deconvolution based perfusion analysis application  (Philips Healthcare)

Journal: Frontiers in Cellular and Infection Microbiology

Article Title: A synergistic multi-omics approach: causal sepsis drivers identified in activated CD4 + T cells by single-cell RNA sequencing and Mendelian randomization

doi: 10.3389/fcimb.2026.1749207

Figure Lengend Snippet: Immune infiltration analysis. (A) Relative proportions of immune cell subpopulations in the control and sepsis groups. (B) Correlation matrix illustrating the associations between different immune cell types. Red indicates a positive correlation, and blue indicates a negative correlation; the intensity of the color and the size of the circles correspond to the strength of the correlation coefficient. (C) Comparison of immune cell abundance between the control and sepsis groups. Blue represents the control group, while yellow represents the sepsis group. (Note: These results are based on CIBERSORT deconvolution of bulk transcriptomic data). (D–G) Correlation analysis between the identified candidate genes and the infiltration levels of various immune cell types. *: p<0.05, **: P<0.01, ***: P<0.001, ns: not significant.

Article Snippet: While we partially addressed this via CIBERSORT deconvolution to explore neutrophil-related correlations, their specific contribution to the ‘Metabolism–Proteostasis–Immunity’ axis may not be fully captured at the single-cell level.

Techniques: Control, Comparison

Journal: Communications Biology

Article Title: Deep-learning deconvolution and segmentation of fluorescent membranes for high-precision bacterial cell-size profiling

doi: 10.1038/s42003-026-10303-y

Figure Lengend Snippet: a Example of deconvolution prediction with FM2FM. Top, input raw fluorescent membrane images (Raw FM 4-64); middle row, true deconvolved fluorescent membrane images (Deconvolved FM 4-64); bottom, FM2FM-predicted image. White squares mark regions zoomed in at right (Block 1 and Block 2). Diagonal dotted lines indicate profile traces in ( b ). Scale bars: full images, 10 µm; zoomed-in blocks, 5 µm. b Fluorescence profiles across the dotted lines in Block 1 (top) and Block 2 (bottom). Light blue, raw FM 4-64; orange, deconvolved FM 4-64; burgundy, FM2FM prediction. c , Structural similarity index measure (SSIM) between FM2FM predicted images and deconvolved images. The distribution of SSIM values across 74 image crops is shown. d Violin plots of cell width, length, surface area, volume, cross-sectional area, convex hull area, eccentricity and solidity calculated from deconvolved FM 4-64 images (orange) and FM2FM-predicted images (burgundy). The white dots and vertical lines within the violin plots represent the medians and standard deviations. P values from statistical comparisons between distributions (ANOVA or Kruskal test; see methods for details) are indicated in each panel. Twenty images (over 1000 cells) were processed and segmented with FMSeg. See the Methods for size calculation details.

Article Snippet: To compare model performance against the membrane deconvolution target (SoftWorx), we conducted qualitative assessments of output images' intensity profiles.

Techniques: Membrane, Blocking Assay, Fluorescence