Tear Analytical Model Based on Raman Microspectroscopy for Investigation of Infectious Diseases of the Ocular Surface

Abstract

Purpose.: To establish a tear analytical model by using Raman microspectroscopy to assess ocular surface diseases associated with infectious pathogens.

Methods.: The authors applied confocal Raman microspectroscopy based on the drop-coating deposition method on Ti/Au-coated glass slides to obtain sample spectra from different types of tears, including simplified synthetic tears (SSTs), SSTs mixed with microbes, and human tears. Raman spectra were sampled by a line-mapping procedure and classified into three groups by three different zones in a dried teardrop. To determine the tear model with optimal discrimination, the spectra of the three zones were compared using spectral morphology and principal component analysis. Finally, the optimal tear model was verified by comparing the Raman spectra of human teardrops of patients with ulcerative keratitis and bacterial infection with those of patients without any identifiable infection.

Results.: Nonhomogeneous intensities of Raman spectra collected by a line-mapping sampling procedure were found in different locations of a dried teardrop. This might have been caused by coffee-ring formation and ferny crystallization phenomena. The normalized spectra in the central zone have better discriminative potential than those in the other zones, ring zone, and transitional zone, tested by pure SSTs or SSTs with microbes. The tear model based on normalized Raman spectra in the central zone was discriminative for patients with ulcerative keratitis in the presence or absence of infection.

Conclusions.: Raman spectra in the central zone of a dried teardrop may serve as useful spectral fingerprints for investigating ocular surface diseases with or without infection.

Infectious diseases of the ocular surface can involve the eyelid, ^1,2 conjunctiva, ^3,4 and cornea. ^5,6 They may cause red eye and ocular irritation in mild symptomatic cases. However, in severe cases, they may cause intense pain and even threaten vision, such as in the case of infectious ulcerative keratitis. These diseases have a wide spectrum of influence on our quality of life at cosmetic, psychophysiologic, and even socioeconomic levels. ^{7 –9} Prompt and corrective diagnosis for the eradication of pathogenic microbes is the best strategy for these diseases. 

Conventional methods for pathogenic diagnosis include cytologic examination and microbial culture. Cytologic examination is a rapid aid for clinical diagnosis, but it has low sensitivity because of its requirement of a high detection limit (≥10⁵ counts/mL). ^{10 –13} Ocular samples yield a low microbial culture rates ^14,15 ; however, microbial culture remains the gold standard for microbial infection diagnosis. ¹⁶ The possible causes are low microbial counts in ocular surface samples, poor sample transportation, ^17,18 and erratic choice of culture media. ^19,20 In addition, some microbial cultures are time consuming. ²¹ Identification of an alternative method to aid clinical diagnosis is necessary to promote the diagnostic rate and efficiency for infectious diseases of the ocular surface. 

Teardrop, which reserves the shed microbes and their metabolites, may be a good target for differential diagnosis of infectious diseases of the ocular surface. Changes in intrinsic tear components and their concentrations may also reflect the defense mechanisms of the ocular surface against microbes. ^{22 –28} One of the defensive proteins in tears, lysozyme, can be used to distinguish patients with herpes simplex virus (HSV) keratitis from subjects without it. ²⁹ Therefore, we attempted to target tears as a scraping-free and less invasive way to diagnose infectious diseases on the ocular surface. 

Raman spectra were first discovered by Raman and Krishnan in 1928. ³⁰ A novel technique to measure Raman spectra has been developed to obtain specific information about the molecular composition of matter through Raman microspectroscopy, as illustrated in Supplementary Figure S1. In brief, the monochromic laser light interacts with the vibrating molecules or the excited electrons in the sample, resulting in the energy of laser photons being upshifted (anti-Stoke Raman scattering) or downshifted (Stoke Raman scattering). The change or shift in energy indicates molecular information about the photon modes in the system. Elastic scattering photons (Rayleigh scattering) are filtered out while the rest of photons (Raman scattering) are dispersed into a photon detector. The molecular information in the sample is displayed by a computerized system for spectral analysis. Raman microspectroscopy can detect in real time the inelastic scattering of light caused by molecular vibration or rotation after laser excitation to qualitatively and quantitatively detect the molecular composition of matter, even in the aqueous state. ³¹ Many studies have pointed out its potential applications in different biomedical investigations ^{32 –41} and in discriminating microbes and their metabolites. ^{42 –47} Raman microspectroscopy was recently shown to be a powerful tool for detecting tear components. Zhang et al. ⁴⁸ have shown that higher efficiency of drop coating deposition Raman (DCDR) can be achieved by deposition of the aqueous samples on the substrates with specific surface properties (e.g., hydrophobic surface). Filik and Stone ^49,50 showed that the Raman signal obtained from 1.5 μL dried teardrop is adequate for detecting major tear components by using the DCDR technique with a high signal-to-noise ratio. 

