In this work, a multiwavelength autofluorescence virtual instant stain (MAVIS) workflow is proposed to provide a multiple virtual staining solution to facilitate clinical diagnosis. Multiple ultraviolet excitation and visible emission wavelengths are used to highlight different biomolecules while a weakly supervised algorithm provides a robust and accurate virtual staining with adjacent tissue slices. The result of MAVIS with three histochemical stains on human tissue slices achieves a multi-scale structural similarity index measure > 0.6, demonstrating the potential of multiple virtual staining as a rapid and low-cost alternative to the current histological workflow.

We present a novel, cost-effective, versatile, and practical frequency-domain fluorescence lifetime imaging/microscopy (FD-FLIM) implementation capable of simultaneous multi-wavelength excitation, simultaneous multispectral detection, and sub-nanosecond to nanosecond fluorescence lifetime estimation. Rigorous validation experiments have demonstrated the capabilities of this novel FD-FLIM implementation to accurately measure fluorescence lifetimes in the range of ~0.4-12 ns. This novel technology has been adopted to implement an FD-FLIM endoscopy system for label-free metabolic imaging of oral lesions based on NADH (375-nm excitation, 484/37-nm emission) and FAD (445-nm excitation, 553/93-nm emission) autofluorescence imaging. Future efforts will focus on developing imaging systems for simultaneous endogenous (UV-VI excitation) and exogenous (NIR excitation) fluorescence lifetime imaging.

Induced pluripotent stem cells (iPSC) can generate patient-specific disease models and drug screening platforms. Heterogeneity in commercial culture substrates causes batch variation in manufactured iPSC cardiomyocyte differentiation and maturation levels. We performed optical metabolic imaging (OMI) of iPSC-cardiomyocytes on polyethylene glycol hydrogels of varying stiffnesses and adhesion ligands to assess the metabolic co-factors NAD(P)H and FAD during differentiation and maturation. Optical redox ratio at 6 days post-differentiation identified hydrogel features most favorable for subsequent cardiomyocyte maturation. These findings indicate that OMI enables rapid, non-invasive screens of favorable culture conditions during early differentiation and may translate to the biomanufacturing industry.

Active neurons experience rapid changes in their metabolic states since they have dynamic energy requirements. In this presentation, we demonstrate fast dual-channel label-free fluorescence lifetime imaging microscopy (FLIM) of NAD(P)H and FAD as a method for neurophysiology by performing computational photon counting in the onboard FPGA of the digitizer. The data throughput is reduced by 4x for each channel by compressing the photocurrents (16 bits) to photon counts (4 bits); the parallel processing on the FPGA ensures no lag. The setup was demonstrated for mammalian stem-cell-derived neurons under chemical stimulation, ion-channel blockers, and optical excitation. Fast FLIM on the FPGA enables dual-channel label-free metabolic optophysiology of neural activity in real time.

Patient-derived cancer organoids (PDCOs) are valuable for patient management and drug development. Label-free widefield optical redox ratio (WF-ORR) imaging measures organoid-level treatment response using the autofluorescence intensity of the metabolic co-enzymes NAD(P)H and FAD. We have developed WF-ORR imaging and analysis tools to maximize the accessibility, speed, sensitivity, and reproducibility of drug response measurements in PDCOs. These tools include leading-edge segmentation methods that isolate ORR changes within the proliferating zone of PDCOs, and single organoid tracking that reduces variability within drug screens. Overall, this approach enables rapid, sensitive, and reproducible measurements of treatment response and heterogeneity in PDCOs.

Two-photon excited fluorescence (TPEF) can provide metabolic information of the cellular microenvironment. Metabolic changes are intimately associated with numerous diseases. It has been suggested that the metabolic interpretations from fresh tissue imaging can be maintained in tissue samples post-fixation. Given the large availability of fixed tissues for various conditions, we sought to evaluate the impact of fixation on several TPEF-extracted metabolic indicators of unstained tissues based on endogenous intensity and lifetime TPEF measurements. Our results indicate that the metabolic interpretations of fixed tissue cannot be used as a representation of the in vivo environment or that of freshly excised specimens.

