UNMIX-ME: spectral and lifetime fluorescence unmixing via deep learning

Live-cell imaging powered by computation

The proliferation of microscopy methods for live-cell imaging offers many new possibilities for users but can also be challenging to navigate. The prevailing challenge in live-cell fluorescence microscopy is capturing intra-cellular dynamics while preserving cell viability. Computational methods can help to address this challenge and are now shifting the boundaries of what is possible to capture in living systems. In this Review, we discuss these computational methods focusing on artificial intelligence-based approaches that can be layered on top of commonly used existing microscopies as well as hybrid methods that integrate computation and microscope hardware. We specifically discuss how computational approaches can improve the signal-to-noise ratio, spatial resolution, temporal resolution and multi-colour capacity of live-cell imaging.

BINGO: a blind unmixing algorithm for ultra-multiplexing fluorescence images

MOTIVATION

Spectral imaging is often used to observe different objects with multiple fluorescent labels to reveal the development of the biological event. As the number of observed objects increases, the spectral overlap between fluorophores becomes more serious, and obtaining a "pure" picture of each fluorophore becomes a major challenge. Here, we propose a blind spectral unmixing algorithm called BINGO (Blind unmixing via SVD-based Initialization Nmf with project Gradient descent and spare cOnstrain), which can extract all kinds of fluorophores more accurately from highly overlapping multichannel data, even if the spectra of the fluorophores are extremely similar or their fluorescence intensity varies greatly.

RESULTS

BINGO can isolate up to 10 fluorophores from spectral imaging data for a single excitation. nine-color living HeLa cells were visualized distinctly with BINGO. It provides an important algorithmic tool for multiplex imaging studies, especially in intravital imaging. BINGO shows great potential in multicolor imaging for biomedical sciences.

AVAILABILITY AND IMPLEMENTATION

The source code used for this paper is available with the test data at https://github.com/Xinyuan555/BINGO_unmixing.

5-ALA induced PpIX fluorescence spectroscopy in neurosurgery: a review

The review begins with an overview of the fundamental principles/physics underlying light, fluorescence, and other light-matter interactions in biological tissues. It then focuses on 5-aminolevulinic acid (5-ALA)-induced protoporphyrin IX (PpIX) fluorescence spectroscopy methods used in neurosurgery (e.g., intensity, time-resolved) and in so doing, describe their specific features (e.g., hardware requirements, main processing methods) as well as their strengths and limitations. Finally, we review current clinical applications and future directions of 5-ALA-induced protoporphyrin IX (PpIX) fluorescence spectroscopy in neurosurgery.

Deep learning assisted classification of spectral photoacoustic imaging of carotid plaques

Spectral photoacoustic imaging (sPAI) is an emerging modality that allows real-time, non-invasive, and radiation-free assessment of tissue, benefiting from their optical contrast. sPAI is ideal for morphology assessment in arterial plaques, where plaque composition provides relevant information on plaque progression and its vulnerability. However, since sPAI is affected by spectral coloring, general spectroscopy unmixing techniques cannot provide reliable identification of such complicated sample composition. In this study, we employ a convolutional neural network (CNN) for the classification of plaque composition using sPAI. For this study, nine carotid endarterectomy plaques were imaged and were then annotated and validated using multiple histological staining. Our results show that a CNN can effectively differentiate constituent regions within plaques without requiring fluence or spectra correction, with the potential to eventually support vulnerability assessment in plaques.

AutoUnmix: an autoencoder-based spectral unmixing method for multi-color fluorescence microscopy imaging

