A particular optical modality provides information due to a specific contrast mechanism. This contrast can, however, have a multitude of origins, resulting in challenges when interpreting the data. Frequently, and especially in OCT and RS, the interpretation of data is based on the experience of the user or a reference analysis, such as a histopathological examination of a biopsy sample. The combination of two complementary modalities has mostly been used to increase the diagnostic value of the underlying pathology but could also bear the potential to better understand and reference the fundamental origin of the signals. For example, while the origin for OCT and the Raman signals is based on different physical principles, i.e. OCT on scattering properties of the tissue and RS on the molecular composition of the tissue, the scattering properties of the tissue also depend on the molecular composition, making the origins of both modalities inherently connected. To investigate this relation, an optical setup was developed, based on two forward-viewing fiber-optical probes, which allows consecutive acquisition of Raman maps and OCT volume stacks from large biopsy samples, Figure 1. For more details on the system, see Materials and Methods section. Human urinary bladder biopsies were used as surrogate for those experiments.
A total of 119 biopsies were imaged in a co-registered way with a forward viewing endoscope-based setup. Because the imaging data for both modalities on biopsies was co-registered, it offers the opportunity to characterize the signal origins for both modalities and qualitatively correlate the data. To demonstrate a potential relation between both modalities the data was divided in separate groups, based on the information present in the Raman data. Depending on the biopsy the molecular signatures of the Raman signal show the presence of certain macromolecules commonly found in tissues. We have therefore divided the biopsies in three groups: lipid-rich, collagen-rich, and epithelial rich samples. For the manuscript a subset of representative biopsies was selected and analyzed. Figure 2, 3 and 5 show healthy bladder tissue, while Figure 4 is based on data from an early tumor stage. Additionally, a comparison to histopathological data for those biopsies was also performed.
Figure 2 displays the co-localized morphological and molecular data for a representative biopsy, exhibiting very strong collagen signal, but only little epithelium and lipid signal in the Raman spectra. Additionally, the histological (H&E) information for the biopsy is presented. As described in the material and methods section, the individual Raman spectra were fitted by three components, representing spectral information of epithelium, lipid, and collagen, and are displayed as distribution enface maps of those components over the sample. For a better differentiation the epithelium was color-coded red (Fig. 2a), collagen green (Fig. 2b) and lipid blue (Fig. 2c). The strongest contribution appears in collagen, which is homogenous throughout the biopsy. The lipid and epithelium signals, on the other hand, are barely visible. Additionally, to the Raman maps, an OCT enface image, based on the maximum intensity projections of ten slices from a depth of 120 -150 µm is also displayed, showing very bright signal features throughout the entire area of the biopsy, Fig. 2d. As reported by Sergeev et al.34, collagen fibers cause an increase in ballistic scattering of photons, resulting in a very bright OCT enface projection. To gain a better understanding of the underlying depth information of the data, B-scans are displayed, Fig. 2e and f, supporting the overall bright appearance of the signal in the top layer of the tissue. The collagen-rich layer has an approximate thickness of 100 µm, as indicated in the image. No other notable features are apparent in the B-scans. For comparison, the Raman signal for the same cross-section for each molecular component was plotted above the B-scan. Here again, the Raman signal and the OCT data correspond quite well, displaying strong features, where the scattering in OCT is also highest. The Raman signals are detectable in depth up to ~300 µm, while the OCT probe provides an efficient collection of signals within 950 µm. To support the morphological and molecular co-localized data, an H&E-stained section from the same biopsy is presented, Fig. 2g. The H&E slide of the biopsy confirms that the tissue is primarily composed of lamina propria, with no apparent presence of epithelium, confirming the observations from RS and OCT. It is possible that the epithelium was detached during handling of the biopsy after resection. The H&E slice does, however, indicate edema in one part of the lamina propria and two representative fibroblasts, having the task of building up the extracellular matrix mainly composed of collagen fibers, are indicated. These results support the notion that at least for collagen-related signals both modalities provide relatable information, and Raman spectroscopy could be used to specify the presence of collagen in the OCT data. This particular information could be a potential biomarker and define the transition from benign to malignant tissue.
