Figure 4.

The top row (A–D) illustrates discrimination ability of DHM when a pixel intensity histogram is used to describe the phase map images. (A) Represents the PC loadings where the most significant contribution comes from number of zero intensity pixels indicating cell size is the biggest discriminating factor. (B–D) Represent PC scatter plots for all cell types using the first 3 PCs. The bottom row (E–H) illustrates discrimination ability of DHM when four texture parameters are used to describe the phase map images. Parameters 1–4 are contrast, correlation, energy and homogeneity respectively. (F–H) Represent the PC scatter plots for all cell types using the first 3 PCs. Each method forms a distinct cluster for monocytes indicating they are easy to identify using DHM, however B cells and CD4+ T cells appear to be more difficult to successfully differentiate. Using higher order PCs can see differences between the more closely related CD4+ and B cells as noticed in (C). Using a pixel intensity histogram appears to be more effective at discriminating between cell types than texture analysis.