Multiplexed fluorescence microscopy imaging is widely used in biomedical applications. However, simultaneous imaging of multiple fluorophores can result in spectral leaks and overlapping, which greatly degrades image quality and subsequent analysis. Existing popular spectral unmixing methods are mainly based on computational intensive linear models, and the performance is heavily dependent on the reference spectra, which may greatly preclude its further applications. In this paper, we propose a deep learning-based blindly spectral unmixing method, termed AutoUnmix, to imitate the physical spectral mixing process. A transfer learning framework is further devised to allow our AutoUnmix to adapt to a variety of imaging systems without retraining the network. Our proposed method has demonstrated real-time unmixing capabilities, surpassing existing methods by up to 100-fold in terms of unmixing speed. We further validate the reconstruction performance on both synthetic datasets and biological samples. The unmixing results of AutoUnmix achieve the highest SSIM of 0.99 in both three- and four-color imaging, with nearly up to 20% higher than other popular unmixing methods. For experiments where spectral profiles and morphology are akin to simulated data, our method realizes the quantitative performance demonstrated above. Due to the desirable property of data independency and superior blind unmixing performance, we believe AutoUnmix is a powerful tool for studying the interaction process of different organelles labeled by multiple fluorophores.

Compact and robust deep learning architecture for fluorescence lifetime imaging and FPGA implementation

This paper reports a bespoke adder-based deep learning network for time-domain fluorescence lifetime imaging (FLIM). By leveraging thel1-norm extraction method, we propose a 1D Fluorescence Lifetime AdderNet (FLAN) without multiplication-based convolutions to reduce the computational complexity. Further, we compressed fluorescence decays in temporal dimension using a log-scale merging technique to discard redundant temporal information derived as log-scaling FLAN (FLAN+LS). FLAN+LS achieves 0.11 and 0.23 compression ratios compared with FLAN and a conventional 1D convolutional neural network (1D CNN) while maintaining high accuracy in retrieving lifetimes. We extensively evaluated FLAN and FLAN+LS using synthetic and real data. A traditional fitting method and other non-fitting, high-accuracy algorithms were compared with our networks for synthetic data. Our networks attained a minor reconstruction error in different photon-count scenarios. For real data, we used fluorescent beads' data acquired by a confocal microscope to validate the effectiveness of real fluorophores, and our networks can differentiate beads with different lifetimes. Additionally, we implemented the network architecture on a field-programmable gate array (FPGA) with a post-quantization technique to shorten the bit-width, thereby improving computing efficiency. FLAN+LS on hardware achieves the highest computing efficiency compared to 1D CNN and FLAN. We also discussed the applicability of our network and hardware architecture for other time-resolved biomedical applications using photon-efficient, time-resolved sensors.

Two-Photon Imaging for Non-Invasive Corneal Examination

Two-photon imaging (TPI) microscopy, namely, two-photon excited fluorescence (TPEF), fluorescence lifetime imaging (FLIM), and second-harmonic generation (SHG) modalities, has emerged in the past years as a powerful tool for the examination of biological tissues. These modalities rely on different contrast mechanisms and are often used simultaneously to provide complementary information on morphology, metabolism, and structural properties of the imaged tissue. The cornea, being a transparent tissue, rich in collagen and with several cellular layers, is well-suited to be imaged by TPI microscopy. In this review, we discuss the physical principles behind TPI as well as its instrumentation. We also provide an overview of the current advances in TPI instrumentation and image analysis. We describe how TPI can be leveraged to retrieve unique information on the cornea and to complement the information provided by current clinical devices. The present state of corneal TPI is outlined. Finally, we discuss the obstacles that must be overcome and offer perspectives and outlooks to make clinical TPI of the human cornea a reality.

Computational macroscopic lifetime imaging and concentration unmixing of autofluorescence

Single-pixel computational imaging can leverage highly sensitive detectors that concurrently acquire data across spectral and temporal domains. For molecular imaging, such methodology enables to collect rich intensity and lifetime multiplexed fluorescence datasets. Herein we report on the application of a single-pixel structured light-based platform for macroscopic imaging of tissue autofluorescence. The super-continuum visible excitation and hyperspectral single-pixel detection allow for parallel characterization of autofluorescence intensity and lifetime. Furthermore, we exploit a deep learning based data processing pipeline, to perform autofluorescence unmixing while yielding the autofluorophores' concentrations. The full scheme (setup and processing) is validated in silico and in vitro with clinically relevant autofluorophores flavin adenine dinucleotide, riboflavin, and protoporphyrin. The presented results demonstrate the potential of the methodology for macroscopically quantifying the intensity and lifetime of autofluorophores, with higher specificity for cases of mixed emissions, which are ubiquitous in autofluorescence and multiplexed in vivo imaging.