In this study, we aimed to establish a tear analytical model based on the DCDR technique to fill the gap between basic tear Raman spectra studies and preclinical application tests, especially for differentiating between ocular surface diseases with infection and those without infection. 

Methods

Experimental Design

We applied confocal Raman microspectroscopy (Thermo Fisher Scientific Inc., Pittsburgh, PA) based on the DCDR method to Ti/Au-coated glass slides to collect sample spectra from different tear samples. The tear samples include simplified synthetic tears (SSTs), SSTs mixed with microbes, and human tears. SSTs with or without microbes could be used as pathologic tear models for ocular surface diseases with or without infection. Raman spectra were formatively sampled using a line-mapping procedure and classified as three groups by three different zones in a dried teardrop. To determine which pathologic tear model had optimal discrimination ability, we compared the spectra of the three zones on the basis of spectral morphology and principal component analysis (PCA) results. Finally, we tested the optimal model by comparing the Raman spectra of teardrops from patients with acute ulcerative keratitis and bacterial infection with those of patients without any identifiable infection. 

Preparation of SSTs

For the construction of pathologic tear models, we applied a simplified formula of tears according to a previous work with slight modification. ⁵¹ SSTs were compared with human tears and are briefly summarized in Table 1. Because lipocalin is not a commercially available product, ⁵² only lactoferrin and lysozyme were used as backbone proteins in SSTs. The activity of these two tear proteins may be representative responses of tears to different ocular surface diseases and may be potential indicators of them. ^{53 –58} In brief, we used sterile deionized water to prepare all the stock solutions for the different SST concentrations. The stock solutions included 20 mg/mL lactoferrin, 20 mg/mL lysozyme, 26 mg/mL albumin, 6 mg/mL IgA, 12 mg/mL urea, and 42 mg/mL sodium bicarbonate. Taking SST3.0 as an example, we mixed 150 μL lactoferrin and 150 μL lysozyme with 50 μL other stock solutions and added 500 μL sterile deionized water. SSTs were mixed with reference strains Pseudomonas aeruginosa ATCC 27853 and Staphylococcus aureus ATCC 14957, which were used as infectious model tears. In brief, we mixed 150 μL lactoferrin and 150 μL lysozyme with 50 μL other stock solutions as described to create the stock synthetic tears. After obtaining fresh bacterial growth on a solid medium (approximately 3 days after inoculation), we vortexed the bacterial suspensions to break up any clumps and create a homogenous bacterial solution. The 2 × 10⁹ cfu/mL bacterial suspensions were prepared using sterile deionized water according to the McFarland standard using a turbidity meter. Finally, we mixed the 2 × 10⁹ cfu/mL bacterial suspensions and the stock synthetic tears in a ratio of 1:1. The infectious model tears were incubated for approximately 20 minutes at 36.5°C before Raman acquisition was performed. 

Table 1.

 

Components of Simplified Synthetic Tears

Components	Human Tears	Synthetic Tears	Chemicals	References
Lactoferrin	1.84–2.73 mg/mL	X = 1.0, 2.0, 3.0 mg/mL	Sigma-L4765	56 –58
Lysozyme	1.6–2.46 mg/mL	X = 1.0, 2.0, 3.0 mg/mL	Sigma-L6876	56, 58, 65
Lipocalin	1.23–2.89 mg/mL	NA	NA	56, 66
Albumin	1.3 mg/mL	1.3 mg/mL	Sigma-A2153	56
IgA	0.23–0.3 mg/mL	0.3 mg/mL	Sigma-19889	56, 62
Urea	6.27–22.7 mM	10 mM	Ameresco-0378	67, 68
Bicarbonate	26 mM	26 mM	Sigma-S5761	68

 NA, not available.

Collection of Clinical Tears

All procedures involving human subjects adhered to the Declaration of Helsinki and were approved by the Committee of Medical Ethics and Human Experiments of Chang Gung Memorial Hospital. Informed consent was obtained from each subject in Chang Gung Memorial Hospital, Kaohsiung Medical Center. Tears were gently collected from all participants under 16× magnification by a slit lamp using disinfected microcapillary tubes (Hirschmann Laborgeräte GmbH & Co. KG, Eberstadt, Germany) with self-modified head curvature. Approximately 5 to 10 μL tear samples were collected and stored at 4°C for up to 2 weeks before spectral acquisition of Raman microscopy was performed. 

