Search

Spectroscopy and spectral imaging have widespread applications in many scientific fields. Deep learning techniques have achieved many successes in recent years across numerous domains. However, the application of deep learning to spectral data remains a complex task due to the need for tailored augmentation routines, specific architectures for spectral data and significant memory requirements. Here we present spectrai, a comprehensive open-source deep learning framework and Python/MATLAB package designed to facilitate the training of neural networks on spectral data. spectrai provides numerous built-in spectral data pre-processing and augmentation methods, neural networks for spectral data including spectral (image) denoising, spectral (image) classification, spectral image segmentation and spectral image super-resolution. spectrai includes both command line and graphical user interface (GUI) tools designed to assist users with model and hyperparameter decisions for a wide range of applications. We demonstrate three case studies of spectral denoising, spectral segmentation and super-resolution. By providing baseline implementations of these functions, spectrai enables wider use of deep learning in spectroscopy and spectral imaging.

Theranostics, the combination of diagnosis and therapy, has long held promise as a means to achieving personalised precision cancer treatments. However, despite its potential, theranostics has yet to realise significant clinical translation, largely due the complexity and overriding toxicity concerns of existing theranostic nanoparticle strategies. Methods: Here, we present an alternative nanoparticle-free theranostic approach based on simultaneous Raman spectroscopy and photodynamic therapy (PDT) in an integrated clinical platform for cancer theranostics. Results: We detail the compatibility of Raman spectroscopy and PDT for cancer theranostics, whereby Raman spectroscopic diagnosis can be performed on PDT photosensitiser-positive cells and tissues without inadvertent photosensitiser activation/photobleaching or impaired diagnostic capacity. We further demonstrate that our theranostic platform enables in vivo tumour diagnosis, treatment, and post-treatment molecular monitoring in real-time. Conclusion: This system thus achieves effective theranostic performance, providing a promising new avenue towards the clinical realisation of theranostics.

I. INTRODUCTION Medicine historically relied on astute history and examination skills. As technology was lacking, ward rounds focused on debate and discussion of diagnoses and possible differential diagnoses based on the history and physical examination. The technology movement into healthcare was never truly predicted. With its occurrence, came the ability to scan a patient from top to toe via computed tomography and magnetic resonance imaging. Technology now serves as our main diagnostic tool (Patel, 2013).

Spontaneous Raman spectroscopy enables non-ionising, non-destructive, and label-free acquisition of a biochemical fingerprint for a given sample. However, the long integration times required largely prohibit high-throughput applications. Here, we present a comprehensive deep learning framework for extreme speed-up of spontaneous Raman imaging. Our deep learning framework enhances Raman imaging two-fold, effectively reconstructing both spectral and spatial information from low spatial resolution, low signal-to-noise ratio images to achieve extreme Raman imaging time speed-ups of 40-90x while mainting high reconstruction fidelity. As such, our framework could enable a host of higher-throughput spontaneous Raman spectroscopy applications across a diverse range of fields.

Osteoarthritis (OA) is a painful, debilitating disease characterized by the degeneration of articular cartilage. We have developed a novel multiplexed polarized, hypodermic-needle-compatible Raman arthroscope probe that can achieve intra-articular assessments of the compositional and structural changes to cartilage associated with early-stage OA, including depletion of glycosaminoglycans from the cartilage superficial regions and changes to superficial zone collagen alignment. Through ex vivo models on human and bovine cartilage, we demonstrate that using multivariate linear regression, this platform can accurately measure superficial zone cartilage GAG depletion. This work shows that Raman needle arthroscopy can provide a practical, minimally invasive, point-of-care clinical tool capable of diagnosing OA before irreparable cartilage degeneration is radiographically evident.

The invention relates to an optical tomography scanning system (1) for 3D imaging of an object (100), for example an experimental rodent or a patient, using Raman scattered light from the object. A laser (10) and an optical configuration (20) for fixating the object and arranged for rotation (Θ) is provided, and an optical receiving part (30) receives Raman scattered light. The optical receiving part has a detector array (35) for receiving the Raman scattered light in two spatial dimensions (X, Z), and optical guide means (37) convey Raman scattered light to a spectrometer (38), where multiple spectra are recorded for the two spatial dimensions for each relative rotational position (Θ) of the object. A 3D image reconstruction of the object based on the multiple spectra at a plurality of relative rotational positions (Θ) of the object enables 3D tomographic imaging of the object, such as an experimental rat.

Zebrafish embryos provide a unique opportunity to visualize complex biological processes, yet conventional imaging modalities are unable to access intricate biomolecular information without compromising the integrity of the embryos. Here, we report the use of confocal Raman spectroscopic imaging for the visualization and multivariate analysis of biomolecular information extracted from unlabeled zebrafish embryos. We outline broad applications of this method in: (i) visualizing the biomolecular distribution of whole embryos in three dimensions, (ii) resolving anatomical features at subcellular spatial resolution, (iii) biomolecular profiling and discrimination of wild type and ΔRD1 mutant Mycobacterium marinum strains in a zebrafish embryo model of tuberculosis and (iv) in vivo temporal monitoring of the wound response in living zebrafish embryos. Overall, this study demonstrates the application of confocal Raman spectroscopic imaging for the comparative bimolecular analysis of fully intact and living zebrafish embryos., Raman spectroscopic imaging (RSI) can provide information on the chemical composition of a sample, but application to living organisms has lacked sufficient spatial resolution and signal strength. Here the authors apply confocal RSI to whole-mount zebrafish embryos to distinguish different infectious bacteria and to living zebrafish embryos to monitor the wound healing process.

