Skip to main content
NCBI home page
Communications Biology logoLink to Communications Biology
. 2026 Mar 3;9:511. doi: 10.1038/s42003-026-09744-2

Label-free in vivo molecular profiling of the human retina by non-resonant Raman spectroscopy

Ryan Sentosa 1, Milana Kendrisic 1, Matthias Salas 1, Matthias Eibl 2, Michael Schmitt 3, Vasyl Shynkar 4, Wim de Jong 5, Heiko Stino 6, Andreas Pollreisz 6, Michael Kempe 2, Jürgen Popp 3,7, Tilman Schmoll 8, Marco Andreana 1, Angelika Unterhuber 1, Wolfgang Drexler 1, Rainer A Leitgeb 1,✉
PMCID: PMC13066014  PMID: 41775856

Abstract

Early detection of retinal molecular biomarkers is crucial for addressing the unmet clinical need to prevent irreversible neural tissue damage in ophthalmic and neurodegenerative diseases. Among emerging molecular sensing techniques, non-resonant Raman spectroscopy stands out as a naturally label-free and noninvasive method, offering rich biochemical information. However, in vivo detection of non-resonant Raman spectra from retinal tissue has proven to be challenging so far. Previous studies have reported conflicting results, likely due to overwhelming pigment autofluorescence. In this study, we identified the optic nerve head as the optimal retinal location for acquiring non-resonant Raman spectra in the molecular fingerprint region. Through longitudinal intra-subject measurements, we revealed dynamic changes in the molecular composition. Furthermore, a comparative study across age groups enabled the identification of molecular alterations associated with aging. These findings establish a critical foundation for utilizing non-resonant Raman spectroscopy as an early diagnostic tool for the detection of molecular biomarkers associated with ophthalmic and neurodegenerative diseases.

Subject terms: Retina, Translational research, Optic nerve diseases, Raman spectroscopy, Optical imaging

In vivo non-resonant Raman spectroscopy of the optic nerve head enables label-free molecular fingerprinting of the human retina, revealing longitudinal and age-related changes toward early diagnosis of ophthalmic and neurodegenerative diseases.

Introduction

The demand for attaining real-time molecular information on the human eye is growing, driven by research and clinical needs1,2. Molecular alterations at the cellular or tissue level can provide subtle hints of early disease progression that precede structural manifestations. For example, the leading cause of visual impairment among the elderly, age-related macular degeneration3, remains asymptomatic until early signs of blurred vision due to drusen formation are detected by ophthalmic examination4. As an extension of the brain, the posterior segment of the eye, the retina, and the optic nerve also serve as an optically accessible “window to the brain”5. There is increasing evidence linking the ocular pathology to central nervous disorders including but not limited to stroke6, multiple sclerosis7, Parkinson’s disease8,9 and Alzheimer’s disease10,11. The current gold standard of ophthalmic examination, optical coherence tomography (OCT)12, has been clinically accepted by physicians13 and provides unprecedented 3D visualization of retinal layers with micrometer resolution. OCT has also been used to detect structural changes (e.g., retinal layer thickness and presence of hyper-reflective retinal spots) associated with aging14, systemic and neurodegenerative diseases15. Despite its remarkable diagnostic value in ophthalmology, it relies on visible structural changes. However, such apparent damage to neural tissues, as is observed in the retina, tends to be irreversible. Therefore, early diagnosis is needed to prevent lifelong vision loss or neural disorders, as well as the related increase in healthcare costs. Furthermore, in the context of neurodegenerative diseases, the observed structural changes of the neural retina lack differential diagnostic strength16. Enhancing OCT with a molecular-specific technique17, while maintaining its noninvasive and label-free characteristics, holds the potential to improve the early differential diagnosis and targeted treatment of ophthalmic and central nervous system (CNS)-related diseases.

One of the most basic yet versatile molecular sensitive methods, Raman spectroscopy (RS), has been recognized as a powerful tool in biomedical science. RS provides a molecular fingerprint of a sample without the need for external labels18. The Raman effect is based on the inelastic scattering of incident photons, which involves an exchange of energy that leads to excitation of molecular vibrational states. Thus, the intensity of the Raman-scattered light is proportional to the concentration of specific chemical bonding of molecules. Therefore, Raman spectroscopy reveals highly complex information at the tissue level due to overlapping signals among macromolecules, i.e., carbohydrates, lipids, proteins and nucleic acids19. However, by analyzing the Raman spectra using chemometric methods20, researchers have demonstrated its potential to detect subtle differences between healthy and diseased tissues, such as in cancer, cardiovascular, neurodegenerative, and ophthalmic diseases21–24.

RS applications in ophthalmology can be categorized according to the excitation wavelength that determines the information content of the acquired Raman spectra25,26. The resonant RS method, which uses excitation wavelengths in resonance with electronic chromophores, displays selective vibrations associated with the resonantly excited electronic chromophore. This resonant RS approach has been demonstrated as a means to study the carotenoid levels (i.e. lutein and zeaxanthin) in the macular region27,28 and the cytochromes and hemoglobin in the optic nerve head region of murine eyes29. Resonant RS, however, leads to simplified Raman spectra dominated by a few molecular vibrations determined by the (a priori known) electronic chromophores. Meanwhile, Raman spectra recorded with an excitation wavelength outside the electronic absorption band of any specific chromophore (non-resonant RS), provide rich information on the molecular composition of tissues. Numerous studies have demonstrated its potential for the study of ex vivo animal30–39 and human32,40,41 retina tissue, but mostly in flat-mounted scenarios. Evidence suggests that the autofluorescence from retinal pigments42 prevents non-resonant RS measurement of intact eyes32,37,41,43. To date, only a few studies have reported in vivo non-resonant RS measurements of the retina43–45. Our research team demonstrated the feasibility of interpreting the non-resonant Raman spectra from a living albino rat retina that lacks retinal pigments, thus minimizing the autofluorescence contribution45. A recent study assessed in vivo molecular changes in the retina of healthy subjects and multiple sclerosis patients using non-resonant RS44. The authors decomposed the acquired Raman spectra from the central retina into 10 specific molecules. Considering the influence of autofluorescence from the retinal pigment epithelium and the complex molecular constituents, without any proof of relevant Raman peaks from the defined molecules and interpretable acquired Raman spectra, the results presented in the study remain questionable.

