About 'res.dF' and 'res.dffMatFav' about aqua

[freemanwyz]

Hi Yue,

Is res.dF A three-dimensional matrix of ΔF/F0 for the entire image sequence.

res.dF is the result of the first step of AQuA, which removes the background (in a rough way). So it is just ΔF.
In more detail, it cut the video to shorter time durations to alleviate the impact of the global trends.

We found it have similar waveform but showed different ΔF/F0 curve in res.dffMatFav(:,:,1) (curve is higher) and averaged res.dF in res.ftsFav.loc.x2D (curve is lower) formed 3D cylinder formed. So, how to derive from res.dF to res.dffMatFav(:,:,1)?

In res.dffMat (as well as res.dffMat), we first fit a polynomial curve to approximate the global trend of each event (using raw data, not based on res.dF). After removing this trend, we use its minimum value (along with the background fluorescence value) as F0, and calculate ΔF/F0 for that event. res.dffMatFav(:,:,1) and res.dffMatFav(:,:,2) differs in the way it deals with the impact of other events. In res.dffMatFav(:,:,2), the influence of other events is removed by setting these values to NaN.

Usually, the base F0 intensity is low, so ΔF/F0 peak value is likely to be higher than ΔF.

In terms of your code, since you use res.dF, it is just ΔF, so you need to divide F0 yourself, maybe from raw data.

Thanks,
Yizhi Wang

[invincibleyue]

Dear Dr. Wang
  Thank you very much for your timely reply!
•  The mean is about 10000 to 50000 if I use one frame of ** res.datOrg** as F0 which has no signal.
•  Did you Normalize or use other calculations in ** res.datOrg** to processing F0？
•       Best Wishes ,
•       Yue 20210930
 

[freemanwyz]

If your movie does not have some global trend (like bleaching), I think using a frame with no signal as F0 is good.

We calculate F0 by fitting a polynomial curve for each event, which may remove the trend. Then we smooth the resulting curve and take the minimum value of the smoothed curve. If the events are very dense, we optionally remove these interfering events from the current event of interest. But if your data does not have these problems, your strategy is just fine.

[invincibleyue]

Thanks a lot
if we want to show a ΔF/F0 image in some frame , What algorithm do you suggest we use to make the ΔF/F0 frame image same as res.dffMatFav numerically? I want to use res.dF divide F0 (But averaged res.datOrg is about 30000 )
Yue

[freemanwyz]

I think your method is fine. Just choose a frame in raw data (res.datOrg) without signal. You don't need to use res.dF. Minus that frame from raw data, then divide by that frame.

To make the values numerically consistent, I suggest calculating per event ΔF/F0 the same way as above. For example, just extract the per event ΔF/F0 from the above-calculated maps. You can just ignore res.dffMatFav (due to the choice of per event F0 and the de-trending, they are not likely to be numerically the same as above).

[invincibleyue]

Thanks a lot !!