FAQ
https://github.com/zhoupc/CNMF_E/wiki/Understand-CNMF-E-results
Q2: what exactly is C_raw and C? How are they connected to the conventional DF/F? How can we get DF/F?
C_raw is a scaled version of the conventional DF, i.e., the change of the fluorescence intensity over time. It is scaled because different pixels within the ROI have different brightness while they share the same temporal profile.
C is the denoised version of C_raw. The denoising is achieved by modeling the calcium trace as a convolution of the calcium events and a temporal kernel. Both the temporal kernel and the spiking events are estimated by fitting the model. The calcium events were given by neuron.S.
In the conventional calcium imaging analysis, DF/F0 is a widely used metric. Since F0 is a constant, C_raw can be seen as a scaled version of DF/F0 as well.
We don't encourage using the DF/F0 for the analysis of microendoscopic data because F0 has no meaning here. In the typical 2p data analysis, F0 can be interpreted as the intracellular calcium concentration in the resting state and DF/F0 gives the relative change of the calcium concentration. While in 1p data analysis, F0 is dominated by the background level, which accumulates the fluorescence signals in a broad area and has no connections with the intracellular calcium concentrations. Thus scaling DF with F0 is meaningless for 1P analysis.
We suggest to use noise-scaled (or z-scored) C_raw or C to replace DF/F. The noise is estimated from C_raw using power-spectrum-density (PSD) method. This metric tells you how large is each calcium transients in relative to the noise level.