Serial Reconstruction-a brief history

Dr. Peter Sterling's lab has long been on the cutting edge (pun fully intended) of serial reconstruction from electron micrographs. As early as 1980, the lab had a number of ingenious-if complex-serial reconstruction methods perfected to the finest detail.

The following two papers chronicle the first two stages of the evolution of our serial reconstruction system:

Stevens, John K., Thomas L. Davis, Neil Friedman, and Peter Sterling."A Systematic Approach to Reconstructing Microcircuitry by Electron Microscopy of Serial Sections". John K. Stevens, Thomas L. Davis, Neil Friedman, and Peter Sterling. Brain Research Reviews, 2 (1980): 265-293.
Smith, Robert G. "Montage: a system for three-dimensional reconstruction by personal computer". Journal of Neuroscience Methods. 21 (1987) 55-69.

A less comprehensive treatment of the history of serial reconstruction is as follows:

John Stevens developed and optomized several techniques that permitted the reconstruction of fine processes in retinal tissue. These are described fully in the above paper, but briefly, his findings were the following: sections of a relatively uniform thickness of 100nm could be produced in series of 200+, which would contain sufficient information to make nearly complete reconstructions of many cells, and partial reconstructions of others. The sections could then be reconstructed in one of two ways: manually or with a "computer-assisted method".

A user using the manual reconstruction method would follow the processes of a cell from one section to the next on a photographic print, tracing the outlines of the processes on to sheets of clear plastic mounted on a cartoonist's jig. In this way, the alignment of the contours could be examined from section to section, and compared with the relative alignments of "fiducial" processes-3 to 4 small dendritic or axonal cross-sections of a different cell used for alignment. By manually correcting distortions between sections, the irregularities caused by the sectioning process can be eliminated, and an aesthetic and accurate visualization of the cell can be produced by an artist.

The, second, computer-aided or "movie" reconstruction system is very complex. The user first produced 35mm film positives from the negatives, correcting the alignment by comparing the superposition of the current negative on the previous. This 35mm film became the input for what was surely one of the first multi-media systems-a z80 controlling a video camera that would, essentially, allow the user to correct for non-biological alignment distortions and trace the processes on-screen.

The Montage package

With the advent of the Montage computer digitization system, the cell reconstruction process has become much simpler. Now the negatives are converted to photographic print positives using standard darkroom techniques, catalogued and analyzed. These positives are then traced by section on to acetate on a cartoonists jig as with the manual method, but are then digitized by means of a Summagraphics SummaSketch bitpad. This produces a data file with a series of parallel sections, each containing a number of irregular polygonal contours. The original Montage package contains several programs for analyzing and displaying this data. The Montage serial reconstruction system, including source and Linux binaries are available by anonymous ftp to retina.anatomy.upenn.edu.

Further work in the area of serial reconstruction is being pursued by Alex Shrom. He is currently working with true three-dimensional visualization schemes that produce triangulated surfaces between the sections.

Comments, questions, observations, etc., should be emailed to "alex@retina.anatomy.upenn.edu". -Alex