Suppression of natural lens fluorescence in fundus autofluorescence measurements: review of hardware solutions

Fluorescence lifetime imaging ophthalmoscopy (FLIO), a technique for investigating metabolic changes in the eye ground, can reveal the first signs of diseases related to metabolism. The fluorescence of the natural lens overlies the fundus fluorescence. Although the influence of natural lens fluorescence can be somewhat decreased with mathematical models, excluding this influence during the measurement by using hardware enables more exact estimation of the fundus fluorescence. Here, we analyze four 1-photon excitation hardware solutions to suppress the influence of natural lens fluorescence: aperture stop separation, confocal scanning laser ophthalmoscopy, combined confocal scanning laser ophthalmoscopy and aperture stop separation, and dual point confocal scanning laser ophthalmoscopy. The effect of each principle is demonstrated in examples. The best suppression is provided by the dual point principle, realized with a confocal scanning laser ophthalmoscope. In this case, in addition to the fluorescence of the whole eye, the fluorescence of the anterior part of the eye is detected from a non-excited spot of the fundus. The intensity and time-resolved fluorescence spectral data of the fundus are derived through the subtraction of the simultaneously measured fluorescence of the excited and non-excited spots. Advantages of future 2-photon fluorescence excitation are also discussed. This study provides the first quantitative evaluation of hardware principles to suppress the fluorescence of the natural lens during measurements of fundus autofluorescence.

In vitro and in vivo NIR fluorescence lifetime imaging with a time-gated SPAD camera

Near-infrared (NIR) fluorescence lifetime imaging (FLI) provides a unique contrast mechanism to monitor biological parameters and molecular events in vivo. Single-photon avalanche diode (SPAD) cameras have been recently demonstrated in FLI microscopy (FLIM) applications, but their suitability for in vivo macroscopic FLI (MFLI) in deep tissues remains to be demonstrated. Herein, we report in vivo NIR MFLI measurement with SwissSPAD2, a large time-gated SPAD camera. We first benchmark its performance in well-controlled in vitro experiments, ranging from monitoring environmental effects on fluorescence lifetime, to quantifying Förster resonant energy transfer (FRET) between dyes. Next, we use it for in vivo studies of target-drug engagement in live and intact tumor xenografts using FRET. Information obtained with SwissSPAD2 was successfully compared to that obtained with a gated intensified charge-coupled device (ICCD) camera, using two different approaches. Our results demonstrate that SPAD cameras offer a powerful technology for in vivo preclinical applications in the NIR window.

Versatile compressive microscope for hyperspectral transmission and fluorescence lifetime imaging

Increasing demand for multimodal characterization and imaging of new materials entails the combination of various methods in a single microscopic setup. Hyperspectral imaging of transmission spectra or photoluminescence (PL) decay imaging count among the most used methods. Nevertheless, these methods require very different working conditions and instrumentation. Therefore, combining the methods into a single microscopic system is seldom implemented. Here we demonstrate a novel versatile microscope based on single-pixel imaging, where we use a simple optical configuration to measure the hyperspectral information, as well as fluorescence lifetime imaging (FLIM). The maps are inherently spatially matched and can be taken with spectral resolution limited by the resolution of the used spectrometer (3 nm) or temporal resolution set by PL decay measurement (120 ps). We verify the system's performance by its comparison to the standard FLIM and non-imaging transmission spectroscopy. Our approach enabled us to switch between a broad field-of-view and micrometer resolution without changing the optical configuration. At the same time, the used design opens the possibility to add a variety of other characterization methods. This article demonstrates a simple, affordable way of complex material studies with huge versatility for the imaging parameters.

