This PDF file contains the front matter associated with SPIE Proceedings Volume 11971, including the Title Page, Copyright information, and Table of Contents.

Multiphoton microscopy (MPM) is the go-to technique for high spatial resolution, deep imaging in scattering biological tissue. The performance of MPM, e.g., imaging speed and volume, depends on the characteristics of the excitation source and is fundamentally limited by the signal photon budget. We developed an adaptive excitation source (AES). By feeding the structural information of the sample to the source, the AES produces on-demand pulses only in the regions of interest (ROIs). The AES therefore transforms a conventional multiphoton microscope into a “random-access” microscope that only excites the ROIs and performs the most photon efficient excitation for functional imaging.

Fast brain-wide imaging with single-cell resolution, high signal-to-noise ratio, and no optical aberrations have the potential to inspire new avenues of investigations in biology. However, such imaging is challenging because of the inevitable tradeoffs among these parameters. Here, we overcome these tradeoffs by combining a resonant scanning system, a large objective with low magnification and high numerical aperture, and highly sensitive large-aperture photodetectors. The result is a practically aberration-free, fast-scanning high optical invariant two-photon microscopy (FASHIO-2PM) that enables calcium imaging from a large network composed of <16,000 neurons at 7.5 Hz froma9mm2 contiguous image plane, including more than 10 sensory-motor and higher-order areas of the cerebral cortex in awake mice. Network analysis based on single-cell activities revealed that the brain exhibits small-world rather than scale-free behavior. FASHIO-2PM will enable revealing biological dynamics by simultaneous monitoring of macroscopic activity and its composing elements. This paper is modified from Ota et al., Neuron 2021.

Spontaneous Raman spectroscopy is a powerful label-free and non-invasive imaging technique for mapping cells and tissues, delivering relevant biochemical information. However, in its standard implementation, the spontaneous Raman cross sections are too low, thus preventing high-speed microscopy, especially in the more informative low wavenumber region, also known as fingerprint region. Coherent Raman microscopy overcomes this hurdle providing several orders of magnitude higher speed thanks to the coherent excitation of the vibrational modes in the laser focus. In this work, we present a novel approach to broadband coherent anti-Stokes Raman scattering (B-CARS) that allows acquiring the entire fingerprint vibrational response of the sample at an unprecedented speed. The system is based on an amplified Ytterbium laser at 2 MHz repetition rate, that provides sufficient pulse energies to generate broadband near-infrared white-light supercontinuum in bulk media that we employ as broadband Stokes pulses. Coupled to narrowband pump pulses at 1035 nm, we demonstrate B-CARS microscopy down to 1 ms pixel dwell time with a diffraction-limited spatial resolution over large field of views. To extract the maximum amount of information, we enhance the signal to noise ratio of the vibrational spectra via artificial intelligence-based methods. In particular, we developed a convolutional neural network trained on data-augmented experimental input-output pairs of B-CARS spectra. Traditional algorithms are then used to remove the non-resonant background, extrapolating the pure vibrational response, and to perform chemometric analysis on the hypercubes. We test the setup performances by imaging heterogeneous biological systems, such as tissue slices of murine spine.

The continuous rotating planar mirror-based focus-shift system can be used for various microscopes including light-sheet, confocal, fluorescent, and others. The uniform rotation provides a fast, quiet and stable operation without the normal vibration and noise produced by reciprocal motion system. The planar mirrors produce images without the aberrations produced by the Electronically Tunable Lens (ETL) systems. In a light sheet illumination system for 3D applications, high speed change in the focal plane is required such that the imaging portion of the system is synchronized to the scanning light sheet with a stationary sample. A series of 2D images at different times are captured by the digital camera, which are converted into z-axis information and displayed as 3D images.

