SCALPEL: EXTRACTING NEURONS FROM CALCIUM IMAGING DATA.

Area of Science:

• Neuroscience
• Computational Biology
• Data Science

Background:

• Recent advancements enable large-scale neuronal activity imaging in behaving animals.
• Publicly available calcium imaging datasets present opportunities for new analytical methods.
• Accurate neuron identification is a critical first step in analyzing calcium imaging data.

Purpose of the Study:

• To develop a robust method for identifying neurons in calcium imaging videos.
• To address the analytical challenges posed by large-scale neuroscience data.
• To provide an accessible tool for the neuroscience community.

Main Methods:

• A dictionary learning approach is proposed for neuron identification.
• Image segmentation is used to create a dictionary of candidate neurons.
• Clustering refines the dictionary, followed by sparse group lasso optimization for neuron selection and activity estimation.

Main Results:

• The proposed method demonstrates effective neuron identification on simulated data.
• The approach was successfully applied to three real-world calcium imaging datasets.
• Performance was validated, showing the utility of the dictionary learning technique.

Conclusions:

• The developed dictionary learning method provides an effective solution for neuron identification in calcium imaging.
• The R package 'scalpel' facilitates the application of this method for analyzing large-scale neuroscience data.
• This work contributes to advancing statistical methods for interpreting complex neural activity.