Artificial Intelligence Augmented Microscopy:

Deep Learning Enabled Digital Immuno-Fluorescence Labeling for Intelligent Cell Phenotyping (Aim-Dif)

FALL 2018 RESEARCH INCUBATION AWARDEE

PI: Lei Tian, Assistant Professor, Electrical and Computer Engineering
Co-PI: Yi Ji, Assistant Professor, Medicine

What is the Challenge?
The specific problem we aim to tackle in this project is the fundamental multiplexing limitation in immunofluorescence (IF) microscopy that restricts imaging only a few types of molecular targets on the same biological sample simultaneously. Our goal is to develop an AIM technology to enable highly multiplexed Digital IF labeling (AIM-DIF) that is scalable and promises intelligent, high-throughput cell phenotyping. 

What is the Solution?
The IF is one of the most broadly used labels/tags used in fundamental biological research. By labeling with specific antibodies and secondary fluorescent tags, distinct molecular signatures can then be visualized to differentiate subcellular structures, identify cell types, and characterize cellular functions and phenotypes. However, most commonly found fluorophores permits at most three colors (e.g., red, green, blue) to be imaged simultaneously due to the overlapping emission spectra in order to obtain more than three IF labels, sequential IF multiplexing scheme is required, which is labor-intensive and time-consuming. 

What is the Process?
To achieve this goal, our multidisciplinary team combines complementary expertise in computational microscopy, machine learning (PI: Lei Tian, ECE), biophotonics and biomedicine (co-PI: Yi Ji, BUMC). Our team has recently developed a novel multi-modal microscopy platform that enables simultaneously capture of IF and label-free multispectral images. This platform is an ideal testbed and will provide abundant data for the proposed AIM-DIF development.