Deep learning models for atypical serotonergic cells recognition.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Machine Learning

Background:

• The serotonergic system uses diverse neurons with varied electrophysiological properties.
• Current methods struggle to identify atypical serotonergic neurons, limiting research into their function.

Purpose of the Study:

• Develop deep learning models to accurately distinguish typical and atypical serotonergic neurons from non-serotonergic cells.
• Enable comprehensive study of the functional relevance of diverse serotonergic neuron activities.

Main Methods:

• Utilized in vitro electrophysiological recordings from identified serotonergic and non-serotonergic neurons.
• Employed convolutional neural networks (CNNs) for classification based on action potential characteristics.
• Trained models on a large dataset of real and synthetic action potential samples.

Main Results:

• Deep learning models achieved high accuracy in discriminating serotonergic neuron subtypes.
• Models demonstrated robustness and reliability when validated on independent, non-homogeneous datasets.
• Achieved over 94% accuracy in recognizing serotonergic neurons, including atypical ones.

Conclusions:

• The developed model reliably identifies both typical and atypical serotonergic neurons.
• The model is ready for experimental use and adaptable to various recording parameters.
• Facilitates broader investigation into the neurobiological roles of serotonergic system diversity.