Real-time digital hologram rendering in ophthalmology faces challenges with high-bitrate image streams. A software called holovibes designed a digital hologram rendering engine, utilizing spatial and temporal signal demodulation operations. It achieves real-time rendering from a 60,000 frames per second stream of 12-bit, 256x256 pixel interferograms, enabling high-quality Doppler imaging and optical coherence tomography (OCT) of the retina. Additionally, sustained real-time 3D OCT image rendering is achieved through short-time analysis of pre-recorded optical interferogram data, transferred at ~7.7 GB/s. Doppler and OCT renderings are done at ~3.8 Giga voxels/s over 512 consecutive frames.

We present rolling shutter speckle imaging (RSSI), a single-shot temporal speckle imaging technique that can quantitatively measure the fast dynamics of scattering media without high speed cameras. Utilizing a rolling shutter image sensor and vertically elongated speckles, RSSI can quantitatively map the speckle dynamics from a single image capture. We discuss the speckle spatiotemporal intensity correlation model for RSSI, which is validated through simulations and phantom experiments. We show in vivo quantitative blood flow imaging of the mouse brain from a snapshot measurement. In addition to imaging, we also present rolling shutter speckle plethysmography for cardiovascular monitoring.

Artificial Intelligence (AI) is an increasingly important tool in the biological sciences. Here, we demonstrate our ongoing efforts to develop AI methods to directly decode the organizational principles of biological systems from large microscopy datasets. Examples include machine learning approaches that can be utilized to build multiscale maps of biological systems, to derive new quantitative insight from timelapse microscopy data and to build predictive models of cell fate in complex biological tissues. In exploring these ideas, we hope to enable a new platform for automated scientific hypothesis generation and directed experimental studies.

Fluorescence microscopy helps observe specific labeled molecules but has some limitations. First, the need for labeling can affect the physiological function of targets and is incompatible with clinical applications. Phototoxicity is not negligible in high-speed, high-resolution imaging. Illumination with longer wavelengths can mitigate phototoxicity, but it results in lower spatial resolution. Photobleaching also restricts the observation duration and quantification of signals. Phase contrast imaging has a long history of label-free, live cell imaging, including Zernike phase contrast (ZPC) and differential interference contrast (DIC) microscopy. ZPC usually has a limited spatial resolution, and fine structures in the cells are often hidden by the halo artifact. Orientation dependency of DIC causes information loss. Moreover, phase shifts obtained by ZPC and DIC are not quantitative. Here, we report our computational approach to quantitative phase microscopy. Our method allows us to observe fine structures and rapid dynamics for a long time with minimal modifications to the optics of commercial microscopes.

Autofluorescence lifetime imaging microscopy is sensitive to functional or treatment-induced metabolic changes in single cells. However, this label-free technique typically relies on laser-scanning microscopes that are not common for single-cell assessments in immunology. Here, we present a microfluidic lifetime-sensitive flow cytometer that measures time-domain decays of the metabolic co-enzyme NAD(P)H from single cells with picosecond temporal resolution using time-correlated single-photon counting (TCSPC) at a throughput of tens of cells per second. The system uses a picosecond-pulsed ultraviolet diode laser, alkali photomultiplier tubes, an FPGA-based time tagger, and provides real-time phasor-based classification (i.e., gating) of flowing cells for applications in immunology.

Multiplex imaging facilities biological studies in multicellular dynamics in living organisms due to its molecular specificity, 3D subcellular resolution, and deep tissue penetration. However, one major bottleneck is detecting and resolving multiplexed signals of weak fluorescence due to a tradeoff between signal throughput and spectral resolution. Here, we demonstrate high-speed, programmable, and broadband excitation encoding to enhance sensitivity without sacrificing signal throughput in multiplex multiphoton imaging. We utilize a 22-kHz programmable digital micromirror device to modulate the spectrum of a high-power broadband laser, achieving versatile excitation encoding schemes with a 750-nm bandwidth in the NIR regime. The proposed method will benefit applications that demand high-speed and high-content performance, including hyperspectral multiphoton microscopy and computational spectroscopy.

In this work, we customized a multi-photon spinning disc unit with a tunable high-power femtosecond laser to demonstrate the first three-photon spinning disk microscopy, achieving ~100Hz frame rate. We demonstrate its pioneering applications on green fluorescent beads, in vivo Drosophila brains, and fixed mouse brains. Furthermore, our findings from brain tissue imaging reveal that three-photon spinning disk imaging exhibits a significantly lower attenuation rate when compared to its two-photon spinning disk counterpart, paving the way of utilizing three-photon excitation on high-speed deep tissue imaging.

In understanding the relationship between neuronal structure and functionality, we leverage calcium/voltage imaging to record the propagation of neuronal signals in vivo. To aptly capture these functional dynamics, we present 2-photon Line Excitation Array Detection (2p-LEAD) microscopy. Through shaping the 1035nm excitation beam into a line, 2p-LEAD can operate at 3000 FPS, using a linear array of PMT’s to collect the resulting fluorescence. With the replacement of the scanning mirror for an acousto-optic deflector, 2p-LEAD holds the potential of achieving kilohertz volumetric imaging over the visual cortex of the mice brain.

