CytoBERT: A Foundation Model for Cytometry Data
CytoBERT is presented as an open-weight foundation model meant to transfer across messy, mismatched cytometry datasets.
The model was pretrained self-supervised on 15 human cytometry datasets covering more than 50 million cells. The authors say marker standardization lets it handle variable marker panels and learn inter-marker relationships within cells. Fine-tuning for sample-level classification is used to show transfer learning across heterogeneous cytometry data is feasible. Code is available, according to the paper. ArXiv · AI/CL/LG's note
The model was pretrained self-supervised on 15 human cytometry datasets covering more than 50 million cells. The authors say marker standardization lets it handle variable marker panels and learn inter-marker relationships within cells. Fine-tuning for sample-level classification is used to show transfer learning across heterogeneous cytometry data is feasible. Code is available, according to the paper. ArXiv · AI/CL/LG's note
score 6