Roadmap

Road map

This will provide the road map documentation for the image viewer.

Pixel/image level features (Phase 0)

• ability to import imaging data from a variety of multiplexed imaging sources (mcd, tiff, .txt, h5py)
• Support for multiple datasets within a session with a common panel
• Editable session metadata (IMC panel labels) allows for flexible channel labels on publication-quality canvas figures
• Canvas blending: ability to blend user-selected number of channel images into a canvas blend
• Channel modification: channel recoloring, pixel intensity adjustment, filters (Gaussian, median), pixel distribution histogram and automatic percentile scaling
• Customized canvas layout/view: user-selected zoom & pan view, legend + scalebar size & position
• Interactive tile gallery: users can view individual channels in a tile gallery and add them to the blend quickly
• Outputs: users can save blend parameters in JSON or h5py format, canvas images can be output as tiff or interactive HTML
• loading on demand: data loading strategy and server side data caching makes image loading and blending very fast, can support large images (ROIs of up to 7000x4000 pixels processed)
• across dataset queries: multi ROI datasets can be queried with the current blend parameters and loaded directly into the current view
• session image caches make image blending rapid and smooth
• streamlined UI and controls make toggling views and navigating among ROIs very efficient

Cell/segmentation/object level features (Phase 1)

• mask overlay: users can import cell or object segmentation masks as tiffs and overlay profiles over images with a range of adjustable visualization controls
• Quantification: import tabular expression channel statistics for each cell and project onto interactive UMAP and summary chart
• Annotations: users can generate annotations using click and freeform shape drawing. Annotations can be output as annotated PDF (to support clinical reporting) or tiff masks with object id lists
• click annotations support rapid cell counting and output as coordinate lists in CSV format with toggle able shapes and overlay grid
• Freeform draw and click annotations using masks will automatically link to a quantification sheet if provided with cell segmentation masks
• custom annotation masks in tiff format can be output for one or more annotation categories using freeform shapes, zoom, and rectangles
• Cross ROI/dataset querying
• generate selectable previews of ROIs based on a random selection or generated from interactive UMAP clustering: select a cluster of cells and identify the ROIs containing the cells of interest, sorted by cell number
• link cell quantification results to images prioritized by cell counts

Upcoming features (Phase 2)

• segmentation and cell/object quantification directly in the browser

• Infrastructure for robust public instances:

  • persistent user sessions (log credentials and saved annotations)
  • job submission queue and brokers
  • dedicated back-end databases for dataset storage
  • public domain instance
  • possible infrastructure extensions: HPC, AWS, Celery, SQL/MongoDB

• support for spatial transcriptomics linked to images: multi-modal datasets using SpatialData (Nanostring, Visium) and 3D tissue datasets (serial sections)

• plugin support for cell type classifiers, registration of IF/IMC images, double detection, and quality control