Raman spectroscopy enables nondestructive, label-free imaging with unprecedented molecular contrast, but is limited by slow data acquisition, largely preventing high-throughput imaging applications. Here, we present a comprehensive framework for higher-throughput molecular imaging via deep-learning-enabled Raman spectroscopy, termed DeepeR, trained on a large data set of hyperspectral Raman images, with over 1.5 million spectra (400 h of acquisition) in total. We first perform denoising and reconstruction of low signal-to-noise ratio Raman molecular signatures via deep learning, with a 10× improvement in the mean-squared error over common Raman filtering methods. Next, we develop a neural network for robust 2-4× spatial super-resolution of hyperspectral Raman images that preserve molecular cellular information. Combining these approaches, we achieve Raman imaging speed-ups of up to 40-90×, enabling good-quality cellular imaging with a high-resolution, high signal-to-noise ratio in under 1 min. We further demonstrate Raman imaging speed-up of 160×, useful for lower resolution imaging applications such as the rapid screening of large areas or for spectral pathology. Finally, transfer learning is applied to extend DeepeR from cell to tissue-scale imaging. DeepeR provides a foundation that will enable a host of higher-throughput Raman spectroscopy and molecular imaging applications across biomedicine. ispartof: ANALYTICAL CHEMISTRY vol:93 issue:48 pages:15850-15860 ispartof: location:United States status: published

Extracellular vesicles (EVs) are biologically derived nanovectors important for intercellular communication and trafficking, yet understanding of their underlying biological mechanisms remains poor. Advances have been hampered by both the complex biological origins of EVs and a lack of suitable imaging techniques. Here, we present a strategy for simultaneous in vitro imaging and molecular characterisation of EVs in 2D and 3D based on Raman spectroscopy and minimally-obstructive metabolic deuterium labelling. Metabolically-incorporated deuterium acts as a bio-orthogonal Raman-active tag for direct Raman identification of EVs and provides insights into their biocomposition and trafficking, with implications for their development as therapeutic delivery vectors.

We report a biodegradable synthetic material that opens a door to repair damaged tissue via boosting cellular energy metabolism., Cellular bioenergetics (CBE) plays a critical role in tissue regeneration. Physiologically, an enhanced metabolic state facilitates anabolic biosynthesis and mitosis to accelerate regeneration. However, the development of approaches to reprogram CBE, toward the treatment of substantial tissue injuries, has been limited thus far. Here, we show that induced repair in a rabbit model of weight-bearing bone defects is greatly enhanced using a bioenergetic-active material (BAM) scaffold compared to commercialized poly(lactic acid) and calcium phosphate ceramic scaffolds. This material was composed of energy-active units that can be released in a sustained degradation-mediated fashion once implanted. By establishing an intramitochondrial metabolic bypass, the internalized energy-active units significantly elevate mitochondrial membrane potential (ΔΨm) to supply increased bioenergetic levels and accelerate bone formation. The ready-to-use material developed here represents a highly efficient and easy-to-implement therapeutic approach toward tissue regeneration, with promise for bench-to-bedside translation.

Label-free surface-enhanced Raman spectroscopy (SERS) can interrogate systems by directly fingerprinting their components’ unique physicochemical properties. In complex biological systems however, this can yield highly overlapping spectra that hinder sample identification. Here, we present an artificial-nose inspired SERS fingerprinting approach where spectral data is obtained as a function of sensor surface chemical functionality. Supported by molecular dynamics modeling, we show that mildly selective self-assembled monolayers can influence the strength and configuration in which analytes interact with plasmonic surfaces, diversifying the resulting SERS fingerprints. Since each sensor generates a modulated signature, the implicit value of increasing the dimensionality of datasets is shown using cell lysates for all possible combinations of up to 9 fingerprints. Reliable improvements in mean discriminatory accuracy towards 100% are achieved with each additional surface functionality. This arrayed label-free platform illustrates the wide-ranging potential of high-dimensionality artificial-nose based sensing systems for more reliable assessment of complex biological matrices., Label-free surface-enhanced Raman spectroscopy is an emergent method for the detection and discrimination of biological analytes. Here, the authors describe SERS sensors with arrayed mildly-selective surface chemistries to give a fingerprint based on different interactions for analysing biological samples.

Dear EditorMedical education has advanced significantly. Decades past, medical doctors in training were required to have a heavy science-based background leading to the rise of the MD PhD. As envir...

The extracellular matrix (ECM) consists of a complex mesh of proteins, glycoproteins, and glycosaminoglycans, and is essential for maintaining the integrity and function of biological tissues. Imaging and biomolecular characterization of the ECM is critical for understanding disease onset and for the development of novel, disease-modifying therapeutics. Recently, there has been a growing interest in the use of Raman spectroscopy to characterize the ECM. Raman spectroscopy is a label-free vibrational technique that offers unique insights into the structure and composition of tissues and cells at the molecular level. This technique can be applied across a broad range of ECM imaging applications, which encompass in vitro, ex vivo, and in vivo analysis. State-of-the-art confocal Raman microscopy imaging now enables label-free assessments of the ECM structure and composition in tissue sections with a remarkably high degree of biomolecular specificity. Further, novel fiber-optic instrumentation has opened up for clinical in vivo ECM diagnostic measurements across a range of tissue systems. A palette of advanced computational methods based on multivariate statistics, spectral unmixing, and machine learning can be applied to Raman data, allowing for the extraction of specific biochemical information of the ECM. Here, we review Raman spectroscopy techniques for ECM characterizations over a variety of exciting applications and tissue systems, including native tissue assessments (bone, cartilage, cardiovascular), regenerative medicine quality assessments, and diagnostics of disease states. We further discuss the challenges in the widespread adoption of Raman spectroscopy in biomedicine. The results of the latest discovery-driven Raman studies are summarized, illustrating the current and potential future applications of Raman spectroscopy in biomedicine.

This study aims to characterize the in vivo Raman spectroscopic properties of normal colorectal tissues and to assess distinctive biomolecular variations of different anatomical locations in the colorectum for cancer diagnosis. We have developed a novel 785 nm excitation fiber-optic Raman endoscope that can simultaneously acquire in vivo fingerprint (FP) spectra (800-1800 cm(-1)) and high-wavenumber (HW) Raman spectra (2800-3600 cm(-1)) from the subsurface of colorectal tissue. We applied the FP/HW Raman endoscope for in vivo tissue Raman measurements of various normal colorectal anatomical locations (i.e., ascending colon (n = 182), transverse colon (n = 249), descending colon (n = 124), sigmoid (n = 212), and rectum (n = 362)) in 50 subjects. Partial least-squares (PLS)-discriminant analysis (DA) was employed to evaluate the interanatomical variability. The normal colorectal tissue showed a subtle interanatomical variability in molecular constituents (i.e., proteins, lipids, and water content) and could be divided into three major clusterings: (1) ascending colon and transverse colon, (2) descending colon, and (3) sigmoid and rectum. The PLS-DA multiclass algorithms were able to identify different tissue sites with varying sensitivities (SE) and specificities (SP) (ascending colon: SE: 1.10%, SP: 91.02; transverse colon: SE: 14.06%, SP: 78.78; descending colon: SE: 40.32%, SP: 81.99; sigmoid: SE: 19.34%, SP: 87.90; rectum: SE: 71.55%, SP: 77.84). The interanatomical molecular variability was orders of magnitude less than neoplastic tissue transformation. Further PLS-DA modeling on in vivo FP/HW tissue Raman spectra yielded a diagnostic accuracy of 88.8% (sensitivity: 93.9% (93/99); specificity 88.3% (997/1129) for colorectal cancer detection. This work discloses that interanatomical Raman spectral variability of normal colorectal tissue is subtle compared to cancer tissue, and the simultaneous FP/HW Raman endoscopic technique has promising potential for real-time, in vivo diagnosis of colorectal cancer at the molecular level.

