Classification of adult human dentate nucleus border neurons: Artificial neural networks and multidimensional approach.

AIMS

Primary aim in this study is to investigate whether external and internal border neurons of adult human dentate nucleus express the same neuromorphological features or belong to a different morphological types i.e. whether can be classified not only by way of their topology as external and internal, but also based on their morphological features or in addition to their topology also by way of their morphology. Secondary aim is to determine and compare various methodologies in order to perform the first aim in a more accurate and efficient manner.

MATERIAL AND METHODS

Blocks of tissue were cut out from the adult human cerebellum and stained according to the Kopsch-Bubenaite method. Border neurons of the dentate nucleus were investigated and digitized under the light microscope and processed thereafter. Seventeen parameters quantifying various aspects of neuron morphology are then measured. They can be categorized as shape, magnitude, complexity, length and branching parameters. Analyzes used are neural networks, separate unifactor, cluster, principal component, discriminant and correlation-comparison analysis.

RESULTS

The external and internal border neurons differ significantly in six of the seventeen parameters investigated, mainly concerning dendritic ramification patterns, overall shape of dendritic tree and dendritic length. All six methodological approaches are in accordance showing slight clustering of data. Classification is based on six parameters: neuron (field) area, dendritic (field) area, total dendrite length, and position of maximal dendritic arborization density. Cluster analysis shows two data clusters. Separate unifactor analysis demonstrates inter-cluster differences with statistical significance (p < 0.05) for all six parameters separately. Principal component, discriminant and correlation-comparison analysis further prove the result on a more factor integrate manner and explain it, respectively. Thus, these neurons can be classified, not only according to their location but also according to some morphological features. Also, the group if internal border neurons is more homogeneous in itself than the other group of external border neurons.

CONCLUSION

Border neurons from adult human dentate nucleus can be divided to external and internal according to its topology and based on neuromorphological computational parameters. This has potentially significant neurofunctional implications but further studies are needed to elucidate that. Multimethodological approach is shown as the best for finding the solution closest to reality. The possible functional meaning of these morphological differences for cerebellar network structure and function are discussed.