Regions of Interest

Regions of Interest set the spatial scope of an analysis.

1. Definition: ROIs come from manual drawing (rectangle, ellipse, custom polygon or freedrawn shape), from Tissue Detection (the Autodetect ROIs button), or from import (regions and annotations brought in with the slides, regions in OME-XML metadata, or QuPath annotations).
2. Events at the boundary: Analysis restricts detection results to the ROI's shape. What happens to an object that sits across the ROI's edge is set per layer by its Events Strategy (Layers Editor → More → Properties), which offers three options: select only events entirely in the ROI, select events entirely or partially in the ROI, or crop events to the ROI's shape.
3. Multi-ROI analysis: Several ROIs on one section produce separate statistical outputs. A section might carry "Tumor Center," "Invasive Margin" and "Stroma" ROIs, each with its own cell counts, phenotype distributions and spatial metrics.
4. Exclusion: Exclusion Areas, a separate kind of region drawn with the Exclusion Area button, mask out artifacts, necrosis or damaged tissue. Their area is subtracted from the analyzed tissue, and no events detected in them are counted.

Howard's brick rule for partial objects (Pawley's Confocal Handbook): A nucleus that is half inside and half outside an ROI could be counted in both regions it touches, in neither, or in one. Counting it in both overestimates; counting it in neither underestimates. Howard's brick rule includes objects touching the top and right boundaries and excludes objects touching the bottom and left, which gives an unbiased count on average, though the variance can be large for small ROIs.

StrataQuest does not impose a fixed rule. The layer's Events Strategy lets you keep only whole events, keep events that touch the ROI, or crop them to its shape. The choice changes counts near the boundary, so it belongs in the methods description.

ROI masking as spatial filtering: In signal-processing terms, applying an ROI multiplies the image by a spatial mask, a binary function that is 1 inside the ROI and 0 outside. It acts as a spatial filter that passes only the signal within the region. The mask's sharp edges can introduce boundary artifacts, which is why the boundary-handling rule matters for accurate quantification.

Biological rationale for spatial stratification: The tumor microenvironment is spatially organized: immune cells at the invasive margin behave differently from those in the tumor center, and PD-L1 expression varies by location. Spatial phenotypingLoading... results change substantially depending on which tissue zone is analyzed. Whole-section averages hide this spatial heterogeneity and ROI-based analysis preserves it, which is why published scoring systems such as the Immunoscore for colorectal cancer measure the tumor center and the invasive margin separately.

Take an Immunoscore analysis of colorectal cancer.

1. An expert reader draws two ROIs: Tumor Center (CT) and Invasive Margin (IM).
2. Nuclei DetectionLoading... and phenotyping identify CD3+ and CD8+ T cells throughout the section.
3. Per-ROI statistics give four densities: CD3+ in CT, CD3+ in IM, CD8+ in CT and CD8+ in IM.
4. The four density values determine the Immunoscore (I0 through I4).

Without ROIs you would get a single whole-section T-cell density, mixing the margin population with the center population. The Immunoscore is built from the two zones measured separately. A single whole-section density cannot show how the two zones differ.

Because detection results are restricted to the ROI's shape, the same engines and the same parameters produce statistics for that region alone, and no time is spent on other areas.

Conceptually, an ROI works like a binary maskLoading... laid over the sample: inside is kept, outside is ignored. An object that sits across the boundary is handled by the layer's Events Strategy, described under How It Works.