This topic came up during the meeting, and it would be valuable to evaluate how the on- and off-morphological features differ before and after QC.
We can use the CFReT dataset for this comparison and assess how the feature space changes across these conditions.
Proposed steps:
- Find and download the non-QC’ed dataset
- Generate a UMAP embedding to visualize differences between pre- and post-QC data
The goal is to assess how Buscar performs on lower-quality data, such as poorly segmented or out-of-focus cells, and to determine how these artifacts impact the on- and off- signatures and scores.