We developed an adaptive optics based multifocal structured illumination microscopy system (AO-MSIM) for deep penetration, superresolution imaging. A spatial light modulator was used to generate a multi-focus excitation lattice, and multi-focus structured illumination imaging is achieved by stacking the phase of the grating. Based on the detected wavefront information, a spatial light modulator and a deformable mirror were used to correct the aberrations of excitation and emission paths respectively. By using pixel relocation and deconvolution image reconstruction algorithm, spatial resolution of ~150nm was achieved at an imaging depth of 500μm. The applications of AO-MSIM technology in imaging thick brain tissue samples will be demonstrated.

We developed a new super resolution multi-molecular optical metabolic imaging platform with Adam optimization-based Pointillism Deconvolution (A-PoD) and penalized reference matching (PRM) algorithms for DO-SRS hyperspectral imaging detection of metabolic changes in cells and animals during aging processes and in diseases conditions.

The trade-off between throughput and image quality is an inherent challenge in microscopy. To improve throughput, compressive imaging under-samples image signals; the images are then computationally reconstructed. However, the information loss in the acquisition process sets the compression bounds. Here we propose differentiable compressive fluorescence microscopy (∂μ) that includes a realistic generalizable forward model with learnable-physical parameters (i.e. illumination patterns), and a novel physics-inspired inverse model. The cascaded model is end-to-end differentiable and can learn optimal compressive sampling schemes through training data. Proposed learned sampling outperforms widely used traditional compressive sampling schemes at higher compressions. We also demonstrate task-aware sampling (e.g. segmentation-aware) with the proposed framework.

Rapid and slide-free cellular imaging with histological contrast and minimal tissue preparation has long been a challenging and yet appealing medical pursuit. We have recently proposed a promising and transformative histological imaging method, coined computational high-throughput autofluorescence microscopy by pattern illumination (CHAMP), which can provide rapid and label-free imaging of thick and unprocessed tissues with large surface irregularity at an acquisition speed of 10 mm^2/10 s with 1.1-µm lateral resolution. CHAMP images can be subsequently transformed into virtually stained histological images (Deep-CHAMP) through unsupervised learning. By incorporating a sectioning vibratome with a similar system configuration, three-dimensional histopathology is also feasible.

We report an integrative unsupervised deep learning approach to translate the complex morphological information of cells into interpretable representations that can be generalizable for downstream single-cell analytics. The method, integrating the respective advantages of generative adversarial network and variational autoencoder, enables faithful prediction of biological processes based on cell morphology read out from different imaging modalities. We demonstrate the generalizability and scalability of this method in a diverse range of applications, including cellular responses to SARS-CoV-2 infection, cell-cycle progression imaged by high-throughput quantitative phase imaging (QPI), and cellular changes during epithelial to mesenchymal transition (EMT) captured by fluorescence imaging.

Molecular self-assembled (MSA) materials such as lipid membranes, collagen, and DNA are common in living organisms and their hierarchical organization is crucial to the ultimate function of many biological systems. To understand the structure-properties correlations of MSA, it is necessary to characterize the micro-level structures and to obtain the chemical information simultaneously. To address this challenge, we coupled polarization resolved collinear vibrational sum frequency generation spectroscopy, an interface and symmetry selective second order nonlinear optical technique, with a line scanning microscopy platform to create a line scanning VSFG hyperspectral microscope which characterizes different chemical environment with spatial fidelity. The microscope was applied to a biomimetic MSA comprised of β-cyclodextrin (β-CD), a cup shaped heptamer of glucose, and sodium dodecyl sulfate (SDS), a common surfactant.

Monitoring spiking activity across large neuronal populations at behaviorally relevant timescales is critical for understanding neural circuit function. Voltage imaging requires kilohertz sampling rates which reduce fluorescence detection to near shot noise levels. High-photon flux excitation can overcome photon-limited shot noise but photo-bleaching and photo-damage restrict the number and duration of simultaneously imaged neurons. We investigated an alternative approach aimed at low two-photon flux, voltage imaging below the shot noise limit with the goal of achieving simultaneous high-speed, deep-tissue imaging of more than one hundred densely labeled neurons over one hour in awake behaving mice.

