Dot Measurements

Take a nucleus with three bright HER2 signals inside it and one more just outside its edge. Dots (FISH/CISH)Loading... has already found all four as dot events. Dot Measurements finds, for each nucleus, the dots that lie inside it, and summarizes them.

The engine sits under the layer's Measurements, Add Measurements (advanced), and takes three inputs. Coded is the coded imageLoading... of master events, such as nuclei from NucleiLoading... detection. Shades is the grayscale image of the dot marker, and Dots Coded is the coded image of the dot events.

For each master event, the engine uses every dot event detected inside it. You select which results to compute:

• Dots Count: the number of dot events inside the master event.
• Dots Mean Intensity: the average gray level of the pixels of all those dots.
• Dots Total Area: the total area of all those dots.
• Dots Sum Intensity: the sum of the gray levels of all their pixels.
• Dots Intensities List: the average intensity of each dot, as a list.
• Dots Areas List: the area of each dot, as a list.

The results belong to the nucleus, not the dot. Right-click a nucleus for its Event Data, or open Raw DataLoading... to see every nucleus at once. The dot outside the edge in this example is counted for no nucleus. If dots and nuclei were detected in different layers, import the dots into the nuclei layer first with the Coded Image engine.

Source: StrataQuest 8.1 Engines Reference, ch. 18 (pp. 422–425), ch. 17 (pp. 417–418), ch. 9 (p. 379), ch. 13 (p. 403); User Manual (pp. 232–234).

Why area can't count merged dots. In light microscopy generally, an emitter much smaller than the resolution limitLoading..., about 200 to 250 nm for visible light at high numerical aperture, images as a blur set by the optics: the point spread functionLoading.... Two such emitters sitting together image as nearly the same blur, only brighter. So a dot's area says little about how many copies it holds.

Why summed intensity can. The total light under a blurred spot still equals what the emitters gave off, so two overlapping spots carry about twice the summed intensity of one. That makes Dots Sum Intensity the output to use for such an estimate, divided by the summed intensity of a dot you know to be single. Dots Mean Intensity is an average over pixels, so it is not the quantity that adds up.

The photon budget. Suppose a single dot delivers 100 photons to the detector, so a merged pair delivers about 200. Photon counts carry Poisson noise equal to their square root: 10 for the single, about 14 for the pair. The 100-photon gap is several times either noise level, so singles and pairs separate cleanly. At 25 photons per dot the noise is 5 and about 7 against a gap of 25, and errors grow, especially for clusters of three or more.

How one dot moves the ratio. In HER2 FISH, the published ASCO/CAP convention uses a HER2/CEP17 ratio of 2.0 as the main dividing line. A cell with 6 HER2 and 3 CEP17 dots sits right on it, and one missed or extra dot moves it to 1.67 or 2.33. That is one reason the scoring rules average over many cells, which software can measure in large numbers with identical settings. Signal-to-noise ratioLoading..., set mostly at acquisition, determines how reliable each count is.

Here is how a HER2 FISH count could be built. The engine names are StrataQuest's; the scoring rule is the published ASCO/CAP convention, not part of the software.

1. Detect nuclei on the DAPI channel.
2. Run Dots (FISH/CISH) on the HER2 channel and again on the CEP17 channel, with the nuclei as Events Mask.
3. Add Dot Measurements for HER2, with Coded set to the nuclei, Shades to the HER2 channel and Dots Coded to the HER2 dots, and select Dots Count. Add a second one for CEP17 the same way.
4. For a ratio per nucleus, divide the two Dots Count measurements with Derived MeasurementsLoading....
5. For the sample-level ratio, add a Global Statistic summing each Dots Count over the layer in the Statistics Report, then a Derived Statistic dividing the HER2 sum by the CEP17 sum.

The published ASCO/CAP convention treats a HER2/CEP17 ratio of 2.0 or more as the main sign of amplification, with further rules based on the average HER2 copy number per cell. Software can score hundreds of nuclei where manual counting covers a few dozen, and it applies the same settings to each one.