Item

Abstract

Neuronal networks are established by cells with different characteristics, for example the ability to produce specific transmitters or to buffer calcium-ions by expressing calcium-binding proteins. To understand neuronal networks, it is essential to know, how many cells express the specific markers, which can be visualized immunohistochemically. Cells numbers then have to be evaluated.
Here we propose a reliable method for automatic counting of cells in 5 \(\mu\)m paraffin sections of the brain, labelled with different antibodies (against NeuN, parvalbumin, GABA and GAD) or stained with Nissl. Images of stained sections are acquired with a CCD-camera on the microscope and are converted to binary images by thresholding.
Clusters of “ON pixels” (value of 1) corresponding to cell bodies are selected based on size. The parameters of the algorithm (intensity range and cluster-size) are adjusted for different methods of staining according to expert knowledge. Cell size is simultaneously determined by the histogram. The presented automatic cell counting method (ACCM) provides correct counting results as demonstrated by a comparison of computational results with counts gained by
human experimenters and with a commercially available image analysis system. On the basis of ACCM counts small and perhaps physiologically relevant differences in the
number of labelled cells can be revealed, as demonstrated here for the GABAergic system after two hours of an electrical stimulation, changing neuronal network
characteristics.