Live-cell fluorescence spectral imaging as a data science challenge

Live-cell fluorescence spectral imaging is an evolving modality of microscopy that uses specific properties of fluorophores, such as excitation or emission spectra, to detect multiple molecules and structures in intact cells. The main challenge of analyzing live-cell fluorescence spectral imaging data is the precise quantification of fluorescent molecules despite the weak signals and high noise found when imaging living cells under non-phototoxic conditions. Beyond the optimization of fluorophores and microscopy setups, quantifying multiple fluorophores requires algorithms that separate or unmix the contributions of the numerous fluorescent signals recorded at the single pixel level. This review aims to provide both the experimental scientist and the data analyst with a straightforward description of the evolution of spectral unmixing algorithms for fluorescence live-cell imaging. We show how the initial systems of linear equations used to determine the concentration of fluorophores in a pixel progressively evolved into matrix factorization, clustering, and deep learning approaches. We outline potential future trends on combining fluorescence spectral imaging with label-free detection methods, fluorescence lifetime imaging, and deep learning image analysis.

Deep learning in macroscopic diffuse optical imaging

SIGNIFICANCE

Biomedical optics system design, image formation, and image analysis have primarily been guided by classical physical modeling and signal processing methodologies. Recently, however, deep learning (DL) has become a major paradigm in computational modeling and has demonstrated utility in numerous scientific domains and various forms of data analysis.

AIM

We aim to comprehensively review the use of DL applied to macroscopic diffuse optical imaging (DOI).

APPROACH

First, we provide a layman introduction to DL. Then, the review summarizes current DL work in some of the most active areas of this field, including optical properties retrieval, fluorescence lifetime imaging, and diffuse optical tomography.

RESULTS

The advantages of using DL for DOI versus conventional inverse solvers cited in the literature reviewed herein are numerous. These include, among others, a decrease in analysis time (often by many orders of magnitude), increased quantitative reconstruction quality, robustness to noise, and the unique capability to learn complex end-to-end relationships.

CONCLUSIONS

The heavily validated capability of DL's use across a wide range of complex inverse solving methodologies has enormous potential to bring novel DOI modalities, otherwise deemed impractical for clinical translation, to the patient's bedside.

3D denoised completion network for deep single-pixel reconstruction of hyperspectral images

Single-pixel imaging acquires an image by measuring its coefficients in a transform domain, thanks to a spatial light modulator. However, as measurements are sequential, only a few coefficients can be measured in the real-time applications. Therefore, single-pixel reconstruction is usually an underdetermined inverse problem that requires regularization to obtain an appropriate solution. Combined with a spectral detector, the concept of single-pixel imaging allows for hyperspectral imaging. While each channel can be reconstructed independently, we propose to exploit the spectral redundancy between channels to regularize the reconstruction problem. In particular, we introduce a denoised completion network that includes 3D convolution filters. Contrary to black-box approaches, our network combines the classical Tikhonov theory with the deep learning methodology, leading to an explainable network. Considering both simulated and experimental data, we demonstrate that the proposed approach yields hyperspectral images with higher quantitative metrics than the approaches developed for grayscale images.

Deep Learning in Biomedical Optics

This article reviews deep learning applications in biomedical optics with a particular emphasis on image formation. The review is organized by imaging domains within biomedical optics and includes microscopy, fluorescence lifetime imaging, in vivo microscopy, widefield endoscopy, optical coherence tomography, photoacoustic imaging, diffuse tomography, and functional optical brain imaging. For each of these domains, we summarize how deep learning has been applied and highlight methods by which deep learning can enable new capabilities for optics in medicine. Challenges and opportunities to improve translation and adoption of deep learning in biomedical optics are also summarized. Lasers Surg. Med. © 2021 Wiley Periodicals LLC.

Recent innovations in fluorescence lifetime imaging microscopy for biology and medicine

SIGNIFICANCE

Fluorescence lifetime imaging microscopy (FLIM) measures the decay rate of fluorophores, thus providing insights into molecular interactions. FLIM is a powerful molecular imaging technique that is widely used in biology and medicine.

AIM

This perspective highlights some of the major advances in FLIM instrumentation, analysis, and biological and clinical applications that we have found impactful over the last year.

