蜜桃传媒app

Machine-learning-assisted microscopy : from smart scanning approaches to the generation of synthetic super-resolution images

Flavie Lavoie-Cardinal, Laval University
Tuesday October 1, 12-1pm
McIntyre Medical Building, room 1034

Abstract:聽Super-resolution microscopy (or optical nanoscopy) techniques allow characterizing molecular interactions inside living cells with unprecedented spatiotemporal resolution. These techniques come with several layers of complex颅ity in their implementation. My research team focuses on transdisciplinary approaches at the interface of molecular neurosciences, multimodal optical nanoscopy, and machine learning to study structure/function relationship of synapses in the brain. We develop machine learning and deep learning tools to increase the adaptability and accessibility of high-end imaging methods (e.g. optical nanoscopy) to complex experimental paradigms. Recently, we implemented a machine learning assisted optimization framework for optical nanoscopy allowing real-time optimization of multi-modal live-cell imaging of synaptic activity and structure proteins. We also implemented diverse deep learning approaches for high throughput microscopy image analysis, allowing us to characterize activity-dependent remodelling of neuronal proteins. We develop weakly supervised deep learning strategies to reduce the burden of extensive labelling of complex images and evaluate how they can be applied to real-time microscopy image analysis. We aim at developing new AI-assisted microscopy techniques that will adapt in real-time to the sample, predict changes in the structures and modify the experimental protocol depending on the measured response to a stimuli.