Documentation(Activity Index): how to implement celeST's Activity Index measure into WF NTP

[jhparkkkk]

User Story

I want to find out how to implement a measure from celeST(Activity Index) into wf ntp script, so that wf_ntp also calculates and returns the same measure as celeST as an output.

Acceptance Criteria

• [x] Understand the definition of Activity Index in CeleST paper (more specificaly the concerned parameter)
• [x] Find out what measures in wf_ntp script could be relevant to calculate activity index
• [x] Write a description of how we calculate the Activity Index in WF_NTP in a markdown docs/features/activity_index.md.
  • [x] theoritical definition of the activity index
  • [x] description of the different parameters in the activity index
  • [x] transcription in a pratical way of the theoritical definition.

Definition of Done

• [x] Define Activity Index from celeST .
• [x] markdown is docs/features/activity_index.md.
• [x] linter is passing

[cnstll]

From what I gathered, Activity index is a measure of the area covered (number of pixels?) over a period of time (per min) by a worm.

Some links I used during my research:

• CeleST website
• CeleST publication paper
• Video representing Activity index measures

[jhparkkkk]

1) Define celeST Activity Index

What is Activity Index ?

Sum of number of pixels that are painted by the body during the time that it takes to an animal to do 2 strokes.

What is it for ?

→ To indicate how vigorously the worm bend while swimming over time. → The bigger the area it takes to swim the more the animal is active (healthy ?)

The Activity Index measure is calculated with the help of another measure : brush strokes which is the area that the body would “paint” (= number of pixels covered) in a single complete stroke.

2) Datas from WF NTP

First, the software identifies individual worms as labeled region giving coordinates of each animal for each frame.

From that, metrics can be defined. Those that could be helpful to use in order to render a measure like Activity Index could be :

• Area / length : size of the worm →in px_to_mm (l. 446)

If we can get the area of each worm for each frame -> We can sum up the total area for a given time

nb pixel / frame -> nb pixel / time

nb frame / sec

[jhparkkkk]

[weekly 3 - 2022.06.30]

CeleST source code

Activity Index is obtained by this formula (cf. CSTCheckResults.m l. 572):

sweeping_noTimeNorm(ff0) = areaCoveredNorm(ff0) temporalFreq(ff02) fileDB(currentVideo).frames_per_second 60

areaCoveredNorm(ff0) = sum(maskAreaCovered(:) > 0) / (averageWormBodyArea / (1+timeWindowForSweptAreas))

[cnstll]

After some research in the source code I can see that the Activity Index is defined with the variable sweeping_noTimeNorm (link to line).

The calculus of sweeping_noTimeNorm is done there : it is a product of a worm area (areaCoveredNorm) by a time frequency (temporalFreq) by the number of frames per min (fps * 60).

temporalFreq seems to be calculated in the function CSTComputeMeasuresFromBody whose definition begins here. ATM, I have a hard time understanding the calculus here as I am missing a lot of mathematical background.

areaCoveredNorm is defined here and depends on maskAreaCovereddefined here and averageWormBodyArea defined here which also depends on averageWormBodyMask.

In a nutshell, the calculus of these variables is still unclear.

[madvid]

don't forget to rebase