The acquisition scenario
StitchIt was written with serial section two-photon tomography (STP) in mind, but should work with any other similar devices which produce raw tiles that need stitching into complete sections.
In the case of an STP system images are acquired as a series of "tiles." A tile is defined as a single image from one channel at one depth. Tiles are acquired in a snaking pattern from the front-left corner (as you look at it standing in front of the microscope), producing a grid of tiles. A small overlap, e.g. 5%, is employed between adjacent tiles.
Typically, multiple optical planes are acquired at each X/Y position. Once the section has been completely imaged, the specimen is moved upwards by one section thickness, it is cut, then the newly exposed face is imaged. The virtue of this process is that all sections are aligned with each other. There is generally no overlap between physical sections in the Z dimension (although some systems, such as the TeraVoxel system do implement Z overlap).
One of the most important decisions the user takes regarding acquisition is the image resolution: the number of microns per pixel, the section thickness, and the number of optical sections (planes) in each physical section. Acquisition time, analysis time, and data size depend on these choices. The user will most likely choose to acquire somewhere between 4 μm/pixel to 0.5 μm/pixel in X/Y. Doubling the resolution quadruples the data size. The time required for many analyses will likely also scale with the data size. It is therefore important to take images at the minimum resolution required to adequately answer the question.
The image below shows the effect of increasing z resolution on acquisition time (left) and data size (right).
Times assume a galvo/galvo scanning 2-photon microscope and are significantly faster for a resonant scanning system.
Red lines are 0.5 μm/pixel and black lines 1 μm/pixel.
Image parameters are 16 by 10 tiles over a depth of 10 mm.
The resolution at which we image much of our data is shown by the black circle.
Where possible, however, we use 2 μm/pixel in order to save time and disk space.