Definition of Ulcerative Keratitis with or without Infection

Ulcerative keratitis with infection is defined as any positive finding identified in the microbial survey protocol for confirmation of infectious keratitis for originally suspected infectious keratitis. This protocol includes cytologic evaluation, microbial culture, and HSV PCR of the corneal scraping samples. For cytologic evaluation, the Gram and acid-fast stains were used for immediate microscopic evaluation. If the diameter or the maximal length of the ulcer was <1.5 mm, the cytologic evaluation was omitted to preserve the samples for microbial culture or HSV PCR (in-house real-time PCR for HSV types 1 and 2). Routine microbial culture includes aerobic culture (blood agar and chocolate blood agar), anaerobic culture (CDC anaerobe 5% sheep blood agar, phenylethyl alcohol blood agar, and bacteroides bile esculin agar), mycobacterial culture (Lowenstein-Jensen slant), and fungal culture (Sabouraud's dextrose agar). On any specific suspicion of atypical corneal ulceration at the initial visit, we also performed acanthamoeba culture (nonnutrient agar with Escherichia coli overlay) and HSV PCR. In the event of an unexpected clinical course of treatment, we repeated the standard microbial survey and added acanthamoeba culture or HSV PCR. The definition of ulcerative keratitis without infection is no positive findings by the above microbial confirmation protocol for presumed microbial keratitis at the initial visit, smooth tapering of antimicrobial agents, and no occurrence of unexpected events until smoothing of the ocular surface. 

In this study, at the model proof stage, we adopted acute ulcerative keratitis with bacterial infection as the infectious model of the ocular surface. The noninfectious model of ocular surface is ulcerative keratitis without infection, as defined. 

Dried Teardrop Preparation for Raman Spectral Acquisition

We prepared dried tear samples for acquisition of Raman spectra based on the DCDR method with a little modification. ^50,55 In brief, 1.5 μL tears was dropped onto a clean glass slide coated with 30/150 nm Ti/Au. To avoid a turbulent effect caused by airflow, the slide was placed in a closed chamber and allowed to dry naturally at room temperature with 48% relative humidity. According to works based on the DCDR theory, a dried teardrop presents various patterns shown in Figure 1. The shape of a dried teardrop may depend on the distribution of superimposed proteins, lipids, and other substances between their edges and centers (Fig. 1a). Therefore, we classified a dried teardrop as being in 1 of 3 zones and collected the spectra through a line-mapping procedure (Fig. 1b; three mapping lines, each with approximately 100 acquisition points, and three groups of spectral data separated by three zones). A 1.5-μL dried SST drop is illustrated in Figure 1c. Using the maxima and minima Raman intensities from each acquisition point, we could outline the association between the dried teardrop locations and the spectral intensities (Fig. 1d). 

Poorly formed dried teardrops were excluded from further spectral collection under the concern of interference from abnormal teardrops morphology, such as accidental air bubble formation during dropping of the teardrop on the Ti/Au glass slide. A collection of tear Raman spectra was conditioned as a 633-nm He-Ne laser for sample excitation, the laser power at sample position around 7 mW, and a 50× objective lens with a connected charge-coupled device (CCD) for scattered light collection. A grating of 600 lines/mm was used to disperse the scattered light. The Raman shift was calibrated using a signal of 520 cm⁻¹ generated from a silicon wafer. The laser exposure time of 10 seconds with signal accumulation of 3 times in a range of 400 to 3400 cm⁻¹ was set for each spectral acquisition point. 

Postprocessing of the spectra was conducted as described elsewhere ^{59 –61} with some modification for the nonhomogeneous display of a dried teardrop, as shown in Figure 1. In brief, the autofluorescence background was removed by subtraction of a fifth-order polynomial fit from the full spectrum. This is one of the acknowledged methods to remove the autofluorescence background of biological samples optimally and simply from the experimental and computational points of view. We applied this method to remove the slowly varying background that is assumed to be autofluorescence in tear samples. Each spectrum was corrected by its root mean square in a range of 1800 to 2000 cm⁻¹. Spectra in the same zone were sorted by spectra integrals in the range of 700 to 1800 cm⁻¹ for removal of the 10% superior and inferior outliner spectra. Under the principle of stratified random sampling, a collection of 10 representative spectra in each zone was obtained for each teardrop. Each representative spectrum was normalized by the intense band at approximately 1002 cm⁻¹, and the mean was centered to get a 0 mean for spectral pattern analyses and statistics, as illustrated in Supplementary Figure S2. 

