Fig. 3: DeepSNiF enhances automated cell phenotyping on human bone marrow IMC data.

a t-SNE plots of DIMR and DeepSNiF with cell phenotyping results. b The relative change in cell phenotypes before and after DeepSNiF. c, d Comparisons of DIMR and DeepSNiF-processed IMC images labeled with different cell markers, and the corresponding cell annotation results. The sub-panels (i)–(iv) in c and the bottom row in d correspond to the white dashed box region selection in their first panels, respectively. The white contours represent the differential phenotyping results between DIMR and DeepSNiF. e DeepSNiF enhances the sensitivity of cell phenotyping. After DeepSNiF processing, the non-specific marker signals reduce while the specific ones enrich in the cell types, respectively. The circle size indicates the positive marker percentage in a particular phenotype of DIMR, and the circle color indicates the relative changes of the positive rate for the particular markers after DeepSNiF enhancement. f DeepSNiF enhances the specificity of cell phenotyping. With DeepSNiF denoising, the ratios of specific phenotypes increase while those of non-specific phenotypes decrease in the positive markers. The relative change is the difference in percentage composition of each cell type before and after DeepSNiF enhancement. Scale bar: c 110 μm. d Top: 145 μm, bottom: 50 μm.