APPROACH

Innovations in FLIM instrumentation resulted in faster acquisition speeds, rapid imaging over large fields of view, and integration with complementary modalities such as single-molecule microscopy or light-sheet microscopy. There were significant developments in FLIM analysis with machine learning approaches to enhance processing speeds, fit-free techniques to analyze images without a priori knowledge, and open-source analysis resources. The advantages and limitations of these recent instrumentation and analysis techniques are summarized. Finally, applications of FLIM in the last year include label-free imaging in biology, ophthalmology, and intraoperative imaging, FLIM of new fluorescent probes, and lifetime-based Förster resonance energy transfer measurements.

CONCLUSIONS

A large number of high-quality publications over the last year signifies the growing interest in FLIM and ensures continued technological improvements and expanding applications in biomedical research.

Single-pixel image reconstruction from experimental data using neural networks

Single-pixel cameras that measure image coefficients have various promising applications, in particular for hyper-spectral imaging. Here, we investigate deep neural networks that when fed with experimental data can output high-quality images in real time. Assuming that the measurements are corrupted by mixed Poisson-Gaussian noise, we propose to map the raw data from the measurement domain to the image domain based on a Tikhonov regularization. This step can be implemented as the first layer of a deep neural network, followed by any architecture of layers that acts in the image domain. We also describe a framework for training the network in the presence of noise. In particular, our approach includes an estimation of the image intensity and experimental parameters, together with a normalization scheme that allows varying noise levels to be handled during training and testing. Finally, we present results from simulations and experimental acquisitions with varying noise levels. Our approach yields images with improved peak signal-to-noise ratios, even for noise levels that were foreseen during the training of the networks, which makes the approach particularly suitable to deal with experimental data. Furthermore, while this approach focuses on single-pixel imaging, it can be adapted for other computational optics problems.

Fluorophore unmixing based on bleaching and recovery kinetics using MCR-ALS

Fluorescence microscopy is a key technology in the life sciences, though its performance is constrained by the number of labels that can be recorded. We propose to use the kinetics of fluorophore photodestruction and subsequent fluorescence recovery to distinguish multiple spectrally-overlapping emitters in fixed cells, thus enhancing the information that can be obtained from a single measurement. We show that the data can be directly processed using multivariate curve resolution - alternating least squares (MCR-ALS) to deliver distinct images for each fluorophore in their local environment, and apply this methodology to membrane imaging using DiBAC₄(3) and concanavalin A - Alexa Fluor 488 as the fluorophores. We find that the DiBAC₄(3) displays two distinct degradation/recovery kinetics that correspond to two different label distributions, allowing us to simultaneously distinguish three different fluorescence distributions from two spectrally overlapping fluorophores. We expect that our approach will scale to other dynamically-binding dyes, leading to similarly increased multiplexing capability.

Luminescence lifetime imaging of three-dimensional biological objects

A major focus of current biological studies is to fill the knowledge gaps between cell, tissue and organism scales. To this end, a wide array of contemporary optical analytical tools enable multiparameter quantitative imaging of live and fixed cells, three-dimensional (3D) systems, tissues, organs and organisms in the context of their complex spatiotemporal biological and molecular features. In particular, the modalities of luminescence lifetime imaging, comprising fluorescence lifetime imaging (FLI) and phosphorescence lifetime imaging microscopy (PLIM), in synergy with Förster resonance energy transfer (FRET) assays, provide a wealth of information. On the application side, the luminescence lifetime of endogenous molecules inside cells and tissues, overexpressed fluorescent protein fusion biosensor constructs or probes delivered externally provide molecular insights at multiple scales into protein-protein interaction networks, cellular metabolism, dynamics of molecular oxygen and hypoxia, physiologically important ions, and other physical and physiological parameters. Luminescence lifetime imaging offers a unique window into the physiological and structural environment of cells and tissues, enabling a new level of functional and molecular analysis in addition to providing 3D spatially resolved and longitudinal measurements that can range from microscopic to macroscopic scale. We provide an overview of luminescence lifetime imaging and summarize key biological applications from cells and tissues to organisms.

