tissue (Fig. 2). A typical probe may consist of a central optical fiber delivering excitation light to tissue and a surrounding circular array of optical fibers collecting the emitted fluorescence. The light source can be a laser or a wavelength-filtered lamp. The collected fluorescence is separated into its component wavelengths and corresponding intensities by a spectrograph coupled to a detector and computer, and the resultant fluorescence spectrum of tissue is displayed as an intensity-versus-wavelength curve on the computer monitor. Fluorescence spectra (and other types of spectra) can be analyzed by a variety of statistical/computational means to extract relevant spectral features or regions that discriminate between tissue types. These methods may use simple empiric procedures, such as taking the intensity ratio of two fluorescence emission wavelength bands to determine a threshold that optimally differentiate between tissue categories or use sophisticated multivariate techniques of spectral analysis, such as principal component analysis, linear discriminant analysis, and artificial neural networks. Diagnostic algorithms derived from these methods can be automated, so that the collection of an unknown tissue spectrum at the time of endoscopy and input of this spectrum into the computerized algorithm provides a rapid diagnosis of tissue based on its spectral characteristics. Spectral diagnosis is then compared with histologic diagnosis to determine sensitivity and specificity of the algorithm. The entire fluorescence point spectroscopic system can be integrated into a compact, mobile, and user-friendly device. An example of this is the WavSTAT system (SpectraScience, Inc, Minneapolis, MN). Moreover, a unique and helpful feature of this spectroscopic device is the incorporation of the optical fiber within a standard biopsy forceps. The obvious advantage here is the ability to accurately perform a biopsy at the site immediately following fluorescence measurement. The WavSTAT system uses 337-nm excitation and a proprietary algorithm to analyze the spectra, with the result displayed in a few seconds on an liquid crystal display (LCD) panel. At the excitation wavelength or wavelengths typically used, the sampling depth of the light into tissue is approximately 500 mm, which effectively probes
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
283
the mucosa, the critical layer of interest. The obvious limitation with the ‘‘pointand-shoot’’ spectroscopic technique, however, is that the optical probe samples only a small volume of tissue ( 1mm3) and is directed to the Barrett’s mucosa in a somewhat random fashion, akin to conventional biopsy. Multiple fluorescence measurements (‘‘optical biopsies’’), however, can be collected in the time it takes to obtain a single pinch biopsy, thus increasing the sampling yield. To circumvent the sampling limitations imposed by point spectroscopy, much effort has been devoted toward designing fluorescence imaging systems, which are simply an extension of the spectroscopic technique whereby a two-dimensional surface map of tissue fluorescence is generated. Clearly, the advantage is the visualization of a much larger surface area of the mucosa, with fields of view approaching those obtained with a standard endoscope. The goal here is to detect dysplastic or early cancerous lesions otherwise occult to WLE. Although no new physical concepts are involved, there is a considerable increase in technological complexity and cost relative to spectroscopy, especially if multiwavelength imaging is considered. Several fluorescence imaging systems are currently undergoing evaluation in BE and other sites [4,5]. For example, the second-generation light-induced fluorescence endoscopy system (LIFE-II; Xillix Technologies Corporation, Richmond, BC, Canada) uses blue-light excitation (400 – 450 nm) and two intensified charged-coupled device (CCD) cameras to detect selected fluorescence emission bands in the green (490 –560 nm) and in the red (630 – 750 nm; Fig. 3). LIFE-II has the two CCD cameras incorporated in a detachable lightweight module, which connects to the eyepiece of a conventional fiberoptic endoscope. The two fluorescence images are digitized and combined to produce a single composite pseudocolor image in real time, where normal tissue typically appears green/cyan and abnormal tissue appears red. The system can alternate rapidly between standard WLE and endoscopic fluorescence imaging by means of a simple switch ( 4 s). Sensitivity and specificity values are determined by correlating positive (red fluorescence) and negative (green/cyan fluorescence) images with histologic diagnosis of corresponding biopsy samples. The Storz system is another fluorescence imaging device, which consists of an incoherent light source (D-Light, Storz, Tuttlingen, Germany) and a special camera (Endovision Telecam SL, Storz, Tuttlingen, Germany) specifically tailored for the detection of protoporphyrin IX (PpIX) fluorescence induced by 5-ALA [6,7]. Violet-blue light (375 –440 nm) serves as the excitation source, with a red channel displaying the PpIX fluorescence, a green channel displaying tissue autofluorescence, and an optional blue channel displaying a portion of the excitation light reflected from the tissue surface. Lesion detection can be accomplished through observation of increased red-to-green or red-to-blue contrast. Endoscopic applications Autofluorescence spectroscopy In 1995, Panjehpour et al [8] assessed an in vivo point spectroscopic device with 410-nm excitation and a spectral classification algorithm based on linear
284
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Fig. 3. Schematic of an investigational fluorescence imaging system for in vivo endoscopic applications (Xillix LIFE-II). A combination of green and red fluorescence channels produces the composite false-colored fluorescence image. Note the difference in the red-to-green fluorescence spectral ratio of normal and tumor sites, and how these differences are depicted in the composite image (See also Color Plate 5). (From Wong Kee Song LM, Marcon NE. Novel optical diagnostic techniques for the recognition of metaplasia and dysplasia. In: Sharma P, Sampliner RE, editors. Barrett’s esophagus and esophageal adenocarcinoma. Boston: Blackwell Science; 2001. p. 123 – 36; with permission.)
