Before accumulating the image stack, the user must decide upon the size of the image to be captured for each optical section
(typical image sizes are 256*256, 512*512 and 768*512, but may also be set to custom dimensions), the number of these
images to be captured, and the desired Z-step size between collection of succesive optical sections. The selection of the
former is usually dependent on desired size of the final 3-D dataset. For example, an image stack consisting of 50, 512*512 8-bit images ocupies 12.8 MB of storage space. The 3-D size of the image stack not only has implications on storage spage, but also has adverse effects on the processing, display and manipulation of this data.

When selecting the Z-step between optical sections, several things must be considered. These include the fineness of the
structure to be studied, the mechanics of the focusing movement and the resolution of the objective. An important
consideration when planning 3-D reconstructions from optical sections is choosing a step size in the Z-plane such that the
resulting X, Y and Z-axes of the reconstructed specimen reflect the correct proportions. One way of doing this is to make
sure that the sampling in the Z plane of the specimen is made to match the sampling in the X/Y planes. This can be done by
adjusting the Z-step of the focusing motor to the resolution of the X/Y planes prior to image capture, or through interpolation
of the Z plane using computer software after the images have been captured. If the latter method is to be used, the best results
are obtained when a linear interpolation is used. With this method, each intermediate "missing" optical section is generated
from 2 successive optical sections. It is quite obvious that this method is not as good as initially matching the Z-step to the
X/Y resolution, as additional information is added to the data set. However, in some cases where there is very little
difference in intensity between optical sections gathered, linear interpolation may prove practical.

Perhaps the most serious problem in collecting a useful image stack is that confocal microscopy is highly destructive. As the
laser scans the specimen, it is selectively bleached in the X/Y plane and nonselectively bleached in the Z plane. If the
specimen is scanned one plane at a time, not only is that plane bleached, but all other cross-sectional planes above and
below that plane are also bleached. If the laser power is set to its maximum, the bottom slice in an image stack would have
received as much irradiation as if all of the images were captured at the same plane. It has been suggested that this
nonselective bleaching effect in the Z plane can be reduced by a process called Z-axis distributed averaging (Stevens, 1994). In this method, bleaching is distributed over the entire 3-D volume rather than within one image of the stack. The method
involves making several passes of the optical sections comprising an image stack using low laser power, and then averaging
the optical sections obtained with each pass. Even with this method, the author acknowledges that there is a "total slice
capacity", or fixed number of images that can be sampled before the data become artifactual. This capacity must be tested for each new set of scanning conditions, including the number of slices to be captured, slice thickness, fluorophore used, etc. In
general, it is recommended that the total slice capacity is estimated by determining the point in nondistributed sampling
at which bleaching becomes highly destructive, doubling this number and testing this with distributed sampling.

Another factor in data acquisition using a LSCM is the signal-to-noise ratio (SNR) in the optical sections of the specimen. In
this respect, the adjustment of the detector pinhole is of critical importance. For example if the SNR is too low, the diameter
of the pinhole can be increased, thereby allowing increased detection of signal at the expense of resolution. However, if the
pinhole is made too large, both the SNR and resolution will decrease. Conversely, if the pinhole is too small, the signal can be lost.

When capturing optical sections of a specimen using a single laser line and a fixed filter set, the alignment of the images in the image stack is usually preserved in the X, Y and Z axes. However, when dealing with multi-labelled specimens where image stacks are gathered with separate filters and/or laser lines, the image stacks gathered for each fluorophore in the same area of the specimen may be misaligned if the filter sets are switched manually. This presents a problem if a single 3-D reconstruction consisting of both of these image stacks is desired.

Misalignments can usually be corrected using computer software after the stacks have been collected, and can be in the form
of translations (shifts in the X and/or Y plane) or rotations in the X/Y plane. While the latter is more complex to correct and
can introduce sampling variations in the data, correction of both forms of misalignment have the effect of reducing the
bounding area of the original X/Y image size that contributes to the 3-D volume. Some newer confocal microscopes have
automated methods of changing filters which greatly reduces or eliminates alignment problems for multilabel imaging. 

   Back to Heading