In this study, we demystified the feasibility of using non-resonant RS for in vivo retinal measurements in humans. To address the inconsistencies in the current literature, we analyzed the Raman spectra from different locations on the retina. We found that the autofluorescence from the pigmented area of the retina prevents further interpretation of the Raman spectra. In contrast, the optic nerve head (ONH) region revealed rich Raman-related information from biological tissues, i.e. carbohydrates, proteins, lipids and nucleic acids (Fig. 1A). Additionally, we investigated the temporal variability of the acquired non-resonant Raman spectra through intra-subject measurements. To achieve this, we consistently measured the Raman spectra at the temporal horizontal sector of the neuroretinal rim of the ONH46. Despite an overall high reproducibility of Raman features in a single subject, we identified the spectral regions and the associated molecular components that changed significantly over time. In a study involving a larger cohort, we observed age-related molecular alterations of the neural ocular tissue, exemplifying the sensitivity provided by Raman spectroscopy. Our findings are crucial to consider in the design of future studies targeting molecular biomarkers in ophthalmic and CNS-related diseases.

Fig. 1. Measurement concept of the in vivo non-resonant Raman spectroscopy of the human retina.

Fig. 1

Open in a new tab

A Label-free and non-invasive examination of the eye by using a monochromatic laser source to study the molecular composition of the retina. The illustration shows a Raman excitation spot targeted on the optic nerve head, revealing the molecular composition of unmyelinated axons. This illustration is based on an original digital artwork (PENUP) by Marie C. Nguyen. B The developed device for in vivo measurement of the human retina. C The optomechanical scheme of the measurement head, excluding the bulk OCT/Raman components (laser sources, interferometer and detectors). The measurement device combines IR fundus, fixation light, OCT and Raman spectroscopy modules. The sub-images show a typical OCT cross section of human retina and a typical Raman spectrum from biological tissue. The optomechanical scheme is created using Solid Edge. Credit: Thorlabs Inc. (RC08APC-P01, DMLP1000R, PF05-03-P01, F280FC-780, P3-1064Y-FC, M42L01), and Edmund Optics GmbH (#49-956).

Results

Identification of a suitable measurement site for in vivo non-resonant Raman spectroscopy of the human retina

The measurement concept is depicted in Fig. 1. Reliable measurements were achieved by integrating the RS system into an OCT and infrared (IR) fundus imaging device to facilitate subject alignment. Our measurement device complies with international laser safety standards and is approved for clinical use under the European Union Medical Device Regulation (2017/745).

To test our hypothesis that no non-resonant Raman-related information can be obtained from the central retina, we analyzed Raman spectra from a single subject (healthy male, 54 years old (yo)) at different locations across the retina. The selection of the measurement location was guided by using the en face OCT image as shown in Fig. 2A. We minimized eye motion of the subject by using a fixation target. The Raman measurements started at position #1 and ended at position #20 (#1 to #10 in horizontal steps and #11 to #20 in vertical steps), with 3 consecutive acquisitions at each position. The measurement positions #6 and #16 are located at the same position, being the fovea. The measurement position #1 (ONH) has been selected carefully to avoid large blood vessels, which are clearly visible on the OCT en face image (see Supplementary Fig. S1). Each Raman spectrum in Fig. 2B is an average of three accumulations (@10 s integration time). The Raman spectrum from all measurement locations outside of the ONH region shows a prominent autofluorescence background. Meanwhile, the Raman spectrum at the ONH region (position #1) shows the least autofluorescence (see Supplementary Fig. S2). Next, we acquired Raman spectra from one position outside of the ONH region (star-marked in Fig. 2A) and from the ONH region (measurement position #1 in Fig. 2A) with high accumulation numbers (30 ×20 s at each position). We recorded and subtracted the background Raman signal (black curve, Fig. 2C) from each Raman spectrum to remove the contribution of the optical system, dark current, and fixed pattern noise. Figure 2C shows the averaged Raman spectra from these two positions after smoothing using a Savitzky-Golay filter (5th order, 51 data points). Figure 2D shows the Raman spectra after we subtracted the autofluorescence contribution by using a 5th order polynomial baseline correction47,48. The averaged Raman spectrum from outside of the ONH region (top curve, Fig. 2D) does not show distinguishable Raman signals typically expected in biological tissue, such as the CH₂ deformation mode (~1450 cm–1) or the C = C stretching mode in unsaturated lipids and amide I mode of proteins (~1660 cm–1). Those peaks have been observed in our previous ex vivo study on retinal tissue with the autofluorescence-inducing pigment structures removed41. The remaining features observed in the spectrum of the central region cannot be reliably assigned and is likely a result of subtraction artifacts due to dominant fluorescence background. Meanwhile, referring to our previous studies37,41,45, the typical biological tissue bands mentioned above can be easily identified in the averaged Raman spectrum from the ONH region (lower curve, Fig. 2D). This finding suggests that the dominance of autofluorescence prevents the identification of non-resonant Raman information from spectra obtained from outside of the ONH region by using 785 nm excitation wavelength. For this reason, we strategically focused our following studies of the in vivo application of non-resonant Raman spectroscopy on the ONH region. We describe the molecular origin of the obtained Raman spectra from the ONH region in Supplementary Text and the tentative assignment of the Raman bands in Supplementary Table S1.

Fig. 2. In vivo non-resonant Raman spectroscopy across the human retina.

Fig. 2

Open in a new tab

A En face OCT (~60° FOV) of a healthy volunteer, with 10 horizontal and 10 vertical points representing the Raman measurement locations. Position #1 represents the optic nerve head region (ONH). B Intensity-corrected Raman spectra of the 20 measurement points shown in (A). Position #1 shows the least autofluorescence background. Meanwhile, positions #2 to #20 show significantly higher autofluorescence. C Mean Raman spectra (n = 30) from the star-marked region in (A), the ONH region and the background signal. The dashed lines show the baseline correction estimation. D Processed Raman spectra of the star-marked region and the ONH region. Vertical dashed lines indicate the typical biological tissue bands of phenylalanine (~1003 cm–1), CH2 deformation (~1450 cm–1) and amide I (~1658 cm–1), plotted as the mean and standard deviation. Scale bar: 1 mm.

