Skip to Main Content
Advances in microscopy and biochemistry now allow investigators to image the calcium dynamics of hundreds to thousands of neurons in awake behaving animals. However, as speed and resolution of such techniques rapidly increase, so do the dimension and complexity of the data collected. ICA has been widely employed to reveal independent non-Gaussian sources underlying large data sets consisting of mixed sources. We apply a recently developed sparse regression method, the Elastic Net (ENET), to the columns of the mixing matrix of a independent component analysis (ICA) procedure. This regression method automatically selects only those columns of the mixing matrix relevant to a dependent variable of interest. Further, because ICA is a linear operator, we can easily project the ldquorelevance filteredrdquo data back into the native data space for interpretation. We demonstrate the utility of this method on 3D calcium imaging data collected from the optic tectum of an awake behaving larval zebrafish watching a prey-like stimulus.