One-shot Active Neuron Localization with in vivo Fluorescence Image Sequences.

Area of Science:

• Neuroscience
• Computational Biology
• Machine Learning

Background:

• Active neuron localization in vivo is crucial for understanding neural activity.
• Manual observation of fluorescence image sequences is time-consuming and labor-intensive.

Purpose of the Study:

• To develop an automated, one-shot active neuron localization method.
• To reduce manual intervention and labor costs in neuron activity analysis.

Main Methods:

• Formulated the task as a one-shot multiple instance learning (MIL) problem.
• Developed an MIL-based framework integrating sequence proposals and temporal filtering.
• Utilized a deep neural network trained with self-supervision for feature extraction.
• Classified sequences and identified neuron positions by comparing template and proposal features in an MIL fashion.

Main Results:

• The proposed method demonstrated a clear performance advantage over existing comparison methods.
• Achieved accurate active neuron localization with minimal user intervention.

Conclusions:

• The novel MIL-based framework provides an efficient and effective solution for active neuron localization.
• This work lays the foundation for advanced physiological analysis of neuron activity.