Subsampling Flow Cytometry Data Using t-SNE Binning

The patent segments the high-dimensional flow cytometric data space into multiple lower-dimensional subspaces using dimensionality reduction functions. Each subspace captures different aspects of the data, allowing selective subsampling from each subspace while collectively preserving rare cell populations across all subspaces. This segmentation enables efficient processing of each subspace independently while maintaining overall data integrity.

The patent transforms data from a higher-dimensional space to a lower-dimensional space using dimensionality reduction functions like t-SNE. This dimensionality change allows the data to be projected into multiple lower-dimensional subspaces where rare events can be captured more effectively, enabling subsampling that preserves rare cell populations while reducing computational burden.

The patent divides the dimensionality reduction process into multiple segments, creating several lower-dimensional subspaces rather than a single reduced space. Each subspace maintains specific relationships and patterns from the original high-dimensional data, allowing the system to preserve measurement precision for different event characteristics while collectively reducing overall data volume.

The patent changes the dimensional parameters of the data by applying dimensionality reduction functions that transform high-dimensional event data into lower-dimensional representations. This parameter change reduces data volume while the multi-subspace approach ensures that critical measurement precision is maintained by distributing information across multiple reduced subspaces.