Figure 2.

Label-free subcellular structure identification and cellular and functional classification using AIM. (a) CellNet^Nuc identified cell nuclei from DIC images of MDA-MD-231 and SK-BR-3 cells. (b) The CellNet^Nuc performance was evaluated based on recall, BF score, and IoU. Demonstrations of (c) CellNet^Mito and (d) CellNet^Actin for mitochondrial and actin fiber identification from DIC images, respectively. ClassNet identified the (e) cell type, (h) cell viability, and (k) maturation stage of dendritic cells. The ClassNet classification performance was evaluated using (f,i,l) confusion matrixes and (g,j,m) receiver operating characteristic curves (ROCs) for cell type (f,g), cell viability (I,j), and maturation stage identification for dendritic cells (l,m). Scale bar: (a,c,d) 100 µm, (e,h,k) 20 µm.