To further extend the capability of super-resolution in XYZλθT 6 dimensions, we developed polarized structured illumination microscopy (Polar-SIM), which achieves super-resolution imaging of dipoles by interpreting the dipoles in spatio-angular (XYZθ) hyperspace. Further, the multi-wavelength excitation enables study of the dynamic interactions of subcellular organelles. We demonstrate the application of Polar-SIM on a series of biological filamentous systems, such as cytoskeleton networks and λ-DNA, and report the dynamics of short actin sliding across a myosin-coated surface. With the modulation of the incident angle, we can generate TIRF- or HiLo- illumiation for Polar-SIM. Also, for thick specimen, the structured illumination can be used with lower frequency to generate grid confocal, enabling the optical sectioning capability.

This presentation will center on the investigation of two distinct relaxation mechanisms: photothermal relaxation and vibrational dephasing. We introduce a fluorescence-independent super-resolution technique, photothermal relaxation localization (PEARL) microscopy. PEARL microscopy leverages the evaluation of absorption-induced local refractive index modulation to extract intricate spatial details. Notably, PEARL is universally applicable to a broad range of absorbers across various spectral windows, encompassing both vibrational and electronic absorptions. Additionally, we employ a femtosecond Time-Resolved Coherent Anti-Stokes Raman Scattering (T-CARS) to scrutinize the temporal characteristics of molecular vibrations. Our findings reveal vibrational dephasing as an emerging indicator capable of discerning subtle alterations in molecular structure and dynamics. This dual approach of extracting temporal features from diverse optical processes opens new vistas for detection and sensing across an array of scientific domains.

Free-space angular-chirp-enhanced delay (FACED) imaging is proven to offer ultrafast imaging speed needed in a wide range of bioimaging applications. However, this superior performance critically requires stringent hardware mechanical stability, which limits it wide applicability for robust long-term operation. We develop FACED 2.0, that integrates diffractive optical elements with the infinity mirrors for preserving the capability of high-density laser-scanning foci generation, and more importantly, and enabling extended stable continuous operation over days or even a week. As a proof-of-concept demonstration, we showcased the applicability of FACED 2.0 in high-throughput imaging flow cytometry at the throughput equivalent to ~10,000 cells/sec.

Real-time imaging of rapid events or fast-evolving processes at the microscale is paramount in applications ranging from neuroscience to industrial optical inspection. However, current microscopy architectures can be limited in speed by the camera frame rate or the need for mechanical scanning. Here, we propose a scanless method that exploits the intrinsic high speed of single-pixel detectors to achieve two-dimensional diffraction-limited imaging at rates as high as 5 kHz. Based on interfering orthogonal light stripes from two acousto-optic deflectors to generate a grid of flickering points - each with its unique modulation frequency -, our approach offers high spatiotemporal resolution compatible with a wide spectral range.

Visualizing the diverse anatomical and functional traits within the densely packed cellular space and over large heterogeneous populations unfolds critical details of the fundamentals of living organisms. In this talk, I will present my laboratory’s recently developed advanced microscopy techniques for high-throughput extraction of molecular information in cells and tissues with ultrahigh-spatiotemporal resolution and accessibility.

The imaging speed of current mid-infrared photothermal (MIP) microscope is limited to tens of seconds per frame due to the long pixel dwell time and slow sample scanning process, which is insufficient for capturing dynamics inside living systems. In this work, we developed a video-rate MIP microscope by employing a lock-in free demodulation scheme to resolve single IR pulse induced contrast. We further developed a synchronous pump-probe Galvo scanning for reaching a line rate over 2kHz. With such scheme, the system is capable of resolving chemical dynamics of various biomolecules in living organisms at multiple scales.

The technique of in vivo flow cytometry was developed for the detection and quantification of fluorescent cells in blood circulation without the need for drawing blood samples. In its simplest form, a stationary excitation slit was focused onto a blood vessel and the fluorescence of individual cells was detected through a confocal slit. It has become a very useful tool for noninvasive analysis of circulating leukocytes or tumor cells in animal models where the cell populations of interest can be fluorescently tagged. Translation to human use, however, requires a new label-free approach, which will be discussed in this presentation.

Thrombosis remains a global health concern, necessitating research into its underlying mechanisms. Utilizing a high-speed bright-field microscope based on optical frequency-division multiplexing and microfluidics, we performed image-based single-cell profiling and temporal monitoring of circulating platelet aggregates that are the precursors to thrombosis. Our analysis encompassed 41 thrombosis patients, 110 COVID-19 patients, and 11 healthy individuals. By investigating the morphological changes of platelet aggregates under the influence of thrombosis, COVID-19, and COVID-19 vaccination, we observed distinct morphological alterations in platelet aggregates across different conditions, which shed light on the interplay between platelet aggregation and thrombotic events.

