Imaging Devices for Healthcare Applications

COVID-19 Detector

Large-scale testing is one of the top strategies to prevent the spread of COVID-19; however, its application is limited by the cost and processing time of the standard Polymerase Chain Reaction (PCR) tests. New rapid saliva-based tests promise to overcome those limitations. We collaborated with the group of Prof. Peter Preiser (SBS, NTU), who work on such a rapid test, which can give results on-site in 5 minutes or less. Our contribution lay in the development of an easy-to-use reader to evluate changes in colour of the test strips and, by using artificial intelligence, to provide the answers whether the tested subject was COVID-19 positive or not.

Team: Peter TÖRÖK, Josep RELAT GOBERNA, Sean KRUPP
Collaborators: Peter PREISER (NTU), Chen Change LOY (NTU)

AI in Image Analysis

Deep-learning analysis of bacteria behaviour in multi-species communities

We aim to time-resolve how multi-species bacteria communities colonize wet surfaces with single-cell resolution. During early-stage biofilm formation, we use continuous live-cell imaging to capture every cell event as they land, spin-walk-swim, divide, leave or explore around the surface as they meet other same-species or different-species cells, all the while leaving extracellular materials and signals along their trails. With full spatial-and-temporal causal understanding behind how biofilms develop, there is a transformative potential to untangle the myraid complex correlations observed in microbiology and genomics studies. We hope to tap this to study how bacteria infections fundamentally develop. We collaborate with Professor Gerard Wong at UCLA Bioengineering, whose team pioneered the field of quantitative bacteria tracking for single species. Our main challenge is to apply this rigor to multi-species communities where bacteria species may look identical but behave differently together as they coexist or fight against each other, giving rise to unexpected emergent patterns at very different timescales. The bacteria species have to be genetically engineered to express different fluorescent reporters. Continuous flowcell environments are inoculated with bacteria and continuously imaged under bright-field with high-temporal frequency and intermittent fluorescent imaging, and we develop deep learning techniques to help us tackle the huge terabyte datasets of detailed bacteria cell activity on the surface.

Team: LAI Ghee Hwee
Collaborators: Gerard Wong (UCLA)

BACK TO TOP