Differentiating atherosclerotic plaque burden in arterial tissues using femtosecond CARS-based multimodal nonlinear optical imaging

Development of an image classification pipeline for atherosclerotic plaques assessment using supervised machine learning

Background

During atherosclerosis, the narrowing of the arterial lumen is observed through the accumulation of bio compounds and the formation of plaque within artery walls. A non-linear optical imaging modality (NLOM), coherent anti-stokes Raman scattering (CARS) microscopy, can be used to image lipid-rich structures commonly found in atherosclerotic plaques. By matching the lipid’s molecular vibrational frequencies (CH bonds), it is possible to map the accumulation of lipid-rich structures without the need for exogenous labelling and/or processing of the samples. CARS allows for the visualization of the morphological features of plaque. In combination with supervised machine learning, CARS imaged morphological features can be used to characterize the progression of atherosclerotic plaques. 

Results

Based on a set of label-free CARS images of atherosclerotic plaques (i.e. foam cell clusters) from a Watanabe heritable hyperlipidemic rabbit model, we developed an automated pipeline to classify atherosclerotic lesions based on their major morphological features. Our method uses image preprocessing to first improve the quality of the CARS-imaged plaque, followed by the segmentation of the plaque using Otsu thresholding, marker-controlled watershed, K-means segmentation and a novel independent foam cell thresholding segmentation. To define relevant morphological features, 27 quantitative features were extracted and further refined by a novel coefficient of variation feature refinement method in accordance with filter-type feature selection. Refined morphological features were supplied into three supervised machine learning algorithms; K-nearest neighbour, support vector machine and decision tree classifier. The classification pipeline showcased the ability to exploit relevant plaque morphological features to accurately classify 3 pre-defined stages of atherosclerosis: early fatty streak development (EFS) and advancing atheroma (AA) with a greater than 85% class accuracy

Conclusions

Through the combination of CARS microscopy and computational methods, a powerful classification tool was developed to identify the progression of atherosclerotic plaque in an automated manner. Using a curated dataset, the classification pipeline demonstrated the ability to differentiate between EFS, EF and AA. Thus, presenting the opportunity to classify the onset of atherosclerosis at an earlier stage of development

Label-free multiphoton microscopy as a tool to investigate alterations of cerebral aneurysms

Cerebral aneurysms are abnormal focal dilatations of arterial vessel walls with pathological vessel structure alterations. Sudden rupture can lead to a subarachnoid hemorrhage, which is associated with a high mortality. Therefore, the origin of cerebral aneurysms as well as the progression to the point of rupture needs to be further investigated. Label-free multimodal multiphoton microscopy (MPM) was performed on resected human aneurysm domes and integrated three modalities: coherent anti-Stokes Raman scattering, endogenous two-photon fluorescence and second harmonic generation. We showed that MPM is a completely label-free and real-time powerful tool to detect pathognomonic histopathological changes in aneurysms, e.g. thickening and thinning of vessel walls, intimal hyperplasia, intra-wall haemorrhage, calcification as well as atherosclerotic changes. In particular, the loss or fragmentation of elastin as well as fibromatous wall remodelling appeared very distinct. Remarkably, cholesterol and lipid deposits were clearly visible in the multiphoton images. MPM provides morphological and biochemical information that are crucial for understanding the mechanisms of aneurysm formation and progression.

Fractal dimension and directional analysis of elastic and collagen fiber arrangement in unsectioned arterial tissues affected by atherosclerosis and aging

