The LSDCAS data acquisition algorithm involves the sequential cycling through a number of stage (x,y) coordinates: for each microscope field, it is necessary to re-focus the image. Although hardware auto-focus modules are available (and we have used them in the past), these modules require an analog signal input, and most new high-resolution cooled CCD cameras produce digital output only (using a firewire data bus). Thus, software controlled auto-focus processing is required for the new cameras; having more control over auto-focus has also led to improved reliability over hardware auto-focus. In order to perform automatic focusing it is necessary to define an image property that can accurately describe the quality of being in focus. It has been previously described that image contrast is capable of describing this quality. Furthermore, it has been shown that determining contrast via correlative methods is an appropriate method for automatic focusing. The algorithm we have implemented is based on Vollath, D: "Automatic Focusing by Correlative Methods", J Micros 147, 279-288, 1987.
Focusing begins by acquiring images along the z-axis of the microscope. Using the last known focus position, the stage is moved slightly out of focus and then sweeps through the focus plane while acquiring images. The focus function is then calculated for each image, and the image with the best value is selected as the in focus image. Although this limits resolution to the step size used along the z-axis, in practice we have found that the the difference between two steps is undetectable by the eye. In the future if sub-sample focusing is needed, standard numerical analysis techniques can provide this.

Spatial calibration of the microscope images was accomplished using a calibration specimen obtained from <em>Electron Microscopy Sciences</em>. This specimen, a 5 mm x 5 mm square of single crystal silicon was photo-etched with a repeated pattern of 10 µm x 10 µm squares, which are further subdivided into grids with 1.9 µm wide lines. After acquisition of an image of the grid (in a culture flask containing cell-culture medium) using LSDCAS, the image was analyzed in the frequency domain using Fourier analysis. Subsequent registration of the grid by rotation in the frequency domain (to correct for non-orthogonal placement of the specimen), yielded the frequencies of intensities corresponding to the grid lines. Reverse transformation of the frequencies yielded the grid separations in pixels in the X- and Y-dimensions, which were then used to provide calibration constants for the overall image magnification for the given microscope objective.