Accelerating Cancer Histopathology Workflows with Chemical Imaging and Machine Learning.

Histopathology has remained a cornerstone for biomedical tissue assessment for over a century, with a resource-intensive workflow involving biopsy or excision, gross examination, sampling, tissue processing to snap frozen or formalin-fixed paraffin-embedded blocks, sectioning, staining, optical imaging, and microscopic assessment. Emerging chemical imaging approaches, including stimulated Raman scattering (SRS) microscopy, can directly measure inherent molecular composition in tissue (thereby dispensing with the need for tissue processing, sectioning, and using dyes) and can use artificial intelligence (AI) algorithms to provide high-quality images. Here we show the integration of SRS microscopy in a pathology workflow to rapidly record chemical information from minimally processed fresh-frozen prostate tissue. Instead of using thin sections, we record data from intact thick tissues and use optical sectioning to generate images from multiple planes. We use a deep learning­based processing pipeline to generate virtual hematoxylin and eosin images. Next, we extend the computational method to generate archival-quality images in minutes, which are equivalent to those obtained from hours/days-long formalin-fixed, paraffin-embedded processing. We assessed the quality of images from the perspective of enabling pathologists to make decisions, demonstrating that the virtual stained image quality was diagnostically useful and the interpathologist agreement on prostate cancer grade was not impacted. Finally, because this method does not wash away lipids and small molecules, we assessed the utility of lipid chemical composition in determining grade. Together, the combination of chemical imaging and AI provides novel capabilities for rapid assessments in pathology by reducing the complexity and burden of current workflows. SIGNIFICANCE: Archival-quality (formalin-fixed paraffin-embedded), thin-section diagnostic images are obtained from thick-cut, fresh-frozen prostate tissues without dyes or stains to expedite cancer histopathology by combining SRS microscopy and machine learning.

Phasor Representation Approach for Rapid Exploratory Analysis of Large Infrared Spectroscopic Imaging Data Sets.

Infrared (IR) spectroscopic imaging is potentially useful for digital histopathology as it provides spatially resolved molecular absorption spectra, which can subsequently yield useful information by powerful artificial intelligence methods. A typical analysis pipeline in using IR imaging data for chemical pathology often involves iterative processes of segmentation, evaluation, and analysis that necessitate rapid data exploration. Here, we present a fast, reliable, and intuitive method based on a phasor representation of spectra and discuss its unique applicability for IR imaging data. We simulate different features extant in IR spectra and discuss their influence on the phasor waveforms; similarly, we undertake IR image analysis in the transform space to understand spectral similarity and variance. We demonstrate the potential of phasor analysis for biomedical tissue imaging using a variety of samples, using fresh frozen surgical prostate resections and formalin-fixed paraffin-embedded breast cancer tissue microarray samples as model systems that span common histopathology practice. To demonstrate further generalizability of this approach, we apply the method to data from different experimental conditions─including standard (5.5 µm × 5.5 µm pixel size) and high-definition (1.1 µm × 1.1 µm pixel size) Fourier transform IR (FTIR) spectroscopic imaging using transmission and transflection modes. Quantitative segmentation results from our approach are compared to previous studies, showing good agreement and quick visualization. The presented method is rapid, easy to use, and highly capable of deciphering compositional differences, presenting a convenient tool for exploratory analysis of IR imaging data.

Phase-Sensitive Vibrationally Resonant Sum-Frequency Generation Microscopy in Multiplex Configuration at 80 MHz Repetition Rate.

Vibrationally resonant sum-frequency generation (VR SFG) microscopy is an advanced imaging technique that can map out the intensity contrast of infrared and Raman active vibrational modes with micron to submicron lateral resolution. To broaden its applications and to obtain a molecular level of understanding, further technical advancement is needed to enable high-speed measurements of VR SFG microspectra at every pixel. In this study, we demonstrate a new VR SFG hyperspectral imaging platform combined with an ultrafast laser system operated at a repetition rate of 80 MHz. The multiplex configuration with broadband mid-infrared pulses makes it possible to measure a single microspectrum of CH/CH2 stretching modes in biological samples, such as starch granules and type I collagen tissue, with an exposure time of hundreds of milliseconds. Switching from the homodyne- to heterodyne-detected VR SFG hyperspectral imaging can be achieved by inserting a pair of optics into the beam path for local oscillator generation and delay time adjustment, which enables self-phase-stabilized spectral interferometry. We investigate the relationship between phase images of several different C-H modes and the relative orientation of collagen triple-helix in fibril bundles. The results show that the new multiplex VR SFG microscope operated at a high repetition rate is a powerful approach to probe the structural features and spatial arrangements of biological systems in detail.

Detection and Quantification of Myocardial Fibrosis Using Stain-Free Infrared Spectroscopic Imaging.

