Alberto Diaspro

Alberto Diaspro

Dept.Physics, University of Genoa - Nanoscopy, Italian Institute of Technology

Advanced fluorescence light microscopy deals with designing and developing optical microscopes able to produce images that are rich sources of quantitative information at the nanoscale. This fact offers an unprecedented large amount of data about the molecular mechanisms that govern and determine the fate of living cells. Like in the Plato’s cavern we see on a wall the results of light-matter interactions in the real world. In this scenario, multimodal optical microscopy is a growing attitude boosted by artificial intelligence. The microscope enables the realization of a kind of artificial environment that offers potential advantages in analysing data. It is like an "infinite" space where visualising data. In the era of super-resolved fluorescence microscopy, considering the lessons of Giuliano Toraldo di Francia and Gregorio Weber, the photochemical parameters, from brightness to lifetime, can be merged with the ones coming from label-free approaches like Mueller matrix microscopy. Here, we report on the potential of convolutional neural networks (CNN) and independent component analysis (ICA) algorithms in the domain of machine learning and deep learning with the ambitious target to create a robust virtual environment "to see "what we could not perceive before. We will report about the development of a multimodal optical microscope reinforced by the artificial intelligence components to address a fundamental problem in biology: how does chromatin organization in the nucleus rule the compaction and function of the human genome in the interphase of cells and mitotic chromosomes?  Our approach is in tune with the point raised by the photographer Gianni Berengo Garden having in mind the difference between a photography and an image:” You walk into a place, you see there could be a great picture, but something is missing. If that something comes, it can be a good photo - if not, it's nothing.”