Laser scanning multiphoton microscopy is widely used for in vivo imaging applications. Commonly employed systems can offer an imaging rate of tens of Hz. With the emergence of high-speed function indicators, an imaging rate at hundreds of Hz is required. Recently, we developed the optical gearbox, which can convert commonly employed imaging systems and sources for high-speed imaging with a frame rate ranging from tens of Hz to 1,000 Hz. Here, we present our latest work which showcases the performance of optical gearbox based two-photon laser scanning microscopy for in vivo imaging of neuronal and vascular dynamics in mouse brain.

Nonlinear label-free microscopy techniques have shown great promise for analyzing biological tissues with an unmatched level of information. These techniques are capable of identifying intrinsic optical signals generated in response to ultrafast laser pulses. A roadblock in fast label-free imaging is the limited signal-to-noise ratio (SNR), due to the low number of photons close to the noise level. We implemented heterodyne detection of the third-order signals to overcome this limit and surpass the 1/f noise. Our method increases the SNR by a factor of 2.3, leading to a factor of 5 reduction in imaging time.

To catalyze a digital pathology transformation to improve clinical decisions, a novel technological approach is needed that offers significant advantages over traditional “gold-standard” histopathology in terms of accuracy and throughput. We have developed an open-top light-sheet (OTLS) microscopy platform for slide-free 3D pathology of large clinical specimens, enabling whole biopsies and surgical specimens to be non-destructively imaged in toto. Using machine-learning techniques, we are quantifying 3D spatial and molecular biomarkers for prognosticating patient outcomes (indolent vs. aggressive disease) and for guiding treatment decisions. These non-destructive large-volume digital pathology methods are synergistic with the growing fields of radiomics and genomics.

Oblique plane light-sheet microscopy provides versatile and rapid volumetric functional and structural fluorescence imaging for biology. A single microscope objective illuminates a tilted plane in the specimen and collects emission. Descanned emission is relayed and tilt-corrected through two additional microscope arms to a camera. Due to relative system complexity, estimating required hardware for diverse applications and system alignment can be challenging. Towards overcoming this obstacle and maximizing access, we present an open access scanned oblique plane microscopy platform, along with a design GUI, detailed alignment protocols, control software, application examples in zebrafish and mice, as well as possible systems extensions.

We present a high-throughput computational imaging system capable of performing dense, volumetric fluorescence imaging of freely moving organisms at up to 120 volumes per second. Our method, termed 2pi Fourier light field tomography (2pi-FLIFT), consists of a planar array of 54 cameras and a parabolic mirror serving as the primary objective that allows for synchronized multi-view video capture over ~2pi steradians. 2pi-FLIFT features a novel 3D reconstruction algorithm that recovers both the 3D fluorescence distribution and attenuation map of dynamic samples. We demonstrate 2pi-FLIFT on important, freely moving model organisms, such as zebrafish and fruit fly larvae.

Compressed sensing has emerged as an appealing option for single-shot high-speed imaging. Recent work proposed light field tomography (LIFT) to acquire en-face one-dimension (1-D) projections instead of 2-D images and reformulate as a computed tomography problem. The light field with reduced dimension enables 3D imaging at temporal resolution less than 10 picoseconds. We hereby propose a spectrum encoding method to increase the number of projections in LIFT within a snapshot. We multiply projections by the number of spectral channels for better reconstruction quality while maintaining the high-speed 3-D imaging capability. This mitigates the sparse-view problem in reconstruction.

In this talk, I will introduce “Convex Lens-induced Confinement (CLiC)” imaging, which works by mechanically confining single-molecules to an optical microscope’s field of view, isolating them in nanofabricated features, and eliminates the complexity and potential biases inherent to tethering molecules. I will discuss applications of our single-molecule platform to study structure-mediated DNA interactions in out-of-equilibrium conditions (Shaheen et. al., Nucl. Acid Research 2022), and how CLiC is helping develop emerging classes of genetic medicines by studying their interactions with nucleic acids. Finally, I will highlight current and future applications to connect our observations from single molecule to single cells.

Label-free optical observation of transparent nano-objects is challenging. Here, we demonstrate PlAsmonic NanO-apeRture lAbel-free iMAging (PANORAMA) that addresses existing issues for both SPR and LSPR imaging. PANORAMA involves a standard commercial bright-field microscope without the need for laser or interferometric detection. PANORAMA can image, size, count, and monitor a single polystyrene nanoparticle down to 25 nm at a millisecond timescale. The extrapolated size limit of detection is sub-10 nm. Higher imaging speed is achievable with high-speed cameras. Molecular imaging is envisioned with functionalized substrates for single nanoparticle analysis for extracellular vesicles (e.g., exosomes) and pathogens (e.g., viruses).