We report a novel method making use of multivariate reference signals of fused silica and sapphire Raman signals generated from a ball-lens fiber-optic Raman probe for quantitative analysis of in vivo tissue Raman measurements in real time. Partial least-squares (PLS) regression modeling is applied to extract the characteristic internal reference Raman signals (e.g., shoulder of the prominent fused silica boson peak (~130 cm(-1)); distinct sapphire ball-lens peaks (380, 417, 646, and 751 cm(-1))) from the ball-lens fiber-optic Raman probe for quantitative analysis of fiber-optic Raman spectroscopy. To evaluate the analytical value of this novel multivariate reference technique, a rapid Raman spectroscopy system coupled with a ball-lens fiber-optic Raman probe is used for in vivo oral tissue Raman measurements (n = 25 subjects) under 785 nm laser excitation powers ranging from 5 to 65 mW. An accurate linear relationship (R(2) = 0.981) with a root-mean-square error of cross validation (RMSECV) of 2.5 mW can be obtained for predicting the laser excitation power changes based on a leave-one-subject-out cross-validation, which is superior to the normal univariate reference method (RMSE = 6.2 mW). A root-mean-square error of prediction (RMSEP) of 2.4 mW (R(2) = 0.985) can also be achieved for laser power prediction in real time when we applied the multivariate method independently on the five new subjects (n = 166 spectra). We further apply the multivariate reference technique for quantitative analysis of gelatin tissue phantoms that gives rise to an RMSEP of ~2.0% (R(2) = 0.998) independent of laser excitation power variations. This work demonstrates that multivariate reference technique can be advantageously used to monitor and correct the variations of laser excitation power and fiber coupling efficiency in situ for standardizing the tissue Raman intensity to realize quantitative analysis of tissue Raman measurements in vivo, which is particularly appealing in challenging Raman endoscopic applications.

Tissue architecture is intimately linked with its functions, and loss of tissue organization is often associated with pathologies. The intricate depth-dependent extracellular matrix (ECM) arrangement in articular cartilage is critical to its biomechanical functions. In this study, we developed a Raman spectroscopic imaging approach to gain new insight into the depth-dependent arrangement of native and tissue-engineered articular cartilage using bovine tissues and cells. Our results revealed previously unreported tissue complexity into at least six zones above the tidemark based on a principal component analysis and k-means clustering analysis of the distribution and orientation of the main ECM components. Correlation of nanoindentation and Raman spectroscopic data suggested that the biomechanics across the tissue depth are influenced by ECM microstructure rather than composition. Further, Raman spectroscopy together with multivariate analysis revealed changes in the collagen, glycosaminoglycan, and water distributions in tissue-engineered constructs over time. These changes were assessed using simple metrics that promise to instruct efforts toward the regeneration of a broad range of tissues with native zonal complexity and functional performance., Raman imaging was used to map biochemical composition and collagen orientation of native and engineered articular cartilage. This approach provides insight into tissue complexity and can be harnessed to evaluate the quality of engineered constructs.

Imaging by Raman spectroscopy enables unparalleled label-free insights into cell and tissue composition at the molecular level. With established approaches limited to single image analysis, there are currently no general guidelines or consensus on how to quantify biochemical components across multiple Raman images. Here, we describe a broadly applicable methodology for the combination of multiple Raman images into a single image for analysis. This is achieved by removing image specific background interference, unfolding the series of Raman images into a single dataset, and normalisation of each Raman spectrum to render comparable Raman images. Multivariate image analysis is finally applied to derive the contributing 'pure' biochemical spectra for relative quantification. We present our methodology using four independently measured Raman images of control cells and four images of cells treated with strontium ions from substituted bioactive glass. We show that the relative biochemical distribution per area of the cells can be quantified. In addition, using k-means clustering, we are able to discriminate between the two cell types over multiple Raman images. This study shows a streamlined quantitative multi-image analysis tool for improving cell/tissue characterisation and opens new avenues in biomedical Raman spectroscopic imaging.

Raman spectroscopy is an inelastic light scattering technique that is capable of probing biochemical and biomolecular structures and conformations of tissue. This study aims to characterize the in vivo Raman spectroscopic properties of different normal oral tissues in the fingerprint region (800–1800 cm−1) and to assess distinctive biochemical variations of different anatomical regions in the oral cavity. A specially designed fiber-optic Raman probe with a ball lens was utilized for real-time, in vivo Raman measurements of various oral tissue sites (i.e. inner lip, attached gingiva, floor, dorsal tongue, ventral tongue, hard palate, soft palate, and buccal). The semiquantitative non-negativity-constrained least squares minimization fitting of reference biochemicals representing oral tissue constituents (i.e. hydroxyapatite, keratin, collagen, DNA, and oleic acid) and partial least squares-discriminant analysis (PLS-DA) were employed to assess the significance of inter-anatomical variability. A total of 402 high-quality in vivo oral Raman spectra were acquired from 20 subjects. The histological characteristics of different oral tissues were found to have influence on the in vivo Raman spectra and could be grossly divided into three major clusterings: (1) buccal, inner lip, and soft palate; (2) dorsal, ventral tongue, and floor; (3) gingiva and hard palate. The PLS-DA multiclass algorithms were able to identify different tissue sites with varying accuracies (inner lip 83.1%, attached gingiva 91.3%, floor 86.1%, dorsal tongue 88.8%, ventral tongue 83.1%, hard palate 87.6%, soft palate 83.3%, and buccal mucosa 85.3%), bringing out the similarities among different oral tissues at the biomolecular level. This study discloses that inter-anatomical variability is significant and should be considered as an important parameter in the interpretation and rendering of Raman diagnostic algorithms for oral tissue diagnosis and characterization. Copyright © 2011 John Wiley & Sons, Ltd.