We apply the label-free hyperspectral light field tomographic (Hyper-LIFT) technology in eye imaging application. A 5D (x, y, θ, φ, λ) (x, y, spatial coordinates; θ, φ, angular coordinates; λ, wavelength) 300×300×4×4×90 plenoptic datacube were acquired with a spectrum range from 530nm to 570nm in a snapshot. The capability of capturing light field enables digital refocusing which extend the depth of field. We demonstrate the spectral imaging performance with a color checker and artificial eye model.

Multiphoton microscopy was used to collect five spectral channels of two-photon excited fluorescence (2PEF) from clinical duodenal neuroendocrine tumor specimens. As clinical use of a device capable of collecting 2PEF images would prohibit such extensive data collection, the optimal spatial and spectral data required for tissue classification was determined. Tissue classifiers were created using linear discriminant analysis of texture feature data from 2PEF images. Classifier accuracy, and the features that provided the highest accuracy, were compared at varying stages of resolution downsampling. Results suggest optimal imaging parameters for a system designed for the automatic detection of cancerous lesions using 2PEF.

We present a novel broadband coherent anti-Stokes Raman scattering (B-CARS) approach, employing a femtosecond fiber laser at 1035 nm and 2 MHz repetition rate. These characteristics of the driving laser enable to generate broadband Stokes pulses through white-light supercontinuum in bulk media and to use higher pulse energy to shine the sample, thus producing stronger signals. In this way, we demonstrate state-of-the-art acquisition speed (<1 ms/pixel) with unprecedented sensitivity (≈14.1 mmol/L), covering the whole fingerprint vibrational region. The B-CARS microscope, coupled with a deep learning-based post-processing pipeline, delivers high-quality hyperspectral images identifying the main Raman features of heterogeneous biological tissues.

Water is the most basic requirement of life, acting as a medium of exchange, transport, homeostasis and diffusion. We present a technique which leverages quantitative phase imaging (QPI) and a two-component mixture model to quantify, in absolute terms, the intracellular water fraction (by mass and volume) as well as the intracellular water volume in femtoliters. We demonstrate our technique by applying a single microsecond-duration electric pulse, and observe water influx in physiological buffer, water efflux in hyperosmotic buffer, and no movement across the membrane of synthetic vesicles. Dose response curves and channel blockers are employed to better understand underlying phenomena.

We have developed a polychromatic digital holographic microscope (P-DHM) as a quasi-speckle-free quantitative-phase imaging, having preserved its nanometric sensitivity from interferometric detection, for live-cell imaging. With this approach, we can visualize tiny subcellular structures and monitor various fine cellular dynamic events including neuronal branching and connectivity rearrangement, lamellipodia movements and cellular trafficking. We are working to reach a higher frame-rate to image faster-acting cellular phenomena like neuronal excitation.

At Langevin Institute, ESPCI Paris, we have developed a label-free approach to detect and track viruses in aquatic solutions using an interferometric microscopy approach working in transmission. By improving the experimental setup design, we have increased the sensitivity to detect viruses as small as 20nm in diameter, AAV viruses, which are of paramount importance for gene therapy. Furthermore, based on the variation of the interferometric signal along the axial position, we introduce a new way to perform single-shot volumetric imaging which enables high-precision 3D tracking.

Hundreds of thousands of people are diagnosed with bladder cancer yearly, with recurrence rates of 60% in the first year. We propose a new label-free imaging technique for noninvasive and automated individual cell processing with high discriminating power in detecting cancer cells in urine samples. We analyzed urine samples from bladder cancer patients by acquiring holograms of cells during flow. We then extracted highly discriminative features and classified the cells to their types. This noninvasive label-free technique allows us to monitor and diagnose cancer progression from a simple urine sample and has the potential to substitute the invasive cystoscopy procedure.

We have integrated tissue quantitative phase imaging with Raman spectroscopy, and used it for analyzing benign and malignant cancer tissue samples without staining. We used the acquired stain-free OPD-based texture features of the tissues and deep learning to segment the urothelium layers, where cancer starts from, with the stained tissue as a ground-truth for training. Then, location-guided Raman spectroscopy measurements were acquired from the urothelium according to the segmentation results. We then classified the tissue type based on the location-guided Raman spectroscopy measurement with higher accuracy compared to classifying them without urothelium segmentation.

