Regensburg 2022 – wissenschaftliches Programm

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

BP: Fachverband Biologische Physik

BP 7: Poster 1

BP 7.41: Poster

Montag, 5. September 2022, 18:00–20:00, P1

Deep learning for single particle tracking in noisy data — •Mattias Luber, Mohammad Amin Eskandari, and Timo Betz — University of Goettingen, Goettingen, Germany

The quantitative analysis of particle motion critically depends on the quality of particle trajectory detection. Especially the position detection of particles in fluorescence microscopy images is an important task faced in biophysics. Trajectories are used to study processes like intra-cellular transport protein diffusion within and through membranes and the reconstruction of force fields driving the particle motion. In such settings, high spatial and temporal resolution are be desired. However, in practice those factors have contradictory measurement requirements. High temporal resolution requires short exposure times, which limit the photon budget and thus lead to low signal to noise ratios. We developed an approach to reconstruct the particle position from noisy images, by applying U-NET based deep learning models to fluorescence microscopy images. Using this we can successfully track particles with shorter exposure times, compared to traditional denoising techniques.