RE: [NMusers] How to model observed quantity tha is an average over compartments

Thanks, Doug, four your input


I wrote:
"As a first step, assuming that T2 is the only observed quantity gives an
excellent fit, so the mix-in is probably low."
As a reminder for others: here is the model again:

|
C <> T1 <> T2
|

-----

You wrote:

Model 1: Assume mix-in is low -> excellent fit
Model 2: Assume mix-in is high -> bad fit

---
I don't even need a computer to see that this is true. Our observed data are
so perfectly bi-exponential that they could serve as a textbook example;
assuming that we observe T1 alone would pass reviewers without problems when
we only think mathematics. If the mix-in where high, we would have a more
mono-exponential looking curve (or am I of the track here?).

The crux is that we KNOW that T2 is observed from biology, it must be in the
region of interest of MRI, not somewhere in other parts of the body.
Therefore, we are not really happy with the perfect fit. Under the
assumption that only T1 is observed, the computed volume of T2 is about 10
times higher than that of T1, which partly explains the low contribution,
and is the implicit assumption made by single-fitters.

We are looking for some excuse (=rejection of alternate model) to justify
the "T1 is only observed is good enough" idea.

In reality, we have additional data from simultaneously measured reference
tissues, so the overall model is numerically well locked.

Dieter




Received on Fri Jul 30 2010 - 05:21:43 EDT