Search

appear to be implemented by a large, highly-interconnected network of neurons [4]. Determination of neural interconnections is therefore necessary to gain an understanding of the function of particular neural pathways as desired by neurophysiologists and anatomists, and of disease and phamacological mechanisms which is of interest to neurologists, pathologists and neuropharmacologists. Neural interconnections are generally determined from extremely time-consuming, manual three-dimensional reconstructions of neural tissue [2,1]. The requirement for large numbers of reconstructions, the very large numbers of neurons and interconnections in even the smallest functional unit, and the inter- and intra-technician variability due fatigue and training make a fully-automated reconstruction system desireable. We present a new algorithm to detect neural elements (soma, dendrites, axons) in images from serial optical sections, as required for a three- dimensional reconstruction system. The algorithm is also being used in the study of diabetic retinopathy and is applicable to angiography. Determination of element depth may be found in [3].