Micro-Optical Coherence Tomography (μOCT) enables non-invasive, label-free cross-sectional imaging with subcellular resolution and extended depth of focus. However, μOCT provides limited sensitivity and specificity to the molecular and genomic signatures. The combination of µOCT and Fluorescence Confocal Microscopy (FCM) would improve our capability to interrogate biological specimens. We built and characterized a µOCT-FCM that enables colocalized imaging of µOCT and fluorescence signals. Depth scanning in the fluorescence channel was enabled by using an electronically tunable lens. We report the optical design, system-level integration, and characterization, and demonstrate the quantification of pH of airway mucus in a spatially resolved manner in cell cultures and ex-vivo swine tissue.

Sudden cardiac death being the most likely reason of mortality in western countries is associated with preliminary occurring myocardial infarctions causing variations in structure, texture, metabolism and molecular composition. Common therapy methods lacking sufficient resolution for precise differentiation between fibrotic and physiological cardiac tissue. Novel multimodal diagnostic approaches are required and explored with our developed ultrahigh-resolution multimodal optical imaging platform including optical coherence tomography, multiphoton microscopy and line scan Raman spectroscopy in a label-free manner at cellular resolution. Co-registration of all modalities and integrated multiparametric analysis and radiomics provide complementary morphological, metabolic and molecular contrast for unprecedented cardiac tissue classification.

Multimodal imaging that includes optical coherence tomography and a secondary imaging modality in small single-channel endoscopes is often implemented using double-clad fiber (DCF). Unfortunately, the properties of DCF cladding modes generate multipath OCT artifacts degrading image quality. Curiously, the en face mean intensity projection of these multipath artifacts is a high quality image. The differential scattering of en face projections from the image and artifact could be used as an additional imaging modality, sensitive to sub-resolution features. Multipath artifacts are inherent to DCF-based OCT, meaning a wealth of previously acquired data could be explored using this technique.

The triple-negative breast cancer is an aggressive subtype that has a high rate of relapse and a poor five-year survival rate. The Tumor Microenviroment influences the behaviors such as proliferation, migration and the formation of metastasis. We present the complementary imaging and subsequent analysis of the Tumor Microenviroment using Lightsheet Microscopy, Scanning Laser Optical Tomography and TPEF. The samples were imaged in different size and resolution scales and the three-dimensional information is obtained using the SHG-signal of collagen and fluorescence of labeled key markers. The results presented here may help to create a patient specific model of breast cancer.

Esophageal cancer has a low survival rate, which is significantly improved through early detection, as well as monitoring of a common pre-cursor, Barrett’s esophagus. Optical Coherence Tomography is a low coherence interferometry technique which produces three-dimensional depth scans of tissues. OCT provides morphology but lacks in molecular specificity, and can thus be combined with targeted Near Infrared Fluorescence imaging. The viability of this dual-modality technique for use in detecting BE and esophageal cancer is assessed using topically and intravenously administered Bevacizumab in combination with an OCT-NIRF system, to produce ex vivo tissue scans, which are then histology matched.

We hypothesize that there may be cancer-sensitive image biomarkers present in a novel image processing technique. We leverage the multipath artifacts derived from higher order modes that present in double clad fiber-based OCT systems, which are sensitive to scattering angle and as such may be sensitive to sub-resolution features such as nuclear density and size. This work explores multipath contrast in previously collected clinical imaging data; preliminary work has found that this technique can distinguish cancerous and non-cancerous fallopian tube specimens.

We previously introduced Photo-Magnetic Imaging (PMI) as a true multi-modality imaging technique for non-invasive, high resolution optical tomography. With a simple add-on of continuous wave lasers to an existing MRI system, PMI converts the laser induced temperature increase measured by MR thermometry into tissue optical absorption map utilizing a dedicated Finite Element Model (FEM) based image reconstruction algorithm. The newly implemented multi-wavelength capabilities allow PMI to recover 3D tissue oxygenation maps as well as exogenous contrast agent distribution such as indocyanine green or gold nanoparticles with high-resolution and quantitative accuracy.

