Deep Learning-Based Cellular Activity Recognition in Live-Cell Imaging

Abstract

Over the past two decades, two-photon intra-vital microscopy (2P-IVM) has emerged as the gold standard technology for the real-time investigations of interactions between host immune defenses and pathogens. However, despite its significance, analytical tools dedicated to this platform remain scarce, with quantitative measurements primarily relying on cell trajectory analysis. The aim of this doctoral thesis is to address this gap by providing a comprehensive analysis of data generated in vivo. The main objective of this work is the development of novel cell activity recognition (CAR) tools to distinguish and quantify cellular behaviors related to the immune system and cell death. To achieve this goal, I explored the application of deep learning architectures designed for natural language processing and computer vision tasks. Simultaneously, I curated and made public datasets for the application of supervised learning techniques. Finally, I provided a biological validation confirming the usability of the generated CAR tools in the context of the immune system. In conclusion, this doctoral thesis presents a pioneering approach to measuring cellular behavior. In the future, the development of CAR tools will enable the prediction and characterization of complex cellular activities, contributing to the advancement of fundamental research and therapeutic strategies.