This PDF file contains the front matter associated with SPIE Proceedings Volume 12392, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

Mid-infrared (mid-IR) spectroscopy and optoacoustic/optothermal (OA/OT) imaging are perfectly complementary technologies to each other. Vibrational molecular excitations by mid-IR absorption are utterly de-excited in the form of heat while efficient OA/OT signal generation primarily depends on heat deposition. This synergy allows overcoming the (otherwise) persistent limitations of traditional mid-IR spectroscopy and imaging in live-cell/fresh-tissue applications— i.e., sample opacity due to water absorption. Combination of mid-IR excitation and OA/OT detection has resulted in new tools for label-free live-cell, tissues, and in vivo metabolic research. Here we discuss basic principles on mid-IR detection for spectroscopy and imaging as well as the most recent developments on mid-IR OA and OT microscopy that overcome the limitations of conventional vibrational spectroscopy for biosensing and label-free metabolic microscopy.

Infrared photothermal heterodyne imaging (IR-PHI) is an ultrasensitive technique capable of achieving super-resolution chemical and morphological characterization of specimens via absorption of mid-infrared light. While early iterations of IR-PHI have involved point-by-point raster-scanning, here, we introduce a widefield modality to IR-PHI that utilizes ns-timescale infrared pump pulses synchronized to an ultrafast complementary metal-oxide-semiconductor camera to parallelize data acquisition. A 300-fold decrease in image acquisition time is realized, falling from 20 minutes to four seconds.

Radiation therapy (RT) is a standard treatment for ~50% of cancers worldwide. The major goal in RT is to maximize tumor cell kill while minimizing radiation-induced damage to healthy tissue. However, several fundamental questions remain unanswered related to the radiation-matter interactions, and the ensuing radiobiological responses in cells and tissue. Stimulated Raman scattering (SRS) microscopy is a promising technology for imaging the heterogeneous metabolic responses in cells exposed to ionizing radiation. In this work, we demonstrate SRS imaging of endogenous macromolecules such as lipids and proteins in the C-H stretching region (2800 to 3100 cm^-1) in MCF-7 breast cancer cells in vitro. We compare the response of control (unirradiated) cells and cells irradiated at 30 Gy (120 kVp x-rays), at three time points of 24, 48, and 72 hours post exposure to ionizing radiation, and find significant changes between irradiated and control MCF-7 cells over time.

The clinical potential for Raman microscopic systems is well established for early diagnosis via cytology. Although Raman systems offer a complementary diagnostic tool providing molecular information, it is not yet utilised substantially in clinics. A few challenges for the clinical implementation of Raman spectroscopy are system and user variability. In this study, we asked how much variability occurs due to different Raman systems or users. To address these questions, we measured the same set of cells using two different Raman microscopes and by two different users. And classification models were generated using multivariate partial least squares discriminant analysis (PLS-DA) and analysed for clinical implementation. Raman spectra were measured from single exfoliated cells (n=400) from ThinPrep samples with negative cytology (n=10) and high-grade cytology (n=10). Raman spectra were acquired from the same set of cells via two identical HORIBA Jobin Yvon XploRA^TM systems (Villeneuve d'Ascq, France), as well as two different users. The Raman data was subjected to PLS-DA and cross-validated via leave-one-patient out. The study's findings suggest that the data acquired from the two Raman systems are 99% identical. However, the observed classification accuracy for the data obtained by user-one was 92%, whereas by user-two was 99%.

Stimulated Raman scattering (SRS) microscopy is a powerful technique that allows for the visualization of molecular vibrational images for label-free imaging, metabolic imaging, and supermultiplex imaging. However, its sensitivity is mainly limited by the shot noise of laser pulses, and long acquisition times are sometimes required to detect weak SRS signals hidden by the shot noise. To overcome this limitation, quantum-enhanced (QE-) SRS microscopy has been demonstrated, while the low optical power of squeezed pulses limits the sensitivity. In this work, we present QE-SRS microscopy using quantum-enhanced balanced detection (QE-BD) scheme, where the sensitivity of balanced detection SRS microscope is enhanced by injecting squeezed vacuum, allowing for QE-SRS imaging with high power SRS pump pulses (typically several tens of milliwatt), while balanced detection causes 3 dB drawback in the signal-to-noise ratio. We experimentally demonstrate QE-SRS imaging with 2.6 dB noise reduction compared with shot-noise-limited balanced detection SRS. We also demonstrate hyperspectral QE-SRS imaging by fast wavelength tuning of Stokes pulses. These results show the potential feasibility of high-power QE-SRS whose sensitivity is beyond that of classical shot-noise-limited SRS microscopes.