discriminant analysis to separate esophageal cancer from normal squamous mucosa. Using this technique, malignant tissues (squamous cell cancer and adenocarcinoma) were differentiated from normal mucosa with a sensitivity of 100% and a specificity of 98%. Although these malignant lesions were endoscopically obvious, this study demonstrated for the first time the diagnostic potential of laser-induced autofluorescence spectroscopy for the differentiation of esophageal tissues in vivo. This group then described the use of a simpler algorithm for spectral analysis, the differential normalized fluorescence (DNF) index, which could differentiate between normal esophagus and esophageal cancer with greater than 98% accuracy [9,10]. In this technique, each fluorescence spectrum was first normalized by dividing the intensity at each wavelength by the integrated area under the spectrum. This normalization procedure allowed comparisons to be made between spectra, independent of the variable intensity factor, and amplified spectral differences between normal and malignant tissues. A baseline curve was then established as the mean average of normalized fluorescence spectra from a selected number of normal tissue samples. Finally, the DNF index for a given sample was determined by the relative intensity difference between its normalized fluorescence spectrum and the baseline curve.
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
285
The DNF indices at 480 nm and 660 nm were selected because normal and malignant tissues exhibited significant spectral differences around these wavelengths. Using 410-nm excitation and the DNF algorithm, Panjehpour et al [11] subsequently reported the first in vivo study of autofluorescence spectroscopy in BE. Fluorescence spectra were collected at various spots within the Barrett’s segment (N = 36 patients), followed by conventional biopsy for corresponding histologic diagnosis. Analysis of the fluorescence spectra using the DNF index at 480 nm showed that 96% of nondysplastic Barrett’s (NDB) samples and all lowgrade dysplasia (LGD) samples were classified as benign, 90% of high-grade dysplasia (HGD) samples as premalignant, but only 28% of LGD samples with focal HGD were classified as premalignant. When both DNF indices (480 nm and 660 nm) were used concurrently, all patients with any HGD were classified as premalignant, but 7 of 23 NDB patients were classified as having premalignant lesions, accounting for a 30% false-positive rate. Hence, this technique was sensitive at detecting HGD but not LGD, based on two DNF indices. Moreover, the inability of fluorescence to reliably differentiate between reactive/inflammatory and dysplastic changes is a well-known phenomenon [4], which likely partially contributed to the significant false-positive rate observed in this report. Despite these shortcomings, this important study demonstrated the potential of autofluorescence point spectroscopy to identify endoscopically occult but clinically important lesions, such as HGD, in BE, and to enable targeted biopsies of spectroscopically suspicious sites. In addition, six to seven ‘‘optical biopsies’’ could be collected during the amount of time taken to obtain a standard pinch biopsy specimen. Preliminary findings using the WavSTAT system in BE were recently reported [12]. A total of 87 patients were studied and 326 independent optical biopsy samples were analyzed. Similar to the findings by Panjehpour et al [11], this particular spectroscopic modality demonstrated a sensitivity of 95% and a specificity of 80% in differentiating HGD from LGD and NDB samples. As an alternative to a laser, Mayinger et al [13 – 15] used a filtered light source (Storz GmbH, Tuttlingen, Germany) capable of delivering either white light or violetblue light directly by means of the endoscope, followed by collection of the emitted autofluorescence by an optical probe placed in contact with tissue. A total of 129 spectra were acquired from normal mucosa and malignant lesions in nine patients with squamous cell cancer and four with esophageal adenocarcinoma. Using a simple algorithm based on intensity ratios, a sensitivity of 97% and a specificity of 95% were obtained for the diagnosis of esophageal carcinoma [15]. Georgakoudi et al [16] recently described the use of a multimodal in vivo optical spectroscopic system in BE, combining the techniques of fluorescence, reflectance, and elastic scattering. Fluorescence spectra were acquired at 11 different excitation wavelengths (between 337 –620 nm) using a fast excitationemission matrix (EEM) instrument. Reflectance spectra were also collected from the same site using white light (350 –700 nm) coupled into the same probe. The reflectance spectra were used to correct the measured fluorescence spectra for distortions induced by changes in tissue absorption or scattering, rather than by
286
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
