Transformers Research in Medicine

Cytometric gating is a core step in immunology analysis, but it’s time-consuming and variable across operators. As a research assistant in Professor Li Shen's lab at the University of Pennsylvania, I'm working on a follow-up of my research mentor's recent paper, which showed that cytometric gating, which is essentially an image segmentation problem, could be automated effectively using a UNet model partly due to its property of translational invariance.

In our current project, we are exploring whether transformer-based models can provide similarly impressive results. So far, I've reviewed and validated ~4,000 gated cell populations across 135 cytometry files using OMIQ to use as training data for the model.