aind-ophys-dff

This capsule calculates $\Delta F/F$ using the dff function from aind-ophys-utils, which executes the following steps:

• Estimate the noise standard deviation $\sigma$ and initial baseline $b$
• Mask active frames, i.e. outliers where $F > b + 3\sigma$
• Obtain $F_0$ by median-filtering the trace using only inactive frames and interpolation
• Calculate $\Delta F/F = (F - F_0) / F_0$

Input

All parameters are passed to dff.py using python dff.py [parameters]. All parameters are defined in main using argparse. The most important one is 'input-dir' which should point to a directory containing an HDF5 file extraction.h5 with the dataset 'traces/corrected', a 2D array of neuropil-corrected traces for each ROI.

Output

The main output is the dff.h5 file. It contains 4 datasets:

data: Baseline-corrected fluorescence traces $\Delta F/F$
baseline: Estimated baselines $F_0$
noise: Estimated standard deviation of the noise in the input traces
skewness: The skewness of the $\Delta F/F$ traces