Measurement of cartilage sub-component distributions through the surface by Raman spectroscopy-based multivariate analysis

Articular cartilage posesses unique material properties due to a complex depth-dependent composition of sub-components. Raman spectroscopy has proven valuable in quantifying this composition through cartilage cross-sections. However, cross-sectioning requires tissue destruction and is not practical in situ. In this work, Raman spectroscopy-based multivariate curve resolution (MCR) was employed in porcine cartilage samples (n = 12) to measure collagen, glycosaminoglycan, and water distributions through the surface for the first time; these were compared against cross-section standards. Through the surface Raman measurements proved reliable in predicting composition distribution up to a depth of approximately 0.5 mm. A fructose-based optical clearing agent (OCA) was also used in an attempt to further improve depth of resolution of this measurement method. However, it did not; mainly due to a high-spectral overlap with the Raman spectra of main cartilage sub-components. This measurement technique potentially could be used in situ, to better understand the etiology of joint diseases such as osteoarthritis (OA).     

Determination of inclusion depth in ex vivo animal tissues using surface enhanced deep Raman spectroscopy

This work presents recent developments in spatially offset and transmission Raman spectroscopy for noninvasive detection and depth prediction of a single SERS inclusion located deep inside ex vivo biological tissues. The concept exploits the differential attenuation of Raman bands brought about by their different absorption due to tissue constituents enabling to predict the inclusion depth. Four different calibration models are tested and evaluated to predict the depth of surface enhanced Raman scattering labelled nanoparticles, within an up to 40 mm slab of porcine tissue. An external measurement carried out in transmission mode, with a noninvasively built model on the analysed sample, is shown to be insensitive to variations of the overall thickness of the tissue yielding an average root‐mean‐square error of prediction of 6.7%. The results pave the way for future noninvasive deep Raman spectroscopy in vivo enabling to localise cancer biomarkers for an early diagnosis of multiple diseases.      

Quantitative determination of concentration profiles of skin components and topically applied oils by tailored multivariate curve resolution-alternating least squares using in vivo confocal Raman micro-spectroscopy

The main components of the stratum corneum (SC), water, lipids, and proteins, are non-homogeneously distributed throughout the depth. The quantitative determination of their concentration profiles and penetration depth of topically applied substances are urgent topics of dermatological and cosmetic research. Confocal Raman micro-spectroscopy has distinct advantages when determining semi-quantitative concentrations of SC components and topically applied substances non-invasively and in vivo. In this work, we applied a tailored multivariate curve resolution-alternating least squares (tMCR-ALS) method to analyze Raman spectra of the SC in the 2000–4000 cm⁻¹ region for quantitatively determining the concentrations of water, lipids, proteins, and topically applied oils using substance-related spectral loadings which were allowed to change depth-dependently from the SC's surface toward its bottom. tMCR-ALS makes matching of depth-dependent signal attenuation, that is, the normalization on keratin, unnecessary and requires only a few additional experiments for calibration – Raman spectra of the pure materials and their densities.     

An optical system for noninvasive microscopy of psoriatic mice in vivo

Psoriasis is a chronic inflammatory skin disease involved with both complex morphological changes of skin and immune processes. The clinical diagnostics and research of psoriasis often require invasive biopsy which lacks their real-time dynamics in vivo. Here we report a noninvasive microscopic system developed by combining in vivo fluorescent microscopy, optical clearing, and immunolabeling to enable real-time imaging of immune cells and cytokines in blood flow in psoriatic animal models. The vascular morphology and time-lapse kinetics of interleukin (IL)-23, IL-17, tumor necrosis factor-α, and CD4+ cells in blood are captured at submicron resolution through the thickening epidermis and opaque scales during the development of psoriasis in vivo. Our data suggest IL-23 recruits CD4+ cells to release IL-17 in blood that further leaks out in the psoriatic skin area. This optical system enables noninvasive and real-time assessment of immune molecules and cells in vivo, providing good potential for medical researches on psoriasis.     

Deep-skin multiphoton microscopy in vivo excited at 1600 nm: A comparative investigation with silicone oil and deuterium dioxide immersion

Multiphoton microscopy (MPM) excited at the 1700-nm window has enabled deep-tissue penetration in biological tissue, especially brain. MPM of skin may also benefit from this deep-penetration capability. Skin is a layered structure with varying refractive index (from 1.34 to 1.5). Consequently, proper immersion medium should be selected when imaging with high numerical aperture objective lens. To provide guidelines for immersion medium selection for skin MPM, here we demonstrate comparative experimental investigation of deep-skin MPM excited at 1600 nm in vivo, using both silicone oil and deuterium dioxide (D₂O) immersion. We specifically characterize imaging depths, signal levels and spatial resolution. Our results show that both immersion media give similar performance in imaging depth and spatial resolution, while signal levels are slightly better with silicone oil immersion. We also demonstrate that local injection of fluorescent beads into the skin is a viable technique for spatial resolution characterization in vivo.      

Raman spectroscopy of human skin for kidney failure detection

The object of this paper is in vivo study of skin spectral-characteristics in patients with kidney failure by conventional Raman spectroscopy in near infrared region. The experimental dataset was subjected to discriminant analysis with the projection on latent structures (PLS-DA). Application of Raman spectroscopy to investigate the forearm skin in 85 adult patients with kidney failure (90 spectra) and 40 healthy adult volunteers (80 spectra) has yielded the accuracy of 0.96, sensitivity of 0.94 and specificity of 0.99 in terms of identifying the target subjects with kidney failure. The autofluorescence analysis in the near infrared region identified the patients with kidney failure among healthy volunteers of the same age group with specificity, sensitivity, and accuracy of 0.91, 0.84, and 0.88, respectively. When classifying subjects by the presence of kidney failure using the PLS-DA method, the most informative Raman spectral bands are 1315 to 1330, 1450 to 1460, 1700 to 1800 cm⁻¹. In general, the performed study demonstrates that for in vivo skin analysis, the conventional Raman spectroscopy can provide the basis for cost-effective and accurate detection of kidney failure and associated metabolic changes in the skin.     

In vivo detection of human cutaneous beta‐carotene using computational optical clearing

The content of dermal beta‐carotene can be a good indicator showing the body health. Because, it is involved in production of vitamin A maintaining healthy skin and mucous membranes. Also, it reduces the risk of cardiovascular diseases and its antioxidant capacity prevents the formation of cancerous cells. In this work, we use Raman spectroscopy and a low‐cost diffuse reflectance spectroscopy (DRS) to detect the dermal beta‐carotene spectra. We apply computational optical clearing (OC) method to in vivo evaluation the concentration of this chromophore. The results show that Raman spectroscopy is a good tool for in vitro detection of carotenoids but is not able to clearly discriminate the individual carotenoids in skin tissue in vivo. The results also show that using OC enhances the ability of low‐cost diffuse reflection spectroscopy for in vivo detection of dermal beta‐carotene in humans. This method can be used as a low‐cost and portable device to screening the concentration of chromophores such as melanin and carotenoid molecules for oncological studies.     

Unique spectral signature of human cutaneous squamous cell carcinoma by photoacoustic imaging

Cutaneous squamous cell carcinoma (cSCC) is a common skin cancer with metastatic potential. To reduce reoperations due to nonradical excision, there is a need to develop a technique for identification of tumor margins preoperatively. Photoacoustic (PA) imaging is a novel imaging technology that combines the strengths of laser optics and ultrasound. Our aim was to determine the spectral signature of cSCC using PA imaging and to use this signature to visualize tumor architecture and borders. Two‐dimensional PA images of 33 cSCCs and surrounding healthy skin were acquired ex vivo, using 59 excitation wavelengths from 680 to 970 nm. The spectral response of the cSCCs was compared to healthy tissue, and the difference was found to be greatest at wavelengths in the range 765 to 960 nm (P < .05). Three‐dimensional PA images were constructed from spectra obtained in the y‐z plane using a linear stepper motor moving along the x‐plane. Spectral unmixing was then performed which provided a clear three‐dimensional view of the distribution of tumor masses and their borders.     

Deep learning protocol for improved photoacoustic brain imaging

One of the key limitations for the clinical translation of photoacoustic imaging is penetration depth that is linked to the tissue maximum permissible exposures (MPE) recommended by the American National Standards Institute (ANSI). Here, we propose a method based on deep learning to virtually increase the MPE in order to enhance the signal‐to‐noise ratio of deep structures in the brain tissue. The proposed method is evaluated in an in vivo sheep brain imaging experiment. We believe this method can facilitate clinical translation of photoacoustic technique in brain imaging, especially in transfontanelle brain imaging in neonates.     

A representation learning approach for recovering scatter-corrected spectra from Fourier-transform infrared spectra of tissue samples

Infrared spectra obtained from cell or tissue specimen have commonly been observed to involve a significant degree of scattering effects, often Mie scattering, which probably overshadows biochemically relevant spectral information by a nonlinear, nonadditive spectral component in Fourier transform infrared (FTIR) spectroscopic measurements. Correspondingly, many successful machine learning approaches for FTIR spectra have relied on preprocessing procedures that computationally remove the scattering components from an infrared spectrum. We propose an approach to approximate this complex preprocessing function using deep neural networks. As we demonstrate, the resulting model is not just several orders of magnitudes faster, which is important for real-time clinical applications, but also generalizes strongly across different tissue types. Using Bayesian machine learning approaches, our approach unveils model uncertainty that coincides with a band shift in the amide I region that occurs when scattering is removed computationally based on an established physical model. Furthermore, our proposed method overcomes the trade-off between computation time and the corrected spectrum being biased towards an artificial reference spectrum.     

Diseased thyroid tissue classification in OCT images using deep learning: Towards surgical decision support

Intraoperative guidance tools for thyroid surgery based on optical coherence tomography (OCT) could aid distinguish between normal and diseased tissue. However, OCT images are difficult to interpret, thus, real-time automatic analysis could support the clinical decision-making. In this study, several deep learning models were investigated for thyroid disease classification on 2D and 3D OCT data obtained from ex vivo specimens of 22 patients undergoing surgery and diagnosed with several thyroid pathologies. Additionally, two open-access datasets were used to evaluate the custom models. On the thyroid dataset, the best performance was achieved by the 3D vision transformer model with a Matthew's correlation coefficient (MCC) of 0.79 (accuracy = 0.90) for the normal-versus-abnormal classification. On the open-access datasets, the custom models achieved the best performance (MCC > 0.88, accuracy > 0.96). Results obtained for the normal-versus-abnormal classification suggest OCT, complemented with deep learning-based analysis, as a tool for real-time automatic diseased tissue identification in thyroid surgery.     

Detecting mouse squamous cell carcinoma from submicron full-field optical coherence tomography images by deep learning

The standard medical practice for cancer diagnosis requires histopathology, which is an invasive and time-consuming procedure. Optical coherence tomography (OCT) is an alternative that is relatively fast, noninvasive, and able to capture three-dimensional structures of epithelial tissue. Unlike most previous OCT systems, which cannot capture crucial cellular-level information for squamous cell carcinoma (SCC) diagnosis, the full-field OCT (FF-OCT) technology used in this paper is able to produce images at sub-micron resolution and thereby facilitates the development of a deep learning algorithm for SCC detection. Experimental results show that the SCC detection algorithm can achieve a classification accuracy of 80% for mouse skin. Using the sub-micron FF-OCT imaging system, the proposed SCC detection algorithm has the potential for in-vivo applications.     

