Can Deep Learning restore the fast particle dynamics from noisy fluorescent images? Fadil Iqbal 

This video presentation was based on my Masters work at IUPUI in 2021. It was presented at YSCMR - organized by the National Institute of Fundamental Studies (NIFS), Sri Lanka
Abstract:
Single Molecule Tracking (SMT) is an imaging-based experiment to trace the motion of a molecule/particle under a microscope, and knowing the exact position of the molecule bears answers to fundamental research questions in cell biology and molecular biology. In order to track these fast-moving molecules one needs to go for higher frame rates - which in-turn reduces exposure time of the camera and thereby reduces signal to noise ratio (SNR) in the images. This makes the tracking impossible. Recent advancements in Deep Learning have shown great progress in restoring noisy fluorescent images, but it wasn’t clear if it is applicable on SMT movies. We intend to evaluate the performance of two Deep Learning packages - one that uses a supervised approach and the other, an unsupervised approach, on the SMT experiment. We applied them on the following examples: 1. A synthetic bead model where noise can be added artificially - we were able to restore low SNR movies and restore their individual trajectories by both supervised and unsupervised methods. 2. Experimentally we applied these techniques on nucleosomes and chromatin microdomains, to verify if deep learning can restore stationary and dynamic data. It was noted that the supervised approach delivered better results. However there are instances where these algorithms produce deceiving results, despite ‘looking good’ to the naked eye. We anticipate broad application of this approach to critically evaluate artificial intelligence solutions for quantitative microscopy.
IQ Labs covers a wide range of topics. Please check my playlists - I try my best to keep things organized.
PhDs and Scholarships
• PhDs and Scholarships ...
Research Advice
• Research Advice (Sinhala)
Communication Advice
• Communication Advice (...
#ScienceTalks
• Science Talks
Python for Scientists
• Python for Scientists
How I make these videos? Check this video:
• How to make Teaching V...
My equipment:
• අපිට හරියන Graphics Ta...
#PhDAdvice #SriLanka #Research #Sinhala #Science #Engineering #ResearchAdvice