Optical spectrometers are widely used scientific equipment with many applications involving material characterization, chemical analysis, disease diagnostics, surveillance, etc. Emerging applications in biomedical and communication fields have boosted the research in the miniaturization of spectrometers. Recently, reconstruction-based spectrometers have gained popularity for their compact size, easy maneuverability, and versatile utilities. These devices exploit the superior computational capabilities of recent computers to reconstruct hyperspectral images using detectors with distinct responsivity to different wavelengths. In this paper, we propose a CMOS compatible reconstruction-based on-chip spectrometer pixels capable of spectrally resolving the visible spectrum with 1 nm spectral resolution maintaining high accuracy (<95 %) and low footprint (8 μm × 8 μm), all without the use of any additional filters. A single spectrometer pixel is formed by an array of silicon photodiodes, each having a distinct absorption spectrum due to their integrated nanostructures, this allows us to computationally reconstruct the hyperspectral image. To achieve distinct responsivity, we utilize random photon-trapping nanostructures per photodiode with different dimensions and shapes that modify the coupling of light at different wavelengths. This also reduces the spectrometer pixel footprint (comparable to conventional camera pixels), thus improving spatial resolution. Moreover, deep trench isolation (DTI) reduces the crosstalk between adjacent photodiodes. This miniaturized spectrometer can be utilized for real-time in-situ biomedical applications such as Fluorescence Lifetime Imaging Microscopy (FLIM), pulse oximetry, disease diagnostics, and surgical guidance.

Optical coherence tomography angiography (OCTA) has become an essential tool in clinics for structural and functional microvasculature imaging. However, a primary setback for OCTA is its imaging speed. The current protocols require high sampling density from raster scanning and multiple cross-sectional B-scan acquisitions to form a single image frame, limiting the acquisition speed. Although advanced ultrafast imaging systems have been proposed, extensive hardware adjustments are cost-prohibitive and pose limitations for practical implementations. Herein, we present an integrated deep learning (DL) method to simultaneously tackle the sampling density and the B-scan repetition process, thus improving the imaging speed while preserving quality. We designed an end-to-end deep neural network (DNN) framework with a two-staged adversarial training scheme to reconstruct fully sampled, high quality (8 repeated B-scans) angiograms from their corresponding undersampled, low quality (2 repeated B-scans) counterparts by successively enhancing the pixel resolution and the image quality. We evaluate our proposed framework using an in-vivo mouse brain vasculature dataset and demonstrate that our method can enhance the OCTA acquisition speed while achieving superior reconstruction performance than conventional methods. Our DL-based framework can accelerate the OCTA imaging speed from 16 to 256× while preserving the image quality and thus provides a convenient software-only solution to aid preclinical and clinical studies.

Time-domain tomographic image reconstruction is typically based on an iterative process that requires repeated solving of the forward model of time-dependent light propagation in tissue. As a result, image reconstruction times remain relatively high. This has been one of the main obstacles in the practical use of time-domain data, for example, for realtime monitoring of brain function, in which case results have to be displayed in less than a second. To overcome this problem, we have developed a neural-network-based approach that promises to deliver image reconstructions in the subseconds range. The inputs to this network are parameterized data derived from the Mellin and Laplace transforms of the time of flight (ToF) distribution. In this study, we specifically focused on three data types: the integrated intensity (E), the mean time of flight (<t<), and the exponential feature (L). The network tested consisted of an input layer, three hidden layers, and an output layer that represents the spatial distribution of absorption values for the medium. We trained the parameters of the network with simulated brain diffuse optical tomography data. The inverse problem is then solved with a single-feed forward pass through the network. We demonstrate that this network, once trained, can recover single and multiple inclusions in a 3D medium with accurate localization within milliseconds and outperforms constrained iterative reconstruction methods.

We present here a new fluorescence molecular tomographic model that can provide ultrahigh spatial and temporal resolution reconstruction through sparsity constrained dimensional reduction. The new method implements a novel sparsity function to asymptotically enforce the sparsest representation of fluorescent targets while reducing the problem dimension based correlation between sensing matrix and measurement. Parameterized temporal data (TD) 𝐿(𝑠), available from the Laplace transform, is used here as input to the inverse model for their computational efficiency and accuracy and robustness to noise. We use radiative transfer equation (RTE) as a light propagation model as it provides more accurate predictions of light propagation in small-volume tissue. The performance of this new method is evaluated through numerical phantoms, focusing on spatial resolution and computational speed. The results show that the sparsity constrained dimensional reduction inverse model can achieve near cellular resolution (~1mm spatial resolution) at depth of 70 mean free paths (MFPs) within ~25 milliseconds.

Cells react highly sensitive to mechanical and structural influences of their environment by mechanosensing. Micropatterning allows the precise regulation of forces by changing the shape of the cell. In order to record the adaptation to the pattern shape and the changes in viscoelastic properties in detail and over a longer period of time, the VELOMIR sensor is ideally suited. The live cell measurements were recorded at 3000 frames per second over 10 minutes. Here we show the difference in passive microrheology measurements of mouse embryonic fibroblasts with incorporated 1 μm polystyrene beads cultured on three differently patterned shapes in a range of sizes.