Membrane Measurements

Take a nucleus with a brown membrane ring around it, broken on one side. You've detected the nucleus as an event. Separately, the MembraneLoading... engine has turned the membrane stain into a mask. Membrane Measurements relates the two: it measures how much of the nucleus's surroundings the mask occupies, and how far out.

Both versions sit under the layer's Measurements, Add Measurements (advanced), and the mask can come from the current layerLoading... or a different one. The local version takes Coded (master) and Mask (membranes). It places a virtual disk around each master event, of radius Disk Radius, split into the number of sectors set by Sectors count.

The global version takes coded (the events), proximity mask (the membrane), shades (the image for intensity), mask (to restrict where measurements are computed) and proximity distance. It splits the area within the proximity distance into roughly equal sectors around the event. The reference notes the limit at small sizes: a 1-pixel event with a 1-pixel distance has at most 8 sectors, about 45° each.

Source: StrataQuest 8.1 Engines Reference, ch. 29 (pp. 465–468), ch. 30 (pp. 469–470), ch. 31 (pp. 471–478).

The membrane looks wider than it is. In light microscopy generally, a lipid bilayer under 10 nm thick images as a band roughly as wide as the resolution limitLoading..., around 250 nm or more. A membrane intensity therefore mixes true membrane signal with some from the cytoplasm and the space on either side. How much depends on the optics and on the marker, not on the software. Mean thickness measures the mask, so it reports the imaged and detected width, not the bilayer's.

Why measure from the nucleus. In tissue analysis generally, segmenting every cell outline is hard where membranes are faint, broken or shared between neighbors, while nuclei are compact and usually well stained. Anchoring each cell at its nucleus avoids that problem but assumes the membrane lies within the distance you set. Too small a proximity distance misses membrane; too large picks up the neighbor's. Region tools such as Grow (global)Loading... define a ring around each nucleus, but this engine needs a mask of where membrane stain was actually detected.

Completeness in HER2 scoring. The published ASCO/CAP convention for HER2 IHC rests on both the intensity and the completeness of membrane staining in invasive tumor cells. A 3+ requires complete, intense, circumferential staining in more than 10% of tumor cells. Angle of staining gives completeness as a number and Mean intensity gives strength. How the two combine into a score is a decision for you and the convention your study follows, not an engine output.

Angle versus area. Angle of staining counts a sector once, whether the membrane in it is a hairline or a thick band. Percentage of neighborhood area counts covered area, so a thick membrane raises it. Reading the two together tells a thin complete ring from a thick partial one.

Here is how per-cell inputs for HER2 IHC scoring could be built. The engine names are StrataQuest's; the category boundaries are yours to set against the published convention.

1. Separate the DAB and hematoxylin stains with Color SeparationLoading....
2. Detect nuclei on the hematoxylin channel.
3. Run MembraneLoading... on the DAB channel to get the membrane mask (Membrane thick result).
4. Add Membrane Measurements (global) with coded set to the nuclei, proximity mask to the membrane mask, shades to the DAB channel, and a proximity distance suited to your cell size. Enable Angle of staining and Staining – Mean intensity.
5. Plot Angle of staining against Mean intensity on a scattergramLoading... and set cutoffsLoading... for complete versus incomplete staining and for intensity.

Completeness is in degrees here, not percent: 90% of the outline is about 324°. One caveat before trusting any cutoff. The angle comes from the mask, so membrane too faint for the Membrane engine to detect counts as absent, and the angle can be no better than that detection.