Table 2.

Highest ranked morphological features for lymphocyte classification. The table shows the most important features used by gradient boosting for lymphocyte classification using random undersampling

Feature	Channel
Std intensity edge	Brightfield
Lower quartile intensity	Brightfield
MeanFrac Radial Distribution 4of4	Brightfield
Mean intensity	Brightfield
Integrated intensity edge	Darkfield
Granularity 1	Brightfield
FracAtD Radial Distribution 4of4	Brightfield
Granularity 1	Darkfield
DifferenceVariance Texture 3_0	Brightfield
Granularity 3	Brightfield

The first column contains the feature names and the second column contains the associated channel. Features were measured in the entire cell (no subcompartments of cells were defined). The features were sorted by their frequencies in 10 random undersampling test runs.