Statistical Analysis

We collected Raman signals in the range 700 to 1800 cm⁻¹ for the following analysis. We used technical computing software (MatLab 2008a; The MathWorks Inc., Natick, MA) as the multivariable analytic tool for PCA to characterize the differences among different tear groups ^50,52,54 and a graphics tool (Office Excel 2007; Microsoft Corporation, Redmond, WA) for spectral visualization and comparison. For clinical tears, we used two-factor analysis of variance for hypothetical tests for different sampled spectra from the same teardrop. 

Results

A fused macroscopic image of a dried human teardrop, a spectral sampling and classification blueprint for a teardrop, and a fused image of a simplified synthetic tear (SST1.0) along with its line-mapping Raman spectra by extreme intensities are illustrated in Figure 1. According to the DCDR theory and earlier reports, ^50,55 a dried teardrop presents a coffee-ring band in its margin and many ferny crystals internal to the inner ring edge. From our observation of 1.5-μL dried drops from human tears (Fig. 1a) and SSTs (Fig. 1c), we estimated the length ratio of two ring zones in the diameter of a dried teardrop to be approximately 1/10. From the maxima spectral intensity by line-mapping sampling, a basin-like gross intensity profile with nonhomogeneous fine patterns was observed (Fig. 1d). By investigating the relationship between the anatomy of a dried teardrop and the extreme intensity profile of the spectra, we may understand them through linking bands straddling Figures 1c and 1d. Both coffee-ring and ferny crystals may influence spectral intensities. The junction between the ring and the center of a dried teardrop showed an intermediate intensity profile between those in the ring and those in the center. Therefore, we further stratified the portion of ferny crystals as having one center zone and two transitional zones by diameter ratios of 3/10 and 6/10 for each teardrop (Fig. 1b). 

Figure 1.

 

(a) Fused photo of a dried teardrop obtained from a healthy volunteer. (b) Spectral sampling of a dried teardrop. C, central or C zone (pink); T, transitional or T zone (white); R, ring or R zone (yellow). Red solid circle: boundaries between different zones. Red dashed line: line-mapping sampling points. (c) A fused image of a dried teardrop from SST1.0. (d) Two spectra graphed by maxima (solid line) and minima (dashed line) of all sampled spectra in a line-mapping sampling procedure for tear Raman spectra. Linking bands straddle (c) and (d) to highlight the positional relationship between the Raman spectral intensities and the tear zones.

A standard spectra sampling procedure using three mapping lines was used, and 10 representative spectra in the three zones of two synthetic tears are shown in Figure 2. The three zones in order of the highest to the lowest gross spectral intensity (the integral of Raman intensity in the range 700-1799 cm⁻¹) are the R zone, the T zone, and the C zone. The synthetic tears in the R zone of the higher concentration pairs (SST3.0) showed greater intensities in the range of 1200 to 1375 cm⁻¹ than those of the lower concentration pairs (SST1.0). We further compared the normalized spectra for SST3.0 (Fig. 3). The gross spectral patterns are more similar in the three zones than those before normalization, as shown in Figure 2. Comparing the smoothness of a spectral profile, spectra in the R zone (Fig. 3c) are the smoothest but those in the C zone (Fig. 3b) are the most uneven. The spectra in the C zone seem to show many more tiny featured peaks burst from the smoothing spectral profile than those in the other zones. By applying the PCA, we found that the aggregation effect of the spectra illustrated by PC1 and PC2 was better for the R zone and C zone than for the T zone (Fig. 3d). The intrazonal spectral variation was greater in the T zone than in the other zones; thus, it may be not an appropriate fingerprint zone for comparison between different teardrops. 

Figure 2.

 

Comparison of the sampled tear Raman spectra in the R zone (a, d), T zone (b, e), and C zone (c, f) for SST1.0 (a–c) and SST3.0 (d–f). The 1200- to 1375-cm⁻¹ range highlighted in subfigures (a) and (d) was used to aid the visual comparison. R.I., relative intensity; R.S., Raman shift (cm⁻¹).

Figure 3.

 

Normalized spectra in the T zone (a), C zone (b), and R zone (c) of a dried teardrop from SST3.0. PCA of the three zones (d). R.I., relative intensity; R.S., Raman shift (cm⁻¹).