Intra-subject measurements of the optic nerve head region using non-resonant Raman spectroscopy

After clarifying the proper measurement location, we tested the temporal variability of the measured Raman spectra from the human ONH, by acquiring Raman spectra from a single subject (healthy male, 56 yo). For this task, we performed seven measurements over four months. We acquired at least 30 Raman spectra (@ 20 s integration time) for each measurement time point. Raman spectra that showed significant contribution of autofluorescence, which were caused by subject’s fixation loss were excluded for further analysis. Then, we averaged 3 consecutive Raman spectra to improve the signal to noise ratio. Therefore, the dataset consisted of 7 days x 10 Raman spectra (ntotal = 70 Raman spectra). We processed the Raman spectra using the extended multiplicative signal correction (EMSC) method (6th order)48, using the average Raman spectrum across all measurements as the reference spectrum (smoothed and baseline corrected based on a 5th order polynomial). Additionally, we removed the Raman spectra that were missing relevant Raman peaks around the phenylalanine, amide III, CH2 deformation, and amide I bands (see “Methods” section). The linear correlation coefficient value of these 70 Raman spectra is shown in Fig. 3A. The correlation value of the valid Raman spectra is 0.81 ± 0.04, demonstrating good reproducibility of the Raman acquisition. The number of remaining Raman spectra for each measurement (n, @20 s integration time) is shown in Fig. 3B. We then averaged the processed Raman spectra and calculated the standard deviation for each measurement set at a given time point as shown in the upper plot in Fig. 3B. The Raman bands associated with significant differences across the seven measurement time points are highlighted in the lower plot in Fig. 3B (Kruskal-Wallis test, p<0.05 in gray and p<0.01 in red). The gray and red dashed horizontal lines show the p-value below 0.05 and 0.01, respectively. The box plots in Fig. 3C show the relative intensity variation of the significant Raman bands from seven measurement times (calculated as relative percentage to the median value of the first measurement). The tentative molecular origin for each band (p<0.01, with width>1 cm–1) is shown in Fig. 3C.

Fig. 3. Intra-subject analysis of the non-resonant Raman spectra from the optic nerve head (ONH) region.

Fig. 3

Open in a new tab

A Correlation coefficient of 70 Raman spectra collected from 7 measurement days (each spectrum is equivalent to 3 times average) of a single subject. B Mean Raman spectra and its standard deviation from the ONH region acquired from a single subject at seven different measurement times. Measurement number shows the total of acquired Raman spectra without averaging. The Raman bands showing significant changes (Kruskal-Wallis test) are outlined in gray (p < 0.05) and red shades (p < 0.01). The overall p-value is plotted in logarithmic scale. C The relative Raman intensity (relative to the median value of the first measurement) of the selected significant bands (p < 0.01, with >1 cm–1 width) in (B) and the assigned molecular origin (taken from Supplementary Table S1). ACTG: adenine, cytosine, thymine, guanine. Each circular marker represents an average of 3 spectra. Each box plot is shown as median (central mark), 25th and 75th percentiles. Its whiskers extend to the most extreme data points, excluding the outlier. Plus-sign markers represent outliers.

In the higher wavenumber region (~1000–1800 cm–1), the Raman band in the amide I wavenumber region contributes to some variation (lower plot, Fig. 3B) with pvalue<0.05. This band cannot be assigned exclusively to the conformational modes of proteins (contribution of β-sheet, α-helix, and β-turn). The vibrational modes of the carbonyl group of the nucleotides cytosine, thymine and uridine are also present around 1647 cm–1, 1664 cm–1, and 1681 cm–1, respectively49. Additionally, the contribution of C = C stretching vibration from unsaturated lipids can also be found ~1655–1670 cm–1 50,51. Only carbohydrates show no signature between 1500 cm–1 and 1800 cm–1 52. The Raman bands associated with ester bonds (~1720–40 cm–1) can be assigned exclusively to lipids50, particularly phospholipids which are the building blocks of the phospholipid bilayer found in the axonal plasma membrane (maximum variation ~1718 cm–1 in Fig. 3C). The variation around ~1349 cm–1 can be associated with the Fermi resonance doublet from tryptophan53,54. The Raman band associated to CH2 scissoring mode of carbohydrates, proteins and lipids is present around 1450 cm–1.

In the lower wavenumber region (<1000 cm–1), a high degree of variation can be observed from the ~568 cm–1 band in Fig. 3B, which originates from the skeletal bending modes, involving carbohydrates, lipids, proteins and nucleic acids49,51–53. Meanwhile, the Raman band ~883 cm–1 has a complex origin, including β-saccharides52, tryptophan54, and sphingomyelin51,53.

In vivo non-resonant Raman spectroscopy analysis of the human optic nerve head region for different age groups

Next, we investigated whether the measured Raman spectra of the ONH reveal molecular changes associated with aging. We acquired Raman spectra from 3 different age groups: below 45 yo (nsubject = 11 (4 females, 7 males), nspectra = 70, nvalid = 53), between 45 and 65 yo (nsubject = 3 (males), nspectra = 119, nvalid = 108), and above 65 yo (nsubject = 7 (2 females, 5 males), nspectra = 70, nvalid = 59). Each measurement (nvalid) represents a 3 × 20 s acquisition. We processed this dataset using the EMSC algorithm (6th order), using the mean spectrum of all valid measurements as the reference spectrum (baseline corrected using a 5th order polynomial). Figure 4A shows the mean and standard deviation of the Raman spectra for each age group, as well as the mean intensity difference of each group pair. The Raman bands that show significant differences between each group pair (Mann–Whitney U Test) are shown as magenta (p < 0.001, positive difference), cyan (p < 0.001, negative difference), and gray (p < 0.05) on top of the respective difference spectra.

Fig. 4. In vivo non-resonant Raman spectroscopy measurements from different age groups.

Fig. 4

Open in a new tab

A Mean and standard deviation of the Raman spectra from the age groups below 45 years (in dark blue), between 45 and 65 years (in light blue), and above 65 years (in yellow). The mean spectral difference of each pair is shown with its statistical significance (Mann–Whitney U test, p < 0.05 in gray, p < 0.001 in cyan and magenta, for negative and positive differences, respectively). Each measurement number is equivalent to 3 × 20 s acquisition. B Principal component analysis (PCA) score plot of the 2nd and 3rd principal components based on the subject’s age group. C Box plots of the PCA scores of the 2nd and 3rd principal components based on the age group (Mann–Whitney U Test, n.s. ≡ p > 0.05, *** p < 0.001). Each box plot is shown as median (central mark), 25th and 75th percentiles. Its whiskers extend to the most extreme data points, excluding the outlier. Plus-sign markers represent outliers. D Raman intensity of the selected Raman peaks that show high correlation according to the actual age (Spearman’s rank correlation (ρ), ***p < 0.001). Each data point is equivalent to 3 × 20 s of measurement. The red curves that represent linear fitting are not analyzed and serve only for illustration. The subplots show the mean intensity of the 3 age groups around the selected significant peaks. n.s. not significant.