Clearly, the sole presence of collagen features is undeniably a special case for this sample and in most biopsies more than one component can be found. Fig. 3a and b present an example were the enface Raman maps not only display collagen features, but also signal contributions from epithelium tissue. Both molecular maps show that the constituents are distributed homogeneously throughout the biopsy, except for the central location where the collagen density is increased, while the epithelium contribution is reduced. Again, no lipid signal can be observed in the Raman map, Fig. 3c. The corresponding, enface image of OCT is displayed in Fig. 3d. Because the data was normalized over the entire data set for both modalities it is also possible to compare the observations between the different samples (Fig. 2-5). Hence, taking into consideration the observation from Fig. 2, i.e. a high collagen presence resulted in a very bright OCT signal, here, one should see overall an apparent reduction in the OCT signal intensity, except in the location where the collagen Raman map shows higher signals. Indeed, the OCT enface image Fig. 3d has a reduction in brightness when compared to the OCT enface image Fig. 1d. More specifically, the signal is not homogenous throughout the entire sample, but has a central location where the OCT signal exhibits a particular high brightness, while the other positions have considerably lower intensities. When comparing the OCT enface projection to the Raman maps of the same locations that display a strong OCT signal, an increase in collagen contributions in the Raman image can be seen. The reduced OCT signal, on the other hand, correlates with a reduction in collagen, but an increase in epithelium signal in RS. To better comprehend the local morphological differences in depth a cross-sectional representations for the indicated positions of the enface OCT are displayed in Fig. 3e and f. Additionally, the molecular signals of the epithelium, collagen and lipid from Raman spectroscopy for that particular cross-section are displayed above the image. Here again, it is possible to relate the OCT depth profiles, which show a heterogeneous variation of the signal intensity, to the molecular variation from the enface Raman maps for the presence of collagen and epithelium tissue. It can be seen that the signals from both modalities co-localize extremely well, indicating that the pre-processing and the co-registration worked appropriately. Regions of more collagen (lamina propria) are giving a brighter OCT signal in depths of 100 - 200 µm, which was also previously observed by Sergeev et al.34, while regions where the OCT signal is reduced a stronger epithelium signal in Raman is visible. To support this information, an H&E slice for this biopsy is presented, Fig. 3g. The bladder wall with an epithelial layer and lamina propria can be seen, confirming the observations provided by the cross-linked OCT and RS information. Furthermore, mechanical artifacts and the deeper layer of the lamina propria are visible.
An additional example for a mixed epithelium and collagen sample is shown in Figure 4. Here, a similar relation between the decomposed Raman signals and the OCT signal, but in a more heterogeneously way, as described in Fig. 3, is present. For example, the B-scans unveil transition zones of lamina propria to epithelium, Fig. 4e and f, and the co-localized molecular signals from the RS also indicate such transition zones. The collagen rich layer is extending in depth from ~0 - 50 µm up to ~0 - 250 µm, Fig. 4f and e, respectively. On the right side of Fig. 4e, the OCT signal exhibits a homogenous intensity over the entire depth and the molecular contributions to the Raman signal are mainly from the epithelium tissue. The histopathological image, Fig. 4g, shows thickened epithelium on the right-hand side and lamina propria on the left-hand side. This may support the finding in Fig. 4e of the thick epithelium layer and its homogenous appearance. Additionally, vessels are indicated in Fig. 4g.
There are also cases where other signal features can occur in both modalities. Figure 5 shows a representative biopsy with more heterogenous behavior, i.e. where besides the collagen and epithelium signals also substantial contributions of lipids are present, Fig. 5 a-c. In comparison to previous cases, Fig. 2-4, the amount of lipid signals is substantial and well defined, and results in a decrease in epithelium and collagen signals in those particular locations. As was previously shown, the signal intensity/brightness corresponds well to the scattering property of the tissue, i.e. collagen contributions resulted in a high OCT signal, while epithelium tissue resulted in a reduced brightness in comparison to collagen, Fig. 2-4. Taking into consideration that lipid accumulations are highly homogeneous and low scattering37, the observed OCT signal should be even more reduced, as in comparison to epithelium tissue. Fig. 5d shows the corresponding enface OCT image, which has next to the readily observed variation of signal intensity attributed to variation of epithelium and collagen signals, very apparent and distinct black voids. Taking only the OCT information into account, those voids could also easily be assigned to mechanical rupture of the tissue, which frequently occurs through stress during the handling. However, by taking the information extracted from the Raman data those voids can be clearly related to accumulations of esterified lipids. The present lipid signal (Fig. 5c) is significantly higher than present in the previous biopsies in Fig. 2-4 and overlaps with the region of the black features in the OCT image. Based on the depth information of OCT, the lipid signals are co-localized with the black voids in the OCT B-scans in depth (Fig. 5e and f). RS is giving evidence, that those black voids are not caused by mechanical forces but are filled with esterified lipids. Furthermore, the beneficial depth information of OCT enables the detection of the RS lipid intense signal regions in a depth of ~100 - 250 µm. Additionally, the RS collagen intense signals are co-localized with the bright OCT signals at a depth of ~0 - 50 µm. Through the morpho-molecular cross-linking of OCT and RS, the signal origin and the understanding of the present biological tissue is substantially enhanced. Additionally, the H&E slice support the presence of lipid pools within the bladder wall and also epithelium and lamina propria, Fig. 4g. Besides mechanical artifacts, the detrusor muscle is visible. Those results strongly support the notion that, while the signal origin of both modalities is complementary, they both provide a very beneficial information that assists to explain the corresponding modality.