General Electric Receives NIH Grant for Multi-Scale 3-D Imaging Analytics for High Dimensional Spatial Mapping of Normal Tissues
General Electric Receives a 2019 NIH Grant for $587,413 for Multi-Scale 3-D Imaging Analytics for High Dimensional Spatial Mapping of Normal Tissues. The principal investigator is Yousef Al-Lofahi. Below is a summary of the proposed work.
The overall goal of the proposed project is to develop open-source software and algorithms for 3-D reconstruc- tion and multi-scale mapping of normal tissues. Another significant goal is to evaluate effects of aging and envi- ronmental factors on molecular and structural architecture of skin. We will leverage our mature (TRL8) technol- ogy for multiplexed 2-D imaging (Cell DIVE™), and our vast experience in 2-D image analytics and machine learning. We have selected normal skin as the organ to develop these tools for several reasons, a) clinical sam- ples from different age groups are more readily available, b) it is a good model to independently capture changes in extracellular matrix (ECM) due to age and normal exposure to environmental factors as well as a variety of pathogenic insults. While the ECM, cellular and intracellular molecular composition varies considerably among various organs, we believe many of the tools developed under this program will be applicable to reconstruct and map other organ models at high (cellular/subcellular) resolution. This proposal will focus on developing algo- rithms and a framework for multi-scale mapping of 3-D tissue images, which will address HuBMAP priorities around quantitative 3-D image analysis/mapping, including automated 3-D image segmentation, feature ex- traction, and image annotation. High-resolution (subcellular) mapping of biomolecules will be implemented us- ing 2-D multiplexed images that are used to reconstruct the 3-D tissue and linked to a lower resolution 3-D optical coherence tomography (OCT) image of the normal tissue. Other cell-level omic data (e.g., RNA FISH) will be mapped in the same way. The low-resolution image is mapped back to a higher-level landmark (e.g., organ) as defined by the HuBMAP common coordinate framework (CCF). As outlined, our proposed technologies will in- clude several key features that are significant and complimentary to existing HuBMAP consortium projects and will advance the state of the art in 3-D tissue analysis. The proposed algorithms will have several key innova- tions that will advance the state of the art in 3-D multiplexed tissue image analysis. First, given the large vol- umes to be analyzed, high throughput will be a key requirement of each image analysis algorithm. This will be supported by our extensive experience in parallelizing single cell analysis pipelines. Second, the proposed algo- rithms will segment the images at multiple scales. The third area of innovation will focus on efficient multi- channel analysis. The proposed project will include creation of an easy-to-use software tool for assembling and visualizing multiscale tissue data called Tissue Atlas Navigation Graphical Overview (TANGO) .