We report an arrayed optofluidic imaging platform that allows multiplexed image-based cell assay at ultrahigh imaging throughput. The assay platform, consisting of 96 fluidic sample chambers arranged in a circular symmetry, operates in a reconfigurable spinning motion synchronized with an ultrafast laser-scanning microscope (a line-scan rate >10 MHz). Based on a stable through-focus spinning mechanism, the assay platform allows ultralarge field-of-view imaging (18.8 cm2, >160 Gpixels) with high image fidelity and subcellular resolution (in both the bright-field and quantitative phase image contrasts). We further validated that the continuous spinning operation has minimal impact on the cell morphology and viability.

It has long been known that quantum photon correlations have the potential to improve microscope performance, enhancing the information that can be extracted at fixed photon budget. Here, we use them to demonstrate absolute quantum advantage in bioimaging. We show that they enable signal-to-noise beyond the photodamage-free capacity of conventional microscopy. We achieve this in a coherent Raman microscope, which we use to image molecular bonds within a cell with quantum-enhanced contrast. This allows imaging of biological structures that are otherwise inaccessible. Our work provides a path towards order-of-magnitude improvements in both sensitivity and imaging speed.

In this talk, I will present a few hyper spectral IR imaging techniques that allows us to characterize biological and material samples in temporal, spatial and energy domains. We show that IR spectral imaging is sensitive to local chemical enviornments, self-assembled structures, while the imaging reveals the overall morphology samples. With this technique, we can learn local hydration levels and structural ordering of self-assembled material and biological tissues. The correlations revealed among the spectral, spatial and temporal information can lead to comprehensive understanding of the materials when combined with informational techniques.

We develop a high-speed compressive Raman imaging technology using a programmable binary spectral filter and a single channel detector to perform fast Raman imaging for detection and concentration estimation of know species over millimeters field of view. The technology enables Raman imaging with a pixel dwell time as short as few hundreds of microseconds. We report fast Raman imaging of pharmaceutical tablets and micro-plastics. We also present a novel fast line scan compressive Raman imaging scheme using the 2D digital micro-mirror device (DMD) to encode both space and frequency.

Compressive Raman imaging has emerged as a promising technique to speed up chemical imaging by compressing the data during acquisition. Yet, current scanning imaging speed is fundamentally limited by the sensors pixel dwell times of at best 1 µs. Here, we introduce a compressive Raman spectrometer layout equipped with a novel parallelized spatial acquisition using a single-photon avalanche detector array. We show imaging with pixel dwell times of <10µs using the otherwise weak spontaneous Raman effect, thereby reaching real-time imaging.

We propose a novel approach to broadband coherent anti-Stokes Raman scattering (B-CARS) based on a femtosecond laser at 1035 nm and 2 MHz repetition rate. These features of the driving laser enable white-light continuum generation in bulk media, employed as broadband Stokes. In this way, we demonstrate state-of-the-art acquisition speed (<1 ms/pixel) with unprecedented sensitivity (≈14.1 mmol/L) when detecting dimethyl-sulfoxide in water, covering the whole fingerprint region. To further enhance the performance of the system, we designed an innovative spectral denoiser based on a convolutional neural network, coupled with a post-processing pipeline to distinguish different chemical species of biological tissues.

Recent advances in optical microscopy have revolutionized our ability to visualize multifaceted morphological signatures (biophysical & molecular) of cells without the onerous and destructive sample handling commonly used in molecular profiling. This talk will cover our latest developments in combining ultrahigh-throughput microfluidic single-cell imaging (at a throughput up to 100,000 cells/sec) with different supervised/unsupervised image analytic strategies for crafting large-scale morphological “fingerprints” (profiles) of single cells. The talk will further discuss the new opportunities and challenges of how to integrate this rapidly growing image-based repertoire with the single-cell multi-omics data - altogether formulating holistic and integrative single-cell analysis strategies for dissecting the complex biological system.