We discuss label-free high-throughput phase imaging, allowing the identification of platelet-platelet microaggregates and their components at single-platelet resolution in COVID-19 patients. The sample preparation-free measurement can be performed in 2 minutes with a benchtop-sized flow cytometer from a diluted blood sample matching requirement for point-of-care-testing in the ICU. Most importantly, the microfluidic solution allows for low shear stress to minimize loss of microaggregates yet achieving high throughput. We validated a positive correlation of (1) increased platelet-platelet concentration, (2) composition, and (3) size distribution with severity of COVID-19. Most importantly, the microaggregates we observe are in fact in a size regime of a single leukocyte, which allows for a sufficiently long circulation time in the peripheral blood system for quantification.

Label-free detection techniques are widely used in biological and environmental sensing. In particular, whispering gallery mode microtoroid resonators have been used to detect single nanoparticles and molecules by relying on the ultrahigh quality factor of these resonators together with frequency locking for ultra-precise resonance tracking. However, material property characterization on a particle-by-particle basis remains a challenge in these label-free schemes. Here we characterize single particle material properties using microtoroid optical resonators combined with spectroscopic techniques.

This conference presentation was prepared for SPIE BiOS, 2023.

Photoacoustic imaging (PAI) is a technique for detecting ultrasonic signals occurred from the optical absorption properties of biological tissues. The contrast of PAI is determined by the degree of optical absorption properties of biological tissues. PAI uses absorption contrast to distinguish endogenous factors to image biological tissues. Among various endogenous contrast agents, a lipid with high absorption at 1.7 μm is one of the main agents for identifying diseases such as arteriosclerosis and myocardial infarction. We present a high-speed wavelength-switchable active mode-locked Raman fiber laser using active mode-locking (AML) and stimulated Raman scattering (SRS) in a 1.7 um band for photoacoustic generation of lipids. the wavelength of 1.7 um band is oscillated to generate high pulse energy on the order of hundreds of nano-Joules. PA signals of lipids can be acquired to prove the feasibility of label-free PAI of lipids.

Photoacoustic imaging (PAI) is a label-free biomedical imaging technique based on photoacoustic effects. PAI can differentiate various tissues by the different absorption coefficients of each molecule. Therefore, the wavelength band around 1700 nm is advantageous for finding lipid tissue because lipids have a higher absorption coefficient than water at around 1700 nm. In this paper, we demonstrate a stimulated Raman scattering (SRS) pulsed fiber laser with a variable repetition rate for PA generation in 1700 nm. The pulse repetition rate and the pulse energy are inversely proportional under the given optical gain. Therefore, by utilizing a pulse laser with variable repetition rate, we can determine the optimal pulse energy and measurement speed required by the system.

Histological analysis is crucial for the diagnosis of a wide variety of diseases. However, labelling of thin tissue sections can alter tissue chemistry and is greatly influenced by pre-analytic variables. Furthermore, biopsies provide a limited number of tissue sections and thus stains must be used sparingly. Total Absorption Photoacoustic Remote Sensing (TA-PARS) is an all-optical and label-free technique capable of capturing both radiative and non-radiative endogenous contrasts in cells, tissues and biomolecules. Here we present an automated full-slide TA-PARS scanning system capable of providing label-free virtually stained whole slide images with sufficient resolution (~300nm) to recover subcellular diagnostic characteristics.

We present photoacoustic remote sensing and autofluorescence microscopy with surface excitation (PARS-AMUSE) system that combines several methodologies of optical imaging including absorption, scattering and radiative emission. We generate virtual hematoxylin and eosin (H&E) images from absorption and scattering data using a generative adversarial network for margin assessment and cancer grading with simultaneous metabolic measurements from the autofluorescence of NADH and FAD. This system utilizes a single excitation source for all three imaging modalities to minimize complexity and cost. Using this system, we aim to profile specific cancers and predict cancer aggression.

Raman spectroscopy is expected as a non-invasive and effective method for the accurate identification of peripheral nerves. However, the discrimination basis of the Raman spectroscopic method is sometimes ambiguous due to the partial and complicated information of tissue molecules reflected in Raman spectra. In this study, we developed a method for identifying spectral features in Raman spectroscopic detection of peripheral nerves by utilizing a support vector machine (SVM). Raman spectral features for the discrimination of tissue species were extracted by analyzing the feature weight obtained from the linear SVM classifier.

