Projection

Suppose a section holds nuclei at several depths. Focus on the top ones and the deep ones blur. Projection, in a layer's Pre-Processing section under Add Operations (advanced), takes the images you tick in List of images and combines them into one. The Engines Reference says the result has a greater depth of field than any of the source images, and that it fits any case where individual images have a very shallow depth of field.

The Operation setting picks one of nine methods. Sum adds corresponding pixels, Max keeps the largest, Min the smallest, and Average the mean. The other five fuse through a named wavelet: Daubechies 1, Daubechies 2, Biorthogonal 1.5, Reverse biorthogonal 1.5 and Discrete Meyer. Levels sets the wavelet decomposition level.

The reference contradicts itself about two of the four simple methods. Its method list describes Min and Average as pixel operations, like Sum and Max. A sentence just below says that except Sum and Max, the rest of the methods use wavelet decomposition. Until TissueGnostics settles it, treat Sum and Max as pixel-wise, the five wavelets as wavelet fusion, and Min and Average as uncertain.

Depth of field. In general, depth of field shrinks as numerical apertureLoading... grows, so the objectives that resolve the finest detail image the thinnest slab. A z-stack buys depth back by sampling several planes. The number of planesLoading... and their spacing determine how much of the section is covered.

Why Max for fluorescence and Min for brightfield. In fluorescence, objects are bright on a dark background, and an object usually looks brightest in the plane where it is sharp, so keeping the maximum tends to keep it. In brightfield, stained objects are dark on a bright background, so the minimum plays the same part. TissueFAXS says as much for its own Extended Focus: it recommends SUM and MAX for fluorescence, especially confocal, and MIN for brightfield only. The drawback of Max is that it keeps the brightest value whatever its source, so out-of-focus glow and bright noise survive too.

Wavelet fusion, as usually done. In image processing generally, a wavelet transform splits an image into a coarse approximation and detail coefficients at several scales; the decomposition level is the number of scales. The common fusion rule keeps, at each position and scale, the detail coefficient with the largest magnitude, combines the approximations, and transforms back. Sharp structure gives large detail coefficients, so the rule favors whichever slice is in focus at that spot. The Engines Reference names the wavelets and Levels but not the rule, so read this as how wavelet fusion usually works, not as a description of the engine.

Confocal and widefield stacks. In general, each confocalLoading... plane holds mostly in-focus light, while a widefield plane also carries blur from above and below. Any projection of a widefield stack inherits some of that haze.

There is no Z range setting. To leave out planes that are empty or wholly out of focus, untick them in List of images. TissueFAXS can also remove planes from an acquired region with Remove Z-Stack Layers, which sets how many steps above and below to delete.

Source: StrataQuest 8.1 Engines Reference, ch. 35 (pp. 495–499); User Manual, Z-Stack display control (p. 69); TissueFAXS 8.1 User Manual §5.1.3.4 (pp. 182–186).

The Engines Reference example fuses a fluorescence z-stack of 15 levels per channel, shown for DAPI and Cy5, and overlays the four projected channels.

1. Add a Projection and tick one channel's slices in List of images. Repeat for each channel you need, since each Projection produces one image.
2. For fluorescence, Max matches what TissueFAXS recommends for its own Extended Focus. A wavelet method at Levels 2 is the alternative to compare.
3. Run NucleiLoading... detection on the projected DAPI image and measure the other projected channels with Standard MeasurementsLoading....
4. In a small ROILoading..., compare against single planes. Nuclei lying over one another at different depths can merge in a projection.

Measured intensities depend on the method. Sum grows with the number of slices, and Max reports peak values. Keep the same Operation and the same slices for every sample you compare.