Advancements in segmentation of 3D SEM images of neurons: pre-processing steps

Reconstructing neurons from serial EM micrographs and localizing synapses manually is extremely time consuming [1][2]. Advanced, semi-automatic selection tools for segmenting are faster, but require noise reduction while keeping membrane integrity and structural features of synapses. We developed pre-processing steps to make the micrographs suitable for semi-automatic selection.
Serial electron micrographs of identified neurons of the locust, Locusta migratoria were produced using serial block face scanning electron microscopy (Fig. 1A). Several image processing steps were performed sequentially on these stacks, using ImageJ software.
First, a histogram matching algorithm was used to homogenize brightness levels, followed by a low pass Fourier transform band-filter to isolate small structures like cell-membranes, a rolling ball algorithm to give the cell membranes a higher contrast, and an automated Otsu threshold. The thresholded stacks were analyzed with the particle analyzer to filter out small irregular components. In parallel, the particle analyzer was used once more with settings that enabled keeping the large particles such as mitochondria. The result was subtracted from the result of the first round with the particle analyzer.
Second, mitochondria were removed, for this, the histogram - matched stacks were filtered with a median filter after a contrasting step, followed by a Huang´s auto threshold to get a binary image (that contained the mitochondria) which was subtracted from the first preprocessing step.
The first step successfully removed noise and small structures, but mitochondria attached to the membranes were still visible, compromising the results of segmenting (Fig. 1B).
During the second preprocessing step most of the structures inside the cells were deleted. However, at certain locations the membranes appeared discontinuous. Nevertheless combining both steps with fast selection tools significantly enhanced segmentation efficiency compared to a traditional manual approach.