Current report considers development of a real-time biophysically-based 3D rendering framework for the realistic visualization of translucent materials appearance such as human skin. Our numerical models achieve simultaneous solution of the Radiative Transport and Rendering Equations for complex multilayered materials for the first time using two stochastic Monte Carlo based techniques. Our realizations utilize a combination of two powerful techniques working in synergy: Compute Unified Device Architecture (CUDA) and Real-Time Raytracing (RTX) using specialized NVIDIA Graphics Processing Units (GPUs). We present live interactive demos and comprehensive examples of translucent material renderings in comparison with in vivo data obtained during clinical studies and previously developed offline rendering approaches.

The novel Time-Domain Mesoscopic Fluorescence Molecular Tomography (TD-MFMT) imaging modality is aimed to resolve intra-tumoral heterogeneity and monitoring target engagement at the mesoscopic regime; however, its spatial resolution is limited by the scan strategy employed for imaging. Imaging in the Short-Wave Infrared (SWIR) can improve resolution due to reduced scattering effect. We therefore implement data-fusion of low-resolution fluorescence lifetime images with high-resolution SWIR intensity images via deep learning (DL). This approach was validated through phantom study where results show resolution enhancement of fluorescence lifetime imaging in scattering media, demonstrating potential for expansion to preclinical studies.

We present a deep learning-based framework to virtually transfer images of H&E-stained tissue to other stain types using cascaded deep neural networks. This method, termed C-DNN, was trained in a cascaded manner: label-free autofluorescence images were fed to the first generator as input and transformed into H&E stained images. These virtually stained H&E images were then transformed into Periodic acid–Schiff (PAS) stain by the second generator. We trained and tested C-DNN on kidney needle-core biopsy tissue, and its output images showed better color accuracy and higher contrast on various histological features compared to other stain transfer models.

Label-free multimodal optical bioimaging allows non-perturbative profiling of biological samples based on their intrinsic optical molecular properties. In this study, we utilized SLAM and FLIM microscopy to identify CHO cell lines with favorable process performance for the production of therapeutic monoclonal antibodies and proteins. Here, a single-cell analysis pipeline was developed to quantitatively characterize CHO cell lines based on their phenotypes. To perceive the rich information in the multi-modal bioimages, a custom-built multi-task deep neural network was built, which can extract features from different aspects of the optical and molecular properties of the sample. This work demonstrated the potential of ML-assisted multi-modal optical imaging in the identification of cell lines with desirable characteristics for biopharmaceutical production at earlier time points.

The stochastic nature of 3-D Monte Carlo (MC) photon transport simulations requires simulating a large number of photons to achieve stable solutions. In this work, we explore state-of-the-art deep-learning (DL) based image denoising techniques, including the proposal of cascaded DnCNN and UNet denoising networks, aiming at significantly reducing the stochastic noise in low-photon MC simulations to achieve both high speed and high image quality. We demonstrate that all tested DL based denoisiers are significantly more effective compared to model-based denoising methods. In our benchmarks, our cascaded denoisier has achieved a signal enhancement equivalent to running 25x-78x more photons.

Screening colonoscopy is used to detect and remove lesions prior to progressing to colorectal cancer, but some lesions go undetected due to poor visual contrast in white light endoscopy. We present a retrofit clinical colonoscope capable of multispectral, topographic, and blood flow imaging for improving lesion contrast. We develop a custom fiber bundle to enable simultaneous illumination with commercial and research light sources. The research light source consists of nine wavelengths (405nm-659nm) for multispectral imaging and a high-coherence source for speckle-flow imaging. Point sources circling the image sensor are individually toggled to generate topographic maps with photometric stereo.

We propose an end-to-end reconstruction approach for Mesoscopic Fluorescence Molecular Tomography (MFMT) using deep learning. Herein, an optimized deep network based on back-projection with Residual Channel Attention Mechanism architecture is implemented to directly output 3D reconstruction from 2D measurements and diminish the computational burden while overcoming the limitation of the PC's memory during reconstruction. The network is trained by producing a large synthetic dataset through Monte Carlo simulation and validated with in silico data and a phantom experiment. Our results suggest that this approach can reconstruct fluorescence inclusions in scattering media at a mesoscopic scale.