tissue biochemistry, because reflectance is affected by the same absorption and scattering processes. The intrinsic (undistorted) fluorescence spectra were then analyzed using multivariate techniques of principal component analysis and logistic regression. In this study, the use of intrinsic autofluorescence of tissue differentiated HGD from LGD and NDB with a sensitivity and a specificity of 100% and 97%, respectively. The differentiation of LGD and HGD from NDB, however, was less accurate, with a sensitivity of 79% and a specificity of 88%. Moreover, the combination of all three spectroscopic modalities significantly improved the diagnostic accuracy, achieving a sensitivity of 100% and a specificity of 100% for identifying HGD from NDB and LGD, and a sensitivity and specificity of 93% and 100%, respectively, for differentiating dysplastic (LGD and HGD) from nondysplastic Barrett’s samples. These remarkable results, however, were obtained in the context of a limited number of patients and samples (26 NDB, 7 LGD, 7 HGD) and require confirmation on a prospective and much larger sample size. Nevertheless, this study hinted at a multimodal optical technique as the ideal approach for optimizing diagnostic accuracy in the differentiation of Barrett’s tissue. The use of ultraviolet (UV) light for fluorescence excitation in the esophagus has also been reported [17]. Using fluorescence intensity ratio thresholds (I390nm/ I550nm) for tissue classification, the technique differentiated dysplastic/neoplastic tissue (HGD, intramucosal squamous cell cancer, and adenocarcinoma) from normal esophageal and nondysplastic Barrett’s mucosa with a sensitivity and a specificity of 86% and 95%, respectively. The major caveat with the use of UV light, however, is its mutagenic potential, which likely will hinder its application in endoscopic practice. In summary, autofluorescence point spectroscopy appears promising in the identification of Barrett’s lesions that are HGD or frankly neoplastic in nature. The technique seems limited in its ability to identify LGD lesions, or to differentiate between dysplastic and reactive changes. Despite the encouraging results discussed previously, much work remains to be done to determine the true potential of autofluorescence point spectroscopy in BE. For instance, the optimal excitation wavelength or wavelengths for autofluorescence detection of dysplasia or early cancer in BE remain to be determined; wavelength selection is currently done in a somewhat empiric fashion. Mechanistic studies using techniques such as confocal microscopy, microspectrofluorimetry, and fast EEM instruments may shed some light on the origins of the spectral differences observed between nondysplastic and dysplastic Barrett’s tissues and optimize the selection of wavelengths for fluorescence diagnostics in BE [18]. Caution is warranted, however, because the translation of findings obtained ex vivo may not apply in vivo because of changes in the biochemical state of tissue. Autofluorescence imaging Prospective trials are currently underway in assessing the utility of autofluorescence imaging in BE. A preliminary report by Haringsma and Tytgat [19] demonstrated the potential of the Xillix LIFE-II imaging system in 111 patients
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
287
with biopsy-proven BE. Lesions seen within the Barrett’s segment at either WLE or LIFE-II were judged to be dysplastic or nondysplastic, and targeted biopsy samples were taken of these abnormalities, in addition to standard four-quadrant surveillance biopsy samples at each 2-cm interval. LIFE-II detected many more HGD (20 of 24) and early cancerous (17 of 17) lesions than WLE (11 of 24 HGD and 16 of 17 cancers). Similar to autofluorescence spectroscopy, the technique was insensitive at detecting LGD. In the current authors’ experience [20], the LIFE-II system occasionally diagnosed lesions containing HGD that were not recognized by WLE, but the columnar-lined esophagus (with or without dysplasia) routinely produced a falsepositive (red) fluorescence image (Fig. 4), making identification of dysplastic lesions difficult. Moreover, the present system is configured to be used with fiberoptic endoscopes, instead of video endoscopes, which limits its practical utility. Another limitation is the contamination of fluorescence images by the presence of blood in the field of view (following biopsy), because blood is highly fluorescent. The system is currently undergoing technical modifications specifically addressing these issues. Drug-induced fluorescence spectroscopy A limited number of studies are available regarding the use of exogenous fluorophores with fluorescence spectroscopy in BE. These studies to date have mainly exploited photosensitizers used in photodynamic therapy, many of which
Fig. 4. Endoscopic fluorescence imaging of Barrett’s esophagus using the Xillix LIFE-II system. Areas of high-grade dysplasia appear brick red (A), which may be difficult to discriminate from an already red fluorescent metaplastic background (B). Areas were confirmed by histology. This figure appears in color at the front of this issue (See also Color Plate 6). (From Wong Kee Song LM, Marcon NE. Novel optical diagnostic techniques for the recognition of metaplasia and dysplasia. In: Sharma P, Sampliner RE, editors. Barrett’s esophagus and esophageal adenocarcinoma. Boston: Blackwell Science; 2001. p. 123 – 36; with permission.)
