Skip to main content

Data import and configuration

rakaia supports a variety of data imports, which includes channel images, segmentation masks, configuration files, quantification results, etc. Below is a brief list of supported file formats for the different import categories, as well as the input options for dataset import.

Data formats​

Raw image files​

rakaia supports the following formats for importing raw images:

  • mcd
  • tiff/ome.tiff (multi-channel: each page of the tiff is assumed to be a channel)
  • txt files (redundant storage files that come from mcd files)
  • h5py

Raw image files should hold the images as 2-dimensional arrays in greyscale format. Conversely, certain formats will hold 3-dimensional arrays where the third dimension represents the individual greyscale channels.

Masks​

Masks should be imported as single-channel greyscale tiffs. The tiff should represent a 2-dimensional array where values are integer-based and represent different objects, starting at 1. For example, the region in a mask where values are all 1 corresponds to ab object that is identified by 1.

Masks should have matching dimensions to at least one of the imaging datasets that have been imported into the session. When a mask is selected, rakaia will verify the dimensions against the current data loaded, and will warn the user if incompatible dimensions exist. Read more under Masking

Important: masks should be generated as 32-byte dtype arrays for an appropriate processing pipeline. Failure to cast the mask with the appropriate byte size may result in the loss of mask objects that can interfere with proper quantification and object detection.

Measurements results​

measurements such as summarized intensity statistics for mask objects for one or more ROIs may be imported in the following formats:

  • CSV
  • Anndata (h5ad)

Visit the Measuring/Detecting portion of the documentation for additional information on the specific format of quantified results that is required.

Data import components​

rakaia provides two options for importing raw imaging data from file sources. Users can use either the drag and drop uploader to make a copy of imaging files in the session, or by copy and pasting either a filepath or local directory and reading the files directly from the local source.

Option 1: Drag and drop uploader​

rakaia enables the import of data files into the session using a Javascript-enabled drag and drop component as the topmost component in the rakaia data configuration tab:

This component uses an underlying flow.js library to copy the data into a local temporary file cache. It is important to know that this component will duplicate the data on local runs, which will be slow for large file uploads and not optimal for systems that lack a large amount of free disk space. For large files or systems without sufficient disk storage for data duplication, the local file dialog below should be used.

Option 2: Local file or directory dialog​

users can read data files directly from their systems on local runs using the direct file reader below the drag and drop uploader:

Users may choose to read either a copied filepath string when filepath is enabled, or read all of the compatible files in a directory when directory is selected.

Data loading​

In rakaia sessions, the unit of analysis is the region of interest, or ROI, which corresponds to a series of channel arrays of identical dimensions that can be layered together using additive blending. These channels represent a single tissue region where imaging has occurred. Typically, each channel will represent a marker of interest measured using either florescence or imaging mass cytometry.

At any given moment, users can have a single ROI loaded into the session and the canvas, where an additive blended image can be created. Users may toggle among multiple ROIs within one session provided that the channel number/panel is identical across the ROIs. ROIs that have different channels represent different experiments, and therefore cannot be analyzed within the same session.

Users may view and select ROIs from the dataset information table once data have been successfully imported, or select an ROI identifier from the data collectino dropdown menu:

Once an ROI has been selected, users will be able to populate the channel tile gallery and begin to create a blended image in the main canvas.

Main canvas​

The main tab of rakaia contains the components required to create blended multiplexed images, the msot important of which is the main canvas. Once an ROI has been selected, users may additively blend one or more channels together by selecting channels from the Channel selection dropdown menu:

More information about creating additive blended images and performing pixel-level analysis can be found in the Pixel level analysis portion of this documentation.

Adjacent to the main image anlaysis tab is the Image gallery tab, which provides a small thumbnail preview of each channel in the current ROI:

Each thumbnail is a downbsampled version of the greyscale channel image available for blending, autoscaled by percentile scaling (99%) so that a reasonable proportion of the channel signal is available by default. Users may browser and scroll this tile gallery to get a general impression of the expression patterns in each channel before adding one or more channels to the main canvas. Users may also add channels directly from the tile gallery by selecting Add to canvas for any of the channels shown. In turn this will redirect the view to the main canvas where the selected channel will be added to the blend.