We report a novel simultaneous fingerprint (FP) and high-wavenumber (HW) fiber-optic Raman spectroscopy developed for in vivo diagnosis of intestinal metaplasia (IM) in the stomach under wide-field endoscopic imaging. The FP/HW Raman endoscopy technique was performed to differentiate IM from normal tissues with sensitivity of 89% and specificity of 83%. This study shows the great potential of the FP/HW Raman endoscopic technique for early diagnosis of non-neoplastic gastric disease in vivo during routine endoscopic examination.

This work aims to apply a fiber-optic Raman spectroscopy technique for in vivo assessment of adenomatous polyps during clinical colonoscopy. We have developed a fiber-optic Raman endoscopic technique capable of simultaneously acquiring both the fingerprint (FP) (i.e., 800–1800 cm -1 ) and high-wavenumber (HW) (i.e., 2800–3600 cm -1 ) Raman spectra for real-time assessment of colorectal carcinogenesis. Simultaneous FP/HW Raman endoscopy provides a diagnostic sensitivity of 90.9% and specificity of 83.3% for differentiating adenoma from hyperplastic polyps, which is superior to either the FP or HW Raman technique alone. This study shows that simultaneous FP/HW Raman spectroscopy has the potential for improving early diagnosis of adenomatous polyps in vivo during colonoscopic examination.

Colorectal cancer can be prevented if detected early (e.g., precancerous polyps-adenoma). Endoscopic differential diagnosis of hyperplastic polyps (that have little or no risk of malignant transformation) and adenomas (that have prominent malignant latency) remains an unambiguous clinical challenge. Raman spectroscopy is an optical vibrational technique capable of probing biomolecular changes of tissue associated with neoplastic transformation. This work aims to apply a fiber-optic simultaneous fingerprint (FP) and high wavenumber (HW) Raman spectroscopy technique for real-time in vivo assessment of adenomatous polyps during clinical colonoscopy. We have developed a fiber-optic Raman endoscopic technique capable of simultaneously acquiring both the FP (i.e., 800-1800 cm(-1)) and HW (i.e., 2800-3600 cm(-1)) Raman spectra from colorectal tissue subsurface (

Potato is one of the most important food in the world. Rapid and noninvasive identification of potato cultivars plays a important role in the better use of varieties. In this study, The identification ability of optical spectroscopy techniques, including near-infrared (NIR) Raman spectroscopy and NIR fluorescence spectroscopy, for invasive detection of potato cultivars was evaluated. A rapid NIR Raman spectroscopy system was applied to measure the composite Raman and NIR fluorescence spectroscopy of 3 different species of potatoes (98 samples in total) under 785 nm laser light excitation. Then pure Raman and NIR fluorescence spectroscopy were abstracted from the composite spectroscopy, respectively. At last, the partial least squares-discriminant analysis (PLS-DA) was utilized to analyze and classify Raman spectra of 3 different types of potatoes. All the samples were divided into two sets at random: the calibration set (74samples) and prediction set (24 samples), the model was validated using a leave-one-out, cross-validation method. The results showed that both the NIR-excited fluorescence spectra and pure Raman spectra could be used to identify three cultivars of potatoes. The fluorescence spectrum could distinguish the Favorita variety well (sensitivity: 1, specificity: 0.86 and accuracy: 0.92), but the result for Diamant (sensitivity: 0.75, specificity: 0.75 and accuracy: 0. 75) and Granola (sensitivity: 0.16, specificity: 0.89 and accuracy: 0.71) cultivars identification were a bit poorer. We demonstrated that Raman spectroscopy uncovered the main biochemical compositions contained in potato species, and provided a better classification sensitivity, specificity and accuracy (sensitivity: 1, specificity: 1 and accuracy: 1 for all 3 potato cultivars identification) among the three types of potatoes as compared to fluorescence spectroscopy.

Raman spectroscopy represents a unique optical vibrational technique based on the fundamental premise of inelastic light scattering. Raman spectroscopy enables histopathological assessment at the biomolecular level. We have developed a fiber-optic depth-resolved near-infrared (NIR) Raman endoscopy technique integrated with on-line diagnostic algorithms for in vivo real-time epithelial diagnostics under multimodal wide-field imaging (i.e., white light reflectance (WLR), narrow-band imaging (NBI), autofluorescence imaging (AFI)) modalities. A selection of 450 patients who previously underwent Raman endoscopy (n=1900 spectra) was used to render diagnostic models for identifying gastric precancer (i.e., dysplasia) based on probabilistic partial least squares (PLS) - discriminant analysis (DA). The on-line Raman endoscopy technique was tested prospectively on (n=5) patients for real-time in vivo gastric epithelium tissue diagnosis. The fiber-optic confocal Raman endoscopic technique developed could prospectively identify gastric dysplasia in real-time with a sensitivity: 81.3% (61/75) and specificity 88.3% (188/213) on spectrum basis. On lesion basis, all dysplastic lesions were identified. This study successfully demonstrates for the first time the prospective real-time in vivo diagnosis of gastric precancer using depth-resolved Raman endoscopy.

Molecular profiling of tissue using near-infrared (NIR) Raman spectroscopy has shown great promise for in vivo detection and prognostication of cancer. The Raman spectra measured from the tissue generally contain fundamental information about the absolute biomolecular concentrations in tissue and its changes associated with disease transformation. However, producing analogues tissue Raman spectra present a great technical challenge. In this preliminary study, we propose a method to ensure the reproducible tissue Raman measurements and validated with the in vivo Raman spectra (n=150) of inner lip acquired using different laser powers (i.e., 30 and 60 mW). A rapid Raman spectroscopy system coupled with a ball-lens fiber-optic Raman probe was utilized for tissue Raman measurements. The investigational results showed that the variations between the spectra measured with different laser powers are almost negligible, facilitating the quantitative analysis of tissue Raman measurements in vivo.

