SpectralSegmentation

Spectralsegmentation is a pipeline that can be used to detect active neurons and dendrites in ca-imageing data. A series of steps are defined to achieve this. After image stabilization and transposing an image sequence to the (time x pixel) space, Cross-spectral analysis is applied to low frequency(<1Hz) decimated pixel traces. This results in images representing the cross spectral power of each pixel with it's surrounding pixels at increasing frequency components (0.017Hz steps - 0.4Hz). These images are used to define preliminary ROIs using morphological criteria. The ROIs are then constrained to contain only pixels with possitive correlations.

The pipeline includes a graphical user interface to edit the automatically extracted ROIs, to add new ones or delete ROIs by further constraining their properties.