Photoacoustic imaging of in vivo hemodynamic responses to sodium nitroprusside

The in vivo hemodynamic impact of sodium nitroprusside (SNP), a widely used antihypertensive agent, has not been well studied. Here, we applied functional optical-resolution photoacoustic microscopy (OR-PAM) to study the hemodynamic responses to SNP in mice in vivo. As expected, after the application of SNP, the systemic blood pressure (BP) was reduced by 53%. The OR-PAM results show that SNP induced an arterial vasodilation of 24% and 23% in the brain and skin, respectively. A weaker venous vasodilation of 9% and 5% was also observed in the brain and skin, respectively. The results show two different types of blood oxygenation response. In mice with decreased blood oxygenation, the arterial and venous oxygenation was respectively reduced by 6% and 13% in the brain, as well as by 7% and 18% in the skin. In mice with increased blood oxygenation, arterial and venous oxygenation was raised by 4% and 22% in the brain, as well as by 1% and 9% in the skin. We observed venous change clearly lagged the arterial change in the skin, but not in the brain. Our results collectively show a correlation among SNP induced changes in systemic BP, vessel size and blood oxygenation.     

Multispectral near infrared absorption imaging for histology of skin cancer

Multispectral imaging combines the spectral resolution of spectroscopy with the spatial resolution of imaging and is therefore very useful for biomedical applications. Currently, histological diagnostics use mainly stainings with standard dyes (eg, hematoxylin + eosin) to identify tumors. This method is not applicable in vivo and provides low amounts of chemical information. Biomolecules absorb near infrared light (NIR, 800‐1700 nm) at different wavelengths, which could be used to fingerprint tissue. Here, we built a NIR multispectral absorption imaging setup to study skin tissue samples. NIR light (900‐1500 nm) was used for homogenous wide‐field transmission illumination and detected by a cooled InGaAs camera. In this setup, images I(x, y, λ) from dermatological samples (melanoma, nodular basal‐cell carcinoma, squamous‐cell carcinoma) were acquired to distinguish healthy from diseased tissue regions. In summary, we show the potential of multispectral NIR imaging for cancer diagnostics.      

Optimization of spatial resolution and scattering effects for biomedical fluorescence imaging by using sub-regions of the shortwave infrared spectrum

We evaluated the impact of light-scattering effects on spatial resolution in different shortwave infrared (SWIR) sub-regions by analyzing two SWIR emissive phantoms made of polydimethylsiloxane (PDMS)-gold nanoclusters (Au NCs) composite covered with mice skin, or capillary tubes filled with Au NCs or IRDye 800CW at different depth in intralipids and finally, after administration of the Au NCs intravenously in mice. Our findings highlighted the benefit of working at the highest tested spectral range of the SWIR region with a 50% enhancement of spatial resolution measured in artificial model when moving from NIR-II (1000-1300 nm) to NIR-IIa (1300-1450 nm) region, and a 25% reduction of the scattering from the skin determined by point spread function analysis from the NIR-II to NIR-IIb region (1500-1700 nm). We also confirmed that a series of Monte Carlo restoration of images significantly improved the spatial resolution in vivo in mice in deep tissues both in the NIR-II and NIR-IIa spectral windows.     

Miniaturized all fiber probe for optical coherence tomography and pH detection of biological tissue

We present a novel all-fiber probe with 710-μm outside diameter for combined optical coherence tomography and pH detection. In cancer surgery, a significant challenge is how to completely remove the malignant tumor without cutting too much normal tissue. The difference between cancer tissue and normal tissue not only lies in morphology and structure but also in tissue pH, where malignant tissue has a lower pH. This dual-modality probe combined optical coherence tomography and pH detection of biological tissue, is expected to determine whether the tissue is cancerous quickly and accurately. The probe utilizes a typical three-segment structure (double-clad fiber - no-core fiber - graded-index fiber). We obtained a lateral resolution of ~10.6 μm, a working distance of ~506 μm and a pH measurement accuracy of 0.01 pH unit for the probe. The performance of the all-fiber probe was verified through an ex vivo experiment using the porcine brain specimen.     