CONTEXT.­: Myocardial fibrosis underpins a number of cardiovascular conditions and is difficult to identify with standard histologic techniques. Challenges include imaging, defining an objective threshold for classifying fibrosis as mild or severe, and understanding the molecular basis for these changes. OBJECTIVE.­: To develop a novel, rapid, label-free approach to accurately measure and quantify the extent of fibrosis in cardiac tissue using infrared spectroscopic imaging. DESIGN.­: We performed infrared spectroscopic imaging and combined that with advanced machine learning-based algorithms to assess fibrosis in 15 samples from patients belonging to the following 3 classes: (1) patients with nonpathologic (control) donor hearts, (2) patients undergoing transplant, and (3) patients undergoing implantation of a ventricular assist device. RESULTS.­: Our results show excellent sensitivity and accuracy for detecting myocardial fibrosis, as demonstrated by a high area under the curve of 0.998 in the receiver operating characteristic curve measured from infrared imaging. Fibrosis of various morphologic subtypes were demonstrated with virtually generated picrosirius red images, which showed good visual and quantitative agreement (correlation coefficient = 0.92, ρ = 7.76 × 10-15) with stained images of the same sections. Underlying molecular composition of the different subtypes was investigated with infrared spectra showing reproducible differences presumably arising from differences in collagen subtypes and/or crosslinking. CONCLUSIONS.­: Infrared imaging can be a powerful tool in studying myocardial fibrosis and gleaning insights into the underlying chemical changes that accompany it. Emerging methods suggest that the proposed approach is compatible with conventional optical microscopy, and its consistency makes it translatable to the clinical setting for real-time diagnoses as well as for objective and quantitative research.

Fully absorptive 3D IR spectroscopy using a dual mid-infrared pulse shaper.

This paper presents the implementation of 3D IR spectroscopy by adding a second pump beam to a two-beam 2D IR spectrometer. An independent mid-IR pulse shaper is used for each pump beam, which can be programmed to collect its corresponding dimension in either the frequency or time-domains. Due to the phase matching geometry employed here, absorptive 3D IR spectra are automatically obtained, since all four of the rephasing and non-rephasing signals necessary to generate absorptive spectra are collected simultaneously. Phase cycling is used to isolate the fifth-order from the third-order signals. The method is demonstrated on tungsten hexacarbonyl (W(CO)6) and dicarbonylacetylacetonato rhodium (I), for which the eigenstates are extracted up to the third excited state. Pulse shaping affords a high degree of control over 3D IR experiments by making possible mixed time- and frequency-domain experiments, fast data acquisition and straightforward implementation.

A strongly absorbing class of non-natural labels for probing protein electrostatics and solvation with FTIR and 2D IR spectroscopies.

A series of non-natural infrared probes is reported that consist of a metal-tricarbonyl modified with a -(CH2)n- linker and cysteine-specific leaving group. They can be site-specifically attached to proteins using mutagenesis and similar protocols for EPR spin labels, which have the same leaving group. We characterize the label's frequencies and lifetimes using 2D IR spectroscopy in solvents of varying dielectric. The frequency range spans 10 cm(-1), and the variation in lifetimes ranges from 6 to 19 ps, indicating that these probes are very sensitive to their environments. Also, we attached probes with -(CH2)-, -(CH2)3-, and -(CH2)4- linkers to ubiquitin at positions 6 and 63 and collected spectra in aqueous buffer. The frequencies and lifetimes were correlated for 3C and 4C linkers, as they were in the solvents, but did not correlate for the 1C linker. We conclude that lifetime measures solvation, whereas frequency reflects the electrostatics of the environment, which in the case of the 1C linker is a measure of the protein electrostatic field. We also labeled V71C α-synuclein in buffer and membrane-bound. Unlike most other infrared labels, this label has extremely strong cross sections and thus can be measured with 2D IR spectroscopy at sub-millimolar concentrations. We expect that these labels will find use in studying the structure and dynamics of membrane-bound, aggregated, and kinetically evolving proteins for which high signal-to-noise at low protein concentrations is imperative.

Simplified and economical 2D IR spectrometer design using a dual acousto-optic modulator.

Over the last decade two-dimensional infrared (2D IR) spectroscopy has proven to be a very useful extension of infrared spectroscopy, yet the technique remains restricted to a small group of specialized researchers because of its experimental complexity and high equipment cost. We report on a spectrometer that is compact, mechanically robust, and is much less expensive than previous designs because it uses a single pixel MCT detector rather than an array detector. Moreover, each axis of the spectrum can be collected in either the time or frequency domain via computer programming. We discuss pulse sequences for scanning the probe axis, which were not previously possible. We present spectra on metal carbonyl compounds at 5 µm and a model peptide at 6 µm. Data collection with a single pixel MCT takes longer than using an array detector, but publishable quality data are still achieved with only a few minutes of averaging.

Residue-specific structural kinetics of proteins through the union of isotope labeling, mid-IR pulse shaping, and coherent 2D IR spectroscopy.

We describe a methodology for studying protein kinetics using a rapid-scan technology for collecting 2D IR spectra. In conjunction with isotope labeling, 2D IR spectroscopy is able to probe the secondary structure and environment of individual residues in polypeptides and proteins. It is particularly useful for membrane and aggregate proteins. Our rapid-scan technology relies on a mid-IR pulse shaper that computer generates the pulse shapes, much like in an NMR spectrometer. With this device, data collection is faster, easier, and more accurate. We describe our 2D IR spectrometer, as well as protocols for (13)C(18)O isotope labeling, and then illustrate the technique with an application to the aggregation of the human islet amyloid polypeptide implicated in type 2 diabetes.