Eduardo Fernandenz^a, William D. Eldred^b, Josef Ammermüller^c, Arthur Block^d, Werner von Bloh^d, and Helga Kolb^e

Journal of Comparative Neurology 347, 397-408 (1994)

^aInstitute of Neurosciences, University of Alicante, 03080 Spain

^bDepartment of Biology, Boston University, Massachusetts

^cDepartment of Neurobiology, University of Oldenburg, Germany

^dDepartment of Physics, University of Oldenburg, Germany

^eDepartment of Physiology, University of Utah, Salt Lake City

Abstract

In the present study we have evaluated the complexity and scaling properties of the morphology of retinal neurons using fractal dimension as quantitative parameter. We examined a large number of cells from Pseudomys scripta and Mauremys caspica turtles that had been labeled using Golgi-impregnation techniques, intracellular injection of Lucifer Yellow followed by photooxidation, intracellular injection of rhodamine conjugated horseradish peroxidase, or intracellular injection of Lucifer Yellow or horseradish peroxidase alone. The fractal dimensions of two-dimensional projections of the cells were calculated using a box counting method. Discriminant analysis revealed fractal dimension to be a significant classification parameter among several other parameters typically used for placing turtle retinal neurons in different cell classes. The fractal dimension of amacrine cells was significantly correlated with dendritic field parameters, while the fractal dimensions of ganglion cells did not vary with dendritic field span. There were no significant differences between the same cell types in two different turtle species, or between the same types of neurons in the same species after labeling with different techniques. The application of fractal dimension, as a quantitative measure of complexity and scaling properties and as a classification criterion of neuronal types, appears to be useful and may have wide applicability to other parts of the central nervous system.

Keywords: fractal, neurons