Confocal Raman microspectroscopy imaging reveals structural protein reorganization in human intracranial aneurysm tissue Authors: Maria Gollwitzer, David Schäffl, Christian Angerer, Claudia Leimhofer, Serge Weis, Vanessa Mazanec, Matthias Gmeiner, Andreas Gruber, Sabine Hild Published: 2025 ABSTRACT Intracranial aneurysms are pathological dilatations of cerebral arteries that may rupture and cause subarachnoid hemorrhage. Extracellular matrix remodeling, particularly of Collagen and Elastin, is central to aneurysm progression, but detailed molecular insights remain limited. In this proof-of-concept study, we applied confocal Raman microspectroscopy to matched aneurysmal and healthy basilar artery tissue from the same patient, complemented by an additional control from a second patient. Raman imaging was performed on cryo-sectioned samples using a 532 nm excitation laser. Data were processed using a non-negative matrix factorization–based component analysis with reference spectra of Collagen Types I, III, IV and Elastin. Spectral differences were analyzed by Amide I peak deconvolution and Principal Component Analysis. Aneurysmal tissue showed reduced Collagen and Elastin signals and altered spatial organization. Amide I deconvolution indicated a shift from β-sheet–dominated to α-helical and disordered protein structures. Principal Component Analysis revealed consistent spectral markers separating healthy from diseased tissue. These results demonstrate the potential of confocal Raman microspectroscopy to detect extracellular matrix remodeling and protein conformational changes in intracranial aneurysms, supporting the development of spectroscopic markers for aneurysm wall integrity and rupture risk. Introduction Intracranial aneurysms (IAs) are pathological dilatations of cerebral arteries that affect approximately 2–3% of the global adult population1,2. While unruptured IAs are often asymptomatic and discovered incidentally, they carry the risk of rupture, which can lead to subarachnoid hemorrhage (SAH)—a life-threatening condition with a rupture probability of approximately 0.25% per year. SAH is associated with poor outcomes: a mortality rate of around 20% within the first 3 months, increasing to 30% within 5 years, and a morbidity rate of around 50% among survivors3,4. The management of unruptured IAs remains challenging, as the risk of rupture must be weighed against the potential complications of surgical or endovascular interventions. While screening methods are available, they are not routinely recommended and are typically reserved for patients with elevated risk profiles1,2.The etiology and progression of IAs involve a multifactorial interplay of genetic predisposition5, hemodynamic forces6, and morphological7 factors. Current models suggest that disturbed blood flow and chronic inflammation drive aneurysm formation and rupture8. In particular, abnormal wall shear stress can activate proinflammatory pathways, leading to apoptosis or phenotypic switching of vascular smooth muscle cells (VSMCs). This switching, from a contractile to a synthetic phenotype, promotes extracellular matrix (ECM) remodeling, Collagen synthesis, fibrosis, oxidative stress, calcification and degradation of structural proteins9–12. While structural remodeling has been extensively studied in abdominal aortic aneurysms, detailed information on these processes in intracranial aneurysms is still scarce and translating the molecular pathways underlying aneurysm development into clinical markers remains difficult due to the complexity of the involved biological mechanisms13. Although techniques like mass spectrometry and Confocal Raman Microspectroscopy (CRM) have been applied in aortic aneurysm research, detailed spectral characterization of IAs—particularly concerning ECM proteins such as Collagen and Elastin—is still lacking14–16. CRM combines biochemical, histological and spatial information and has proven effective in visualizing tissue remodeling processes. The phenotypic switch of VSMCs involves the expression of new procollagen, which is enzymatically modified into specific Collagen isoforms contributing to the structural integrity of the remodeled vessel wall. Recent findings challenge the notion that IAs are stable over decades. Using radiocarbon (C14)dating, Etminan et al.17, demonstrated ongoing Collagen turnover in aneurysm walls, suggesting active remodeling. Hackenberg et al.18, further associated immature Collagen presence with modifiable risk factors such as hypertension and smoking, reinforcing the link between hemodynamics and ECM instability. High Collagen turnover correlates with increased wall instability and a higher risk of rupture. Furthermore, ruptured and enlarging aneurysms display more complex hemodynamic and morphological characteristics than stable ones19. Advanced computational models, including fluid–structure interaction analyses, have been employed to simulate the dynamic interaction between hemodynamic forces and wall mechanics13. Integrating such models with molecular-level-imaging -such as CRM- could help identify wall vulnerability, but this requires validation in larger datasets. Traditional imaging methods like CT and MRI are standard in aneurysm diagnostics but offer limited biochemical information20,21. Other molecular imaging approaches, such as mass spectrometry, have been explored22 and early studies using CRM to investigate thoracic aneurysms have shown promising results in mapping structural protein distribution14–16,23. CRM utilizes inelastic scattering of monochromatic light—commonly a 532 nm laser—to capture vibrational spectra of molecules (Raman scattering)24. Coupled with a confocal microscope and scanning stage, this method generates hyperspectral datasets, enabling the identification and spatial mapping of biochemical components. Dimensionality reduction techniques such as Principal Component Analysis (PCA) and True Component Analysis (TCA, WITec), an NMF-based spectral unmixing method, allow for effective analysis of complex data and detection of disease-specific spectral markers16. The aim of this study is to investigate structural and molecular changes in the walls of intracranial aneurysms using high-resolution CRM, histological staining and quantitative image analysis. We focus on alterations in the spatial distribution and secondary structure of proteins—especially Collagen and Elastin—between healthy and aneurysmal tissue from the same patient. The Amide I band in Raman spectra is a key indicator of protein secondary structure, with lower wavenumbers typically corresponding to ordered \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta$$\end{document}-sheet structures and higher wavenumbers indicating \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha$$\end{document}-helices or disordered conformations25–33. This proof-of-principle study aims to demonstrate that CRM can generate analyzable datasets to assess protein remodeling processes. The insights gained may contribute to a better understanding of aneurysm progression and ultimately improve risk stratification and therapeutic decision-making. Future studies will expand this methodology to larger patient cohorts. Results Histological assessment & confocal Raman microspectroscopy imaging on healthy arterial cross sections Classical histological staining using Hematoxylin & Eosin (H&E) was first performed to enable morphological interpretation of healthy arterial tissue (Fig. 1). The sample, derived from a non-aneurysmal section of the basilar