Structural proteins like collagen and elastin are major constituents of the extracellular matrix (ECM). ECM degradation and remodeling in diseases significantly impact the microorganization of these structural proteins. Therefore, tracking the changes of collagen and elastin fiber morphological features within ECM impacted by disease progression could provide valuable insight into pathological processes such as tissue fibrosis and atherosclerosis. Benefiting from its intrinsic high-resolution imaging power and superior biochemical specificity, nonlinear optical microscopy (NLOM) is capable of providing information critical to the understanding of ECM remodeling. In this study, alterations of structural fibrillar proteins such as collagen and elastin in arteries excised from atherosclerotic rabbits were assessed by the combination of NLOM images and textural analysis methods such as fractal dimension (FD) and directional analysis (DA). FD and DA were tested for their performance in tracking the changes of extracellular elastin and fibrillar collagen remodeling resulting from atherosclerosis progression/aging. Although other methods of image analysis to study the organization of elastin and collagen structures have been reported, the simplified calculations of FD and DA presented in this work prove that they are viable strategies for extracting and analyzing fiber-related morphology from disease-impacted tissues. Furthermore, this study also demonstrates the potential utility of FD and DA in studying ECM remodeling caused by other pathological processes such as respiratory diseases, several skin conditions, or even cancer. NEW & NOTEWORTHY Textural analyses such as fractal dimension (FD) and directional analysis (DA) are straightforward and computationally viable strategies to extract fiber-related morphological data from optical images. Therefore, objective, quantitative, and automated characterization of protein fiber morphology in extracellular matrix can be realized by using these methods in combination with digital imaging techniques such as nonlinear optical microscopy (NLOM), a highly effective visualization tool for fibrillar collagen and elastic network. Combining FD and DA with NLOM is an innovative approach to track alterations of structural fibrillar proteins. The results illustrated in this study not only prove the effectiveness of FD and DA methods in extracellular protein characterization but also demonstrate their potential value in clinical and basic biomedical research where protein microstructure characterization is critical.

Imaging Collagen in Scar Tissue: Developments in Second Harmonic Generation Microscopy for Biomedical Applications

The ability to respond to injury with tissue repair is a fundamental property of all multicellular organisms. The extracellular matrix (ECM), composed of fibrillar collagens as well as a number of other components is dis-regulated during repair in many organs. In many tissues, scaring results when the balance is lost between ECM synthesis and degradation. Investigating what disrupts this balance and what effect this can have on tissue function remains an active area of research. Recent advances in the imaging of fibrillar collagen using second harmonic generation (SHG) imaging have proven useful in enhancing our understanding of the supramolecular changes that occur during scar formation and disease progression. Here, we review the physical properties of SHG, and the current nonlinear optical microscopy imaging (NLOM) systems that are used for SHG imaging. We provide an extensive review of studies that have used SHG in skin, lung, cardiovascular, tendon and ligaments, and eye tissue to understand alterations in fibrillar collagens in scar tissue. Lastly, we review the current methods of image analysis that are used to extract important information about the role of fibrillar collagens in scar formation.

Multimodal optoacoustic and multiphoton microscopy of human carotid atheroma

Carotid artery atherosclerosis is a main cause of stroke. Understanding atherosclerosis biology is critical in the development of targeted prevention and treatment strategies. Consequently, there is demand for advanced tools investigating atheroma pathology. We consider hybrid optoacoustic and multiphoton microscopy for the integrated and complementary interrogation of plaque tissue constituents and their mutual interactions. Herein, we visualize human carotid plaque using a hybrid multimodal imaging system that combines optical resolution optoacoustic (photoacoustic) microscopy, second and third harmonic generation microscopy, and two-photon excitation fluorescence microscopy. Our data suggest more comprehensive insights in the pathophysiology of atheroma formation and destabilization, by enabling congruent visualization of structural and biological features critical for the atherosclerotic process and its acute complications, such as red blood cells and collagen.

(Second) Harmonic Disharmony: Nonlinear Microscopy Shines New Light on the Pathology of Atherosclerosis

There has been increasing interest in second harmonic generation (SHG) imaging approaches for the investigation of atherosclerosis due to the deep penetration and three-dimensional sectioning capabilities of the nonlinear optical microscope. Atherosclerosis involves remodeling or alteration of the collagenous framework in affected vessels. The disease is often characterized by excessive collagen deposition and altered collagen organization. SHG has the capability to accurately characterize collagen structure, which is an essential component in understanding atherosclerotic lesion development and progression. As a structure-based imaging modality, SHG is most impactful in atherosclerosis evaluation in conjunction with other, chemically specific nonlinear optics (NLO) techniques to identify additional components of the lesion. These include the use of coherent anti-Stokes Raman scattering and two-photon excitation fluorescence for studying atherosclerosis burden, and application of stimulated Raman scattering to image cholesterol crystals. However, very few NLO studies have attempted to quantitate differences in control versus atherosclerotic states or to correlate the application to clinical situations. This review highlights the potential of SHG imaging to directly and indirectly describe atherosclerosis as a pathological condition.