To visualize the separation among the 3 groups, we used the Raman bands with the highest significance (p < 0.001) shown in Fig. 4A for the principal component analysis (PCA). The first principal component, which describes the mean value of the input spectra, is not analyzed. The PCA score plot of the 2nd and 3rd principal components in Fig. 4B shows a clear centroid shift according to age. Some points representing the (45–65) yo group overlap more with the other groups. These changes can also be visualized by the box plots of the 2nd and 3rd principal component scores from each age group, as shown in Fig. 4C. The first 10 components describe ~95% of the data (Supplementary Table S2). The PCA score plots of the 4th and 5th principal components are shown in Supplementary Fig. S3, showing no clear separation among the 3 groups. The 2nd and 3rd principal component (PC) loadings describe the majority of the intensity difference between the (45–65) yo and >65 yo groups and between the <45 yo and (45–65) yo groups, respectively. Meanwhile, the mean intensity difference between the >65 yo and <45 yo groups can only be explained by a combination of both the 2nd and 3rd principal components (see Supplementary Fig. S4). Finally, we used Spearman’s rank correlation test to examine the relationship between age and the significant Raman peaks shown in Fig. 4A. The complete analysis is given in Supplementary Table S3. The mean Raman spectra of the selected peaks with high correlation values to age are shown in Fig. 4D. The Raman bands associated to lipids around ~603 cm–1 (i.e., cholesterol and phosphatidylserine) and ~1745 cm–1 (ester bond in lipids and phospholipids) show a positive correlation with age50,51. Meanwhile, in the non-resonant regime, the Raman band around ~751 cm–1 can be assigned to the phenylalanine ring breathing, C-N stretching of phospholipids, and O-P-O vibration mode of DNA53. The Raman band around ~1660 cm–1 can be assigned to amide I of proteins and C = C stretching of unsaturated lipids49,55. Both ~751 cm–1 and ~1660 cm–1 Raman bands show negative correlation with age.

Discussion

Despite its potential, the ophthalmic application of non-resonant RS has not been widely exploited in clinical settings. A possible hurdle is the contribution of autofluorescence from the retinal pigments, as observed in previous studies37,41. The approach we present here, examining first the Raman spectra from different locations on the human retina in vivo (Fig. 2D), contradicts recent findings44. Our results suggest that the non-resonant Raman spectra from the pigmented part of the retina do not carry interpretable Raman information. An equally important finding is the unveiled molecular information content of Raman spectra from the ONH region. Anatomically, the retinal pigment epithelium and the pigment-rich choroid do not extend to the ONH region56. The acquired Raman spectra from this region show considerable complexity but are highly relevant to the molecular constituents of the ONH and the neural tissue itself.

In addition to the identification of a potential measurement site for non-resonant RS on the retina (the ONH), we further found that the Raman signal can be reliably reproduced from a single subject with a high correlation value among repeated measurements (Fig. 3A). With an estimated Raman spot diameter of 100 µm, we expected an averaging effect that might improve reproducibility. To support this assumption, we equipped the system with a fixation target to help the subjects stabilize their eye during the measurement. Undeniably, Raman signal variation is inherent to the dynamic nature of any living tissue. The axons at the optic nerve head region connect retinal ganglion cells to the soma (lateral geniculate nucleus) and transport molecules such as filamentous proteins, mitochondria, secretory granules and multivesicular bodies57. The Raman bands associated with carbohydrates (Fig. 3B, C) might also suggest the presence of astrocytes, which act as glycogen storage on the ONH57. The ONH region exhibits a high density of mitochondria to support the high energy demands of retinal ganglion cells58. Any significant axonal transport disruption or mitochondrial dysfunction might alter the local molecular composition at the ONH and might be associated with disease or aging59,60.

Since aging is associated with molecular changes in tissues61–63, there is increasing interest in using Raman spectroscopy for age-related analysis of biological tissues64. Therefore, we investigated the influence of age on the acquired molecular fingerprint spectra as shown in Fig. 4. The observed increase in Raman signals associated with lipids (e.g., cholesterol and phospholipids) with age may suggest a shift in the lipid composition of retinal tissue (Fig. 4D). During the measurement, we avoided the locations with major blood vessels as clearly seen from the fundus image (see Supplementary Fig. S1), thus minimizing potential contributions of blood constituents. Hence, we may safely hypothesize that the major contribution of our Raman spectrum at this specific band originates from cellular lipids rather than blood lipids. Imbalances or alterations of membrane lipid composition might affect the cellular signaling pathways that lead to disease progression and aging65,66. Given the complex origin of the two Raman bands around 751 cm–1 and 1660 cm–1, we can only hypothesize that the other contributors (i.e., protein complexes and nucleic acids) apart from lipids are the cause of the observed negative trend with age. In addition to age, we also expect the influence of the gender on the acquired Raman spectra67,68, which has to be investigated in future studies involving a large subject cohort.

Non-resonant Raman spectroscopy probing of neural tissue at the ONH offers several practical advantages. First, we can ensure the accuracy of the measurement position by using the fluorescence background of the acquired spectra as a navigation tool. If the subject changes the direction of gaze and the Raman excitation spot is directed toward the retinal pigments, this is easily detected by the significant increase in the autofluorescence background after each acquisition of 20 seconds. The change of gaze might also happen involuntarily, even when using a fixation target, due to drift and randomly-oriented microsaccades. However, the related motion amplitudes are below 0.3° (of visual field) which is approximately the spot size of the Raman excitation beam (~100 µm)69. When we observed an increase in autofluorescence background, we shortly realigned the system and asked the subject to re-fixate. Then, we resumed the measurements. In practice, we needed to repeat up to 5 measurements for the subjects >65 yo due to fixation loss and autofluorescence increase, which was still well tolerated. Meanwhile, this number was relatively low (less than 3) for subjects <45 yo. During data collection, we also discarded Raman spectra with different shapes of autofluorescence background, which might originate from the optic cup or blood vessels. Second, the ONH area is also known as the blind spot due to the absence of photoreceptors. In this study, the measurement subjects confirmed that their comfort was not compromised because the Raman excitation laser appeared as dim scattered light. Moreover, we found that a total accumulation time of up to 10 min was still well tolerated by the study participants.

Apart from the advantages described above, our approach has some limitations. First, RS at the tissue level provides complex information. A specific assignment of the Raman bands to certain molecules might be too speculative given the rich molecular composition of the tissue and the overlapping Raman bands. Careful selection of Raman bands helps reduce this complexity. For example, the Raman peak around 603 cm–1 might be safely assigned to specific molecules, i.e., cholesterol, phosphatidylserine, and cytosine. The Raman peaks between 1700 cm–1 and 1750 cm–1 might strongly be attributed to phospholipids. A reasonable approach is to spectrally decompose the Raman spectra into the major macromolecules (carbohydrates, lipids, proteins, and nucleic acids)70. However, this approach might benefit more from a small excitation spot size in a Raman imaging setting. Furthermore, more detailed decomposition should be supported by reasonable evidence of molecular concentration, such as a correlation analysis with mass spectrometry data. Second, we hypothesize that our Raman spectra include the contribution of the vitreous humor (see Supplementary Text for the depth of field estimation). This can be observed from the broadening of amide I band due to the Raman active bending mode at ~1645 cm–1 from water (H-O-H bending)71,72. Therefore, we could not confidently decompose the amide I band into its secondary structures (α-helix, β-sheet, β-turns, β-strands, and random coil)73. Third, by focusing our study on the ONH region, we lost the rich molecular information hidden in the central retina itself. Autofluorescence suppression by using a time-gated approach in the non-resonant regime might be a suitable solution for assessing the central retina74, once the potential light hazard of pulsed laser excitation can be mitigated to ensure patient safety in a clinical setting.