artery of a patient with a basilar aneurysm, showed three expected arterial wall layers—tunica intima, media and adventitia—with preserved histoarchitecture. In the H&E stain, basophilic structures, such as nuclei, ribosomes and the endoplasmatic reticulum appear blue due to hematoxylin binding, confirming the presence of VSMCs. Eosinophilic components, primarily intermediate filaments and myofilaments, stain pink, reflecting34. Following histological assessment, CRM was performed on cryo-sectioned samples of healthy tissue. Raman spectroscopy, based on the inelastic scattering of monochromatic light (here using a 532 nm laser), yields vibrational spectra representative of molecular bonds within the tissue. Through hyperspectral imaging and raster scanning, spatially resolved chemical maps were generated. Area scans covering the arterial cross section from the tunica adventitia to tunica intima revealed the molecular composition and structural protein organization in non-pathological tissue. Particular focus was given to the Amide I band (~ 1600–1700 cm⁻1, shaded in the left panel of Fig. 1), which arises primarily from double-bonded oxygen-carbon stretching vibrations and is highly sensitive to protein secondary structure. The corresponding zoomed views of this shaded region with bi-Gaussian peak deconvolution for each anatomical layer are shown in the right panel of Fig. 1. Layer-specific differences were evident: the intima displayed relatively higher contributions from α-helical and random coil structures, the media was enriched in β-sheet components consistent with its dense smooth muscle and Collagen content and the adventitia showed mixed signals of fibrillar Collagen and Elastin. These distinct secondary structure signatures underline the ability of CRM to resolve biochemical heterogeneity across the vessel wall. In our measurements, the apparent maximum of the Amide I band was observed at slightly lower wavenumbers (~ 1600 cm⁻1) compared to typical literature values for hydrated tissue (~ 1650 cm⁻1). This downshift is attributable to the dehydrated state of the fixed, cryo-sectioned samples, the 532 nm excitation wavelength and overlapping contributions from aromatic amino acid side chains and β-sheet–associated modes within the Amide I envelope. The reported values correspond to peak positions obtained from deconvoluted component fits rather than unprocessed spectra, which can further influence the apparent maximum without affecting the identification of characteristic secondary structure elements. Complementary information was gathered from the Amide III band (~ 1300 cm⁻1) and where detectable, the Amide II band (~ 1550 cm⁻1, though the latter typically shows weak Raman activity). An overview of wavenumber assignment is summarized in Table 1.Fig. 1H&E staining of a healthy arterial wall, with the tunica adventitia located in the upper left corner of the image. The molecular origin of the Amide I is illustrated using the peptide bond structure, highlighting the key atomic vibrational contributions (C=O stretching and N–H bending). Average Raman spectra were obtained from each of the three anatomical layers—adventitia, media and intima—with the Amide I band region (1600–1700 cm−1, shaded in the left panel) used for secondary-structure analysis. The corresponding zoomed views of this region with bi-Gaussian peak deconvolution for each anatomical layer are shown in the right panel. Spectral comparisons with Table 1 indicate the presence of both ordered and disordered α-helices as well as parallel β-strands, with relative abundances varying by anatomical region. In our measurements, the Amide I maximum appears at slightly lower wavenumbers (~ 1600 cm⁻1) than typical hydrated tissue values (~ 1650 cm⁻1), due to the dehydrated state of the fixed cryo-sections, the 532 nm excitation wavelength, and overlapping contributions from aromatic amino acids and β-sheet modes; this does not affect secondary structure assignments. Notably, a progressive change in relative α-helical and β-sheet contributions was observed from the adventitia toward the intima, as determined by Amide I band deconvolution (see also Table 1). The Raman intensity map for CH-stretching vibrations at 2950 cm⁻1 reflects overall protein content and distribution across the arterial wall.Table 1Overview of characteristic Raman shifts of Amide bands and their interpretation in relation to protein secondary structure.Amide band shiftsReference systemsExperimentalInterpretation & ReferencesNo.Approximate range in known system35/rel. cm−1BandAssignment(Deconvolution, healthy → diseased shift / rel. cm−1)Positve correlation to aneurysms detected with PCA / rel. cm−1–11620Amide IUnordered α-helices1583 → 1591–16,3621640–1655Amide IOrdered α-helices1611 → 1635164329,3331670–1690Amide IAntiparallel β-strands1673 → 1669–23,29,3341650–1675Amide IParallel β-strands1673 → 1669166951550Amide IIParallel & antiparallel β-sheets–––61265–1300Amide IIIα-helices–1311 (peak in broad pos. Range)Not robust for deconvolution, instead PCA approach, cf.14,3171229–1235Amide IIIβ-sheets–1239 (broad pos. range)81243–1253Amide IIIRandom coil–1239 (broad pos. range)The table summarizes typical wavenumber ranges for Amide I, II and III bands and their respective associations with α-helices, β-sheets and unordered (random coil) confirmations. These shifts provide key spectral indicators for assessing structural protein organization in biological tissues27,35–37. H&E staining of a healthy arterial wall, with the tunica adventitia located in the upper left corner of the image. The molecular origin of the Amide I is illustrated using the peptide bond structure, highlighting the key atomic vibrational contributions (C=O stretching and N–H bending). Average Raman spectra were obtained from each of the three anatomical layers—adventitia, media and intima—with the Amide I band region (1600–1700 cm−1, shaded in the left panel) used for secondary-structure analysis. The corresponding zoomed views of this region with bi-Gaussian peak deconvolution for each anatomical layer are shown in the right panel. Spectral comparisons with Table 1 indicate the presence of both ordered and disordered α-helices as well as parallel β-strands, with relative abundances varying by anatomical region. In our measurements, the Amide I maximum appears at slightly lower wavenumbers (~ 1600 cm⁻1) than typical hydrated tissue values (~ 1650 cm⁻1), due to the dehydrated state of the fixed cryo-sections, the 532 nm excitation wavelength, and overlapping contributions from aromatic amino acids and β-sheet modes; this does not affect secondary structure assignments. Notably, a progressive change in relative α-helical and β-sheet contributions was observed from the adventitia toward the intima, as determined by Amide I band deconvolution (see also Table 1). The Raman intensity map for CH-stretching vibrations at 2950 cm⁻1 reflects overall protein content and distribution across the arterial wall. Table 1Overview of characteristic Raman shifts of Amide bands and their interpretation in relation to protein secondary structure.Amide band shiftsReference systemsExperimentalInterpretation & ReferencesNo.Approximate range in known system35/rel. cm−1BandAssignment(Deconvolution, healthy → diseased shift / rel. cm−1)Positve correlation to aneurysms detected with PCA / rel. cm−1–11620Amide IUnordered α-helices1583 → 1591–16,3621640–1655Amide IOrdered α-helices1611 → 1635164329,3331670–1690Amide IAntiparallel β-strands1673 → 1669–23,29,3341650–1675Amide IParallel β-strands1673 → 1669166951550Amide IIParallel & antiparallel β-sheets–––61265–1300Amide IIIα-helices–1311 (peak in broad pos. Range)Not robust for deconvolution, instead PCA approach, cf.14,3171229–1235Amide IIIβ-sheets–1239 (broad pos. range)81243–1253Amide IIIRandom coil–1239 (broad pos. range)The table summarizes typical wavenumber ranges for Amide I, II and III bands and their respective associations with α-helices, β-sheets and unordered (random coil) confirmations. These shifts provide key spectral indicators for assessing structural protein organization in biological tissues27,35–37. Overview of characteristic Raman shifts of Amide bands and their interpretation in relation to protein secondary structure. The table summarizes typical wavenumber ranges for Amide I, II and III bands and their respective associations with α-helices, β-sheets and unordered (random coil) confirmations. These shifts provide key spectral indicators for assessing structural protein organization in biological tissues27,35–37. Reference substance spectra and spectral differentation To facilitate component identification in the tissue, Raman reference spectra for Collagen Types I,III, IV and Elastin were acquired using the same parameters as tissue scans. These spectra are shown in Fig. 2, both in full range and zoomed into the Amide I band region (1550–1750 cm−1). Spectral differeces between the Collagen types and Elastin are primarily observed in the fingerprint and CH-stretch regions. Collagen Type I additionally shows fluorescence-related intensity between 1700–2700 cm−1. Minor variations in the exact position of the Amide I maximum compared to some published spectra were observed. These shifts did not interfere with the identification of characteristic marker bands or with the subsequent spectral deconvolution. Although some differences appear subtle, NMF and PCA enable robust spectral separation and tissue classification. The average aneurysmal tissue spectrum represents a heterogeneous composite across structurally distinct regions and is therefore not expected to match any single reference spectrum; instead, it reflects the weighted sum of multiple biochemical constituents. To further support the interpretation of tissue-derived Raman signals, Amide I band deconvolution was also performed on the reference substance spectra. This allowed identification of characteristic vibrational contributions associated with secondary protein structures. Among the analyzed compontents, Collagen Type I showed distinctly different spectral profile—both across the full Raman range and within the Amide I band—compared to Collagen Type III and IV as well as Elastin. Notably, the spectral differences in the Amide I region reflect variations in dominant secondary structure elements such as α-helices and β-sheets. Collagen Types I and III, Elastin and Collagen Type IV displayed distinct Raman spectral profiles. These differences were evident both in the full Raman range and in the Amide I region and were consistent across replicate measurements. These reference-based spectral assignments form the basis for identifying structural protein changes in aneurysmal tissue and support the interpretation of tissue-specific remodeling processes38–43.Fig. 2Raman spectra of reference substances used to generate spatial distribution maps. Shown are full-range spectra and corresponding enlarged views of the Amide I band region for Collagen Type I (red), Type III (orange) and Type IV (purple) as well as Elastin (green). Deconvolution of the Amide I region highlights characteristic vibrational peaks associated with each protein’s secondary structure, enabling differentiation between fibrillar, load-bearing proteins (Collagen I, III, Elastin) and network-forming Collagen (Type IV). These reference spectra were recorded under identical conditions to tissue scans and served as the spectral basis for TCA. Minor differences in the exact Amide I peak position relative to some published spectra reflect the dry state of the reference proteins (lyophilized for Collagen Types I and IV; vacuum-evaporated from solution for Collagen Type III), differences in hydration state, and the 532 nm excitation wavelength used in this study. These factors do not affect the identification of characteristic marker bands or the interpretation of tissue spectra. Raman spectra of reference substances used to generate spatial distribution maps. Shown are full-range spectra and corresponding enlarged views of the Amide I band region for Collagen Type I (red), Type III (orange) and Type IV (purple) as well as Elastin (green). Deconvolution of the Amide I region highlights characteristic vibrational peaks associated with each protein’s secondary structure, enabling differentiation between fibrillar, load-bearing proteins (Collagen I, III, Elastin) and network-forming Collagen (Type IV). These reference spectra were recorded under identical conditions to tissue scans and served as the spectral basis for TCA. Minor differences in the exact Amide I peak position relative to some published spectra reflect the dry state of the reference proteins (lyophilized for Collagen Types I and IV; vacuum-evaporated from solution for Collagen Type III), differences in hydration state, and the 532 nm excitation wavelength used in this study. These factors do not affect the identification of characteristic marker bands or the interpretation of tissue spectra. H&E staining and Raman cross sectional imaging of the aneurysm wall Aneurysmal tissue was initially assessed using conventional histological staining with H&E, following the same protocol applied to healthy arterial sections. As shown in Fig. 2, the aneurysmal wall is markedly thinner than the healthy counterpart and exhibits early luminal thrombosis, characterized by aggregated cell clusters, as well as acellular regions. Additionally, intimal hypoplasia is evident and highlighted in the image by black boxes and arrows. The three typical anatomical layers are no longer clearly distinguishable in the aneurysmal wall. Subsequently, CRM was performed on cryo-sectioned aneurysmal samples to obtain a cross-sectional hyperspectral dataset. Using the previously recorded reference spectra for structural proteins, True Component Analysis (TCA, WITec), a non-negative matrix factorization (NMF)-based algorithm was applied to generate protein-specific distribution maps. Figure 2 presents representative Raman maps of aneurysmal sections. Collagen Type I, III and Elastin are predominantly localized in the peripheral regions of the wall, corresponding to the remnants of the adventitia and intima. In contrast, Collagen Type IV, a network-forming protein, is more abundant in the former medial region of the vessel. Spectral analysis of the Amide I band revealed distinct spectral shifts in peak position and intensity, indicating changes in protein secondary structure between healthy and diseased tissue. Based on these findings, Amide I band peak deconvolutions were performed on the spectra of both healthy and diseased samples to identify contributing molecular vibrations and assess their relevance for structural protein reorganization. Due to technical artifacts (focal distortions), tissue scans AN3 and AN5 were excluded from further analysis. The dataset labeled ANC corresponds to a basilar artery aneurysm from a second patient, included for comparison. Spectral analysis revealed distinct molecular differences between healthy and aneurysmal arterial walls. In aneurysmal tissue, Raman spectra