Exploring the utility of Deep Red Anthraquinone 5 for digital staining of ex vivo confocal micrographs of optically sectioned skin

We investigated the utility of the fluorescent dye Deep Red Anthraquinone 5 (DRAQ5) for digital staining of optically sectioned skin in comparison to acridine orange (AO). Eight fresh-frozen thawed Mohs discard tissue specimens were stained with AO and DRAQ5, and imaged using an ex vivo confocal microscope at three wavelengths (488 nm and 638 nm for fluorescence, 785 nm for reflectance). Images were overlaid (AO + Reflectance, DRAQ5 + Reflectance), digitally stained, and evaluated by three investigators for perceived image quality (PIQ) and histopathological feature identification. In addition to nuclear staining, AO seemed to stain dermal fibers in a subset of cases in digitally stained images, while DRAQ5 staining was more specific to nuclei. Blinded evaluation showed substantial agreement, favoring DRAQ5 for PIQ (82%, Cl 75%-90%, Gwet's AC 0.74) and for visualization of histopathological features in (81%, Cl 73%-89%, Gwet's AC 0.67), supporting its use in digital staining of multimodal confocal micrographs of skin.     

Reliability of wavefront shaping based on coherent optical adaptive technique in deep tissue focusing

Wavefront shaping can compensate the wavefront distortions in deep tissue focusing, leading to an improved penetration depth. However, when using the backscattered signals as the feedback, unexpected compensation bias may be introduced, resulting in focusing position deviations or even no focus in the illumination focal plane. Here we investigated the reliability of wavefront shaping based on coherent optical adaptive technique in deep tissue focusing by measuring the position deviations between the foci in the illumination focal plane and the epi‐detection plane. The experimental results show that when the penetration depth reaches 150 μm in mouse brain tissue (with scattering coefficient ~22.42 mm^?1) using a 488 nm laser and an objective lens with 0.75 numerical aperture, the center of the real focus will deviate out of one radius range of the Airy disk while the optimized focus in the epi‐detection plane maintained basically at the center. With the penetration depth increases, the peak to background ratio of the focus in the illumination focal plane decreases faster than that in the epi‐detection plane. The results indicate that when the penetration depth reaches 150 μm, feedback based on backscattered signals will make wavefront shaping lose its reliability, which may provide a guidance for applications of non‐invasive precise optogenetics or deep tissue optical stimulation using wavefront shaping methods. A, Intensity distribution in the epi‐detection plane and the illumination focal plane before and after correction, corresponding to brain sections with 250 and 300 μm thickness, respectively. Scale bar is 2 μm. B, Averaged focusing deviations in the epi‐detection plane (optimized) and the illumination focal plane (monitored) after compensation. The unit of the ordinate is one Airy disk diameter. Black dashed line represents one Airy disk radius. Bars represent the SE of each measurement set.      

Deep learning: step forward to high-resolution in vivo shortwave infrared imaging

Shortwave infrared window (SWIR: 1000–1700 nm) represents a major improvement compared to the NIR-I region (700–900 nm) in terms of temporal and spatial resolutions in depths down to 4 mm. SWIR is a fast and cheap alternative to more precise methods such as X-ray and opto-acoustic imaging. Main obstacles in SWIR imaging are the noise and scattering from tissues and skin that reduce the precision of the method. We demonstrate that the combination of SWIR in vivo imaging in the NIR-IIb region (1500–1700 nm) with advanced deep learning image analysis allows to overcome these obstacles and making a large step forward to high resolution imaging: it allows to precisely segment vessels from tissues and noise, provides morphological structure of the vessels network, with learned pseudo-3D shape, their relative position, dynamic information of blood vascularization in depth in small animals and distinguish the vessels types: artieries and veins. For demonstration we use neural network IterNet that exploits structural redundancy of the blood vessels, which provides a useful analysis tool for raw SWIR images.