Barrett’s esophagus (BE) is a metaplastic precursor of esophageal adenocarcinoma (EAC). Since the late 1970s, despite extensive efforts for prevention (eg, periodic surveillance of high-risk BE patients), EAC still has had a substantial rise in incidence rates (>350%), and is growing more rapidly than other cancers in the developed countries. Given the poor therapeutic response of symptomatic EAC, early identification of high-risk lesions (ie, dysplasia) together with therapeutic interventions is the most critical measures to improving survival rates of BE patients. However, dysplastic lesions or grossly inconspicuous cancers are endoscopically indistinguishable from the surrounding benign tissue. This is because conventional endoscopy heavily relies on visual assessment of structural and morphologic changes of the tissue surface, resulting in poor diagnostic accuracy. Existing diagnostic guidelines recommend extensive biopsy samplings (typically 4-quadrant samplings) at every 1to 2-cm interval along suspicious Barrett’s segments during endoscopic inspections of BE patients. This approach produces a vast number of negative biopsies and is clinically labor intensive and a burden to the patients. Because only a minute amount of the mucosa is sampled (as little as 5%), tissue biopsies may not accurately characterize BE segments. Foci of dysplasia in a background of intestinal metaplasia are frequently overlooked, even when the biopsies are diligently performed by the experienced endoscopists using extensive 4-quadrant biopsy protocols. Taken into account the enormous rise in incidence rates of EAC and the existing clinical challenges, the need for new advanced endoscopic modalities has never been greater. The objective targeting of high-risk tissue areas (eg, high-grade dysplasia [HGD]) with a noninvasive or minimally invasive technique could greatly reduce random biopsy sampling errors as well as health care expenses on the patients. Recent attention has thus been directed toward molecular diagnosis using optical spectroscopy and imaging. Raman spectroscopy represents a unique optical vibrational technique based on the fundamental premise of inelastic light scattering for tissue diagnosis and characterization. When an incident laser light induces a polarization change of molecules, a small proportion of incident light photons (w1 in 10) is inelastically scattered with the frequency shifts corresponding to the specific Raman active vibrational modes of the molecules in the sample. Taking advantage of the Raman spectroscopic ability of harvesting a wealth of fingerprint information from interand/or intracellular components (eg, proteins, lipids, and DNA) in cells and tissue, Raman technique has shown great promise for histopathologic assessments (ie, optical biopsy) at the biomolecular level. In the last 2 decades, there has been accumulating evidence on the accurate diagnostic capability of Raman spectroscopy through comprehensive in vitro studies. In vivo Raman endoscopic applications, however, have been limited not only by the difficulty in capturing inherently very weak tissue Raman signals, but also by the slow speed of spectral measurements (>5 s). The miniaturization of flexible fiberoptic Raman probes with depth-resolving capability that can pass down the instrument channel of medical endoscopes for effective tissue Raman light collections presents another technical challenge in endoscopic applications of Raman spectroscopy. To tackle these challenges, we have developed a novel beveled fiberoptic confocal Raman probe coupled with a ball lens capable of enhancing in vivo epithelial tissue Raman measurements at endoscopy. We present this work on in vivo clinical applications of the fiberoptic confocal Raman spectroscopy for real-time objective diagnosis of dysplasia in BE at endoscopy. The direct assessment of the biomolecular contents of epithelial cells and tissue in vivo enables the gastroenterologists to perform noninvasive or minimally invasive optical biopsies in real-time during clinical endoscopy.

We report on the development of a beveled fiber-optic confocal Raman probe coupled with a ball lens for enhancing in vivo epithelial tissue Raman measurements at endoscopy. Our Monte Carlo simulations show that by selecting a proper fiber-ball lens distance and beveled angle of collection fibers, the confocal Raman probe design can be optimized for maximizing shallower tissue Raman measurements in epithelial tissue; in addition, the ratio of epithelium to stromal Raman photons collected using an optimized confocal Raman probe is approximately 19-fold higher than that using a volume-type Raman probe. Further experiments confirm that the confocal Raman endoscopic probe developed is in favor of probing superficial tissue Raman signals from a two-layer tissue phantom as well as esophagus tissue in vivo during endoscopy. This work suggests the great potential of applying the beveled fiber-optic confocal Raman probe for improving in vivo diagnosis of precancer occurring in epithelial tissue at endoscopy.

Raman spectroscopy is a vibrational analytic technique sensitive to the changes in biomolecular composition and conformations occurring in tissue. With our most recent development of depth-resolved near-infrared (NIR) Raman endoscopy integrated with on-line diagnostic algorithms, in vivo real-time epithelial diagnostics has been realized under multimodal wide-field imaging (i.e., white- light reflectance (WLR), narrow-band imaging (NBI), autofluorescence imaging (AFI)) modalities. A selection of 43 patients who previously underwent Raman endoscopy (n=146 spectra) was used to render a robust model based on partial least squares - discriminant analysis (PLS-DA) for diagnosis of dysplasia in Barrett’s esophagus. The Raman endoscopy technique was validated prospectively on 2 new esophageal patients for in vivo tissue diagnosis. The Raman endoscopic technique could identify esophageal high-grade dysplasia in vivo with an accuracy of 85.9% (sensitivity: 91.3% (21/23): specificity 83.3% (40/48)) on spectrum basis. This study realizes for the first time depth-resolved Raman endoscopy for real-time in vivo diagnosis of dysplasia in Barrett’s epithelium at the biomolecular level.

A rapid image-guided Raman endoscopy system integrated with on-line diagnostic scheme is developed for in vivo Raman tissue diagnosis (optical biopsy) in the upper GI during clinical gastrointestinal endoscopy under multimodal wide-field imaging guidance. The real-time Raman endoscopy technique was tested prospectively on new gastric patients (n=4) and could identify dysplasia in vivo with sensitivity of 81.5% (22/27) and specificity of 87.9% (29/33). This study realizes for the first time the novel image-guided Raman endoscopy as a screening tool for real-time, online diagnosis of gastric cancer and precancer in vivo at endoscopy.

In recent years, there has been an increasing interest in the use of optical spectroscopic technologies for early cancer detection in the gastrointestinal system. Raman endoscopy is a unique fiber-optic spectroscopic technology for in vivo histopathological assessment of tissue based on endogenous biomolecules. Recent technical advances have enabled in vivo tissue Raman measurements in real-time during clinical endoscopic examination. The endoscopist can apply Raman endoscope to analyze the biomolecular signature of cells and tissues in vivo, thereby enabling noninvasive optical biopsies of suspicious lesions during ongoing endoscopy. In this article, we review the applications of Raman endoscopy in the gastrointestinal system and also discuss its potential function and role in gastrointestinal endoscopy and therapeutics. The results of the latest Raman clinical trials are summarized, illustrating the great potential of Raman endoscope technology for non-invasive, in vivo diagnosis and detection of pre-cancer and early cancer in the gastrointestinal system.