288
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
fluoresce and demonstrate some selective localization in neoplastic tissues [18]. Fluorescence generated by these drugs is less ambiguous because of distinguishable intensity peaks or bands. The photosensitizer Photofrin (QLT Therapeutics, Vancouver, Canada), for instance, contains porphyrins that emit characteristic fluorescence peaks in the red (630 and 690 nm) and that are generally retained in neoplastic tissue to a higher degree than in normal tissue. Von Holstein et al [21] initially assessed tissue fluorescence induced by low-dose Photofrin (0.35 mg/kg) in a small number of patients with BE. Following 405-nm excitation, fluorescence spectra were collected and intensity ratios calculated as the quotient of Photofrin fluorescence (at 630 nm) divided by the autofluorescence background of tissue (at 500 nm). From initial ex vivo results, a ratio of 0.55 was used as the cut-off level to distinguish malignant from normal mucosa or Barrett’s epithelium in vivo. In this study, there was a high probability that fluorescence ratios greater than 0.55 represented malignancy; ratios less than 0.55 were more difficult to categorize because of considerable overlap between normal mucosa, nondysplastic or dysplastic Barrett’s epithelium, and malignant tissue. This small-scale study was a first step in the assessment of drug-induced fluorescence in the esophagus in vivo, although Photofrin seemed limited in its ability to differentiate between grades of dysplasia in BE. There also has been interest in the use of 5-ALA in combination with fluorescence spectroscopy in BE. Although not a photosensitizer in itself, the drug is intracellularly converted to PpIX, which has a characteristic fluorescence emission in the red between 625 and 725 nm. Because of low ferrochelatase activity in tumor cells, PpIX preferentially accumulates in malignant tissues and, unlike Photofrin, undergoes rapid metabolic breakdown resulting in a low incidence of side effects and short duration of skin photosensitivity [22]. In a preliminary study, 16 patients with BE received 5-ALA (DUSA Pharmaceuticals, Inc., Wilmington, MA) 10 mg/kg orally 3 hours before endoscopy and spectroscopy. The excitation source was 400-nm light, and the normalized PpIX fluorescence intensity peak at 635 nm was quantitatively analyzed and used for tissue differentiation. A total of 98 spectra were collected. Based on whether the normalized PpIX fluorescence intensity of each spectrum was above or below 0.15, HGD was distinguished from NDB samples with sensitivity and specificity of 86% and 89%, respectively. There was significant overlap between spectra from squamous epithelium, NDB, and indefinite/LGD samples, however, suggesting that fluorescence induced by 5-ALA is more appropriate for detection of HGD and early cancerous lesions in BE [23]. Drug-induced fluorescence imaging In the upper gastrointestinal (GI) tract, initial studies of fluorescence imaging aided by exogenous fluorophores have produced encouraging results [6,7, 24 – 26]. Endlicher et al [24] assessed the feasibility of detecting Barrett’s dysplasia or early adenocarcinoma using fluorescence imaging after sensitization with 5-ALA. The study was performed on 47 patients with biopsy-proven BE, 10
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
289
with known LGD or HGD. The 5-ALA was given either orally at different dosages (5, 10, 20, or 30 mg/kg) or sprayed onto the Barrett’s mucosa (500 – 1000 mg). Fluorescence imaging using the Storz system was performed 4 to 6 hours after systemic sensitization or 1 to 2 hours after topical spray. Interim analysis revealed that 5 mg/kg of 5-ALA orally was too low for the detection of dysplasia, whereas 30 mg/kg orally achieved high sensitivity (100%) at the expense of poor specificity (27%). Subsequent patients were sensitized with 10 or 20 mg/kg of 5-ALA orally or 500 mg topically. A total of 243 biopsy specimens from PpIX red fluorescent (n = 113) and nonfluorescent (n = 130) areas were taken; 33 of these were LGD or HGD by histology. Fluorescence imaging at 4 to 6 hours following drug administration (10 – 30 mg/kg orally) showed relatively high sensitivity (80% –100%) but poor to fair specificity (27% – 56%). Fluorescence imaging 1 to 2 hours following 5-ALA topical spray (500 mg) demonstrated moderate sensitivity (60%) and specificity (69%) for the detection of Barrett’s dysplasia. In addition to the 10 patients with known dysplasia, 2 patients with early cancers and 1 patient with dysplasia were solely detected by fluorescence imaging. Adverse drug effects were limited to nausea and vomiting in 4 patients and increased liver enzymes in 2 patients. Mild skin photosensitivity was observed only in patients receiving 20 to 30 mg/kg of 5-ALA. False-positive fluorescence was induced mainly by inflammation, bile, and metaplasia. Another preliminary study using 7.5 to 10 mg/kg of 5-ALA orally and fluorescence evaluation at 3 hours following drug administration reported a sensitivity of 85% and a specificity of 70% for the detection of dysplastic and early cancerous lesions involving the upper GI tract, including BE [25]. Ortner et al [27] evaluated both fluorescence spectroscopy and imaging following topical application of 5-ALA (500 mg) in 42 patients with BE. Of 36 patients with negative WLE, 4 (11%) had dysplastic or malignant lesions identified by fluorescence spectroscopy and imaging. Fluorescence imaging achieved a sensitivity of 92% but a specificity of 32% in detecting these lesions, whereas fluorescence spectroscopy demonstrated a sensitivity of 76% and a specificity of 74%. The complementary use of both techniques appeared to improve the detection rate of these lesions and diminished the number of biopsy specimens required for optimal diagnosis. The initial findings to date suggest that fluorescence imaging aided by 5-ALA enhances the detection of dysplastic lesions, particularly HGD and early cancer, albeit with a decrease in specificity. A relatively high number of false-positive results (red fluorescence) can be attributed to confounding factors such as inflammation and even metaplasia itself, similar to autofluorescence. Regardless of the technique used (spectroscopy or imaging), it remains to be seen whether drug-induced fluorescence will outperform autofluorescence in the detection and grading of dysplasia in BE. Perhaps the development of agents by means of conjugation of a fluorophore dye to a monoclonal antibody or other dysplastic/ tumor-targeting moiety may tip the balance in favor of drug-induced fluorescence, but this principle will require much work demonstrating efficacy and safety for in vivo use.