Widefield mid-infrared photothermal heterodyne (WIPH) microscopy enables sensitive and fast chemical imaging with high spatial resolution. The technique is realized using an external-cavity quantum cascade laser emitting and a digital frequency-domain lock-in filter for simultaneous multi-harmonic demodulation of WIPH signals recorded by individual camera pixels at a frame rate of 20 kHz. The filter allows the use of continuous-wave probe light and the time-resolved detection of photothermal decay curves. The microscope provides <1 µm spatial resolution in a 64x64 µm field of view. Here, we present preliminary results from hyperspectral WIPH imaging of alkyne-tagged palmitic acid (PA), azidetagged PA and perdeuterated PA via their absorption features in the cell-silent spectral region around 2100 cm^-1. The alkyne and azide functional groups and deuterium are promising vibrational probes for selective imaging of biomolecules, such as lipids and proteins, in cells.

Nonalcoholic fatty liver disease (NAFLD) is characterized by the accumulation of lipid in the hepatocytes (steatosis), which can progress to fibrosing steatohepatitis (NASH). The molecular mechanisms driving this progression are insufficiently understood. A leading hypothesis is lipotoxicity, which postulates that specific lipid species can trigger the cascade of inflammation leading to liver damage. In this study, we used label-free stimulated Raman scattering (SRS) microscopy to characterize the distribution of free cholesterol, cholesteryl esters, and triglycerides of different saturation levels in liver tissue from patients with histologically diagnosed NAFLD versus NASH. By probing the intrinsic vibrational frequencies of lipid molecules in the C–H and the fingerprint regions, we can localize and classify different lipid molecules in lipid droplets via spectral unmixing. We report our developed image acquisition and processing pipeline in this paper and demonstrate example applications such as examining the composition of previously described cholesterol crystals. We discovered that most of the birefringent liquid crystals presumed to be free cholesterol crystals in NASH tissues are predominantly composed of saturated cholesteryl esters. Our method provides a detailed characterization of the lipid composition in NAFLD tissues and allows us to probe further into the potential mechanism of NASH development.

Mid-IR imaging combined with machine learning is a powerful combination for non-destructive, label free chemical imaging. Key applications include computational staining and tissue classification. These applications are enabled by information rich mid-IR hyperspectral images and reliable ground truth data. As novel, nano-scale spatial resolution mid-IR spectroscopy techniques are finding broader use we realize that ground truth datasets will be needed at the nano-scale as well. Here, we propose image fusion and registration of nano-scale images as a generic approach for establishing such datasets. We demonstrate the viability of this approach for imaging the sub-cellular distribution of proteins and specific enzymes. Furthermore, we demonstrate that image registration of AFM-IR spectral data is a key step in processing AFM-IR chemical imaging data in general.

The stratum corneum (SC), the outermost layer of the skin, plays a vital role in providing a barrier against dry environments. In order to understand the barrier function and overall condition of the skin, it is essential to assess the ability of the SC to absorb and retain water. In this study, we used stimulated Raman scattering (SRS) imaging to investigate the threedimensional SC structure and water distribution as water was absorbed into dried SC sheets. Our findings indicate that the process of water absorption and retention is dependent on the specific sample and can be spatially heterogeneous. These results demonstrate the potential of SRS imaging in diagnosing skin conditions.

Live-cell mid-infrared (MIR) imaging has always been challenging because of the absorptive nature of water. However, there is a strong drive to image this spectroscopic window–to see the protein and lipid vibrations directly without the help of dyes. Though the dyes are convenient for imaging, they interfere with the biological functions of live cells. In the past two decades, people have relied on attenuated total reflectance (ATR) Fourier transform infrared (FTIR) spectroscopic imaging to probe such systems to reduce the infrared penetration depth to a few microns. In our previous works, we found a way to further restrict the penetration to a hundred nanometers with plasmonic nanoantennas, also known as the metasurfaces. We named the technique-metasurface-enhanced infrared reflection spectroscopy (MEIRS), and used it for either label-free spectroscopy or imaging. We had demonstrated MEIRS in various live-cell drug dynamics studies, including trypsin, cholesterol depleting agents, and chemotherapeutics, of live cells enclosed in microfluidics chambers. With the recent advancement of commercial mid-infrared quantum cascade laser (QCL), we now have a unique opportunity to acquire high-quality single-cell resolution metasurface-enhanced infrared reflection chemical imaging (MIRCI), which reveals the important protein information in real time. We built an inverted QCL microscope setup and cultured the cells on a cell-culture multiwell plate. The bottom of the multiwells is made of infrared-transparent window and with metasurface fabricated on. In this work, we demonstrated two proofs of concept of MIRCI on both fixed cells in water (single-cell resolution and spectroscopy) and live cells (capturing cell adhesion process). The application provides a novel tool to the drug discovery and fundamental cell biology research.