showed increased contributions from random coil structures (1240 cm−1), alongside peaks indicative of parallel β-strands (1671 cm−1) and unordered α-helices (1632 cm−1). Notably, the α-helix-related peak was shifted towards lower wavenumbers, suggesting protein unfolding or loss of structural order (Fig. 3). In VEC, Collagen Types I and III showed a structured, layer-specific distribution, with Type III particularly enriched in the tunica media, and Elastin signals localized mainly to the adventitia. In contrast, ANC exhibited reduced intensity of fibrillar Collagens, irregular Collagen I/III distribution, and a pronounced depletion of Elastin. The former medial layer of the aneurysmal wall appeared almost entirely replaced by network-forming Collagen Type IV, indicating a structural reorganization and loss of normal anatomical boundaries. These spectroscopic alterations corresponded spatially with the disrupted morphology seen in H&E-stained sections. Minor signals detected at tissue borders in Collagen III maps were traced to low-level background scattering from the glass substrate and were excluded from quantitative interpretation (see Methods).Fig. 3H&E staining of the scanned aneurysmal tissue section. The image reveals typical pathological features including luminal thrombosis indicated by cell aggregations, decellularized areas and intimal hypoplasia, consistent with advanced wall degeneration44. The accompanying Raman maps show the spatial distribution of CH-stretching vibrations(2950 cm⁻1) and the relative abundance of Collagen Types I, III, IV and Elastin across multiple aneurysmal tissue slices. Marker bands for visualization were selected from the reference spectra shown in Fig. 2, with a ± 5 cm⁻1 matching tolerance: Collagen I—1574, 1607, 1638 cm⁻1; Collagen III—1615, 1635, 1700 cm⁻1; Collagen IV—1622, 1671, 1676, 1607 cm⁻1; Elastin—1603, 1641, 1663, 1696 cm⁻1. Additionally, an average Raman spectrum of the aneurysmal wall is presented, with enlarged views of the Amide III and I bands. Spectral features indicate increased random coil contributions at 1240 cm−1, alongside prominent peaks associated with parallel β-strands (1671 cm−1) as well as unordered α-helices (1632 cm−1). These shifts suggest substantial alterations in protein secondary structure, reflective of extracellular matrix remodeling and loss of organized Collagen architecture in the diseased vessel wall. H&E staining of the scanned aneurysmal tissue section. The image reveals typical pathological features including luminal thrombosis indicated by cell aggregations, decellularized areas and intimal hypoplasia, consistent with advanced wall degeneration44. The accompanying Raman maps show the spatial distribution of CH-stretching vibrations(2950 cm⁻1) and the relative abundance of Collagen Types I, III, IV and Elastin across multiple aneurysmal tissue slices. Marker bands for visualization were selected from the reference spectra shown in Fig. 2, with a ± 5 cm⁻1 matching tolerance: Collagen I—1574, 1607, 1638 cm⁻1; Collagen III—1615, 1635, 1700 cm⁻1; Collagen IV—1622, 1671, 1676, 1607 cm⁻1; Elastin—1603, 1641, 1663, 1696 cm⁻1. Additionally, an average Raman spectrum of the aneurysmal wall is presented, with enlarged views of the Amide III and I bands. Spectral features indicate increased random coil contributions at 1240 cm−1, alongside prominent peaks associated with parallel β-strands (1671 cm−1) as well as unordered α-helices (1632 cm−1). These shifts suggest substantial alterations in protein secondary structure, reflective of extracellular matrix remodeling and loss of organized Collagen architecture in the diseased vessel wall. Comparison of protein distribution in healthy and aneurysmal arteries We compared cross-sectional CRM distribution maps of healthy basilar arteries with those of basilar artery aneurysms. Figure 4 presents spatial maps of Collagen I, III, IV and Elastin in both tissue types. In aneurysmal walls, the signal intensities of Collagen types were markedly reduced, particularly for Collagen I and III, suggesting ECM degradation. Elastin was also significantly depleted in all aneurysmal sections, consistent with a loss of elastic recoil capacity and structural resilience. Interestingly, Collagen IV, which in healthy vessels is typically localized to the basement membrane of VSMCs within the tunica media, was no longer detectable in its expected locations. Instead, it was found almost exclusively in the former medial region of the aneurysmal wall, suggesting a pathological redistribution and possible compensatory role in the absence of intact fibrillar Collagen.Fig. 4Raman imaging scans of a basilar artery-aneurysm tissue pair, highlighting key components of the extracellular matrix. The left panel shows the healthy vessel (left) and the right panel displays the corresponding aneurysm section (right). Spectral distribution maps are presented for CH-stretching vibrations (2950 cm⁻1) and for Collagen Types I, III, IV and Elastin. The aneurysmal tissue exhibits reduced signal intensities for all structural proteins, particularly Collagen and Elastin, indicating advanced ECM degredation and loss of arterial elasticity. The spectral differences confirm aneurysm-associated Collagen breakdown and Elastin depletion. A comparison of averaged Raman spectra reveals a shift in dominant secondary structure: while the healthy wall is characterized by β-strand-associated signals (1671 cm−1), the aneurysmal section shows increased contributions from unordered α-helices (peak at 1636 cm−1), suggesting protein unfolding and conformational changes in the diseased vessel. Raman imaging scans of a basilar artery-aneurysm tissue pair, highlighting key components of the extracellular matrix. The left panel shows the healthy vessel (left) and the right panel displays the corresponding aneurysm section (right). Spectral distribution maps are presented for CH-stretching vibrations (2950 cm⁻1) and for Collagen Types I, III, IV and Elastin. The aneurysmal tissue exhibits reduced signal intensities for all structural proteins, particularly Collagen and Elastin, indicating advanced ECM degredation and loss of arterial elasticity. The spectral differences confirm aneurysm-associated Collagen breakdown and Elastin depletion. A comparison of averaged Raman spectra reveals a shift in dominant secondary structure: while the healthy wall is characterized by β-strand-associated signals (1671 cm−1), the aneurysmal section shows increased contributions from unordered α-helices (peak at 1636 cm−1), suggesting protein unfolding and conformational changes in the diseased vessel. Amide I peak deconvolution and PCA marker peak identification To assess protein conformational changes, Amide I band deconvolution was performed on averaged Raman spectra (Fig. 5a). Healthy vessel tissue (VE) exhibited stronger \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta$$\end{document}-associated peaks, while spectra from aneurysmal tissue (AN) showed a shift in peak maxima toward regions associated with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha$$\end{document}-helical structures and random coil conformations. These shifts suggest a reorganization of protein secondary structure in the aneurysmal vessel wall. The dataset used for this analysis was generated from all valid imaging scans, excluding sample pairs AN3 & AN5, which were omitted due to focal artifacts. In healthy tissue (VE), the Amide I deconvolution relealed higher peak intensities corresponding to ordered \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta$$\end{document}-sheet structures, while in diseased tissue (AN), peaks shifted toward positions indicative of disordered \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha$$\end{document}-helices and non-structured \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta$$\end{document}-space (i.e. “random coils”), consistent with protein unfolding and structural destabilization. Deconvolution was performed on pooled spectra (Healthy = VE + VEC; Aneurysm = AN + ANC) because both patients exhibited consistent band behavior as verified by PCA25–33.Fig. 5(a) Deconvolution of the Amide I peak region from averaged Raman spectra of healthy (VE, red) and diseased (AN, blue) tissue. Spectra were generated by averaging of all valid scans and Amide I deconvolution of pooled spectra per tissue state (Healthy = VE + VEC, Aneurysm = AN + ANC) showing the shift from 1583/1611/1673 cm⁻1 (healthy) to 1591/1635/1669 cm⁻1 (aneurysmal). The resulting peak profile reveals distinct differences in protein secondary structure. (b) Principal Component Analysis: Plot of Principal Component 2 (PC2) versus PC3, based on spectral data from 650 to 1700 cm−1. Clear clustering along PC2 seperates healthy (VE, red, negative correlated) from aneurysmal (AN, blue, positively correlated) tissue indicating distinct biochemical signatures (VE = healthy vessel from primary patient; AN = aneurysmal tissue from primary patient; VEC = healthy vessel from control patient; ANC = aneurysmal tissue from control patient).. (c) Corresponding PC2 loadings, showing characteristic peaks associated with each tissue type identifying healthy (VE, red, negative correlation) and diseased tissue (AN, blue, positive correlation). Peaks at 1091 cm⁻1 (C–O stretching in carbohydrates/glycoproteins and phosphodiester bonds) and 1507 cm⁻1 (aromatic C = C stretching in phenylalanine and tryptophan side chains) are characteristic for healthy tissue. Positive loadings for aneurysmal tissue correspond to Collagen degradation products and altered secondary structures. (a) Deconvolution of the Amide I peak region from averaged Raman spectra of healthy (VE, red) and diseased (AN, blue) tissue. Spectra were generated by averaging of all valid scans and Amide I deconvolution of pooled spectra per tissue state (Healthy = VE + VEC, Aneurysm = AN + ANC) showing the shift from 1583/1611/1673 cm⁻1 (healthy) to 1591/1635/1669 cm⁻1 (aneurysmal). The resulting peak profile reveals distinct differences in protein secondary structure. (b) Principal Component Analysis: Plot of Principal Component 2 (PC2) versus PC3, based on spectral data from 650 to 1700 cm−1. Clear clustering along PC2 seperates healthy (VE, red, negative correlated) from aneurysmal (AN, blue, positively correlated) tissue indicating distinct biochemical signatures (VE = healthy vessel from primary patient; AN = aneurysmal tissue from primary patient; VEC = healthy vessel from control patient; ANC = aneurysmal tissue from control patient).. (c) Corresponding PC2 loadings, showing characteristic peaks associated with each tissue type identifying healthy (VE, red, negative correlation) and diseased tissue (AN, blue, positive correlation). Peaks at 1091 cm⁻1 (C–O stretching in carbohydrates/glycoproteins and phosphodiester bonds) and 1507 cm⁻1 (aromatic C = C stretching in phenylalanine and tryptophan side chains) are characteristic for healthy tissue. Positive loadings for aneurysmal tissue correspond to Collagen degradation products and altered secondary structures. To identify characteristic spectral markers, PCA analysis was conducted on cross-sectional spectral datasets in the fingerprint region (600–1800 cm−1) using OriginPro (OriginLab). PC1 was excluded from interpretation due to dominant background contributions from the glass substrate. PC2, however, clearly separated (negatively correlated) from aneurysmal (positively correlated) tissue spectra (Fig. 5b), indicating distinct biochemical profiles between the two groups. Spectra from the control patient (VEC, ANC) showed consistent clustering with their respective healthy and aneurysmal counterparts. Figure 5c displays the corresponding PC2 loadings, highlighting spectral marker peaks associated with healthy and aneurysmal tissue. Peaks were matched to known vibrational modes using a ± 5 cm−1 tolerance. A summary of the putative peak assignments is provided in Table 1. In healthy tissue (negative PC2 loadings), characteristic peaks included 1091 cm⁻1 (C–O stretching in carbohydrates/glycoproteins and phosphodiester bonds), 1507 cm⁻1 (aromatic C=C stretching in phenylalanine and tryptophan side chains), and β-sheet–associated bands, consistent with a more ordered ECM composition. In aneurysmal tissue (positive PC2 loadings), spectral markers indicated Collagen I degradation, β-turn structures, hydroxyproline-related modes and increased aromatic residues, consistent with ECM disintegration and secondary structure destabilization. Notably, several peaks showed consistent matches between Amide I band deconvolution and PCA loadings. These include the peak pair at 1611/1614 cm−1, as well as 1669 rel. cm−1 (both in the deconvolution and PCA) matched. Table 1 gives an overview over the identified peaks with putative referenced peak assignments. Considering the pair at 1611/1614 rel. cm−1, increased (C=C) stretching vibrations, associated with higher levels of tyrosine and tryptophane36 are observed, whilst the vibrations at 1669 rel. cm−1 is directly associated with Collagen Type I, β-turns27,29,33,36. Discussion Our results demonstrate a distinct reorganization of Collagen types within IA walls (Figs. 3 and 4), suggesting active structural remodeling that may contribute to wall instability and increased rupture risk. Notably, Collagen Types III and IV were redistributed in aneurysmal tissue (Fig. 3), particularly in regions typically corresponding to the tunica media and intima. Collagen Types I, III and IV all showed reduced signal intensity in diseased tissue (Fig. 3), while signs of calcification and Elastin depletion were also evident. These observations reinforce the central role of ECM disruption in aneurysm pathogenesis, consistent with earlier finding by Ostergaard & Oxlund, 198745. Mapping the spatial distribution of Collagen species offers valuable insights into their biomechanical roles, especially in the context of computational hemodynamic modeling. While healthy arteries show a structured distribution of Collagens, with Collagen Type III particularly abundant, aneurysmal tissues display irregular organization and a loss of identifiable anatomical regions. Collagen Type III appears to occupy parts of altered medial layer, possibly as a compensatory remodeling response. In human arteries, the main structural Collagens are Type I and Type III, which are fibrillar and load-bearing, providing tensile strength and elasticity to the vessel wall. Elastin, likewise a fibrous protein, contributes primarily to elastic recoil. In aneurysmal tissue, the former medial layer was predominantly composed of network-forming Collagen Type IV, replacing much of the fibrillar Collagen normally present. This redistribution may represent a compensatory attempt to stabilize the vessel wall in the absence of structural Collagens such as Types I and III. The marked reduction in these fibrillar Collagens, together with a pronounced depletion of Elastin, indicates substantial ECM degradation. Loss of Elastin is biomechanically relevant, as it reduces the vessel’s elastic recoil capacity and compromises structural resilience under pulsatile blood flow. These Raman-based findings are consistent with histological observations from H&E-stained sections (Fig. 4), which showed thinning of the aneurysmal wall, focal loss of cellularity and regions suggestive of early luminal thrombosis. Areas of altered Collagen distribution in the Raman maps corresponded spatially to regions lacking clear layer boundaries in histology, underscoring the capacity of CRM to detect microstructural disorganization beyond what is visible in conventional stains34,35. Control tissue from the same patient (VEC) and from an aneurysmal control of a second patient (ANC) confirmed these distribution patterns, providing an internal and external reference for structural comparisons (Fig. 5). Genetic predisposition and enzymatic degradation are likely to act in concert in driving these ECM changes. Previous studies (e.g., Peters et al.) have shown that genes involved in ECM composition, remodeling, cell adhesion and migration are differentially expressed in aneurysmal tissue46. Of particular relevance are genetic variants in the COL3A1 gene, which can lead to altered structure or reduced expression of Collagen Type III . Given that our dataset demonstrates a marked reduction and redistribution of Collagen Type III (Figs. 3 and 4), these findings support the hypothesis that genetic predisposition and ECM degradation processes act synergistically to impair wall integrity. In parallel, increased proteolytic activity has been reported in IA tissue, notably matrix metalloproteinases (e.g., MMP-2, MT1-MMP) and plasmin which promote degradation of Collagen and Elastin16. Such enzymatic activity could explain the Collagen Type I and III degradation detected in the PCA (Fig. 5c) and Amide I deconvolution analyses (Fig. 5a, b). Whether these enzymes directly contribute to the observed Amide I band shifts remains to be determined. Potential technical limitations must also be acknowledged. In some Raman maps (Fig. 3), minor signal contributions were detected outside the main tissue boundaries, particularly in Collagen Type III heatmaps. These were carefully reviewed and attributed to low-level background scattering from the glass substrate rather than true tissue signals. Such areas were excluded from quantitative evaluation, ensuring that the primary conclusions are based on spectrally and morphologically consistent tissue regions. At the molecular level, PCA loadings in the 702–1700 cm⁻1 range (Fig. 5c) revealed distinct peaks, including a feature at ~ 1091 cm⁻1. This band is consistent with C–O stretching in carbohydrates/glycoproteins and phosphodiester bonds, and may also include contributions from hydroxyproline-containing structures. Hydroxyproline is an integral component of the Collagen helix and stabilizes the fibrillar architecture through cross-linking27. The relative prominence of this feature in regions with reduced Collagen Type III and IV (Fig. 3) may indicate a compensatory stabilization response that cannot fully restore the native structural organization. Although phosphate-associated vibrations at ~ 1091 cm⁻1 were present, their spectral profile did not match crystalline hydroxyapatite, and no sharp mineral peaks were detected29. This suggests that any mineral deposition in these samples is likely in an early or amorphous state, which may nonetheless contribute to localized ECM stiffening and warrants further study in larger cohorts. In our Raman measurements, the α-helix–related Amide I peak in aneurysmal tissue was shifted toward lower wavenumbers, consistent with protein unfolding or reduced structural order (Fig. 5a). Such secondary structure alterations, when combined with compositional changes in the ECM, reflect profound pathological remodeling processes within the aneurysm wall. Deconvolution of the Amide I band in aneurysmal tissue revealed a pronounced shift from β-sheet–associated peaks, dominant in healthy samples, toward spectral regions characteristic of α-helical structures and random coil conformations33. Such changes are consistent with protein unfolding and a loss of structural order within the extracellular matrix. PCA-based marker peak identification (Fig. 5c) further supported these findings, with aneurysmal tissue showing spectral signatures indicative of Collagen Type I degradation, increased β-turn structures and biochemical alterations involving hydroxyproline and aromatic amino acids. These molecular features suggest both compositional and conformational ECM remodeling39. In contrast, healthy tissue was characterized by spectral peaks related to cholesterol, phosphate vibrations and β-sheet–associated modes, consistent with a more ordered ECM architecture. Notably, peaks at 1611/1614 cm⁻1 in aneurysmal tissue point to elevated (C=C) stretching vibrations, likely reflecting increased levels of aromatic residues such as tyrosine and tryptophan, while the 1669 cm⁻1 feature corresponds to Collagen type I β-turns. Together, these spectral markers provide a molecular fingerprint of aneurysm-associated ECM degradation and structural destabilization. In contrast, Collagen Type IV is a network-forming Collagen, predominantly localized in the basement membrane of VSMCs and the endothelial layer. Its structure and function differ significantly from fibrillar Collagens, consistent with the observed spectral distinctions (Fig. 3). Collagen Type I, III and Elastin are predominantly located at the edge sections of the adventitia and intima in aneurysms, while a more uniform distribution corresponding to respective anatomical regions of a healthy vessel is evident (Fig. 4). Collagen Type III is particularly prominent in healthy vessels, and its deficiency in aneurysmal tissue has been reported in multiple earlier studies45,47. These works linked the absence of Collagen Type III to increased vessel extensibility despite unchanged tensile strength, implying compromised resistance under physiological pressure. Our findings are consistent with these observations and extend them by adding spectroscopic evidence of altered secondary structure. Minor variations in the exact position of the Amide I maximum compared to some published spectra are attributable to the lyophilized state of the reference proteins, differences in hydration state, and the use of 532 nm excitation. These factors do not affect the identification of characteristic marker bands or the subsequent spectral deconvolution. The position and shape of the Amide I band can vary depending on intramolecular hydrogen bonding, molecular packing, and protein conformational states. This sensitivity allows the use of Amide I characteristics, in conjunction with Amide II and III bands, to assess dominant secondary structures such as α-helices, β-sheets, and random coils, thereby providing insight into the folding state of structural proteins in both healthy and diseased vessels35. Taken together, the combination of histological and spectroscopic findings (Figs. 2, 3, 4 and 5) highlights a multifactorial remodeling process in IA walls, involving genetic predisposition, enzymatic degradation, early mineral deposition and profound alterations in Collagen secondary structure. These processes appear to act synergistically, progressively weakening the vessel wall and diminishing its capacity to withstand hemodynamic stress. Beyond proteolytic activity and genetic predisposition, chronic inflammation plays a key role in aneurysm wall remodeling. Several studies have demonstrated the infiltration of inflammatory cells—particularly macrophages and T lymphocytes—into ruptured aneurysm walls, where they secrete cytokines and matrix-degrading enzymes, exacerbating ECM disintegration11. This inflammatory activity is often accompanied by oxidative stress, which may also impair Collagen fiber integrity. Expanding CRM analysis with immunohistochemical staining or multiplex molecular profiling could help establish direct correlations between inflammation and the observed structural changes in the extracellular matrix. The application of CRM in vascular research offers unique diagnostic potential by enabling high-resolution, label-free molecular mapping of tissue composition48. In this proof-of-principle study based on a limited dataset, we demonstrated that Collagen degradation and specific spectral shifts in the Amide I band can be detected in aneurysmal tissue. While promising, clinical translation will require validation in larger patient cohorts, methodological standardization and integration with complementary histological or molecular analyses. Finally, the strong correlation between Raman spectral markers, histological features and known biomechanical roles of specific ECM components provides a compelling rationale for extending this approach to larger, clinically diverse cohorts. Such future studies will allow statistical validation of spectral biomarkers and their integration into predictive models for IA stability and rupture risk. Limitations Despite its promising results, the present study is subject to several limitations. First, the sample size is very limited, which constrains the generalizability and statistical power of the findings. This proof-of-principle investigation was based on matched aneurysmal and healthy basilar artery samples from one patient, complemented by a second patient dataset and purified reference substances (Collagen Types I, III, IV and Elastin) to support spectral assignment. While these measures provide important internal and external references, the number of biological replicates remains low and the results should be interpreted as exploratory. Future work will address this by including substantially larger, clinically diverse patient cohorts to enable robust statistical evaluation. Second, only a subset of relevant ECM components, Collagen Types I, III, IV and Elastin, were examined. Other potentially important proteins such as fibronectin or laminin were not included in the current analysis. Subsequent studies will expand the reference library to capture a broader spectrum of ECM constituents. A key methodological lesson from this study is the critical importance of consistent sample preparation in CRM experiments, particularly when analyzing human intracranial aneurysm tissue. Variability in preparation can lead to inconsistencies in spectral quality, data acquisition and interpretation, thereby affecting reproducibility. To mitigate this, future studies will implement standardized protocols for tissue handling, storage and processing, ensuring higher comparability between datasets. In addition to increasing the sample size, planned research will integrate correlative statistical analyses that account for patient-specific factors such as age, sex, aneurysm size, rupture status, presence of multiple aneurysms and degree of atherosclerosis. Examining these parameters will help identify potential associations or causal relationships, thereby providing a more comprehensive understanding of the pathophysiology of intracranial aneurysms. The planned studies will also employ advanced statistical modeling to evaluate how these patient-specific variables influence biochemical composition and structural changes detected by CRM. This multifactorial approach will not only validate the initial observations but also enhance our understanding of how individual differences affect disease progression and therapeutic outcomes. By broadening the patient demographic and integrating molecular, clinical and biomechanical data, future research may reveal population-specific patterns relevant for personalized intervention strategies and improved clinical decision-making. Conclusion & outlook Understanding the cellular degeneration and structural reorganization that precede IA formation is essential for improving risk assessment and therapeutic strategies. This feasibility study was designed to explore the potential of confocal CRM rather than to draw definitive clinical conclusions. In this proof-of-principle study, we demonstrate that confocal CRM can detect both compositional and conformational alterations in extracellular matrix ECM components, particularly Collagen Types I, III, IV and Elastin, and reveal shifts in protein secondary structure associated with aneurysmal remodeling. Future investigations should focus on the role of inflammation in modulating SMC behavior and Collagen turnover. Combining spectroscopic markers of ECM degradation with histological and molecular indicators of inflammation could provide a more complete mechanistic framework. A deeper understanding of inflammation-driven ECM remodeling may reveal novel therapeutic targets aimed at stabilizing aneurysm walls and preventing rupture. A more comprehensive mapping of protein organization and distribution in human IAs may support the development of conservative, individualized treatment approaches, potentially reducing the need for invasive interventions. Integrating Raman-derived molecular fingerprints with clinical risk factors, hemodynamic simulations and biomechanical measurements could improve predictive modeling of aneurysm behavior. While the present single-patient dataset provides promising preliminary evidence, validation in larger, clinically diverse cohorts with rigorous statistical analysis is necessary. Expanding this approach to include longitudinal follow-up, genetic profiling and multimodal imaging will be critical for translating these findings into robust clinical decision-support tools. Methods Tissue preparation and histology A matched tissue pair from a single patient—comprising a healthy cerebral artery (VE) and an intracranial aneurysm of the basilar artery (AN1-6)—was sectioned into six slices, each 5 µm thick. In addition, a second matched tissue pair from a control patient was analyzed, consisting of a healthy cerebral artery (VEC) and an intracranial aneurysm (ANC). Tissue specimens were stored at − 80 °C and sectioned at − 20 °C using a Leica cryo-microtome with polyvinyl alcohol (PVA) based cryo-gel (Leica). This water-soluble medium was additionaly analyzed via Raman spectroscopy to exclude spectral contamination. The first section of each pair was stained with H&E for histological evaluation. Healthy regions of interest (ROIs) for Raman analysis were preselected by examining the adjacent H&E-stained sections for preserved histoarchitecture, including intact tunica intima, media and adventitia. Only regions meeting these morphological criteria were chosen for hyperspectral Raman imaging. The nomenclature for all scanned slices (VE1-6, AN1-6, VEC, ANC)is summerized in Fig. 6. Histological changes in the aneurysmal wall included marked wall thinning, hypocellularity, focal regions of intimal hyperplasia, with aggregated cell clusters and decellularized areas suggestive of early luminal thrombosis (Fig. 2).Fig. 6(a) Schematic illustration of the workflow used to generate a 3D anatomical model of the basilar artery from patient-specific pre-mortem MRI scan data. These imaging data were acquired during routine