A hierarchy of graph-based methods to study the behavior of immune cells in vivo

Decanato - Facoltà di scienze informatiche

Data: 19 Febbraio 2020 / 08:30 - 10:30

USI Lugano Campus, Room A-31, Red Building

You are cordially invited to attend the PhD Dissertation Defense of Diego Pizzagalli on Wednesday February 19th, 2020 at 08:30 in room A-31 (Red building).

The immune system has a critical role in the control of diseases of primary importance such as infections and tumors. Hence, it represents a target for novel therapeutic strategies. However, the immune system relies on a complex network of cell-to-cell interactions which remains largely unexplored and difficult to be interpreted. The combination of microscopy data with computational methods is a promising approach to study these interactions. Indeed, the recently developed 2-photon intravital microscopy technique (2P-IVM) allows capturing videos of cells while interacting in vivo, in organs of living organisms. These interactions are often associated with specific movement patterns. Hence, computer vision methods have the potential to extract biomedical knowledge from these videos by analyzing the movement of cells. Unfortunately, common methods for movement analysis poorly apply to 2P-IVM videos of leukocytes (the main cells of the immune system). This is mainly due to the complex appearance and biomechanical properties of these cells, which look like texture-less shapes moving with high plasticity and frequent contacts. Additionally, images are largely affected by artifacts. A lack of publicly available 2P-IVM datasets further hampers the development of better analysis methods, along with data-driven studies of the immune system. Finally, common measures of cell motility, poorly describe the behavior of leukocytes which is dynamic in space and time. In this thesis, we address these limitations by - Making available the first database of 2P-IVM videos and manually-annotated trajectories of leukocytes. - Modeling as graph the content of 2P-IVM videos, from pixels to biological processes. - Developing, refining, and applying a variety of computational methods to extract biologically relevant information from this graph. - Shifting the analysis of cell motility towards the recognition of cell actions. This combination of microscopy data, graph-based methods, and action-based models allowed us to analyze the dynamic behavior of neutrophils (one type of leukocytes) following influenza vaccination, unraveling three distinct phases associated with distinct biological functions.

Dissertation Committee:
- Prof. Rolf Krause, Università della Svizzera italiana, Switzerland (Research Advisor)
- Prof. Santiago F. Gonzalez, Università della Svizzera italiana, Switzerland (Research co-Advisor)
- Prof. Luca Maria Gambardella, Università della Svizzera italiana, Switzerland (Internal Member)
- Prof. Olaf Schenk, Università della Svizzera italiana, Switzerland (Internal Member)
- Prof. Matteo Matteucci, Politecnico di Milano, Dipartimento di Ingegneria Informatica, Elettronica, Biomedica (External Member)
- Prof. Marcus Thelen, Institute for Research in Biomedicine (External Member)