This study provides a better understanding of the limitations and future potential of non-resonant RS for investigating ophthalmic and CNS-related diseases. Importantly, future studies should also consider the influence of aging on the investigated tissue and avoid misinterpreting the results directly as disease biomarkers, as aging has a significant impact on the local biochemistry.

Methods

Ethics approval and in vivo study protocol

All ethical regulations relevant to human research participants were followed. Our study was approved by the ethics committee of the Medical University of Vienna (ethics committee number: 1322/2021) and complied with the European Union Medical Device Regulation (EU) 2017/745, article 70 (7) (a) (reference number: 100099096). A total of 21 healthy subjects (23–79 years old) recruited from the Department of Ophthalmology of the Medical University of Vienna were included in this study. On the day of the study, the subjects gave their informed consent and underwent a standard ophthalmic examination including slit lamp bio-microscopy. Exclusion criteria were previous ocular trauma, media opacities, poor central vision and fixation capability and a history of retinal disease. If the subject passed the screening test, the pupil was dilated with one drop of Tropicamide (Mydriaticum “Agepha”, Agepha Pharma, Senec, Slovakia). Then, we measured the subjects using our custom-built measurement device as shown in Fig. 1B.

Experimental setup

We adapted the multimodal ophthalmic imaging device from the previous work45, considering the pupil size difference between rat and human eyes. Figure 1C illustrates the main optomechanical components of the device we developed. In short, we have combined IR fundus imaging, optical coherence tomography (OCT) and Raman spectroscopy (RS) systems into a single measurement device. The role of the IR fundus imaging and the OCT system is to support the RS measurement and to guide the measurement location selection. The light from the IR fundus imaging system (730 nm central wavelength) was combined with the OCT/RS system by using a custom-made short-pass dichroic mirror (~750 nm cutoff wavelength). To reduce the motion artifact during the in vivo measurement, we installed a custom-made fixation target (~570 nm central wavelength) inside the IR fundus imaging system. We combined the optical path of the OCT and the RS system by using a long-pass dichroic mirror (DMLP1000R, Thorlabs Inc., ~1000 nm cutoff wavelength). The OCT system was based on a swept source laser (Insight Photonic Solutions, 1060 nm central wavelength, 70 nm bandwidth and 87 kHz sweep rate). We increased the beam size (up to ~2.5 mm) in front of the subject’s eye by using a longer focal length collimator (RC08APC-P01, Thorlabs Inc.), giving a theoretical lateral resolution of ~8 µm on the human retina (see Supplementary Text for the spot size estimation). We used a pair of 14 mm aperture galvanometer scanners (6231H, Cambridge Technology) to raster scan the OCT light and control the excitation spot of the RS system.

The RS system used a diode laser (0785MU0100MF, Innovative Photonic Solutions, Inc., 785 nm) as the Raman excitation light. This wavelength was chosen as a compromise between reduced tissue autofluorescence, Raman scattering intensity (~λ-4) and detector efficiency. We connected the excitation laser to a multimode scrambler (MMS-201, Luna, Inc.) to obtain a uniform illumination beam profile and to avoid mode pattern hot spots, which was particularly important for human eye measurements. The Raman excitation beam was collimated with an aspheric lens (AL2520-B, Thorlabs Inc., 20 mm focal length) and passed through the laser line filter (LL01-785-12.5, Semrock) to reduce any generated background plasma and secondary emissions. We split the Raman excitation and Raman detection beams using an edge filter (LP01-785RU-25, Semrock, operated at ~7° angle of incidence). Using a custom-made iris that provided ~1.3 mm beam size in front of the subject’s eye, we estimated a~100 µm Raman excitation spot on the retina (see Supplementary Text for the spot size estimation). The Raman-scattered photons from the subject’s eye were collected by an achromatic lens (#49-956, Edmund Optics GmbH) and delivered to a Raman spectrometer (iHR320, HORIBA Scientific) by using a graded-index multimode fiber. A scientific grade charge-coupled device CCD camera (Syncerity, HORIBA Scientific) was used as a detector. Using the 300 lines/mm grating and 200 µm slit size, the expected spectral resolution of the Raman spectrometer system according to the datasheet was 20.8 cm–1. We measured an optical power of ~1 mW for the OCT and Raman excitation beam at the subject’s cornea, which complies with the laser safety standard (DIN EN ISO 15004-2:2007).

Raman spectra processing

The acquired Raman spectra were partially processed using LabSpec 6 and MATLAB software. With LabSpec 6, we applied cosmic spike removal and removed the optical system and CCD-related background (laser on, but no sample) from the acquired spectra. We calibrated the signal intensity, by using a white light source (SLS201L, Thorlabs). The background data is unique because it was performed after each in vivo subject measurement. In MATLAB, the spectra were cropped to the Raman wavenumber range of 401 cm–1–2000 cm–1 for further processing. We smoothed the raw Raman spectra using the Savitzky-Golay filter (5th order, 75 kernel size). Next, we implemented the extended multiplicative signal correction (EMSC) method with a baseline-corrected reference spectrum. We found that a 5th order polynomial based on the method described by Lieber and Mahadevan-Jansen was sufficient to correct the baseline of the reference spectrum47. The choice of the reference spectrum for each analysis was explicitly mentioned in the “Results” section. To ensure the reliability of the following analysis, we excluded the Raman spectra that did not show relevant peaks from the main Raman bands, including the phenylalanine (~1003 cm–1), amide III (~1250 cm–1), CH2 deformation (~1450 cm–1) and amide I (~1650 cm–1) bands (using isoutlier function with ‘quartiles’ parameter in MATLAB). The principal component analysis (PCA) used the wavenumber range between 400 cm–1 and 1800 cm–1 to include only the fingerprint region. The PCA algorithm was based on singular value decomposition with 10 principal components.

Statistics and reproducibility

The Raman spectrum from the ONH and from outside of the ONH region (Fig. 2D) is shown as the mean and standard deviation across 30 measurements.

For the intra-subject analysis, the Raman spectrum from each measurement is shown as the mean and standard deviation (the number of measurements for each spectrum is shown in Fig. 3A). Statistical significance of the intra-subject analysis was determined using the Kruskal-Wallis test (kruskalwallis function in MATLAB, p < 0.05 in gray, p < 0.01 in red). The boxplots of the statistically significant bands are plotted relative to the first measurement.

For the age-dependent analysis, the Raman spectrum from each age group is shown as mean and standard deviation. The normality of the data was tested using the Anderson-Darling test (adtest function in MATLAB, see the results in Supplementary Table S4). Statistical significance of each group-pair was determined using the Mann–Whitney U test (ranksum function in MATLAB; Fig. 4A, p < 0.05 in gray, p < 0.001 in red and cyan for positive and negative differences, respectively). Statistical significance of the PCA score of each group-pair was determined using the Mann–Whitney U test (Fig. 4C and Supplementary Table S5). The statistical dependence between the selected Raman bands (defined by the statistical test in Fig. 4A) and age was determined using the Spearman’s rank correlation coefficient (corr function with ‘Type’, ’Spearman’ in MATLAB; Fig. 4D). Linear regression in Fig. 4D was done using the fit function in MATLAB using 1st order polynomial. The difference spectrum fitting (with PC2 and PC3 loadings) in Fig. S4 was done using the lsqcurvefit function in MATLAB.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Transparent Peer Review file (2.9MB, pdf)
Supplementary information (1.1MB, pdf)
42003_2026_9744_MOESM3_ESM.pdf (27.6KB, pdf)

Description of Additional Supplementary Files

Supplementary Data 1 (1.3MB, xlsx)
Reporting Summary (1.4MB, pdf)

Acknowledgements

The authors acknowledge: A. Amelink and P. Bussink from TNO, P. de Bettignies from HORIBA France SAS for technical support; C. Stiebing from Leibniz-IPHT and A. Krause from the Medical University of Vienna for discussions on Raman signal processing; H. Sattmann, A. Hodul and B. Rosenauer from the Medical University of Vienna for electrical and mechanical support; S. Rentz-Chorherr, M. Martin, T. Schneider and K. Memarpour from the Medical University of Vienna for support related to the medical device approval; M.C. Nguyen for the illustration in Fig. 1A. This work was carried out in the framework of project MOON. The project has received funding from: European Union’s Horizon 2020 research and innovation program under grant agreement No 732969 (It is an initiative of the Photonics Public Private Partnership. www.photonics21.org), CD-Labor OPTRAMED (CD10260501), Carl Zeiss Labor (UE60506007) and Medical Scientific Fund of the Mayor of the City of Vienna.

Author contributions

R.A.L., W.D., M.Kem., M.Sch. and J.P. conceptualized the project. R.S. designed, built and integrated the OCT/RS module. M.Sal. developed the OCT acquisition software. M.E. and W.Dj. conceptualized and assisted the integration between the IR fundus imaging module and the OCT/RS module. V.S. conceptualized and assisted the integration of the RS module. R.S., M.Ken. and R.A.L. carried out the investigation and experimental work. R.S. performed the data analysis, visualization and interpretation. R.A.L., M.A. and A.U. assisted the data analysis. H.S. and A.P. recruited the subjects and provided clinical support. R.A.L., T.S. and W.D. supervised the study. R.S. wrote the original manuscript. All authors reviewed and edited the manuscript.

Peer review

Peer review information

Communications Biology thanks Philippe Leproux, Yasuaki Kumamoto and Adriana Adamczyk for their contribution to the peer review of this work. Primary Handling Editor: Ophelia Bu. A peer review file is available.

Data availability

The source data underlying graphs can be obtained from Supplementary Data 1. All other data are available from the corresponding author (or other sources, as applicable) on reasonable request.

Code availability

The raw data and analysis code to generate the results in this study is available at 10.6084/m9.figshare.3123027475.

Competing interests

M.E. and M.Kem. are employed by Carl Zeiss Meditec AG; V.S. was employed by HORIBA France SAS; T.S. is employed by Carl Zeiss Meditec, Inc; All other authors declare that they have no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-09744-2.