As shown in Figure 4, we used synthetic tears with or without microbes to compare the discriminative potential of two candidates in the fingerprint zone of a dried teardrop. From the similarity of gross spectral patterns in three tear formulas, the R zone (Fig. 4b) is more similar than the C zone. Spectral patterns in the R zone may be more difficult to differentiate between different tear groups than those in the C zone (Fig. 4a). From PCA, the separation effect of spectra illustrated by PC1 and PC2 was also better for the C zone than for the R zone. Both showed that the optimal fingerprint zone in a dried teardrop was likely the C zone. 

Figure 4.

 

Normalized spectra (n = 10 for each group). Solid lines: mean spectra. Dashed lines: superior and inferior limits of the 95% CI in the C zone (a) and the R zone (b) for SST3.0, 10⁹ cfu/mL Staphylococcus aureus ATCC 14957 in SST3.0 (SST3.0pSA), and 10⁹ cfu/mL Pseudomonas aureus ATCC 27853 in SST3.0 (SST3.0pPA). For PC1 and PC2 by PCA, the C zone (c) has a better discriminative potential than does the R zone (d). R.I., relative intensity; R.S., Raman shift (cm⁻¹).

As shown in Figure 5, we found that the mean spectrum in the C zone of patients with infectious ulcerative keratitis (IUK group) was different from that of patients with noninfectious ulcerative keratitis (NIUK group). The difference spectra shown in Figure 5a highlighted some gross (absolute difference ≥0.25) and fine (absolute difference <0.25 and confidence line without crossing the 0 difference line) Raman shift segments with discriminant potentials. Gross difference segments included 850 to 860 cm⁻¹, 980 to 1000 cm⁻¹, 1225 to 1485 cm⁻¹, and 1520 to 1680 cm⁻¹. Fine difference segments included 745 to 755 cm⁻¹, 930 to 940 cm⁻¹, 1030 to 1050 cm⁻¹, 1100 to 1110 cm⁻¹, 1120 to 1130 cm⁻¹, 1210 to 1220 cm⁻¹, and 1485 to 1520 cm⁻¹. With the exception of 980 to 1000 cm⁻¹, the mean differences between IUK and NIUK were almost positive in these described segments. From PCA, we may find that the principal vector PC1 in Figure 5b is similar to the gross difference of mean spectra of (IUK-NIUK) in Figure 5a. PC3 is more similar to the fine mean spectra difference than is PC2. PC1 is much more relevant than are the other PCs for explaining most of the variant Raman shifts in all spectra of the two groups (Fig. 5c). We further found that the two groups can be distinguished by the illustration of PC1 and PC3, two principal components with statistical significance (P < 0.0001 for PC1; P = 0.0004 for PC3) (Fig. 5d). In contrast, there was no statistical significance between the two groups in PC2, PC4, PC5, and PC6. 

Figure 5.

 

(a) Normalized spectra (n = 10 for each case) in the C zones of dried teardrops from six patients—three with infectious ulcerative keratitis (IUK; two with IUK caused by Pseudomonas aeruginosa, one with IUK caused by Streptococcus pneumonia) and three patients with noninfectious central or paracentral ulcerative keratitis (NIUK; two with trauma, one a contact lens wearer). (b) Principal component vectors from PC1 to PC6. (c) Relevancies from PC1 to PC6. (d) The two groups can be discriminated through PC1 and PC3 by PCA. R.I., relative intensity; R.S., Raman shift (cm⁻¹).

Discussion

In this study, we proposed a tear analytical model for future preclinical application tests for tear Raman spectra, especially for differentiating ocular surface diseases with or without infection. The diagnostic rate and efficiency of conventional cytologic examination and microbial culture are inappropriate for quick diagnostic aid in clinical practice. Although PCR-based diagnostic techniques for microbial detection promote the diagnostic rate, ^14,15 labor-intensive requirements may limit their clinical application in the diagnosis of atypical cases. In addition, the time cost still cannot be acceptable for rapid diagnostic aid in the clinic. Our approach using tear Raman spectra might be a good alternative diagnostic aid in clinical settings. Its potential to promote diagnostic quality in the clinical setting makes it worthwhile to establish a practical tear Raman model. 