Heterogeneity plays an important role in medicine and biology, which can be investigated by exploiting single cell analysis (SCA). Among SCA methods, imaging cytometry allows the analysis of individual 2D and 3D spatial features. Here we present a femtosecond laser fabricated optofluidic automated platform encompassing a thermo-optic phase shifter, cylindrical lenses and a microfluidic network to generate and shift a dual-color patterned light sheet within a microchannel where the samples of interest flow. The device can be used as add-on and can provide an acquisition rate of about 1 cell/second, or subnuclear resolution at the single cell level.

There is widespread concern about the safety of COVID-19 vaccinations related to platelet hyperactivity. However, their long-term influence on platelet activity remains unknown. We address this issue by applying a high-speed bright-field microscope based on optical frequency-division multiplexing and microfluidics for massive image-based analysis. We performed image-based single-cell profiling and temporal monitoring of circulating platelet aggregates in the blood samples of healthy human participants before and after they received three vaccination doses over a nearly one-year period. The results demonstrate no significant or persistent change in platelet activity after vaccine doses.

Using high-throughput quantitative phase imaging (QPI) flow cytometry, we demonstrate that label-free single-cell image-based analysis allows the classification of the key human T cell subpopulations, CD4+ and CD8+ cells. Going beyond the existing QPI cytometers, we show that the high-dimensional biophysical phenotypic profiles extracted from this large-scale QPI platform display the label-free statistical power to unambiguously reveal the respective activation changes of the CD4+ and CD8+ cells subpopulations which are costimulated with anti-CD3/CD28. The findings are validated with the standard activation marker CD25. This work has further substantiated the potential of adopting label-free QPI cytometry for in-depth functional immune cell profiling.

Precisely characterising and quantifying interactions between tumour cells and their environment to understand metastatic mechanisms requires a multi-dimensional, high-speed imaging system. To this end, we report on the development of a compressive full spectrum fluorescence lifetime microscope that exploits a novel SPAD line sensor and a DMD to enable monitoring of dynamic sub-cellular interactions. At no cost to its temporal performance, the hyperspectral nature of the system helps to improve unmixing and, crucially, can detect the small spectral changes in the emission of fluorescent probes and intrinsic fluorophores that can occur in complex environments.

It is challenging to image sparse samples with fast FLIM since the laser clock is not directly sampled. Instead, it is inferred from the cumulative photon statistics of an entire line. We present a method for registering the photon arrival times to the excitation using time-domain multiplexing of the laser clock to ensure accuracy for fewer photons for fast FLIM. Our technique also does not add to the existing bottleneck of data throughput and eliminates the errors in registering the photons from different frames, yielding more accurate images at faster frame rates.

We report the development of a novel confocal line-scanning microscope capable of acquiring video-frame rate TCSPC-based FLIM. The system consists of a one-dimensional laser beam, which is optically conjugated to a 1024×16 single photon avalanche diode(SPAD) based line-imaging CMOS(1), with 23.78 μm pixel pitch at 49.31% fill factor. Incorporation of on-chip histogramming on the line-sensor facilitates the acquisition of up to 16.5 Giga-photon counts/s, enabling operation 66 times faster than our previously reported bespoke high speed FLIM platforms. We will demonstrate its use in live-cell imaging investigating the roles that PAK proteins play in regulation of cytoskeletal dynamics.

OCT has been widely employed for capturing dynamic processes in biological tissues via (2+1)D imaging, but are currently limited in (3+1)D imaging by both scan time and storage efficiency. Temporal non-uniform compressive sampling has been previously demonstrated on both simulations and existing data with accurate reconstruction of intracellular motility with up to 8-fold compression. Here we present the experimental implementation of this technique on a spectral-domain OCT system, with demonstration of significantly reduced memory overhead and total imaging time at 4- and 8-fold compression, which enables new applications in high-throughput assays employing (3+1)D imaging.

Cellular resolution of optical coherence tomography (OCT) is vital to achieve precise diagnosis by offering high-quality images of virtual biopsy. Currently, the common solution is to apply dynamic focusing to axially translate the focus through the region of interest with a high numerical aperture (N.A.) objective, followed by Z-stacking to rebuild a high-resolution 3D volume. To accelerate the imaging acquisition, this work developed metasurface optical plates to generate multiple foci along axial direction. Two-/three-/seven-foci had been testified with bead phantom using a scanning OCT. Human skin and human brain samples were imaged with cellular resolution.