References

  • 1.Capozzi, M. E., Gordon, A. Y., Penn, J. S. & Jayagopal, A. Molecular imaging of retinal disease. J. Ocul. Pharmacol. Ther.29, 275–286 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Nguyen, V. P., Zhe, J., Hu, J., Ahmed, U. & Paulus, Y. M. Molecular and cellular imaging of the eye. Biomed. Opt. Expr.15, 360 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Wang, Y. et al. Global incidence, progression, and risk factors of age-related macular degeneration and projection of disease statistics in 30 years: A modeling study. Gerontology68, 721–735 (2022). [DOI] [PubMed] [Google Scholar]
  • 4.Kiruthika, M. & Malathi, G. A comprehensive review on early detection of drusen patterns in age-related macular degeneration using deep learning models. Photodiagn. Photodyn. Ther.51, 104454 (2025). [DOI] [PubMed] [Google Scholar]
  • 5.London, A., Benhar, I. & Schwartz, M. The retina as a window to the brain—from eye research to CNS disorders. Nat. Rev. Neurol.9, 44–53 (2013). [DOI] [PubMed] [Google Scholar]
  • 6.Helboe, K. S., Eddelien, H. S. & Kruuse, C. Visual symptoms in acute stroke – A systematic review of observational studies. Clin. Neurol. Neurosurg.229, 107749 (2023). [DOI] [PubMed] [Google Scholar]
  • 7.Pleșa, A. et al. Eyes as windows: Unveiling neuroinflammation in multiple sclerosis via optic neuritis and uhthoff’s phenomenon. Diagnostics14, 2198 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Archibald, N. K., Clarke, M. P., Mosimann, U. P. & Burn, D. J. The retina in Parkinson’s disease. Brain132, 1128–1145 (2009). [DOI] [PubMed] [Google Scholar]
  • 9.Wagner, S. K. et al. Retinal optical coherence tomography features associated with incident and prevalent Parkinson disease. Neurology101, e1581–e1593 (2023). [DOI] [PMC free article] [PubMed]
  • 10.Zhang, J., Shi, L. & Shen, Y. The retina: A window in which to view the pathogenesis of Alzheimer’s disease. Ageing Res. Rev.77, 101590 (2022). [DOI] [PubMed] [Google Scholar]
  • 11.Costanzo, E. et al. Ocular biomarkers for Alzheimer disease dementia: An umbrella review of systematic reviews and meta-analyses. JAMA Ophthalmol.141, 84 (2023). [DOI] [PubMed] [Google Scholar]
  • 12.Bouma, B. E. et al. Optical coherence tomography. Nat. Rev. Methods Prim.2, 79 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Everett, M., Magazzeni, S., Schmoll, T. & Kempe, M. Optical coherence tomography: From technology to applications in ophthalmology. Transl. Biophotonics3, e202000012 (2021). [Google Scholar]
  • 14.Patel, N. B., Lim, M., Gajjar, A., Evans, K. B. & Harwerth, R. S. Age-associated changes in the retinal nerve fiber layer and optic nerve head. Invest. Ophthalmol. Vis. Sci.55, 5134 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Vujosevic, S. et al. Optical coherence tomography as retinal imaging biomarker of neuroinflammation/neurodegeneration in systemic disorders in adults and children. Eye37, 203–219 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Doustar, J., Torbati, T., Black, K. L., Koronyo, Y. & Koronyo-Hamaoui, M. Optical coherence tomography in Alzheimer’s disease and other neurodegenerative diseases. Front. Neurol.8, 701 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Leitgeb, R. A. Multimodal optical medical imaging concepts based on optical coherence tomography. Front. Phys.6, 17 (2018).29675413 [Google Scholar]
  • 18.Krafft, C. et al. Label-free molecular imaging of biological cells and tissues by linear and nonlinear Raman spectroscopic approaches. Angew. Chem. Int. Ed.56, 4392–4430 (2017). [DOI] [PubMed] [Google Scholar]
  • 19.Talari, A. C. S., Movasaghi, Z., Rehman, S. & Rehman, I. ur. Raman spectroscopy of biological tissues. Appl. Spectrosc. Rev.50, 46–111 (2015). [Google Scholar]
  • 20.Guo, S., Popp, J. & Bocklitz, T. Chemometric analysis in Raman spectroscopy from experimental design to machine learning–based modeling. Nat. Protoc.16, 5426–5459 (2021). [DOI] [PubMed] [Google Scholar]
  • 21.Shipp, D. W., Sinjab, F. & Notingher, I. Raman spectroscopy: Techniques and applications in the life sciences. Adv. Opt. Photonics9, 315 (2017). [Google Scholar]
  • 22.Khristoforova, Y., Bratchenko, L. & Bratchenko, I. Raman-based techniques in medical applications for diagnostic tasks: A review. IJMS24, 15605 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Chen, C. et al. Applications of Raman spectroscopy in the diagnosis and monitoring of neurodegenerative diseases. Front. Neurosci.18, 1301107 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Erckens, R. J. et al. Raman spectroscopy in ophthalmology: From experimental tool to applications in vivo. Lasers Med Sci.16, 236–252 (2001). [DOI] [PubMed] [Google Scholar]
  • 25.Salbreiter, M. et al. Lighting the Path: Raman spectroscopy’s journey through the microbial maze. Molecules29, 5956 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Synytsya, A., Judexova, M., Hoskovec, D., Miskovicova, M. & Petruzelka, L. Raman spectroscopy at different excitation wavelengths (1064, 785 and 532 nm) as a tool for diagnosis of colon cancer. J. Raman Spectrosc.45, 903–911 (2014). [Google Scholar]
  • 27.Bernstein, P. S. et al. Resonance Raman measurement of macular carotenoids in normal subjects and in age-related macular degeneration patients. Ophthalmology109, 1780–1787 (2002). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Li, B. et al. Imaging lutein and zeaxanthin in the human retina with confocal resonance Raman microscopy. Proc. Natl. Acad. Sci. USA117, 12352–12358 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Cakir, B. et al. In vivo noninvasive mitochondrial redox assessment of the optic nerve head to predict disease. PNAS Nexus2, pgad148 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Beattie, J. R., Brockbank, S., McGarvey, J. J. & Curry, W. J. Effect of excitation wavelength on the Raman spectroscopy of the porcine photoreceptor layer from the area centralis. Mol. Vis.11, 825–832 (2005). [PubMed] [Google Scholar]
  • 31.Beattie, J. R., Brockbank, S., McGarvey, J. J. & Curry, W. J. Raman microscopy of porcine inner retinal layers from the area centralis. Mol. Vis.13, 1106–1113 (2007). [PMC free article] [PubMed] [Google Scholar]
  • 32.Evans, J. W. et al. Optical coherence tomography and Raman spectroscopy of the ex-vivo retina. J. Biophotonics2, 398–406 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Patil, C. A. et al. Integrated system for combined Raman spectroscopy–spectral domain optical coherence tomography. J. Biomed. Opt.16, 011007 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Wang, Q. et al. Exploring Raman spectroscopy for the evaluation of glaucomatous retinal changes. J. Biomed. Opt.16, 107006 (2011). [DOI] [PubMed] [Google Scholar]
  • 35.Wang, Q. et al. Detection and characterization of glaucoma-like canine retinal tissues using Raman spectroscopy. J. Biomed. Opt.18, 067008 (2013). [DOI] [PubMed] [Google Scholar]
  • 36.Banbury, C. et al. Development of the Self Optimising Kohonen Index Network (SKiNET) for Raman Spectroscopy Based Detection of Anatomical Eye Tissue. Sci. Rep.9, 10812 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Stiebing, C. et al. Biochemical characterization of mouse retina of an Alzheimer’s disease model by Raman spectroscopy. ACS Chem. Neurosci.11, 3301–3308 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Banbury, C. et al. Spectroscopic detection of traumatic brain injury severity and biochemistry from the retina. Biomed. Opt. Express11, 6249 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Banbury, C. et al. Window into the mind: Advanced handheld spectroscopic eye-safe technology for point-of-care neurodiagnostic. Sci. Adv.9, eadg5431 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Glenn, J. V. et al. Confocal Raman microscopy can quantify advanced glycation end product (AGE) modifications in Bruch’s membrane leading to accurate, nondestructive prediction of ocular aging. FASEB J.21, 3542–3552 (2007). [DOI] [PubMed] [Google Scholar]
