Orientation issues

[AbdulAlaqil]

Dear ActiPASS Community,

I hope you are all well.

I have been processing my accelerometry data in ActiPASS (using ProPASS mode) and have encountered some participant data issues. I would like to know if anyone has faced a similar issue or knows how to handle it without losing the data.

The problem is that I have noticed some data showing an unusually high "SleepInterval" (>21 hours) in ActiPASS, while using PAL analysis has yielded different results (please see an example below).

Could someone please provide an explanation?

Thank you very much,

Abdul

[pasan-h]

Dear ActiPASS Community,

I hope you are all well.

I have been processing my accelerometry data in ActiPASS (using ProPASS mode) and have encountered some participant data issues. I would like to know if anyone has faced a similar issue or knows how to handle it without losing the data.

The problem is that I have noticed some data showing an unusually high "SleepInterval" (>21 hours) in ActiPASS, while using PAL analysis has yielded different results (please see an example below).

Hi Abdul, could you please follow the instructions below and post the correct quality check figure and batch-settings file. It's easier to understand what's going on with those extra information.
#55

[AbdulAlaqil]

Hi Pasan,

Please see below the correct quality check figure, and batch setting file.

Quality check

[pasan-h]

Thanks 👍. It seems ActiPASS cannot identify device orientations properly (auto flips/rotations detection has failed). This means whole activity detection is wrong. So this is not just a sleep-interval problem. Also I can see that you have specified both "Flipped" and "Rotated" as true. Why is that?
Could you rerun this file again with "Flipped" and "Rtated" checkboxes unticked and see what happens? For activpal devices we should not normally need to tick "Rotated" and "Flipped" checkboxes.
Also see: https://github.com/Ergo-Tools/ActiPASS/wiki/ActiPASS-FAQ-and-troubleshooting#if-actipass-has-automatic-orientation-detection-why-is-it-necessary-to-specify-orientation

[AbdulAlaqil]

Thanks, Pasan. I believe it works now correctly and I can see more activites. Just to clarify. for ActivPAL devices, do you recommend to uncheck the "Flipped" and "Rated" checkboxes?

[pasan-h]

Yes, for ActivPAL devices both boxes should be unchecked. Perhaps a long time ago, we had to mark "Flipped" checkbox for ActivPAL devices. But not anymore. If we know that devices are worn in other orientations than the defaults, then we can use those checkboxes. For the default orientations see:
https://github.com/Ergo-Tools/ActiPASS/wiki/ActiPASS-accelerometer-default-orientations

[camillebiemans-oss]

Hello everyone,

I'm encountering a similar problem as AbdulAlaqil.

For some of my files, it keeps giving the quality check error 'NoWalking'. When looking at the data in PALanalysis, there did occur walking during the day for these people. However, once processed in ActiPASS, I get the error 'NoWalking'.
PALanalysis screenshot:

I also got the orientation warning: 'Selected default orientations for the batch... Rotated (upside-down) = 100%'.
Once I've changed the orientation towards rotated (screenshot), I still get the 'NoWalking' error and the visual inspections showed no difference. Underneath one example (step count = 0):

Batch Settings with orientation on 'rotated':
ActiPASS_BatchSettings.txt

Batch Settings with orientation on 'default mode' (so unthicked):
ActiPASS_BatchSettings.txt

Could someone please think along with us? That would be great. Thank you in advance!

Camille

[pasan-h]

Hi,
For PAL devices (if you follow manufacture recommended orientation), run ActiPASS with both rotated and flipped check-boxes unticked. First process all files with this default settings. Then find out which files in the batch has wrong-classifications (no-walk etc). Try to process only those files with rotated checkbox clicked (in a separate batch, but use the same project folder).

This behaviour is expected (for a few cases) when some devices are attached up side down. If you cannot fix them (if the classification doesn't make sense) just exclude them from the analysis. Similarly even with PAL-analysis one can expect some misclassifications when users wear devices in awkward orientations (or loose devices), but perhaps they are not always apparent like in ActiPASS.
BR
Pasan

[pasan-h]

Could someone please think along with us? That would be great. Thank you in advance!

Also try enabling advanced mode and then disabling auto orientation corrections (under advanced settings) for these problematic cases.

[camillebiemans-oss]

Dear @pasan-h,

Thank you so much for your reply. I'm trying to run the advanced mode, and use the files that you've sended me. I've followed all steps for activating the advanced license. But unfortunately, the 'advanced options' do not appear, see below:

The files have been updated, and I've restarted the ActiPASS software:

The only difference that I can find with my data and the explanation is that I don't have the file 'ActiPASSQC.conf'. Could that be the problem?

Thank you so much again, really appreciated.

Camille

[pasan-h]

Hi Camille,
Did you disable ProPASS mode? Advanced mode is enabled once you disable ProPASS mode.
See FAQ:
https://github.com/Ergo-Tools/ActiPASS/wiki/ActiPASS-FAQ-and-troubleshooting#how-do-i-enable-actipass-advanced-mode-of-processing

Pasan

[camillebiemans-oss]

Thank you so much!!! This works. @pasan-h

[pasan-h]

@camillebiemans-oss
I've had time to think about the issue you encountered (differences between ActiPASS and PAL analysis). The problem could be also caused by the cohort-profile. If participants do not have a normal walking gait (like healthy adults) but somewhat weak walking posture (elderly or sick), ActiPASS fails to detect walking and thus it's auto reference-position corrections go wrong. This can also lead to the situation seen above.

In this case I think ActiPASS is not capable correctly distinguishing behaviours. Then we can disable some features and use default reference positions like seen below. Otherwise just stick to PAL Analysis 😀.