MorphSys: a branch-aware contrastive learning framework for neuron morphology graphs.

Area of Science:

• Neuroscience
• Computational Biology
• Machine Learning

Background:

• Neuron morphology is crucial for cell identity and function, aiding in cell type classification and neurological disorder diagnosis.
• Challenges in neuron morphology analysis include learning from complex tree structures and the high cost of expert annotation.

Purpose of the Study:

• To introduce MorphSys, a self-supervised contrastive learning framework for effective neuron morphology representation learning.
• To address limitations of existing methods in capturing complex neuron structures and reducing annotation dependency.

Main Methods:

• MorphSys utilizes a Branch-Aware module with Inter-Branch Attention for branch-level representation, capturing inter-branch relationships.
• A Graph Neural Network (GNN)-based module learns local features, with GatedGraphConv and SumPool showing superior performance.

Main Results:

• MorphSys consistently outperforms existing methods in neuron cell type classification across benchmark datasets.
• Achieved 83.99% KNN-Acc on the BIL dataset, surpassing the state-of-the-art by 2.99%.

Conclusions:

• MorphSys is an effective tool for neuron morphology representation learning.
• Demonstrates the power of branch-level features and self-supervised learning for neuronal analysis.