290
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Raman spectroscopy Principles and techniques Among several optical techniques currently under investigation for in vivo endoscopic applications, Raman spectroscopy provides the most detailed information about the molecular composition of tissue. Light can be absorbed or scattered in interacting with tissue molecules. Whereas almost all of the scattered light is of the same wavelength as the incident light (elastic scattering), a small fraction undergoes so-called Raman (inelastic) scattering, in which slight shifts in energy (or wavelength), relative to the incident light, occur from an exchange of energy between the light photons and the molecular structure of tissue. These wavelength shifts correspond to specific vibrations or rotations of particular molecular bonds. A Raman spectrum is then a plot of the intensity of the Raman scattered light as a function of the wavelength shift,1 with each peak corresponding to a specific molecular state (Fig. 5). Compared with fluorescence or elastic scattering, Raman spectra of tissue are spectrally highly detailed, providing the potential for better tissue discrimination. Molecular vibrations originating from proteins, lipids, or nucleic acids exhibit distinct Raman signatures. For instance, the peak at 1002 cm 1 in Fig. 5 corresponds to the ‘‘in and out’’ vibrations of phenyl rings of aromatic amino acids. Raman scattering within tissue is more difficult to measure because the signal is much weaker relative to fluorescence or elastic scattering and is masked by a broad tissue autofluorescence background. In addition, the endoscopic application of Raman spectroscopy using fiberoptic probes has been technically challenging because of spectral contamination from background fluorescence and Raman signals generated in the fiberoptic materials. Recent technological progress in high-throughput spectrographs, high-sensitivity near-infrared detectors, and filtered fiberoptic probes, however, has enabled Raman spectroscopy instruments to be developed for clinical use [28,29]. Shim et al [30] originally demonstrated the endoscopic feasibility of obtaining near-infrared Raman spectra with reasonable signal-to-noise ratio throughout the GI tract, using a system developed in-house. The instrumentation design is similar to that shown in Fig. 2, except that the technology is more expensive because of the spectroscopic components required to extract the weak Raman signal of tissue. Although Raman scattering can be induced by UV, visible, or near-infrared light, the light source used is usually a near-infrared laser (eg, 785 nm), which minimizes excitation of tissue autofluorescence while generating Raman signals sensitive enough to be detected by existing CCD cameras.
1
The shift is usually expressed in terms of the difference in wave numbers (1/wavelength) in units of cm 1 between the incident and scattered light. Here, only the Raman signals corresponding to transfer of energy from the light to the molecules, resulting in the scattered light having less energy and so longer wavelength, are considered.
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
291
Fig. 5. Raman versus fluorescence spectrum of tissue. Note the spectrally detailed features associated with the Raman signal, which may translate into more accurate tissue diagnosis.
The use of UV light to excite Raman scattering in tissue is attractive because it results in fluorescence that occurs outside the spectrum where the Raman bands are observed, which can then be easily filtered out. Furthermore, UV excitation that corresponds to an absorption band of a molecule produces Raman bands that are significantly amplified, a process known as UV-resonance Raman. The mutagenic potential of UV light is a serious concern, however, which is likely to deter the development of UV Raman for clinical applications.
Fig. 6. Average in vivo Raman spectra of nondysplastic (n = 80), low-grade dysplastic (LGD; n = 26), and high-grade dysplastic (HGD; n = 30) Barrett’s samples obtained from 13 patients with Barrett’s esophagus. Typical tissue Raman peaks are noted (labeled). Subtle or visually unapparent (but potentially discriminatory) spectral differences are present throughout, requiring multivariate techniques to sort them out.
