Analysis of advanced optical microscopy data through Artificial Intelligent algorithms

Microscopy has emerged as an indispensable tool of observation and a gateway to the exploration of intricate biological landscapes. As the excitation and emission spectra inherently reflect the structure and composition of fluorescence molecules, fluorescence microscopy emerges not only as an instrument of observation but also as a tool to explore and quantify the intricate relationships between sub-cellular components that govern living organisms. In the biology field, explored with microscopy techniques, there is a compelling case for the application of cutting-edge technologies, particularly the potential of machine learning and inverse problem algorithms. This research seeks to not only overcome inherent limitations but also to pave the way for a paradigm shift in our understanding of biology through the lens of artificial intelligence. Recent deep learning-based methods have remarkably improved the quality of fluorescence microscopy images compared to classic denoising and deconvolution approaches. This PhD endeavour extends the reach of state-of-the-art deep learning and inverse problem methodologies to the domain of microscopic image analysis pipelines. For instance, in Single Molecule Localization Microscopy (SMLM), accurate background subtraction is crucial for precise localization. Modern machine learning algorithms can effectively exploit prior knowledge about the shapes of single molecules to enhance this process. Similarly, in Fluorescence Correlation Spectroscopy (FCS), conventional fitting methods may falter when applied to data with low signal-to-noise ratio. In such cases, leveraging informative inverse problem approaches that incorporate physical models and prior information can outperform standard methods, providing more robust and accurate results.