Spontaneous Raman flow cytometers and cell sorters typically have cell or particle throughputs less than 5 event/s due to long signal integration times. Here we report an up to 10x increase in throughput by coupling constant illumination along the flow path with time delay integration, a technique that counteracts sample motion by matching it to the charge transfer velocity of a CCD device. This allows longer signal integration times while also multiplexing acquisition along the flow path. We demonstrate high spectral bandwidth (600 – 3200 cm-1) Raman acquisitions from flowing particles and mammalian cells at throughputs up to 50 events/s.

Optoacoustic (photoacoustic) tomography (OAT) is a hybrid imaging modality that enables high-resolution optical imaging at depths that exceed the optical diffusion limit by combining optical excitation with acoustic detection. In this talk, I will discuss two techniques for the detection of acoustic signals in OAT that overcome limitations of piezoelectric technology, commonly used in OAT. In the first technique, a novel acoustic detector was developed in a silicon-photonics platform, achieving a record-high resolution of <20 µm, previously achieved only in optoacoustic microscopy setups. In the second technique, we developed cyclic acoustic masks that enable the application of signal-pixel imaging, facilitating semi-isotropic resolution and sensitivity enhancement.

Hyperspectral oblique plane microscopy provides Raman contrast images of biological specimens with unprecedented pixel throughput. Being a light-sheet technique it can capture hundreds of spectra in the time a point scanning system would take to capture a single spectrum. We will demonstrate the utility of this new system by monitoring wound healing in a zebrafish at five minute intervals, Followed by spontaneous Raman imaging of a beating zebrafish heart.

We developed a high-speed two-photon volumetric imaging system with hundreds of axial layers that match a deep-learning denoising model to capture millisecond-level functional changes in individual neurons with high SNR. Compare with general deep-learning methods, the spatial information-based training method not only enhances SNR by 300% but prevents temporal distortion. Our proof-of-concept experiment focused on calcium dynamics in cerebellum Purkinje cells, revealing similar responses in the parallel dendritic layers, yet significant divergence in the somatic area. This sheds light on the intricate signal processing at individual neuron levels, validating our imaging system.

Photoacoustic imaging (PAI) detects ultrasonic signals from tissue optical absorption. Lipids, with strong absorption at 1.7 μm, are valuable for arteriosclerosis and myocardial infarction imaging. We introduce a high-speed wideband wavelength tunable Raman fiber laser for photoacoustic spectroscopy of lipids. Pump lasers adjust from 1560 to 1600 nm, with a 1 ns pulse width and 100 kHz repetition rate. The pump light combines with seed light and passes the Raman fiber. The seed light amplifies by matching the Raman gain spectrum generated through stimulated Raman scattering. We obtained photoacoustic images in the 1670 - 1720 nm range, demonstrating our lasers for lipid imaging.

Score-based generative modeling, informally referred to as diffusion models, continue to grow in popularity across several important domains and tasks. While they provide high-quality and diverse samples from empirical distributions, important questions remain on the reliability and trustworthiness of these sampling procedures for their responsible use in critical scenarios. In this work, we focus on image-to-image regression tasks and we present a generalization of the Risk-Controlling Prediction Sets (RCPS) procedure, that we term K-RCPS, which allows to i) provide entrywise calibrated intervals for future samples of any diffusion model, and ii) control a certain notion of risk with respect to a ground truth image with minimal mean interval length. Differently from existing conformal risk control procedures, ours relies on a novel convex optimization approach that allows for multidimensional risk control while provably minimizing the mean interval length. We illustrate our approach on two real-world image

We developed a large-scale morphological profiling approach that reads out a multitude of high-resolution biophysical fractal properties of single cells, based on our high-throughput quantitative phase imaging (QPI) platform (at 10,000 cells/sec). We showed that the single-cell morphological profile constructed by the sub-cellular fractal characteristics could be harnessed to implicate cell types and states in the context of identifying lung cancer cell line subtypes, assessing cell morphology in response to drug treatment, and tracking cell cycle progression.

Quantum cascade lasers (QCLs) have presented a new chapter for the development of infrared spectroscopy for its high throughput and ability to tune to absorption bands of interest. However, compared to conventional thermal sources, the output stability of QCLs is a limiting factor. Specifically, the noise associated with pulse-to-pulse instability is of specific concern as we design faster and more sensitive spectrometers. Here, we demonstrate a dual-detector approach to reference measurements with the intensity illuminating the sample in real-time per individual laser pulse. We integrate this concept into our custom point-scanning confocal IR microscope. Results show double the SNR and a 10-fold shorter response time with the proposed method versus a single detector approach, all while being lock-in amplifier free. We characterize the controllable parameters and discuss the trade-offs required for a unified system capable of collecting high-quality point spectra as well as whole-slide spectral imaging.