  • 41.Stiebing, C. et al. Nonresonant Raman spectroscopy of isolated human retina samples complying with laser safety regulations for in vivo measurements. Neurophoton6, 1 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Keilhauer, C. N. & Delori, F. C. Near-infrared autofluorescence imaging of the fundus: Visualization of ocular melanin. Invest. Ophthalmol. Vis. Sci.47, 3556 (2006). [DOI] [PubMed] [Google Scholar]
  • 43.Zhang, P. et al. Feasibility study of Raman spectroscopy for investigating the mouse retina in vivo. In Ophthalmic Technologies XXVIII (eds Manns, F., Söderberg, P. G. & Ho, A.) 54 (SPIE, San Francisco, United States, 10.1117/12.2288790. 2018).
  • 44.Alba-Arbalat, S. et al. In vivo molecular changes in the retina of patients with multiple sclerosis. Invest. Ophthalmol. Vis. Sci.62, 11 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Sentosa, R. et al. Development of a label-free, functional, molecular and structural imaging system combining optical coherence tomography and Raman spectroscopy for in vivo measurement of rat retina. Biomed. Opt. Expr.16, 566 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Wang, Y. X., Panda-Jonas, S. & Jonas, J. B. Optic nerve head anatomy in myopia and glaucoma, including parapapillary zones alpha, beta, gamma and delta: Histology and clinical features. Prog. Retinal Eye Res.83, 100933 (2021). [DOI] [PubMed] [Google Scholar]
  • 47.Lieber, C. A. & Mahadevan-Jansen, A. Automated method for subtraction of fluorescence from biological Raman spectra. Appl Spectrosc.57, 1363–1367 (2003). [DOI] [PubMed] [Google Scholar]
  • 48.Afseth, N. K. & Kohler, A. Extended multiplicative signal correction in vibrational spectroscopy, a tutorial. Chemometr. Intell. Lab. Syst.117, 92–99 (2012). [Google Scholar]
  • 49.Krafft, C. Raman spectroscopy of proteins and nucleic acids: From amino acids and nucleotides to large assemblies. In Encyclopedia of Analytical Chemistry (ed. Meyers, R. A.) 1–15 (Wiley, 2018).
  • 50.Krafft, C., Neudert, L., Simat, T. & Salzer, R. Near infrared Raman spectra of human brain lipids. Spectrochimica Acta Part A: Mol. Biomol. Spectrosc.61, 1529–1535 (2005). [DOI] [PubMed] [Google Scholar]
  • 51.Czamara, K. et al. Raman spectroscopy of lipids: a review. J. Raman Spectrosc.46, 4–20 (2015). [Google Scholar]
  • 52.Wiercigroch, E. et al. Raman and infrared spectroscopy of carbohydrates: A review. Spectrochimica Acta Part A: Mol. Biomol. Spectrosc.185, 317–335 (2017). [DOI] [PubMed] [Google Scholar]
  • 53.Pezzotti, G. Raman spectroscopy in cell biology and microbiology. J. Raman Spectrosc.52, 2348–2443 (2021). [Google Scholar]
  • 54.Hernández, B., Pflüger, F., Adenier, A., Kruglik, S. G. & Ghomi, M. Vibrational analysis of amino acids and short peptides in hydrated media. VIII. Amino acids with aromatic side chains: l -Phenylalanine, l -Tyrosine, and l -Tryptophan. J. Phys. Chem. B114, 15319–15330 (2010). [DOI] [PubMed] [Google Scholar]
  • 55.Rygula, A. et al. Raman spectroscopy of proteins: a review: Raman spectroscopy of proteins. J. Raman Spectrosc.44, 1061–1076 (2013). [Google Scholar]
  • 56.Vaughan, D. W. & Hansen, L. A Learning System in Histology: CD-ROM and Guide. (Oxford University Press, Oxford, 2002).
  • 57.J. Salazar, J. et al. Anatomy of the Human Optic Nerve: Structure and Function. in Optic Nerve (ed. Ferreri, F. M.) (IntechOpen, 2019).
  • 58.Muench, N. A. et al. The influence of mitochondrial dynamics and function on retinal ganglion cell susceptibility in optic nerve disease. Cells10, 1593 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Liu, H., Liu, H. & Prokosch, V. The relationship between mitochondria and neurodegeration in the Eye: A review. Appl. Sci.11, 7385 (2021). [Google Scholar]
  • 60.Dias, M. S., Luo, X., Ribas, V. T., Petrs-Silva, H. & Koch, J. C. The role of axonal transport in glaucoma. IJMS23, 3935 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Guo, J. et al. Aging and aging-related diseases: From molecular mechanisms to interventions and treatments. Sig Transduct. Target Ther.7, 391 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Li, Y. et al. Molecular mechanisms of aging and anti-aging strategies. Cell Commun. Signal22, 285 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 63.Coleman-Belin, J. et al. Aging effects on optic nerve neurodegeneration. Int J. Mol. Sci.24, 2573 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 64.Kim, J. H., Yang, D. & Park, S. Raman spectroscopy in cellular and tissue aging research. Aging Cell e14494 10.1111/acel.14494. (2025) [DOI] [PMC free article] [PubMed]
  • 65.Harayama, T. & Riezman, H. Understanding the diversity of membrane lipid composition. Nat. Rev. Mol. Cell Biol.19, 281–296 (2018). [DOI] [PubMed] [Google Scholar]
  • 66.Skowronska-Krawczyk, D. & Budin, I. Aging membranes: Unexplored functions for lipids in the lifespan of the central nervous system. Exp. Gerontol.131, 110817 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 67.Wagner, H., Fink, B. A. & Zadnik, K. Sex- and gender-based differences in healthy and diseased eyes. Optom. - J. Am. Optometric Assoc.79, 636–652 (2008). [DOI] [PubMed] [Google Scholar]
  • 68.Kobayashi, T. et al. Influence of age and gender on the pulse waveform in optic nerve head circulation in healthy men and women. Sci. Rep.9, 17895 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 69.Niehorster, D. C. et al. The best fixation target revisited: New insights from retinal eye tracking. Behav. Res58, 2 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 70.Bergholt, M. S., Serio, A. & Albro, M. B. Raman Spectroscopy: Guiding Light for the Extracellular Matrix. Front. Bioeng. Biotechnol.7, 303 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 71.Tominaga, Y., Fujiwara, A. & Amo, Y. Dynamical structure of water by Raman spectroscopy. Fluid Phase Equilibria144, 323–330 (1998). [Google Scholar]
  • 72.Lang, X., Shi, L., Zhao, Z. & Min, W. Probing the structure of water in individual living cells. Nat. Commun.15, 5271 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 73.Miyamori, D. et al. A Raman algorithm to estimate human age from protein structural variations in autopsy skin samples: A protein biological clock. Sci. Rep.11, 5949 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 74.Kostamovaara, J. et al. Fluorescence suppression in Raman spectroscopy using a time-gated CMOS SPAD. Opt. Expr.21, 31632 (2013). [DOI] [PubMed] [Google Scholar]
  • 75.Sentosa, R. & Leitgeb, R. Analysis code for: Label-free in vivo molecular profiling of the human retina by non-resonant Raman spectroscopy. 10071887 Bytes figshare 10.6084/M9.FIGSHARE.31230274 (2026). [DOI] [PMC free article] [PubMed]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Transparent Peer Review file (2.9MB, pdf)
Supplementary information (1.1MB, pdf)
42003_2026_9744_MOESM3_ESM.pdf (27.6KB, pdf)

Description of Additional Supplementary Files

Supplementary Data 1 (1.3MB, xlsx)
Reporting Summary (1.4MB, pdf)

Data Availability Statement

The source data underlying graphs can be obtained from Supplementary Data 1. All other data are available from the corresponding author (or other sources, as applicable) on reasonable request.

The raw data and analysis code to generate the results in this study is available at 10.6084/m9.figshare.3123027475.

Articles from Communications Biology are provided here courtesy of Nature Publishing Group

RESOURCES