Technology Research and Development Projects

TRD 1: Quantitative phase imaging with computational specificity (PICS) for in-vivo deep-tissue imaging

Gabriel PopescuInvestigator: Gabriel Popescu

TRD 1 aims to translate QPI technology to in-vivo and deep-tissue imaging with specific markers developed via computation and deep-learning. Quantitative phase imaging (QPI) is emerging as a powerful, label-free approach to imaging cells and tissues, especially because it combines qualities found in microscopy, holography, and light scattering techniques: nanoscale sensitivity to morphology and dynamics, 2D, 3D, and 4D (i.e., time-resolved tomography) nondestructive imaging of completely transparent structures, and quantitative signals based on intrinsic contrast. These capabilities have allowed QPI to be successfully applied in numerous biomedical applications, including cancer diagnosis in histopathology and cell therapy. Recently, we have expanded QPI for the first time to thick structures, such as embryos and spheroids, by developing gradient light interference microcopy (GLIM, the 2018 Microscopy Today Method of the Year). However, despite enormous progress, current QPI techniques are virtually absent from in-vivo and POC applications. We will advance the QPI technology to a confocal reflection geometry, thus, boosting the out of focus light rejection and improving high-resolution 3D imaging of thick tissue structures. Specifically, we will target first imaging the 3D orientation of skin collagen in-vivo. We will develop a label-free endoscopic system (eGLIM) capable of sub-micron spatial and millisecond temporal resolution, while maintaining nanometer pathlength sensitivity.

We will advance phase imaging with computational specificity (PICS) to real-time operation on in-vivo data from CPT (Aim 1) and eGLIM (Aim 2). Specifically, in close collaboration with TRD 3, we will develop computational tools for segmenting cellular and subcellular structures, as well as collagen fibers from in-vivo CPT skin data.

TRD2: Quantitative clinical and in-vivo imaging

Stephen BoppartInvestigator: Stephen Boppart, UIUC

It is essential that technological advances in imaging developed in the laboratory find direct translational paths to rapidly demonstrate their clinical utility in patients, and establish the potential for improving detection, diagnosis, and monitoring of disease. While label-based optical imaging modalities have demonstrated potential in intraoperative cancer detection, mapping the microvasculature, and site-specifically targeting of altered metabolism and pathology, to name a few, these approaches often come at a significant time and financial cost due to the associated safety risks and lengthy review processes required for FDA approval of any new contrast agent or probe. Importantly, any new targeted contrast agent or probe inevitably will have some degree of non-specific binding or off-target labeling, as well as a variable degree of uptake or labeling of the targeted cell or site. In the end, the measured or imaged signal levels are always questioned. Is the signal low because the targeted pathology is minimal, or because the targeted efficiency of labeling is low? The importance of label-free imaging is therefore high, and the need for label-free imaging across size scales is great. FIGURE: Multimodal label-free images (SHG, THG, CARS, 2PEF, 3PEF) and composite image of a fresh unperturbed mammary tumor from a carcinogen-injected rat. By identifying robust label-free signals or biomarkers that indicate changes in structure, molecular composition, metabolism, and function, quantitative clinical and in vivo imaging not only becomes a feasible alternative to label-based methods, but also provides a direct and rapid translational path to clinical human studies, since regulatory approval is not additionally needed for a contrast agent or probe. This enables rapid in vivo first-to-human and limited-scale human subjects research studies (and subsequently larger clinical trials) to be performed with the new optical imaging technologies, and make early determination of the clinical utility of the technologies and the new optical biomarkers that they generate. This TRD focuses on the technological development through four specific aims that progress from 1) the sub-cellular and cellular scale, identifying optical biomarkers and signatures that would indicate more systemic disease processes, to 2) tissue sections in an advanced digital pathology platform with artificial intelligence, to 3) computational optical imaging algorithms that extend the depth and performance of optical imaging in thick tissues, and finally to 4) engineered beam delivery systems to widely expand tissue access and application for these label-free optical imaging modalities. Collectively, this project will demonstrate novel technological advances that will find a myriad of applications to advance the biological and medicine sciences, and improve diagnostic and monitoring capabilities in clinical medicine.



TRD3: Computational imaging and intelligent specificity

Mark AnastasioInvestigator: Mark Anastasio, UIUC

In this technology research and development (TRD) project, advanced computational and machine learning methods will be developed that address a variety of needs related to image formation and image analysis in high-resolution label-free optical microscopy. Computational methods are being rapidly deployed that are changing the way that measurement data are acquired and improving the formation and analysis of microscopy images. The potential impact of such methods on the field of label-free microscopy is very high and can optimally leverage inherent endogenous contrast mechanisms in innovative and informative ways. The developed methods will serve as enabling technologies for many projects in the proposed center. The research will be informed by and jointly developed and evaluated with the TRD and driving biological projects. Figure: Results of Efficient U-Net on testing dataset. a. representative SLIM measurements of HeLa cells not used during training. b. The ground truth of viability of frames corresponding to a. c. The PICS prediction shows high level accuracy in segmenting the nuclear regions and inferring viability states. Scale bars: 50 microns A general theme of this work is the integration of imaging science, physics- and deep learning ( DL)-based approaches to circumvent the limitations of label-free phase-contrast imaging, and the use of objective image quality measures to systematically validate and refine the developed methods. Three broad classes of computational methods will be investigated that will enable the (1) image-to-image mapping of label-free images to provide computational specificity, improved semantic segmentation, and/or enhanced spatial resolution; (2) improved reconstruction of images for 3D cellular imaging; and (3) extraction of biologically relevant information from multi-modality label-free image data. The Specific Aims of the project are: Aim 1: Image-to-image translation methods for providing specificity, semantic segmentation, and/or enhanced spatial resolution; Aim 2: Diffraction tomography and inverse scattering methods for 3D imaging; Aim 3: Biomarker discovery and multi-modal DL methods; and Aim 4: Objective image quality (IQ) assessment, refinement of the methods, and reliable use.
This successful completion of this project will result in computational and DL methods that will advance a variety of label-free imaging technologies. These methods will enable improved computational staining, enhance of spatial resolution, semantic segmentation, 3D image formation, and analysis of multi-modality label-free image data. They will be systematically validated for use in the biomedical applications that are within the purview of the proposed P41 center. All source code, trained models and documentation will be made open-source and shared online.