TABLE 1:

Cega thresholding values used for experimental data.

Step	Optimal values estimated for experimental data	How to estimate values	Additional considerations for optimal use
Pixel calibration	Offset = 2293 and Gain = 71	Estimated from series of dark frames from EMCCD camera (see Materials and Methods).	Data must be calibrated before Cega. Calibration is specific to the camera used.
Stationary model estimation	Sliding temporal window median filter = 31 frames.	Duration of >95% of moving motors of interest.	If tracking particles that move for long periods of time or periodically pause, choose a window size encompassing the duration of >95% of the particles of interest.
Motion model estimation	Spatiotemporal gaussian blur = 5 frames	Duration that >95% of moving motors remain within 2 adjacent pixels.	Smaller temporal kernel provides better computational efficiency.
KL divergence	No user defined parameters are required.	No user defined parameters are required.	Stationary and motion models are used as input.
Connectivity filter	3 connected pixels >0.1 nats	Threshold set to 95th percentile of connectivity model.	Additive salt noise results in many false positives. This noise is removed before candidate finding by applying a threshold.
LoG filter	3 × 3 neighborhood pixel sum from connectivity filter >5 nats.	Threshold set to 95th percentile of LoG model.	The LoG image sequence is used to find initial local minima. Then, connectivity image sequence values for these positions are used to threshold local minima appropriately.