Wow, I wasn’t aware of those findings regarding S0 and T2*. That’s really interesting!
I think @handwerkerd, @dowdlelt, and @CesarCaballeroGaudes might be able to weigh in, but my first thought is that you might need more than three echoes for volume-wise T2* and S0 estimation. Also, if your goal is T2*/S0 estimation rather than multi-echo denoising, you may want to use a dataset with TEs that go up to 1.5x the highest T2* value you’d expect to see (that came up in Compare two protocols for TEDANA - #15 by dowdlelt).