Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

BP: Fachverband Biologische Physik

BP 7: Cell Mechanics, Cell Adhesion and Migration, Multicellular Systems

BP 7.2: Vortrag

Donnerstag, 30. September 2021, 15:30–15:45, H6

Traction force microscopy with invertible neural networks — •Johannes Blumberg¹, Timothy Herbst^1,2, Ullrich Koethe², and Ulrich Schwarz¹ — ¹Institute for Theoretical Physics and Bioquant, Heidelberg University — ²Visual Learning Lab, IWR, Heidelberg University

In traction force microscopy (TFM), the mechanical forces of cells adhering to an elastic substrate are estimated from the substrate displacements as measured by the movement of embedded fiducial marker beads. While the direct problem of calculating displacement from forces is well-defined by elasticity theory, the inverse problem of reconstructing forces from displacements is ill-posed. Usually an estimate is obtained by minimizing the mean squared distance between experimentally observed and predicted displacements. The standard method in this regard is Fourier Transform Traction Cytometry (FTTC), whose superior efficiency is based on the convolution theorem in Fourier space. Here we explore if the performance can be improved by using machine learning methods, in particular invertible neural networks, which recently have emerged as powerful method to solve ill-posed inverse problems.