Automating the quantitative analysis of 2-D neural dendritic trees

Abstract

Neurons in the central and peripheral nervous system vary widely in their dendritic branching patterns. Quantification of the morphological characteristics used to identify different classes of neurons and relate neural structure to function requires that accurate metric and non-metric data be obtained from neural images obtained by camera-lucida drawing or from digitized video images made with transmitted, fluorescence or confocal microscopy. This paper describes a largely automated procedure for determining the dendritic tree structure of largely planar cells (such as retinal ganglion cells or cells in tissue culture monolayers) from an initial pictorial representation or digitized image. From this structure, non-metric data (such as the ordered 'tree' of branches) and metric information (such as total dendritic length and dendritic field area) can be automatically computed. The use of this method is specifically illustrated in the capture of the dendritic tree structure of retinal ganglion cells from the rabbit retina.