Optical spectroscopic techniques including reflectance, fluorescence and Raman spectroscopy have shown promising potential for in vivo precancer and cancer diagnostics in a variety of organs. However, data- analysis has mostly been limited to post-processing and off-line algorithm development. In this work, we develop a fully automated on-line Raman spectral diagnostics framework integrated with a multimodal image-guided Raman technique for real-time in vivo cancer detection at endoscopy. A total of 2748 in vivo gastric tissue spectra (2465 normal and 283 cancer) were acquired from 305 patients recruited to construct a spectral database for diag- nostic algorithms development. The novel diagnostic scheme developed implements on-line preprocessing, outlier detection based on principal component analysis statistics (i.e., Hotelling's T 2 and Q-residuals) for tissue Raman spectra verification as well as for organ specific probabilistic diagnostics using different diagnostic algorithms. Free- running optical diagnosis and processing time of

We report for the first time the implementation of transnasal, image-guided Raman endoscopy to directly assess Raman spectral properties of nasopharyngeal and laryngeal tissue in vivo during clinical endoscopic examinations. A rapid 785-nm excitation Raman endoscopy system, coupled with a miniaturized fiber-optic Raman probe, was utilized for real-time, in vivo Raman measurements of different anatomical locations in the head and neck. A total of 874 high-quality in vivo Raman spectra were successfully acquired from different anatomic locations of the nasopharynx and larynx [i.e., posterior nasopharynx (PN) (n=521), the fossa of Rosenmuller (FOR) (n=157), and true laryngeal vocal chords (LVC) (n=196)] in 23 normal subjects at transnasal endoscopy. Difference spectra and principal component analysis (PCA) were employed for tissue characterization, uncovering the tissue variability at the biomolecular level. The PCA-linear discriminant analysis (LDA) provides sensitivity of 77.0% and specificity of 89.2% for differentiation between PN and FOR, and sensitivity of 68.8% and specificity of 76.0% for distinguishing LVC and PN using the leave-one-subject-out, cross-validation method. This work demonstrates that transnasal, image-guided Raman endoscopy can be used to acquire in vivo Raman spectra from the nasopharynx and larynx in real time. Significant Raman spectral differences (p

Intestinal-type gastric carcinogenesis is a complex multi-step disease, and early precursors (e.g. intestinal metaplasia (IM), dysplasia) can be very challenging to identify using conventional white-light endoscopic imaging. This study aims to assess the capability of Raman spectroscopy for multi-class elucidation of intestinal-type gastric carcinogenesis sequence in vivo for improving precancer detection at endoscopy. We employ a novel image-guided Raman endoscopy technique developed for in vivo gastric tissue Raman measurement within 0.5 s during clinical endoscopic examination. We have acquired a total of 1277 in vivo Raman spectra from 83 gastric patients associated with intestinal-type carcinogenesis. In vivo Raman spectroscopy integrated with semi-quantitative spectral modelling (e.g. DNA, lipids, glycoprotein, proteins and blood) reveals the progressive changes of biochemical constituents in gastric tissue associated with preneoplastic and neoplastic transformation (i.e., IM, dysplasia and adenocarcinoma). Multi-class probabilistic partial least squares-discriminant analysis (PLS-DA) diagnostic algorithms based on in vivo Raman spectra are able to identify normal mucosa with sensitivity of 75.88% and specificity of 87.21%; IM with sensitivity of 46.67% and specificity of 87.55%; dysplasia with sensitivity of 83.33%; specificity of 95.80%, and adenocarcinoma with sensitivity of 84.91% and specificity 95.57%, respectively. This work demonstrates that Raman spectroscopy is a sensitive biomolecular probe for monitoring intestinal-type gastric carcinogenesis to realize early diagnosis and detection of precancer and early gastric cancer in vivo during clinical endoscopic examination.

Raman spectroscopy is a vibrational analytic technique sensitive to the changes in biomolecular composition and conformations occurring in tissue. With our most recent development of near-infrared (NIR) Raman endoscopy integrated with diagnostic algorithms, in vivo real-time Raman diagnostics has been realized under multimodal wide-field imaging (i.e., white- light reflectance (WLR), narrow-band imaging (NBI), autofluorescence imaging (AFI)) modalities. A selection of 177 patients who previously underwent Raman endoscopy (n=2510 spectra) was used to render two robust models based on partial least squares - discriminant analysis (PLS-DA) for esophageal and gastric cancer diagnosis. The Raman endoscopy technique was validated prospectively on 4 new gastric and esophageal patients for in vivo tissue diagnosis. The Raman endoscopic technique could identify esophageal cancer in vivo with a sensitivity of 88.9% (8/9) and specificity of 100.0% (11/11) and gastric cancers with a sensitivity of 77.8% (14/18) and specificity of 100.0% (13/13). This study realizes for the first time the image-guided Raman endoscopy for real-time in vivo diagnosis of malignancies in the esophagus and gastric at the biomolecular level.

The aim of this work was to evaluate the biochemical foundation and clinical merit of multimodal image-guided Raman endoscopy technique for real-time in vivo diagnosis of cancer in the esophagus during clinical endoscopic examinations. A novel fiber-optic Raman endoscopy system was utilized for in vivo esophageal Raman measurements at 785 nm laser excitation within 0.5 second under the multimodal wide-field endoscopic imaging (white light reflectance (WLR) imaging, narrow-band imaging (NBI) and autofluorescence imaging (AFI) guidance. A total of 75 esophageal tissue sites from 27 patients were measured, in which 42 in vivo Raman spectra were from normal tissues and 33 in vivo Raman spectra were from malignant tumors as confirmed by histopathology. The biomolecular modeling (non-negativity-constrained least-squares minimization (NNCLSM) utilizing six basis reference spectra from the representative biochemicals ( i.e., actin, collagen, DNA, histones, triolein and glycogen) were employed to estimate the biochemical compositions of esophageal tissue. The resulting diagnostically significant fit coefficients were further utilized through linear discriminant analysis (LDA) and leave-one tissue site-out, cross validation method to develop diagnostic algorithms for esophageal cancer diagnosis. High-quality in vivo Raman spectra in the range of 800–1800 cm−1 can be acquired from normal and cancerous esophageal mucosa in real-time under multimodal endoscopic imaging guidance. Esophageal cancer tissue showed distinct Raman signals mainly associated with cell proliferation, lipid reduction, abnormal nuclear activity and neovasculation. The fit coefficients for actin, DNA, histones, triolein, and glycogen were found to be most significant for construction of the LDA diagnostic model, giving rise to an accuracy of 96.0% ( i.e., sensitivity of 97.0% and specificity of 95.2%) for in vivo diagnosis of esophageal cancer. This study demonstrates that multimodal image-guided Raman endoscopy technique in conjunction with biomolecular modeling has promising potential for the real-time, in vivo diagnosis and detection of esophageal cancer during clinical endoscopic examination.

