#13 · ML / AI methods
Imaging and spatial data tools embed attribution, heterogeneity modeling and benchmarking
The dominant thread is software that turns raw imaging or spatial data into quantitative, evaluable measurements. Virtual staining appears at two scales: DynaCell benchmarks 3D time-lapse live-cell staining with three metric tiers, and QUEST predicts multiplex immunofluorescence from H&E. PixAQ replaces shared-mask quantification with marker-specific probabilistic pixel attribution in imaging mass cytometry, while CardioShift models nested variance to map single-cell Ca2+ and contractility distributions. In spatial transcriptomics, Tracel assigns transcripts to cells via optimal transport against scRNA-seq references, SWIMR integrates multiple references for deconvolution, and TileScope quantifies gene relationships beyond colocalization. MaCADEMIA covers single-molecule localisation quantification. The common feature is that attribution, heterogeneity and evaluation are embedded within the tools.
Published papers
Enhancing cell type deconvolution in spatial transcriptomics through incorporating multiple reference panels.
SWIMR, a statistical framework integrating multiple scRNA-seq references with spatial correlations for cell-type deconvolution in spatial transcriptomics,…
Preprints
From Cross-Section to Cross-Omics: Conditional Cell Nucleus Image Editing in Spatial Transcriptomics
Applies Manifold-Aware Diffusion (MAD) to conditional editing of cell nucleus images and gene expression in CosMx SMI kidney spatial transcriptomics, with…
The DynaCell Evaluation Suite for Dynamic 3D Virtual Staining of Live Cells
DynaCell is a benchmark suite for 3D time-lapse virtual staining of live cells, pairing labeled and unlabeled imaging with baselines and three metric tiers.
TileScope quantifies spatial gene expression relationships beyond colocalization
TileScope is a framework quantifying gene and program spatial relationships within tissue regions via enrichment, correlation, cross-correlation and boundary…
MaCADEMIA: a software toolbox for single-molecule fluorescence image quantification
MaCADEMIA is a MATLAB toolbox for single-molecule localisation microscopy quantification, covering particle detection, photobleaching-based counting, mobility,…
CardioShift: Resolving cell-to-cell functional heterogeneity in hiPSC-derived cardiomyocytes from high-throughput single-cell imaging
CardioShift, an open-source bioimage framework, models nested variance to map single-cell Ca2+ transient and contractility distributions in hiPSC-cardiomyocyte…
Generalizable virtual staining from histopathology to spatial proteomics across biological scales
QUEST, a semantic-query model that virtually stains H&E histology to predict multiplex immunofluorescence protein expression across cellular and tissue scales.
Marker-specific pixel attribution for single-cell protein quantification with PixAQ
PixAQ uses marker-specific probabilistic pixel attribution to improve single-cell protein quantification in imaging mass cytometry over shared-mask methods,…
Unified cell segmentation for spatial transcriptomics via transcript reassignment
Tracel, a two-stage optimal transport method, assigns spatial transcriptomics transcripts or bin counts to cells using matched scRNA-seq references, improving…
Topic history
| Date | Rank |
|---|---|
| #13 | |
| #36 | |
| #40 |