CalTrig: A GUI-based Machine Learning Approach for Decoding Neuronal Calcium Transients in Freely Moving Rodents

Area of Science:

• Neuroscience
• Computational Neuroscience
• Bioinformatics

Background:

• Miniature microscope imaging allows studying single-neuron activity in freely-moving animals.
• Current tools (MiniAN, CalmAn) convert visual signals to numerical data (CalV2N), but analysis challenges persist.
• Analyzing large CalV2N datasets requires integrating data streams, quality evaluation, and efficient transient identification.

Purpose of the Study:

• Introduce CalTrig, an open-source GUI tool for post-CalV2N data processing.
• Address challenges in integrating data streams, evaluating CalV2N output quality, and identifying neural transients.
• Provide a user-friendly and efficient solution for neuroscientists.

Main Methods:

• Developed CalTrig, a graphical user interface (GUI) integrating imaging, neuronal footprints, traces, and behavioral data.
• Implemented capabilities for evaluating CalV2N output quality and synchronized visualization.
• Evaluated four machine learning models (GRU, LSTM, Transformer, Local Transformer) for neural transient detection.

Main Results:

• CalTrig integrates multiple data streams and enables synchronized visualization and efficient transient identification.
• The GRU model demonstrated the highest predictability and computational efficiency for transient detection.
• GRU model achieved stable performance across different training sessions, animals, and brain regions.

Conclusions:

• CalTrig offers a flexible and accurate solution by integrating manual, parameter-based, and machine learning detection methods.
• The tool's user-friendly interface and low computational demands make it accessible to researchers without programming expertise.
• CalTrig facilitates deeper exploration of brain function, hypothesis generation, and understanding of neurological disorders.