This study aims to evaluate the diagnostic utility of the combined near-infrared (NIR) autofluorescence (AF) and Raman spectroscopy for improving in vivo detection of gastric cancer at clinical gastroscopy. A rapid Raman endoscopic technique was employed for in vivo spectroscopic measurements of normal (n=1098) and cancer (n=140) gastric tissues from 81 gastric patients. The composite NIR AF and Raman spectra in the range of 800-1800 cm(-1) were analyzed using principal component analysis (PCA) and linear discriminant (LDA) to extract diagnostic information associated with distinctive spectroscopic processes of gastric malignancies. High quality in vivo composite NIR AF and Raman spectra can routinely be acquired from the gastric within 0.5s. The integrated intensity over the range of 800-1800 cm(-1) established the diagnostic implications (p=1.6E-14) of the change of NIR AF intensity associated with neoplastic transformation. PCA-LDA diagnostic modeling on the in vivo tissue NIR AF and Raman spectra acquired yielded a diagnostic accuracy of 92.2% (sensitivity of 97.9% and specificity of 91.5%) for identifying gastric cancer from normal tissue. The integration area under the receiver operating characteristic (ROC) curve using the combined NIR AF and Raman spectroscopy was 0.985, which is superior to either the Raman spectroscopy or NIR AF spectroscopy alone. This work demonstrates that the complementary Raman and NIR AF spectroscopy techniques can be integrated together for improving the in vivo diagnosis and detection of gastric cancer at endoscopy.

A near-infrared Raman spectroscopy system integrated with multimodal endoscopic imaging has been developed for in vivo diagnosis and detection of gastric malignancies during clinical gastroscopic examinations. In this study, 1238 high-quality in vivo Raman spectra in the range 800–1800 cm -1 were acquired from gastric normal and malignant tissue within 0.5 second from 81 patients under the guidance of white-light (WL), narrow-band (NB) and autofluorescence (AF) imaging during clinical endoscopy. Significant differences in Raman spectral shapes and intensities between normal and malignant gastric mucosal tissue were observed, particularly in the spectral ranges 800–900, 1000–1100, 1250–1450 and 1600–1800 cm -1 , which primarily contain signals related to proteins, nucleic acids and lipids. Partial least squares discriminant analysis (PLS-DA) could identify in vivo Raman spectra of neoplasia with a sensitivity of 82.9% and specificity of 88.9% using leave-one-tissue site out cross validation. This study demonstrates that in vivo Raman spectroscopy in conjunction with multimodal endoscopic imaging modalities holds a great promise for improving the early diagnosis of gastric malignancies.

The aim of this study was to evaluate the clinical utility of an image-guided Raman endoscopy technique for in vivo differential diagnosis of benign and malignant ulcerous lesions in the stomach. A rapid-acquisition image-guided Raman endoscopy system with 785 nm laser excitation has been developed to acquire in vivo gastric tissue Raman spectra within 0.5 s during clinical gastroscopic examinations. A total of 1102 in vivo Raman spectra were acquired from 71 gastric patients, in which 924 Raman spectra were from normal tissue, 111 Raman spectra were from benign ulcers whereas 67 Raman spectra were from ulcerated adenocarcinoma. There were distinctive spectral differences in Raman spectra among normal mucosa, benign ulcers and malignant ulcers, particularly in the spectral ranges of 800-900, 1000-1100, 1245-1335, 1440-1450 and 1500-1800 cm(-1), which primarily contain signals related to proteins, DNA, lipids and blood. The malignant ulcerous lesions showed Raman signals to be mainly associated with abnormal nuclear activity and decrease in lipids as compared to benign ulcers. Partial least squares-discriminant analysis (PLS-DA) was employed to generate multi-class diagnostic algorithms for classification of Raman spectra of different gastric tissue types. The PLS-DA algorithms together with leave-one tissue site-out, cross validation technique yielded diagnostic sensitivities of 90.8%, 84.7%, 82.1%, and specificities of 93.8%, 94.5%, 95.3%, respectively, for classification of normal mucosa, benign and malignant ulcerous lesions in the stomach. This work demonstrates that image-guided Raman endoscopy technique associated with PLS-DA diagnostic algorithms has for the first time promising clinical potential for rapid, in vivo diagnosis and detection of malignant ulcerous gastric lesions at the molecular level.

We first report on the implementation of a novel narrow-band image-guided Raman endoscopy technique for in vivo diagnosis of gastric dysplasia. High-quality in vivo Raman spectra can be acquired from normal and dysplastic gastric mucosal tissue within 0.5 sec under narrow-band image (NBI) guidance at gastroscopy. Significant differences are observed in in vivo Raman spectra between normal (n=54) and dysplastic (n=18) gastric tissue from 30 gastric patients, particularly in the spectral ranges of 825 to 950, 1000 to 1100, 1250 to 1500, and 1600 to 1800 cm(-1), which primarily contain signals related to proteins, nucleic acids, and lipids. The multivariate analysis [i.e., principal components analysis (PCA) and linear discriminant analysis (LDA)], together with the leave-one tissue site-out, cross validation on in vivo gastric Raman spectra yields a diagnostic sensitivity of 94.4% (1718) and specificity of 96.3% (5254) for distinction of gastric dysplastic tissue. This study suggests that narrowband image-guided Raman endoscopy associated with PCA-LDA diagnostic algorithms has potential for the noninvasive, in vivo early diagnosis and detection of gastric precancer during clinical gastroscopic examination.

This study aims to evaluate the clinical utility of image-guided Raman endoscopy for in vivo diagnosis of neoplastic lesions in the stomach at gastroscopy. A rapid-acquisition image-guided Raman endoscopy system with 785-nm excitation has been developed to acquire in vivo gastric tissue Raman spectra within 0.5 sec during clinical gastroscopic examinations. A total of 1,063 in vivo Raman spectra were acquired from 238 tissue sites of 67 gastric patients, in which 934 Raman spectra were from normal tissue whereas 129 Raman spectra were from neoplastic gastric tissue. The swarm intelligence-based algorithm (i.e., ant colony optimization (ACO) integrated with linear discriminant analysis (LDA)) was developed for spectral variables selection to identify the biochemical important Raman bands for differentiation between normal and neoplastic gastric tissue. The ACO-LDA algorithms together with the leave-one tissue site-out, cross validation method identified seven diagnostically important Raman bands in the regions of 850-875, 1,090-1,110, 1,120-1,130, 1,170-1,190, 1,320-1,340, 1,655-1,665 and 1,730-1,745 cm(-1) related to proteins, nucleic acids and lipids of tissue and provided a diagnostic sensitivity of 94.6% and specificity of 94.6% for distinction of gastric neoplasia. The predictive sensitivity of 89.3% and specificity of 97.8% were also achieved for an independent test validation dataset (20% of total dataset). This work demonstrates for the first time that the real-time image-guided Raman endoscopy associated with ACO-LDA diagnostic algorithms has potential for the noninvasive, in vivo diagnosis and detection of gastric neoplasia during clinical gastroscopy.

