Towards a supervised classification of neocortical interneuron morphologies.

Area of Science:

• Neuroscience
• Computational Biology
• Machine Learning

Background:

• Classifying cortical interneurons remains a significant challenge in neuroscience.
• Data-driven classification of neuronal morphology can offer practical value and insights into established types.

Purpose of the Study:

• To develop and evaluate data-driven models for classifying rat somatosensory neocortex interneurons based on their morphology.
• To identify key morphometric features that distinguish different interneuron types.

Main Methods:

• Trained classification models using 217 high-quality interneuron reconstructions.
• Quantified 103 axonal and dendritic morphometrics, including novel features.
• Employed a one-versus-rest classifier strategy with feature selection and sampling techniques.

Main Results:

• Accurate classification achieved for nest basket, Martinotti, and basket cell types, with the Martinotti model outperforming expert neuroscientists.
• Moderate accuracy for double bouquet, small, and large basket types.
• Limited accuracy for chandelier and bitufted types, suggesting areas for future refinement.

Conclusions:

• Approximately 50 high-quality reconstructions are sufficient to learn accurate models for most interneuron types.
• Further improvements may necessitate quantifying complex arborization patterns and bouton-related features.
• The study highlights practical considerations for neuron classification and provides reproducible code and data.