292
First author
Technique
lexc, nm
lem, nm
No. patients No. Spectra/tissue
Diagnostic algorithm Findings/comments
Panjehpour (1995) [8] Vo-Dinh (1995) [9]
AFS
410
430 – 716
32
LDA
AFS
410
430 – 720
48
Panjehpour (1996) [11]
AFS
410
430 – 716
36
Bourg-Heckly (2000) [17]
AFS
330
350 – 650
24
Wang (2001) [12] AFS
337
350 – 800
87
Georgakoudi (2001) [16]
AFS
11 lexc (between 337 – 620)
350 – 750
18
Mayinger (2001) [15]
AFS
375 – 478
478 – 700
13
Von Holstein (1996) [21]
DFS (Photofrin)
405
450 – 750
5
123 normal 36 SCC and ACA > 200 normal and SCC/ACA spectra 216 NDB 36 LGD 46 focal HGD 10 diffuse HGD 66 normal 116 NDB 11 dysplastic 25 SCC and ACA 266 NDB 46 LGD 14 HGD 26 NDB 7 LGD 7 HGD 57 55 17 48 17
normal SCC ACA normal/NDB ACA
DNF
DNF480nm and DNF660nm
I390nm/I550nm
Sn 100% and Sp 98% for SCC/ACA vs normal > 98% accuracy
Sn 100% and Sp 70% for HGD vs rest. Technique insensitive at detecting LGD Sn 86% and Sp 95% for HGD/ACA/SCC vs normal/NDB
Proprietary algorithm
Sn 95% and Sp 80% for HGD vs NDB/LGD
PCA and LRA
Sn 100% and Sp 97% for HGD vs NDB/LGD Sn 79% and Sp 88% for LGD/HGD vs NDB Sn 97% and Sp 95% for SCC/ACA vs normal
I500 – 549nm/ I657 – 700nm I630nm/I500nm
Sn 88% and Sp 94% for ACA vs normal/NDB
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Table 2 Summary of in vivo fluorescence and Raman studies in Barrett’s esophagus
DFS (5-ALA)
400
16
Haringsma (2001) [19]
AFI Blue light (Xillix system) (400 – 450)
490 – 560 (green) 630 – 750 (red)
Mayinger (2001) [26]
DFI (Storz system)
Violet-blue light (375 – 440)
5-ALA induced 22 (6 BE) PpIX fluorescence
Quantitative normalized PpIX fluorescence intensity peak at 635 nm Red fluorescence abnormal; green/ cyan fluorescence normal PpIX red fluorescence
Endlicher (2001) [24]
DFI (Storz system)
Violet-blue light (380 – 440)
5-ALA induced 47 PpIX fluorescence
PpIX red fluorescence
Ortner (2002) [27]
DFS and DFI
505
5-ALA induced 42 PpIX fluorescence
PpIX red fluorescence
Wong Kee Song (2002) [31]
RS
785
900 – 1800 cm
1
16 normal 23 NDB 23 indefinite/LGD 7 HGD
111
13
80 NDB 26 LGD 30 HGD
PCA and LDA
Sn 86% and Sp 89% for HGD vs NDB
Sn 90% and Sp 89% for detecting HGD/ACA Technique insensitive at detecting LGD DFI: Sn 85% and Sp 53%; WLE: Sn 25% and Sp 94% for detecting premalignant/ malignant lesions Dose ranging study Sn 80% – 100% and Sp 27% – 56% (5-ALA 10 – 30 mg/kg p.o.) and Sn 60% and Sp 69% (5-ALA 500 mg topical spray) for detecting dysplasia DFS: Sn 76% and Sp 74% DFI: Sn 92% and Sp 32% for detecting LGD/HGD/ACA Sn 79% and Sp 79% for LGD/HGD vs NDB
293
Abbreviations: ACA, adenocarcinoma in Barrett’s esophagus; AFI, autofluorescence imaging; AFS, autofluorescence spectroscopy; 5-ALA, 5-aminolevulinic acid; BE, Barrett’s esophagus; DFI, drug-induced fluorescence imaging; DFS, drug-induced fluorescence spectroscopy; DNF, differential normalized fluorescence; HGD, highgrade dysplasia; I, fluorescence intensity; LDA, linear discriminant analysis; LGD, low-grade dysplasia; LRA, logistic regression analysis; NDB, nondysplastic Barrett’s; PCA, principal component analysis; PpIX, protoporphyrin IX; RS, Raman spectroscopy; SCC, squamous cell cancer; Sn, sensitivity; Sp, specificity; WLE, white-light endoscopy; lexc, excitation wavelengths; lem, emission wavelengths.