Third-harmonic generation susceptibility spectroscopy in free fatty acids

Lipid-correlated disease such as atherosclerosis has been an important medical research topic for decades. Many new microscopic imaging techniques such as coherent anti-Stokes Raman scattering and third-harmonic generation (THG) microscopy were verified to have the capability to target lipids in vivo. In the case of THG microscopy, biological cell membranes and lipid bodies in cells and tissues have been shown as good sources of contrast with a laser excitation wavelength around 1200 nm. We report the THG excitation spectroscopy study of two pure free fatty acids including oleic acid and linoleic acid from 1090 to 1330 nm. Different pure fatty acids presented slightly-different THG χ(3) spectra. The measured peak values of THG third-order susceptibility χ(3) in both fatty acids were surprisingly found not to match completely with the resonant absorption wavelengths around 1190 to 1210 nm, suggesting possible wavelengths selection for enhanced THG imaging of lipids while avoiding laser light absorption. Along with the recent advancement in THG imaging, this new window between 1240 to 1290 nm may offer tremendous new opportunities for sensitive label-free lipid imaging in biological tissues.

Applications of coherent Raman scattering microscopies to clinical and biological studies

Coherent anti-Stokes Raman scattering (CARS) microscopy and stimulated Raman scattering (SRS) microscopy are two nonlinear optical imaging modalities that are at the frontier of label-free and chemical specific biological and clinical diagnostics. The applications of coherent Raman scattering (CRS) microscopies are multifold, ranging from investigation of basic aspects of cell biology to the label-free detection of pathologies. This review summarizes recent progress of biological and clinical applications of CRS between 2008 and 2014, covering applications such as lipid droplet research, single cell analysis, tissue imaging and multiphoton histopathology of atherosclerosis, myelin sheaths, skin, hair, pharmaceutics, and cancer and surgical margin detection.

Visualization of mouse neuronal ganglia infected by Herpes Simplex Virus 1 (HSV-1) using multimodal non-linear optical microscopy

Herpes simplex virus 1 (HSV-1) is a neurotropic virus that causes skin lesions and goes on to enter a latent state in neurons of the trigeminal ganglia. Following stress, the virus may reactivate from latency leading to recurrent lesions. The in situ study of neuronal infections by HSV-1 is critical to understanding the mechanisms involved in the biology of this virus and how it causes disease; however, this normally requires fixation and sectioning of the target tissues followed by treatment with contrast agents to visualize key structures, which can lead to artifacts. To further our ability to study HSV-1 neuropathogenesis, we have generated a recombinant virus expressing a second generation red fluorescent protein (mCherry), which behaves like the parental virus in vivo. By optimizing the application of a multimodal non-linear optical microscopy platform, we have successfully visualized in unsectioned trigeminal ganglia of mice both infected cells by two-photon fluorescence microscopy, and myelinated axons of uninfected surrounding cells by coherent anti-Stokes Raman scattering (CARS) microscopy. These results represent the first report of CARS microscopy being combined with 2-photon fluorescence microscopy to visualize virus-infected cells deep within unsectioned explanted tissue, and demonstrate the application of multimodal non-linear optical microscopy for high spatial resolution biological imaging of tissues without the use of stains or fixatives.

Biological imaging with coherent Raman scattering microscopy: a tutorial

Coherent Raman scattering (CRS) microscopy is gaining acceptance as a valuable addition to the imaging toolset of biological researchers. Optimal use of this label-free imaging technique benefits from a basic understanding of the physical principles and technical merits of the CRS microscope. This tutorial offers qualitative explanations of the principles behind CRS microscopy and provides information about the applicability of this nonlinear optical imaging approach for biological research.

