MMLL Kolloquium: GateNet: A novel neural network architecture for automated flow cytometry — Lukas Fisch
Flow cytometry is widely used in academic and clinical laboratories to identify cell populations in body liquids. Gating, the process of cell identification, which has traditionally been done manually is subject to automation. While many automatic gating tools have been developed in the recent past, most practitioners never used any of them since they still need significant input from the human operator. To overcome this, we propose GateNet, the first neural network architecture specifically designed to enable end-to-end automated gating. This architecture allows to train the neural network on labeled samples gated by experts and successively gating novel unseen samples without correcting for batch effects beforehand. GateNet reaches F-measures of up to 0.95 in the publicly available datasets used during the FlowCAP I (Critical Assessment of Population Identification Methods) challenges.
Der Vortrag startet um 13:15 Uhr und findet per Zoom statt:
Termin in lokalen Kalender importieren