[Feature] <SNR scaling>

[oliverchampion]

Feature description

As we use the realistic phantom for unit testing, SNR is poorly defined. The phantom uses T1/T2 values to generate realistic tissue contrasts. This means that "noise" is added to the signal, but that this does not mean "SNR" is 1/noise. It also means that SNR changes per organ (depending on their T1/T2 values). It makes the phantom more realistic, but the SNR is harder to define.

Describe the solution

For our paper, I think it is "easier" to have unit tests at fixed SNR-levels, instead of fixed noise levels.
To me there are two solutions;
1: We could simulate a noise-less signal curve using the phantom, and then rescale per tissue and add noise for unit testing purposes.
2: We could have an option to turn off T1/T2 weighting in the phantom and generate the json files with this option turned off.

Describe alternatives

No response

Additional context

No response

Are you working on this?

Yes

[oliverchampion]

@etpeterson I tagged you as the unit testing is more of your IP. I will try and work on option 2 :). Let me know if it is a bad idea :)

[oliverchampion]

This should be solved in my latest pull request :).

[etpeterson]

This should be solved in my latest pull request :).

So this is being worked on? If so, you should assign it to yourself so nobody else starts work on it too.

[etpeterson]

@etpeterson I tagged you as the unit testing is more of your IP. I will try and work on option 2 :). Let me know if it is a bad idea :)

Not actually sure how the two options differ. It sounds like both generate noiseless phantom data and then the noise is added again later. I think maybe option 1 is still generating noisy json data and 2 doesn't? If that's the case then I think I'd prefer 2 so we can scale the noise on the test side more easily.