This study filled the gap between basic tear Raman spectra studies and clinical tests before proceeding toward the development of a formal diagnostic tool, especially in terms of differentiating ocular surface diseases with or without infection. Reyes-Goddard et al. ⁵¹ compared the surface-enhanced efficacy of different substrates in distinguishing synthetic tears with HSV from those without HSV. They then applied linear discriminant analysis to classify different groups on the basis of principal components. The mean diagnostic sensitivity and specificity were estimated between 75% and 80% for the silver mirror reaction glass slides and the gold thin film. Filik and Stone ⁵⁰ showed that 1.5 μL tears from a healthy human subject is sufficient for Raman signal detection with a high signal-to-noise ratio by using the DCDR method. They attempted to analyze the major tear component distribution in a dried teardrop through the correlation between the principal components of PCA and these components, including proteins, urea, bicarbonate, and lipids. They further used least-squares fitting to estimate the distribution of proteins of different sizes in the R zone of pooled tears of four healthy subjects. They proved that the tear DCDR spectra were highly reproducible, though line mapping in the R zone showed significant radial variation, especially toward the outer edge of the ring. We used gold thin film as the 1.5-μL tear DCRD substrate for synthetic tears, with or without pathogenic microbes in the ocular surface, as our model. The C zone in a dried teardrop might be a good spectral fingerprint zone because the spectra in this area showed lower dispersity than did the R zone in the same tear group and higher dispersity than did the R zone in the different tear group. The fact that the R zone is inferior to the C zone for distinguishing infectious diseases from noninfectious diseases may be a result of significant radial variation of the spectra in the R zone, as demonstrated by Filik and Stone. ⁵² We further verified that ulcerative keratitis with or without microbial infection could be differentiated using DCDR spectra of the C zone in six tear samples of clinical patients. 

We were unable to clearly elucidate the mechanism by which tear components rendered tear spectra discriminative in the two groups of ulcerative keratitis. As reviewed in other studies, ^26,50,62 the composition of human tears is very complex, especially in diseased tear samples. There are no perfect model tears or even pooled tears collected from a few healthy human subjects. Direct mixing of model tears with microbes cannot completely simulate the complex tear changes from ocular surface responses to microbial invasion. Hence, we prepared synthetic tears by using a simplified formula composed of available major tear components. The simplified tears may help us to easily investigate the interaction between tears and microbes. It may be difficult to retrieve pure Raman shift from Raman spectra of a dried teardrop for direct discrimination between infected tears and noninfected tears. Any sampled spectrum indicated a composition of different concentrations of all tear components separated in a specific zone or location. ⁵² This study cannot provide direct evidence to explain why spectra in the C zone have optimal discriminant potential. We hypothesize that several low-concentration products may be released from interactions between tears and microbes. These products may be covered by the major high-concentration tear proteins in the R zone because of the coffee-ring effect. They may be relatively uncovered in the C zone. Therefore, several possible aspects may warrant future exploration. More clinical teardrops of patients with ulcerative keratitis and other infectious diseases of the ocular surface should be collected to test this hypothesis and to verify the discriminant power of tear DCDR spectra for these diseases. In addition, the development of a new tear collection system for avoiding possible stimulation during tear sampling and for performing direct in situ analysis is important for future clinical tests or applications. ⁶³ A controllable, rapid, and user-friendly drying procedure should also be developed to obtain a good-quality dried teardrop. Enhancement of the discrimination effect of the C zone through a prefiltering procedure may help retain microbes or products released from the reaction between tear components and microbes. 

In conclusion, this study suggests that the tear DCDR spectra in the C zone might serve as promising spectral fingerprints for discriminating infectious diseases of the ocular surface. Even though most Raman research topics are still at the proof-of-concept stage, this technique shows a great potential for clinical decision-making in the near future. ⁶⁴ The tear analytical model fills the gap between basic tear Raman studies and preclinical application tests based on this novel technology. 

Supplementary Materials

Figure sf01, PDF - Figure sf01, PDF 

Figure sf02, PDF - Figure sf02, PDF 

Footnotes

 Supported by Chang Gung Research Proposal Grant CMRPG880701; National Science Council Grants NSC 99-2218-E-029-004 and NSC99-2314-B-182A-030-MY3; and National Medical Research Proposal of Chang Gung Memorial Hospital Grant NMRPG896031.

Footnotes

 Disclosure: M.-T. Kuo, None; C.-C. Lin, None; H.-Y. Liu, None; H.-C. Chang, None

The authors thank Tsung Chain Chang for supplying bacterial reference strains and the Ministry of Education, Taiwan, under the NCKU Project of Promoting Academic Excellence and Developing World Class Research Centers (R017) for supplying Raman spectroscopy equipment.