The morphology of dendritic spines is highly correlated with the neuron function. Therefore,
it is of positive influence for the research of the dendritic spines. However, it is tried to manually label
the spine types for statistical analysis. In this work, we proposed an approach based on the combination
of wavelet contour analysis for the backbone detection, wavelet packet entropy, and fuzzy support vector
machine for the spine classification. The experiments show that this approach is promising. The
average detection accuracy of “MushRoom” achieves 97.3%, “Stubby” achieves 94.6%, and “Thin”
Keywords: Dendritic spines, discrete wavelet transform, fuzzy support vector machine, wavelet packet entropy.
Rights & PermissionsPrintExport