Skip to main content

Measuring & Detecting

In addition to the projection of segmentation masks over blended canvas images, rakaia provides a host of tools that can analyze and visualize the quantification of objects across ROIs.

Quantification​

Objects such as cells within a mask can be quantified, either in-browser or externally, to provide a per-object summary of expression statistics for each biomarker, as well as additional object parameters such as coordinates, area, etc. These results can be held in tabular CSV formats and imported for visualization and annotation. Many imaging workflows will compute either the mean or median biomarker signal for every channel for every object in a mask, and save the statistics with an ROI identifier in tidt format; in this way, multiple ROI expression profiles can be imported into one session and analyzed together.

Quantification configuration​

rakaia provides in-browser quantification of the current ROI. Users may select quantification under Advanced canvas options -> Measure/cluster -> Quantify current ROI, which will produce a pop-up modal listing the session biomarkers and provide a checklist of markers to be quantified. Quantification requires canvas with at least one biomarker applied, as well as a mask applied to the canvas so that the application can verify the compatibility of the mask prior to computation:

The computation time for individual ROI quantifications will vary significantly depending on deployment architecture as well as the number of cells and channels requested. generally, the computation time will grow linearly as the number of channels to be quantified increases, so larger datasets (> 30 channels) will likely take between 45 seconds to a minute for approximately 1000 cells.

Heatmap & Dimension reduction​

The Quantification/clustering tab provides a heatmap of expression statistics either column normalized along 0-1 (default) or given as raw expression counts, as well as a UMAP projection of the current quantification results.

The UMAP provides a dropdown list of expression and annotation values that can be projected into the UMAP graph for cluster analysis. This includes summary expression statistics for any of the biomarkers that have been quantified, as well as annotations that have been created from the canvas:

Interactively zooming in on the UMAP layout will automatically trigger the heatmap to update with only the cells in the current view, and Show distribution gives a frequency count of the cells in the various categories for the current UMAP projection.

UMAP configuration options​

Under UMAP options, users may upload a custom set of UMAP projections for the quantified objects, in order to eliminate the need to run the UMAP algorithm in the browser (which can be slow for large datasets and prone to browser timeouts). Additionally, users may also annotate the cells in the current view by creating new or using existing annotation categories and providing an annotation type (such as a cell type or descriptive designation). The UMAP window must be zoomed in on a subset of cells for annotation to take place. The xy UMAP coordinates for the current UMAP graph can also be downloaded in CSV format.

Object gating​

rakaia provides different methods for identifying mask objects that are sptially resolved within a canvas image. Gating can be achieved using either quantified parameters or custom ID lists.

Using quantification​

Masks applied to the canvas can be gated on the expression of one or more biomarkers or parameters when compatible quantification results exist for the current mask. Under Advanced canvas options -> Configuration -> Gating, users may select multiple gating parameters and apply them to the mask:

The parameters to be included in the current gate can be set under Set gating categories, and the gating threshold for individual biomarkers can be adjusted by selecting the relevant biomarker and changing the range slider values under Modify gating parameters. All gating parameters are min-max normalized within the quantification results between 0-1, and gating parameters represent a normalized range of expression. users may apply the gate as either the union of all ranges (cells that are in any of the gating parameters) or the intersection (cells that are present in all of the gating parameters). Is uers identify object populations of interest, the objects may be annotated directly under Create gating annotation with the specified annotation category and type, similar to other annotations. These annotations can then be applied to the UMAP layout, or exported in CSV format with the object ids listed for each annotation grouping.

In the canvas above, cells in mask are gated for higher expression of either GATA3 or CK5. Cells not included in the gate are shown with the objet border; this can be togfled off to show just the gated cells (See Masking & Sgmentation for more information)

Custom gating lists​

Alternatively, users may supply a string of comma separated mask IDs for gating. The mask objects with IDs that correspond to those contained within the string are shown. In the example below, the cells with mask ids 100, 200, 300, 400, and 500 are shown filled in relative to all other objects:

Users may toggle the Gate on custom IDs button to switch between the two gating techniques listed above.

Annotation projection from quantification results​

(Visit the section on masking for more information on how to provide cluster projections inside object masks).

rakaia allows users to transfer object annotations into cluster projection frames on a per ROI basis. From the quantification tab, any category that is currently visualized as the UMAP overlay can be applied as a cluster projection. In the example below, a custom gating annotation of objects has been made using the gating feature, and the category and sub-categories are transferred using UMAP Options -> Current overlay -> Cluster.