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Brand (2000) [23]
294
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
Endoscopic applications Wong Kee Song et al [31] recently reported on the in vivo potential of nearinfrared Raman spectroscopy for the identification of dysplasia in BE during endoscopy. A custom-made filtered fiberoptic probe was placed in gentle contact with the Barrett’s mucosa and a single Raman spectrum was obtained in 5 seconds, followed by biopsy of the site for corresponding histologic diagnosis. Prior studies have demonstrated the effects on the spectra of varying the pressure of the probe tip on the tissue and of the probe-tissue angle to be insignificant [30]. Thirteen patients were studied, and 136 spectra and corresponding biopsy specimens (80 IM, 26 LGD, and 30 HGD) were obtained. Raman spectra of BE demonstrated typical tissue Raman peaks, including the protein amide I band, CH2 bending mode, CH2 twisting mode, protein amide III band, and the phenyl ring breathing mode. As in other tissues, the spectral differences among the categories of Barrett’s tissue were subtle (Fig. 6), yet contained diagnostically important information. In this setting, multivariate analytic techniques are well suited, indeed may be essential, for spectral analysis rather than simply using algorithms based on empirically determined peak intensities, widths, or peak ratios. Using the statistical methods of principal component analysis and linear discriminant analysis to extract and analyze the full diagnostic content of Raman spectra, it was possible to differentiate dysplastic (LGD and HGD) from nondysplastic (IM) Barrett’s samples with a sensitivity of 79% and a specificity of 79% [31]. These initial findings were based on a custom-built Raman spectroscopic system, and it is anticipated that further improvements in Raman instrumentation or methods of spectral analysis will increase the diagnostic performance of this technique in BE. Furthermore, the technique is being refined and tested to assess its discriminating power in classifying dysplasia and early cancer (LGD versus HGD versus early adenocarcinoma). Unlike fluorescence, the prospect for a real-time endoscopic Raman imaging system remains doubtful because of the weak Raman signals and technological limitations. Endoscopic Raman diagnosis can thus be accomplished only by means of an optical probe at the moment, with sampling limitations inherent to all point spectroscopic techniques.
Summary Table 2 provides a summary of selected in vivo fluorescence and Raman studies performed in BE. Although the findings from these studies appear promising, these techniques are still under development, and it is anticipated that technological refinements will further enhance their diagnostic accuracy. Ultimately, however, large-scale prospective clinical trials are required to determine their true diagnostic potential in BE and other sites. Ideally, the instrumentation of choice would be a real-time endoscopic system that combines excellent diagnostic accuracy with wide-area sampling. In this regard, fluo-
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
295
rescence imaging is most appealing, although a variety of issues remain to be resolved, including the choice between autofluorescence versus drug-induced fluorescence and the problematic distinction between dysplastic (true positive) and confounding background metaplastic fluorescence (false positive), among others. It is also not clear whether exogenous fluorophores are necessary to achieve clinically useful sensitivity and specificity for lesion detection in BE. Point spectroscopic techniques, either fluorescence or Raman scattering, are inherently limited by the small volume of tissue (biopsy specimen size) they sample, but more detailed information can be extracted from the spectra, which may increase diagnostic accuracy. Moreover, it may be that the optimal system will be a combination of multiple optical spectroscopic or imaging techniques (multimodality approach), as suggested by Georgakoudi et al [16]. For instance, a lesion could be detected by fluorescence imaging and its dysplastic nature characterized (graded) by Raman spectroscopy. In this era of cost containment, however, the critical challenge is to demonstrate whether an increase in diagnostic accuracy merits investment in costly technology, regardless of the technique used.
References [1] Sampliner RE. Updated guidelines for the diagnosis, surveillance, and therapy of Barrett’s esophagus. Am J Gastroenterol 2002;97:1888 – 95. [2] DaCosta RS, Wilson BC, Marcon NE. New optical technologies for earlier endoscopic diagnosis of premalignant gastrointestinal lesions. J Gastroenterol Hepatol 2002;17:S85 – S104. [3] Pfau PR, Sivak Jr MV. Endoscopic diagnostics. Gastroenterology 2001;120:763 – 81. [4] Wagnieres GA, Star WM, Wilson BC. In vivo fluorescence spectroscopy and imaging for oncological applications. Photochem Photobiol 1998;68:603 – 32. [5] Haringsma J, Tytgat GN, Yano H, et al. Autofluorescence endoscopy: feasibility of detection of GI neoplasms unapparent to white light endoscopy with an evolving technology. Gastrointest Endosc 2001;53:642 – 50. [6] Mayinger B, Reh H, Hochberger J, et al. Endoscopic photodynamic diagnosis: oral aminolevulinic acid is a marker of GI cancer and dysplastic lesions. Gastrointest Endosc 1999;50:242 – 6. [7] Messmann H, Knuchel R, Baumler W, et al. Endoscopic fluorescence detection of dysplasia in patients with Barrett’s esophagus, ulcerative colitis, or adenomatous polyps after 5-aminolevulinic acid-induced protoporphyrin IX sensitization. Gastrointest Endosc 1999;49:97 – 101. [8] Panjehpour M, Overholt BF, Schmidhammer JL, et al. Spectroscopic diagnosis of esophageal cancer: new classification model, improved measurement system. Gastrointest Endosc 1995;41: 577 – 81. [9] Vo-Dinh T, Panjehpour M, Overholt BF, et al. In vivo cancer diagnosis of the esophagus using differential normalized fluorescence (DNF) indices. Lasers Surg Med 1995;16:41 – 7. [10] Vo-Dinh T, Panjehpour M, Overholt BF. Laser-induced fluorescence for esophageal cancer and dysplasia diagnosis. Ann N Y Acad Sci 1998;838:116 – 22. [11] Panjehpour M, Overholt BF, Vo-Dinh T, et al. Endoscopic fluorescence detection of high-grade dysplasia in Barrett’s esophagus. Gastroenterology 1996;111:93 – 101. [12] Wang KK, Buttar NS, Wong Kee Song LM, et al. The use of an optical biopsy system in Barrett’s esophagus [abstract]. Gastroenterology 2001;120:A2112. [13] Mayinger B, Horner P, Jordan M, et al. Light-induced autofluorescence spectroscopy for tissue diagnosis of GI lesions. Gastrointest Endosc 2000;52:395 – 400.
296
L-M. Wong Kee Song, N.E. Marcon / Gastrointest Endoscopy Clin 13 (2003) 279–296
[14] Mayinger B, Horner P, Jordan M, et al. Endoscopic fluorescence spectroscopy in the upper GI tract for the detection of GI cancer: initial experience. Am J Gastroenterol 2001;96:2616 – 21. [15] Mayinger B, Horner P, Jordan M, et al. Light-induced autofluorescence spectroscopy for the endoscopic detection of esophageal cancer. Gastrointest Endosc 2001;54:195 – 201. [16] Georgakoudi I, Jacobson BC, Van Dam J, et al. Fluorescence, reflectance, and light-scattering spectroscopy for evaluating dysplasia in patients with Barrett’s esophagus. Gastroenterology 2001; 120:1620 – 9. [17] Bourg-Heckly G, Blais J, Padilla JJ, et al. Endoscopic ultraviolet-induced autofluorescence spectroscopy of the esophagus: tissue characterization and potential for early cancer diagnosis. Endoscopy 2000;32:756 – 65. [18] DaCosta RS, Wilson BC, Marcon NE. Light-induced fluorescence endoscopy of the gastrointestinal tract. Gastrointest Endosc Clin North Am 2000;10:37 – 69. [19] Haringsma J, Tytgat GNJ. Fluorescence imaging of high-grade dysplasia in Barrett’s esophagus [abstract]. Gastrointest Endosc 2001;53:A4132. [20] Wong Kee Song LM, Marcon NE. Novel optical diagnostic techniques for the recognition of metaplasia and dysplasia. In: Sharma P, Sampliner RE, editors. Barrett’s esophagus and esophageal adenocarcinoma. Boston: Blackwell Science; 2001. p. 123 – 36. [21] Von Holstein CS, Nilsson AMK, Andersson-Engels S, et al. Detection of adenocarcinoma in Barrett’s oesophagus by means of laser induced fluorescence. Gut 1996;39:711 – 6. [22] Messmann H. 5-aminolevulinic acid-induced protoporphyrin IX for the detection of gastrointestinal dysplasia. Gastrointest Endosc Clin North Am 2000;10:497 – 512. [23] Brand SA, Wang TD, Schomacker KT, et al. Detection of high-grade dysplasia in Barrett’s esophagus by 5-aminolevulinic acid (ALA) induced protoporphyrin IX (PPIX) fluorescence spectroscopy [abstract]. Gastroenterology 2000;118:A193. [24] Endlicher E, Knuchel R, Hauser T, et al. Endoscopic fluorescence detection of low and high grade dysplasia in Barrett’s oesophagus using systemic or local 5-aminolaevulinic acid sensitisation. Gut 2001;48:314 – 9. [25] Gossner L, Stepp H, Sroka R, et al. Photodynamic diagnosis (PDD) of high-grade dysplasia and early cancer in the upper GI-tract using 5-aminolaevulinic acid (ALA) [abstract]. Gastrointest Endosc 1998;47:AB29. [26] Mayinger B, Neidhardt S, Reh H, et al. Fluorescence induced with 5-aminolevulinic acid for the endoscopic detection and follow-up of esophageal lesions. Gastrointest Endosc 2001;54:572 – 8. [27] Ortner ME, Voderholzer W, Ebert B, et al. Fluorescence imaging and white light endoscopy system with fluorescence spectroscopy are complementary methods in detecting malignant lesions in Barrett’s esophagus [abstract]. Gastroenterology 2002;122:A31. [28] Choo-Smith LP, Edwards HG, Endtz HP, et al. Medical applications of Raman spectroscopy: from proof of principle to clinical implementation. Biopolymers 2002;67:1 – 9. [29] Hanlon EB, Manoharan R, Koo TW, et al. Prospects for in vivo Raman spectroscopy. Phys Med Biol 2000;45:1 – 59. [30] Shim MG, Wong Kee Song LM, Marcon NE, et al. In vivo near-infrared Raman spectroscopy: demonstration of feasibility during clinical gastrointestinal endoscopy. Photochem Photobiol 2000;72:146 – 50. [31] Wong Kee Song LM, Molckovsky A, Wang K, et al. Raman spectroscopy for in vivo classification of Barrett’s tissue [abstract]. Gastroenterology 2002;122:A288.