Journal: bioRxiv

Article Title: Denoising image-based spatial transcriptomics data with DenoIST

doi: 10.1101/2025.11.13.688387

Figure Lengend Snippet: a) Shown here is a region of the healthy lung sample assayed using Xenium, with the boundary expansion segmentation from 10x Xenium Ranger (Top) and Proseg (Bottom) segmentation. Each dot is a transcript molecule, molecules of lineage marker genes are coloured by their respective lineages. Black lines are segmentation boundaries. b) Log CPM (counts per million) normalised pseudobulked gene expression of 4 selected contaminating genes using 4 annotated immune cell types in the lung fibrosis Xenium data. Each data point is a sample and a lowess curve is fitted over all the samples. c) A high level schematic of the application of DenoIST.

Article Snippet: To address this research gap, we present DenoIST (Denoising Image-based Spatial Transcriptomics), a Poisson mixture model tailored for denoising IST data by reducing the effects of transcript contamination in downstream analysis tasks.

Techniques: Marker, Gene Expression

Journal: bioRxiv

Article Title: Denoising image-based spatial transcriptomics data with DenoIST

doi: 10.1101/2025.11.13.688387

Figure Lengend Snippet: Expression of ACTA2 from a zoomed-in section in Xenium human breast cancer dataset. Each dot is a segmented cell using 10x boundary expansion method, colour shows the log count of ACTA2 . Cells with 0 count are greyed out for visual clarity. b) Heatmap visualisation of gene expression of annotated cell types before (top) and after DenoIST (bottom). Columns are selected genes, annotated by the cell type they mark. Row are cell types. The log(mean count + 1) for each cell type is shown here. c) MECR before (top) and after (bottom) applying DenoIST to Xenium human breast cancer dataset. Rows and columns denote genes and each entry is the MECR of the corresponding pair. Note that genes that mark the same cell type are not expected to be mutually exclusive, but are shown here for positive control.

Article Snippet: To address this research gap, we present DenoIST (Denoising Image-based Spatial Transcriptomics), a Poisson mixture model tailored for denoising IST data by reducing the effects of transcript contamination in downstream analysis tasks.

Techniques: Expressing, Gene Expression, Positive Control

Journal: bioRxiv

Article Title: Denoising image-based spatial transcriptomics data with DenoIST

doi: 10.1101/2025.11.13.688387

Figure Lengend Snippet: a) UMAP visualisation of lung fibrosis data after applying DenoIST. Sample TILD028MA is shown here. Cells with 0 count are greyed out for visual clarity. b) An airway section from fibrotic sample VUILD110. Each dot is a cell. Cells with 0 count are greyed out for visual clarity. Annotated airway cell types and gene expression (raw counts and DenoIST-adjusted counts) for KRT5 and MUC5B are shown. c) Proportions of RCTD classification using raw counts and DenoIST-adjusted counts in healthy sample VUHD116A. d) RCTD assignment weights of the second highest lineage of each cell in healthy sample VUHD116A, stratified by their manually annotated lineages. Cells with a pure identity should have low weights for the incorrect lineages.

Article Snippet: To address this research gap, we present DenoIST (Denoising Image-based Spatial Transcriptomics), a Poisson mixture model tailored for denoising IST data by reducing the effects of transcript contamination in downstream analysis tasks.

Techniques: Gene Expression

Journal: Cancers

Article Title: Advances in the Use of Deep Learning for the Analysis of Magnetic Resonance Image in Neuro-Oncology

doi: 10.3390/cancers16020300

Figure Lengend Snippet: Data Overview: Comprehensive overview of the datasets examined within the DL literature review centered on brain tumor classification tasks and MRI data. Essential information regarding dimensionality, sample size, anatomical plane, MRI modalities, and pre-processing methods are summarized.

Article Snippet: 115 , Vankdothu et al. [ ] (2022) , 2D , Kaggle-II , - , 3264 , - , - , Grayscaling, rotation, denoising, tumor ROI , -.

Techniques: Transformation Assay, Sampling, Shear

Journal: BMC Medical Imaging

Article Title: VER-Net: a hybrid transfer learning model for lung cancer detection using CT scan images

doi: 10.1186/s12880-024-01238-z

Figure Lengend Snippet: Comparing VER-Net with recent literature

Article Snippet: Worku et al. [ ] , Denoising first two-path CNN (DFD-Net) , Kaggle Data Science Bowl 2017 challenge (KDSB) and LUNA 16 , 87.80%.

Techniques: