Josef Lorenz Rumberger: Improving cell phenotyping in multiplexed imaging data by automated classification of cellular marker expression patterns
-
Trained on a vast dataset of 197 million noisy annotations, a U-Net classifies cells into marker positive / negative for a variety of marker proteins, tissue types and imaging platforms without requiring retraining for each new dataset. The output of the model integrates with popular clustering algorithms used for cell phenotyping and can simply be dropped into existing pipeline. Github: github.com/angelolab/Nimbus-Inference.
Venue
MDC (BIMSB)
Hannoversche Straße 28
Room 1.04
10115 Berlin
Germany
Time
-
Organizers
Bioinformatics Social Meetings Berlin