Noise modeling

[jeremieglt]

Hi !

I have been using the SIMPA software for a research project for quite a while, but there is something that you use and on which I can't find any documentation : when you add noise to the initial pressure (post MCX simulation), it is added in a multiplicative fashion, but when it's added to the sinograms, it is additive (e.g. in simpa examples > optical_and_acoustic_simulation). It must come from the physical modeling of the MCX and kWave simulations but I never found a paper to explain that. Would you have an explanation ?

Thanks in advance,
Jérémie

[kdreher]

Hi Jérémie,

the different ways of incorporating noise come from practical considerations. In our case:

• Initial pressure: We assume that the noise comes from variations in the laser input energy. Meaning that maybe the laser has a nominal energy of let's say 9 mJ but it varies to some degree (maybe +- 0.1 mJ). This variation in input energy will affect the initial pressure via the fluence in p0 ~ mu_a * fluence. Since this is multiplicative, we multiply by Gaussian noise.
• Time series: We assume the source of noise is the dark current in the detector which can for example be measured in a water bath or similar. To simulate this, we introduce an additive offset of Gaussian noise.

Cheers,
Kris

[jeremieglt]

Ok thanks !

[kdreher]

By the way, if you like simpa, then leave a star for the repo! 😉