A near-infrared Raman spectroscopy system integrated with multimodal endoscopic imaging has been developed for the early noninvasive in vivo diagnosis and detection of gastric malignancies. High-quality in vivo Raman spectra in the range 800-1800 cm -1 can be acquired from gastric normal and premalignant (dysplastic) mucosal tissue within 1 second under the guidance of white-light and narrow-band gastroscopic imaging during clinical gastroscopy. Prominent differences in Raman spectral shapes and intensities are observed between normal and dysplastic gastric mucosal tissue, particularly in the spectral ranges 800-900, 1250-1450 and 1600-1800 cm -1 , which primarily contain signals related to proteins, nucleic acids and lipids. The empirical intensity ratio algorithm I 875 /I 1450 classifies in vivo Raman spectra of dysplasia with a sensitivity of 100% and specificity of 100%. Our initial investigations show that in vivo Raman spectroscopy in conjunction with multimodal endoscopic imaging modalities holds a great promise for improving the early diagnosis of gastric malignancies.

We report on the development of a novel multiplexing Raman spectroscopy technique using a single laser light together with a volume phase holographic (VPH) grating that simultaneously acquires both fingerprint (FP) and high wavenumber (HW) tissue Raman spectra at endoscopy. We utilize a customized VPH dual-transmission grating, which disperses the incident Raman scattered light vertically onto two separate segments (i.e., −150 to 1950 cm −1 ; 1750 to 3600 cm −1 ) of a charge-coupled device camera. We demonstrate that the multiplexing Raman technique can acquire high quality in vivo tissue Raman spectra ranging from 800 to 3600 cm −1 within 1.0 s with a spectral resolution of 3 to 6 cm −1 during clinical endoscopy. The rapid multiplexing Raman spectroscopy technique covering both FP and HW ranges developed in this work has potential for improving in vivo tissue diagnosis and characterization at endoscopy.

Objective: Raman spectroscopy is a vibrational spectroscopic technique capable of probing the changes of biochemical compositions and conformations occurring in tissue. The main aims of this research are to develop a novel image-guided Raman endoscopic technique to conduct In Vivo optical biopsy and to evaluate its diagnostic potential for improving early diagnosis and detection of gastric precancer and cancer In Vivo. Methods: With our successful development of an integrated Raman endoscopy and multimodal wide-field imaging modalities (white-light reflectance (WLR), narrow-band imaging (NBI), autofluorescence imaging (AFI)) technique, a total of 427 In Vivo gastric sites have been measured from 61 gastric patients during clinical gastroscopy. Histopathological examination revealed that 281 Raman spectra were from normal mucosa, 17 spectra were from dysplasia and 129 spectra were from neoplasia. The partial least squares (PLS) discriminant analysis (DA) was employed to develop multi-class diagnostic algorithms for tissue diagnosis and classification. Results: High quality In Vivo Raman spectra can be acquired in real-time within 0.5 sec from the stomach during clinical endoscopic examination. Significant differences in Raman spectra between normal, dysplastic and neoplastic gastric tissue are observed reflecting the pathological transformation associated with cancer (e.g., increased mitotic activity and elevated DNA content). The PLS-DA modeling on the In Vivo gastric Raman spectra achieved the diagnostic sensitivities of 95.1%, 100.0%, and 89.8%; and specificities of 93.0%, 95.0%, and 93.8%; respectively, for identifying normal, dysplastic and neoplastic gastric tissue In Vivo. Conclusions: This study demonstrates that the image-guided Raman endoscopy technology has promising clinical potential for rapid In Vivo diagnosis and detection of gastric premalignancies and malignancies during routine gastroscopy. ________________________________ *Correspondence to: Dr. Zhiwei Huang, Department of Bioengineering, Faculty of Engineering, National University of Singapore. Tel: +656516-8856; Fax: +656872-3069; E-mail: biehzw@nus.edu.sg

Raman spectroscopy is an optical vibrational technology capable of probing biomolecular changes of tissue associated with cancer transformation. This study aimed to characterize in vivo Raman spectroscopic properties of tissues belonging to different anatomical regions in the upper gastrointestinal (GI) tract and explore the implications for early detection of neoplastic lesions during clinical gastroscopy. A novel fiber-optic Raman endoscopy technique was utilized for real-time in vivo tissue Raman measurements of normal esophageal (distal, middle, and proximal), gastric (antrum, body, and cardia) as well as cancerous esophagous and gastric tissues from 107 patients who underwent endoscopic examinations. The non-negativity-constrained least squares minimization coupled with a reference database of Raman active biochemicals (i.e., actin, histones, collagen, DNA, and triolein) was employed for semiquantitative biomolecular modeling of tissue constituents in the upper GI. A total of 1189 in vivo Raman spectra were acquired from different locations in the upper GI. The Raman spectra among the distal, middle, and proximal sites of the esophagus showed no significant interanatomical variability. The interanatomical variability of Raman spectra among normal gastric tissue (antrum, body, and cardia) was subtle compared to cancerous tissue transformation, whereas biomolecular modeling revealed significant differences between the two organs, particularly in the gastroesophageal junction associated with proteins, DNA, and lipids. Cancerous tissues can be identified across interanatomical regions with accuracies of 89.3% [sensitivity of 92.6% (162∕175); specificity of 88.6% (665∕751)], and of 94.7% [sensitivity of 90.9% (30∕33); specificity of 93.9% (216∕230)] in the gastric and esophagus, respectively, using partial least squares-discriminant analysis together with the leave-one tissue site-out, cross validation. This work demonstrates that Raman endoscopy technique has promising clinical potential for real-time, in vivo diagnosis and detection of malignancies in the upper GI at the molecular level.