Morphological Neuron Classification Using Machine Learning.

Area of Science:

• Neuroscience
• Computational Biology
• Machine Learning

Background:

• Accurate classification of neuronal morphologies is crucial for understanding brain structure and function.
• Current methods for neuronal characterization face challenges in defining cell classes and selecting optimal features.
• Morphological data is vital for anatomical comparisons, morphometric analysis, and brain modeling.

Purpose of the Study:

• To develop and integrate a computational pipeline for extracting features and classifying neuronal morphologies.
• To assess and compare the performance of various machine learning algorithms for neuron classification.
• To identify the most effective algorithms for classifying rat somatosensory cortex neuron morphologies.

Main Methods:

• Utilized a dataset of 430 digitally reconstructed neurons from rat somatosensory cortex.
• Employed a pipeline integrating morphological feature extraction and classification.
• Trained and compared supervised (Linear Discriminant Analysis) and unsupervised (Affinity Propagation, Ward) machine learning algorithms.

Main Results:

• Linear Discriminant Analysis achieved superior classification accuracy among supervised algorithms.
• Affinity Propagation and Ward algorithms demonstrated slightly better performance in unsupervised classification.
• The developed pipeline successfully enabled quantitative characterization and classification of neuronal morphologies.

Conclusions:

• The study presents a validated pipeline for neuronal morphology classification.
• Linear Discriminant Analysis is recommended for supervised classification of these neuron types.
• Unsupervised methods like Affinity Propagation and Ward offer viable alternatives for neuron morphology classification.