Quantification of Trastuzumab-HER2 Engagement In Vitro and In Vivo

Human EGF Receptor 2 (HER2) is an important oncogene driving aggressive metastatic growth in up to 20% of breast cancer tumors. At the same time, it presents a target for passive immunotherapy such as trastuzumab (TZM). Although TZM has been widely used clinically since 1998, not all eligible patients benefit from this therapy due to primary and acquired drug resistance as well as potentially lack of drug exposure. Hence, it is critical to directly quantify TZM–HER2 binding dynamics, also known as cellular target engagement, in undisturbed tumor environments in live, intact tumor xenograft models. Herein, we report the direct measurement of TZM–HER2 binding in HER2-positive human breast cancer cells and tumor xenografts using fluorescence lifetime Forster Resonance Energy Transfer (FLI-FRET) via near-infrared (NIR) microscopy (FLIM-FRET) as well as macroscopy (MFLI-FRET) approaches. By sensing the reduction of fluorescence lifetime of donor-labeled TZM in the presence of acceptor-labeled TZM, we successfully quantified the fraction of HER2-bound and internalized TZM immunoconjugate both in cell culture and tumor xenografts in live animals. Ex vivo immunohistological analysis of tumors confirmed the binding and internalization of TZM–HER2 complex in breast cancer cells. Thus, FLI-FRET imaging presents a powerful analytical tool to monitor and quantify cellular target engagement and subsequent intracellular drug delivery in live HER2-positive tumor xenografts.

Hyperspectral multiphoton microscopy for in vivo visualization of multiple, spectrally overlapped fluorescent labels

The insensitivity of multiphoton microscopy to optical scattering enables high-resolution, high-contrast imaging deep into tissue, including in live animals. Scattering does, however, severely limit the use of spectral dispersion techniques to improve spectral resolution. In practice, this limited spectral resolution together with the need for multiple excitation wavelengths to excite different fluorophores limits multiphoton microscopy to imaging a few, spectrally-distinct fluorescent labels at a time, restricting the complexity of biological processes that can be studied. Here, we demonstrate a hyperspectral multiphoton microscope that utilizes three different wavelength excitation sources together with multiplexed fluorescence emission detection using angle-tuned bandpass filters. This microscope maintains scattering insensitivity, while providing high enough spectral resolution on the emitted fluorescence and capitalizing on the wavelength-dependent nonlinear excitation of fluorescent dyes to enable clean separation of multiple, spectrally overlapping labels, in vivo. We demonstrated the utility of this instrument for spectral separation of closely-overlapped fluorophores in samples containing ten different colors of fluorescent beads, live cells expressing up to seven different fluorescent protein fusion constructs, and in multiple in vivo preparations in mouse cortex and inflamed skin with up to eight different cell types or tissue structures distinguished.

High compression deep learning based single-pixel hyperspectral macroscopic fluorescence lifetime imaging in vivo

Single pixel imaging frameworks facilitate the acquisition of high-dimensional optical data in biological applications with photon starved conditions. However, they are still limited to slow acquisition times and low pixel resolution. Herein, we propose a convolutional neural network for fluorescence lifetime imaging with compressed sensing at high compression (NetFLICS-CR), which enables in vivo applications at enhanced resolution, acquisition and processing speeds, without the need for experimental training datasets. NetFLICS-CR produces intensity and lifetime reconstructions at 128 × 128 pixel resolution over 16 spectral channels while using only up to 1% of the required measurements, therefore reducing acquisition times from ∼2.5 hours at 50% compression to ∼3 minutes at 99% compression. Its potential is demonstrated in silico, in vitro and for mice in vivo through the monitoring of receptor-ligand interactions in liver and bladder and further imaging of intracellular delivery of the clinical drug Trastuzumab to HER2-positive breast tumor xenografts. The data acquisition time and resolution improvement through NetFLICS-CR, facilitate the translation of single pixel macroscopic flurorescence lifetime imaging (SP-MFLI) for in vivo monitoring of lifetime properties and drug uptake.