Collagen morphology and texture analysis: from statistics to classification

In this study we present an image analysis methodology capable of quantifying morphological changes in tissue collagen fibril organization caused by pathological conditions. Texture analysis based on first-order statistics (FOS) and second-order statistics such as gray level co-occurrence matrix (GLCM) was explored to extract second-harmonic generation (SHG) image features that are associated with the structural and biochemical changes of tissue collagen networks. Based on these extracted quantitative parameters, multi-group classification of SHG images was performed. With combined FOS and GLCM texture values, we achieved reliable classification of SHG collagen images acquired from atherosclerosis arteries with >90% accuracy, sensitivity and specificity. The proposed methodology can be applied to a wide range of conditions involving collagen re-modeling, such as in skin disorders, different types of fibrosis and muscular-skeletal diseases affecting ligaments and cartilage.

Quantitative nonlinear optical assessment of atherosclerosis progression in rabbits

Quantification of atherosclerosis has been a challenging task owing to its complex pathology. In this study, we validated a quantitative approach for assessing atherosclerosis progression in a rabbit model using a numerical matrix, optical index for plaque burden, derived directly from the nonlinear optical microscopic images captured on the atherosclerosis-affected blood vessel. A positive correlation between this optical index and the severity of atherosclerotic lesions, represented by the age of the rabbits, was established based on data collected from 21 myocardial infarction-prone Watanabe heritable hyperlipidemic rabbits with age ranging between new-born and 27 months old. The same optical index also accurately identified high-risk locations for atherosclerotic plaque formation along the entire aorta, which was validated by immunohistochemical fluorescence imaging.

Nonlinear optical microscopy in decoding arterial diseases

Pathological understanding of arterial diseases is mainly attributable to histological observations based on conventional tissue staining protocols. The emerging development of nonlinear optical microscopy (NLOM), particularly in second-harmonic generation, two-photon excited fluorescence and coherent Raman scattering, provides a new venue to visualize pathological changes in the extracellular matrix caused by atherosclerosis progression. These techniques in general require minimal tissue preparation and offer rapid three-dimensional imaging. The capability of label-free microscopic imaging enables disease impact to be studied directly on the bulk artery tissue, thus minimally perturbing the sample. In this review, we look at recent progress in applications related to arterial disease imaging using various forms of NLOM.

Identification of cholesterol crystals in plaques of atherosclerotic mice using hyperspectral CARS imaging

The accumulation of lipids, including cholesterol, in the arterial wall plays a key role in the pathogenesis of atherosclerosis. Although several advances have been made in the detection and imaging of these lipid structures in plaque lesions, their morphology and composition have yet to be fully elucidated, particularly in different animal models of disease. To address this issue, we analyzed lipid morphology and composition in the atherosclerotic plaques of two animal models of disease, the low density lipoprotein receptor-deficient (LDLR(-/-)) mouse and the ApoE lipoprotein-deficient (ApoE(-/-)) mouse, utilizing hyperspectral coherent anti-Stokes Raman scattering (CARS) microscopy in combination with principal component analysis (PCA). Hyperspectral CARS imaging revealed lipid-rich macrophage cells and condensed needle-shaped and plate-shaped lipid crystal structures in both mice. Spectral analysis with PCA and comparison to spectra of pure cholesterol and cholesteryl ester derivatives further revealed these lipid structures to be pure cholesterol crystals, which were predominantly observed in the ApoE(-/-) mouse model. These results illustrate the ability of hyperspectral CARS imaging in combination with multivariate analysis to characterize atherosclerotic lipid morphology and composition with chemical specificity, and consequently, provide new insight into the formation of cholesterol crystal structures in atherosclerotic plaque lesions.

Multimodal CARS microscopy of structured carbohydrate biopolymers

We demonstrate the utility of multimodal coherent anti-Stokes Raman scattering (CARS) microscopy for the study of structured condensed carbohydrate systems. Simultaneous second-harmonic generation (SHG) and spectrally-scanned CARS microscopy was used to elucidate structure, alignment, and density in cellulose cotton fibers and in starch grains undergoing rapid heat-moisture swelling. Our results suggest that CARS response of the O-H stretch region (3000 cm(-1)-3400 cm(-1)), together with the commonly-measured C-H stretch (2750 cm(-1)-2970 cm(-1)) and SHG provide potentially important structural information and contrast in these materials.