Current methods for cancer screening are invasive and not accurate in the early stages of the disease. A revolutionary development in the diagnosis of cancer is the analysis of liquid biopsies via Raman spectroscopy. We propose a method for the screening and prognosis of cancer combining Raman spectroscopy with decision algorithms. We tested the efficiency of our method probing the spectra of blood plasma samples from cancer patients and healthy controls, proceeding with a Support Vector Machine classification and a leave-one out cross-validation, obtaining sensitivity and specificity results higher than 90%.

Fiber-based laser sources for Coherent Anti-Stokes Raman scattering (CARS) microscopy and Stimulated Raman Spectroscopy (SRS) have been successfully used for label-free bio-imaging in the past. We present an improved design of a dual-wavelength source for CARS and SRS enabling a wider frequency-range from <630cm^-1 up to 2250cm^-1 while the tuning time remains in the range of a few seconds over the full range. The system is fiber-based and thus resistant to detrimental thermal effects even at output powers in the Watt-range with improved tunability of Stokes and pump power. Repetition rates up to 10MHz are possible and the spectral width of the signals is <15cm^-1 enabling most applications.

We present our recent developments in utilizing a compact and portable light source for high-speed multicolor stimulated Raman scattering imaging (SRS) in biomedical and medical environments. The source combines a rapid and wide tunability for accessing Raman bands between 700 and 3300 cm-1 with high stability in terms of power (deviation < 0.3 %) and wavelength (deviation < 0.5 pm) over more than 100 h. We were able to shorten the Stokes pulse duration to below 3ps, resulting in a twofold increase in SRS signal intensity compared to our previous version (7ps).

We report label-free, in vivo virtual histology of skin using reflectance confocal microscopy (RCM). We trained a deep neural network to transform in vivo RCM images of unstained skin into virtually stained H&E-like microscopic images with nuclear contrast. This framework successfully generalized to diverse skin conditions, e.g., normal skin, basal cell carcinoma, and melanocytic nevi, as well as distinct skin layers, including the epidermis, dermal-epidermal junction, and superficial dermis layers. This label-free in vivo skin virtual histology framework can be transformative for faster and more accurate diagnosis of malignant skin neoplasms, with the potential to significantly reduce unnecessary skin biopsies.

We demonstrate a high-speed confocal-based interferometric scattering (iSCAT) microscopy for label-free live cell imaging. Using a spinning disk scanning unit, our wide-field iSCAT confocal microscope reaches a high image acquisition rate of 1,000 frames per second. Nano-sized objects, as small as 10 nm gold nanoparticles, are directly visualized by the iSCAT confocal microscope in scattering. With the high spatiotemporal resolutions and the high sensitivity, we explore the complex organization and dynamics in live cells.

In this study, the SERS technique was used to detect Thiram, Ferbam and Ziram in Dithiocarbamate by electrochemical cyclic voltammetry. The SERS substrate was prepared first. It was found that the best SERS effect was obtained when using 1 M sulfuric acid as the electrolyte and scanning for six times. The growth of nanoparticles is very uniform, and the SERS substrate using this condition can detect 5 ppm of Thiram, 10 ppm of Ferbam, and 5 ppm of Ziram, respectively. Since it can be seen that the Raman spectra of the Dithiocarbamate pesticides are similar, in this study, with the assistance of artificial intelligence, the Raman spectra combined with artificial intelligence and machine learning algorithms were analyzed, and the prediction accuracy reached 96.7% respectively. With a prediction accuracy of 95%, it proves the feasibility of combining SERS substrates with artificial intelligence, and develops a fast, accurate and inexpensive pesticide detection system.

The intraoperative assessment of tumor margins of head and neck cancer is crucial for complete tumor resection and patient outcome. The current standard is to take tumor biopsies during surgery for frozen section analysis by a pathologist after H&E staining. This evaluation is time-consuming, subjective, methodologically limited. Optical methods like hyperspectral imaging (HSI) are therefore of high interest to overcome these limitations. We present an approach, that enables delineation of tumor margins with label-free HSI-based histopathological information during surgery using deep learning. We show accuracy on par with traditional intraoperative tumor margin assessment on a data set of seven patients.

