Towards the automatic classification of neurons

Abstract

The classification of neurons into types has been much debated since the inception of modern neuroscience. Recent experimental advances are accelerating the pace of data collection. The resulting growth of information about morphological, physiological, and molecular properties encourages efforts to automate neuronal classification by powerful machine learning techniques. We review state-of-the-art analysis approaches and the availability of suitable data and resources, highlighting prominent challenges and opportunities. The effective solution of the neuronal classification problem will require continuous development of computational methods, high-throughput data production, and systematic metadata organization to enable cross-laboratory integration.

Keywords: big data; machine learning; metadata; neural classification; standardization.

Figure 1

Basic dimensions of neuronal characterization: morphology (yellow), physiology (green), and biochemistry (blue). These feature domains are tightly interrelated with other fundamental aspects of neural identity, such as connectivity, development, and plasticity.

Figure 2

Major classification approaches with representative families of algorithms. Examples of implementations with references and links to available resources are provided as Supplementary Material.

Figure 3

Optimal classification balance of neuronal lumping and splitting. Neuron types can be defined at higher or lower resolution in each data domain. The best explanatory power maximizes the trade-off between description complexity and captured generalization.