We describe a simple convolutional network for blind unmixing of transient absorption microscopy data along with a model ensembling strategy. Our network is based on an autoencoder previously developed for blind unmixing of hyperspectral satellite images. Its advantages are (a) that it learns to unmix spectra by unsupervised learning, i.e. by learning to reconstruct imaging data, without knowledge of the underlying spectra or their abundances and (b) that the endmember spectra are directly encoded by the output layer’s coefficients. Extensive modifications to both the network architecture and training loss functions were necessary to produce reasonable performance on transient absorption data. We demonstrate results from blind unmixing of transient absorption images of unstained muscle fibers, acquired at 520 nm pump and 620 nm probe, training an ensemble of 500 different networks (i.e. unmixing models), each starting from a different random initialization. Variability among resulting models was analyzed by principal component analysis on the recovered endmembers from all models, deriving from the projections a model probability density function. We found consistent models (predicting similar endmembers and abundance maps) near the most likely model and surrounding high-probability region, with more variability in low-probability regions. Then, a permutation-aligned average of the ensemble produced much better results than an unweighted ensemble average, or simple selection of one model based on maximum likelihood or best fit. We anticipate this approach of parametrizing models and ensembling based on relative probability to have applications in other chemical imaging modalities such as FLIM, Raman microscopy and mass spectroscopic imaging.

Myopia, a global public health problem, is recognized by the World Health Organization as the leading cause of visual impairment in uncorrected people. At present, a large number of reports focus on the pathological manifestations of retinal level myopia. However, corneal histological changes that may be associated with myopia have not been thoroughly investigated. Raman spectroscopy is a rapid and non-destructive analytical technique with the advantages including label-free, non-invasive and highly specific, providing detailed information at the molecular level. Important components of all the human tissue (proteins, nucleic acids, lipids, etc.) have corresponding Raman spectral characteristic peaks, which contribute to the study of myopia at the molecular level. In this study, a microscopic confocal Raman system (MCRS) was built to collect Raman spectrum of corneal stromal samples, which was obtained through femtosecond laser small incision corneal stroma lens extraction (SMLIE). One hundred fifty-nine corneal stromal Raman spectrum data were collected (54 low myopia, 69 moderate myopia and 36 high myopia). Ten characteristic peaks and corresponding components were further identified. K-nearest neighbor (KNN) was used with principal component analysis (PCA) to classify the samples and the classifications were validated by k-fold cross-validation. Three types of samples with different degrees of myopia were differentiated under the PCA-KNN model with an accuracy of 93.1%. Two characteristic peaks (1099 cm^-1 and 2940 cm^-1) show greatly contribution to the classification results. The results provide that Raman spectroscopy combined with PCA-KNN analysis can effectively distinguish the degree of myopia and is expected to explore the potential causes of myopia.

Flavonoids are natural compounds with diverse structures. This type of nature product is considered to possess a wide range of health beneficial effects. Different skeleton structures and substituent groups lead to different Raman spectral features. In this work, we developed three Raman spectrum analysis methods based on artificial intelligence to classify 18 flavonoids. Firstly, applying principal component analysis (PCA) as dimension reduction method, we compress the 1300cm^-1 -1600cm^-1 spectral band into several important variables. The results obtained by the preprocessing methods were combined with K-Nearest Neighbor algorithm (KNN), support vector machine (SVM) for classification. Secondly, the combination of relevant features was taken by advanced machine learning method of random forest (RF). In terms of the accuracy of the results, all the methods achieved acceptable classification accuracy, which was almost over 84% on the test set. The experimental results demonstrated that the Raman spectroscopy study based on corresponding unique vibration mode exhibited application prospects in chemical structure classification and pharmacological activity prediction.