Gastric cancer is a deadly disease afflicting millions annually, with overall poor prognosis due to lack of effective detection and localization tools. Optical coherence elastography is a promising technique for assessing mechanical properties of tissue, which could be valuable for early detection of stomach cancer, which can influence the stiffness of tissue. We evaluate the suitability of OCE for assessing mechanical properties of tissue for clinical specimens of normal and stomach cancer, finding that the stiffness is increased by up to twofold in cases of cancer. These results show promise for validating OCE as a tool for localizing stomach cancer.

The zebrafish has shown to be an essential preclinical animal model, especially in the field of oncology. A non-invasive, high-resolution, and three-dimensional imaging modality is required to identify disease related changes in this model organism. The presented polarization-sensitive Jones matrix optical coherence tomography (JM-OCT) prototype was utilized in three different studies, covering in vivo imaging of wildtype zebrafish, an adult postmortem tumor model and a longitudinal xenograft tumor zebrafish investigation. This work highlights the potential of JM-OCT as a non-invasive, label-free, and three-dimensional imaging tool for preclinical cancer research based on zebrafish models.

We present a new dynamic full-field optical coherence tomography (D-FF-OCT) imaging technique, which is self-referenced, vibration-insensitive and overcomes D-FF-OCT limitations near specular surfaces, named interface self-referencing D-FF-OCT (iSR-D-FF-OCT). In fibroblasts, retinal organoids, and corneal explants, we demonstrate imaging in the vicinity of reflective interfaces without fringe artefacts, in a live, label-free, non-invasive manner with sub-diffraction limited sectioning capacity within 3D samples. iSR-(D)-FF-OCT has a homogeneous contrast across the imaging field enabling flawless mosaicking. iSR-(D)-FF-OCT complements and is easily implemented on existing D-FF-OCT set-ups to cover a zero to several hundred micrometer depth range.

We introduce volumetric differential contrast (VDC) imaging using optical coherence tomography (OCT). This method was designed based on a new theoretical model of OCT, the disperse scatterer model (DSM). VDC gives the differential image of “the product of the refractive index distribution and the scatterer density” through complex numerical processing of OCT signals. The method was validated by in-vitro and in-vivo samples measured by spectral domain OCT. Differential contrast images with arbitrary shear amount and shear direction were obtained at arbitrary depth positions by a single measurement by numerically applying defocus by holographic signal processing after the signal acquisition.

In this study, an ultrahigh resolution OCT(UHR-OCT) system is constructed, with 1.5 um axial resolution and 1.9 um lateral resolution in enamel. Despite the inverse relationship of clinical process, early-stage dental caries and calculus are both imaged as bright zones on the top of enamel yet cannot be distinguished in conventional OCT. The results indicate that early-stage caries and calculus can be detected and distinguished clearly in our developed UHR-OCT. We also compared the UHR-OCT images with the results obtained from commercial OCT to clarify the system performance. This work shows the potential for using UHR-OCT in early-stage dental detection.

To collect the volumetric image of the spinal cord using optical coherence tomography, the traditional way is to successively repeat the process of cutting the sample and imaging the revealed part. This common method is quite lengthy and laborious. Here, we designed circular scanning optical coherence tomography that has a customized ring-shaped silver-coated mirror reflecting incoming light in a circular pattern around the spinal cord. The sample is carried through that mirror by the linear stage during image acquisition. The results suggest that our system is suitable for the study of the spinal cord as well as applicable to other bigger samples when adapting clearing methods.

We introduce cell dynamic activity analysis method-based combination of dynamic full-field optical coherence tomography (DFFOCT) and machine learning (ML) models. DFFOCT can monitor intracellular migration label-free by capturing scatters movement inside of cells. Since ML builds classification criteria through learning a lot of data, based on the intracellular scatter migration observed through DFFOCT, it is possible to judge abnormal signs of cells regardless of changes in the external experimental environment. We compared the suggested analysis method and staining analysis method for the change of state of HeLa cells (including cell data) and verified the validity.

This conference presentation was prepared for the Label-free Biomedical Imaging and Sensing (LBIS) 2023 conference at SPIE BiOS, 2023.