clinical diagnostics prior to death and served to localize the aneurysm morphology in vivo. (b) and (c) Localization and selection of the target region within the 3D model for subsequent histological and spectroscopic analysis. Image evaluation was performed with OsiriX Lite (Pixmeo, v.12.0, https://www.osirix-viewer.com) and segmentation with 3D Slicer (v.5.6, https://www.slicer.org, see ref. 56). (d) Preparation of mounted cryo-sectioned tissue slices derived from the identified region, including aneurysmal control (ANC) and vessel control (VEC) samples. (a) Schematic illustration of the workflow used to generate a 3D anatomical model of the basilar artery from patient-specific pre-mortem MRI scan data. These imaging data were acquired during routine clinical diagnostics prior to death and served to localize the aneurysm morphology in vivo. (b) and (c) Localization and selection of the target region within the 3D model for subsequent histological and spectroscopic analysis. Image evaluation was performed with OsiriX Lite (Pixmeo, v.12.0, https://www.osirix-viewer.com) and segmentation with 3D Slicer (v.5.6, https://www.slicer.org, see ref. 56). (d) Preparation of mounted cryo-sectioned tissue slices derived from the identified region, including aneurysmal control (ANC) and vessel control (VEC) samples. MRI data and 3D reconstruction Preoperative MRI data were obtained from routine clinical imaging. Contrast-enhanced magnetic resonance angiography (CE-MRA) was performed on a MAGNETOM Aera 1.5 T scanner (Siemens Healthineers) using gadobutrol (Gadovist) as contrast agent. DICOM data were analyzed using OsiriX Lite (Pixmeo) and image segmentation was carried out using 3D Slicer47. A 3D model of the aneurysm was reconstructed to enable precise spatial co-registration of the in vivo aneurysm morphology with the vascular segment sampled during autopsy for histological and CRM analysis (Fig. 6). This approach ensured that all analyzed tissue slices originated from the anatomically defined aneurysm location documented in the imaging data11,44. Confocal Raman imaging, marker-peak identification and deconvolution Raman imaging was conducted using a WITec alpha300 RA confocal Raman microscope with a 532 nm excitation laser. Spectral acquisition was performed at 8.5 mW laser power with 0.5 s intergration time per pixel. Autofocus steps covered a 6 µm focal height range with a minimum integration time of 0.1 s. For aneurysmal tissue, images were acquired at full spatial resolution (1 µm/pixel) over areas of 249 × 116 µm, yielding approximately 28.800 spectra per scan. For the larger healthy vessel segments, scans covered 1100 × 100 µm at half resolution (2 µm/pixel), resulting in around 55.000 spectra per scan. To identify potential biomarkers, cross-sectional line scans of the vessel wall were acquired using the same laser wavelength of 532 nm at a higher power (25 mW) and cooled sample temperature to approximately 2 °C. These conditions (25 mW at 532 nm with a cooled stage at ~ 2 °C) were chosen to enhance the signal-to-noise ratio (SNR) while limiting the thermal load on the sample. Thermal stability was verified by the absence of emergent carbonization/graphitic bands, lack of progressive spectral drift across repeated acquisitions, and post-scan microscopic inspection to confirm no visible damage. Line scans were recorded at half resolution (2 µm/pixel), with measurement points per line ranging from 171 to 210 pixels, depending on section length. Spectral preprocessing and component analysis Raw spectra were preprocessed in Project FIVE (WITec) following a defined sequence of steps. First, cosmic ray artifacts were removed. Background fluorescence was then subtracted using a rolling-ball algorithm, followed by Savitzky–Golay smoothing. All spectra were subsequently normalized to the C–H stretching region (2950–3050 cm⁻1), which was consistent across all measurements. Processed Raman maps were subjected to TCA, a NMF–based algorithm implemented in Project FIVE (WITec), to identify five dominant spectral components corresponding to structural ECM constituents: Collagen Type I (Merck, C7774-5MG, powder from human placenta), Collagen Type III (Merck, CC054, 1 mg/mL in acetic acid, pH 2.5, extracted from human placenta), Collagen Type IV (Merck, C5533-5MG, powder from human placenta) and Elastin (Merck, E6902-1MG, lyophilized powder extracted from human aorta). Reference spectra for these components were acquired under identical measurement parameters. Collagen Types I and IV, as well as Elastin, were measured in bulk lyophilized form, whereas Collagen Type III was obtained by vacuum evaporation (0.075 mbar, Schlenk line) from acidic solution until a solid residue remained for analysis. Differences in hydration state between reference materials and native tissue can cause minor shifts in the Amide I maximum relative to hydrated Collagen in the literature, due to altered hydrogen-bonding environments. All reference substances were pre-evaluated using RamanLIGHT49 in MATLAB (MathWorks). To minimize spurious substrate signals or low-intensity edge artifacts, component maps were masked using the C–H stretch intensity to define on-tissue regions and low-intensity pixels at the scan margins were excluded from quantitative interpretation. This procedure reduced background scattering, particularly evident at tissue boundaries for certain components. Following TCA, extracted component spectra were averaged to account for NMF-induced dimensionality reduction. The resulting average spectra represent heterogeneous composites of structurally distinct tissue regions and are not expected to match any single reference spectrum; rather, they reflect the weighted sum of multiple biochemical constituents in the scanned area. Scans from tissue pairs AN3 and AN5 were excluded from the analysis due to irregular focal artifacts, low signal-to-noise ratio and dominant substrate features, in line with pre-specified exclusion criteria. Spectral deconvolution and statistical analysis Average spectra were calculated from area scans for both healthy and diseased samples. Baseline correction was performed in two stages. First, a rolling-ball algorithm was applied in Project FIVE (WITec) at the individual spectrum level to remove broad fluorescence background. Subsequently, an alternating least squares (ALS) baseline correction (OriginPro, standard parameters) was applied to the averaged spectra from each group to minimize residual low-frequency curvature prior to peak fitting. This second step was applied identically to all spectra (healthy, aneurysmal and reference materials) and improves the accuracy of peak position estimation without altering relative peak ordering. Residuals were visually inspected to ensure no over-correction occurred. Bigaussian peak deconvolution of the Amide I region (1600–1700 cm⁻1) was then performed using OriginPro, applying 100 fitting iterations. Hidden peaks were detected based on either the second derivative or the residuals after first derivative smoothing, depending on signal complexity and baseline behavior. PCA was conducted at the per-scan level in an exploratory manner, with PC1 (substrate signal) excluded from interpretation. Loadings were interpreted as indicators of contribution to group separation rather than effect sizes and characteristic peaks were matched to known vibrational modes within ± 5 cm⁻1. No classifier was trained due to the limited and imbalanced dataset and the analysis was intended to generate hypotheses for future studies.