Previous Topic

Next Topic

Book Contents

Book Index

Ichise's Multilinear Reference Tissue Model (MRTM)

To reduce noise-related bias effects arising in the MRTM0 method Ichise et al. [33] applied a strategy known to be effective in reducing the noise-induced bias for the models requiring blood data. To this end the equation of the MRTM0 was rearranged to remove the noisy tissue radioactivity term C(t) from the independent variables. This approach resulted in a new method called MRTM with operational equation

Equation MRTM

C(t) is the TAC from a receptor-rich region (k3>0), and C'(t) the TAC from a region without receptors (k3=0 in the 2-Tissue compartment model).

The multi-linear relationship above can be solved using multi-linear regression, yielding three regression coefficients. The binding potential can then be calculated from the first two regression coefficients g1 and g2 by

Equation MRTM BP

Furthermore, division of the first by the third regression coefficient yields an estimate of k2' .

Implementation Notes

After switching to the Ichise NonInvasive MRTM in PKIN a suitable reference region must be selected. It allows to fit a multilinear regression within a range starting at the parameters Start Lin. The results are the three regression coefficients, and the derived binding potential BP.

There is also an error criterion Max Err. to fit Start Lin. For instance, if Max Err. is set to 10% and the fit box of Start Lin. is checked, the model searches the earliest sample so that the deviation between the regression and all measurements is less than 10%. Samples earlier than the Start Lin. time are disregarded for regression and thus painted in gray.

Note: The reference methods MRTM2 and SRTM2 require k'2 as an input parameter. The k2' resulting from the MRTM method above is a suitable estimate. Therefore, when switching in PKIN from the MRTM model to MRTM2 or SRTM2, k2' is automatically copied from MRTM, as long as Model conversion in the Configuration menu is enabled. See also.

Abstract [33]

"We developed and applied two new linearized reference tissue models for parametric images of binding potential (BP) and relative delivery (R1) for [11C]DASB PET imaging of 5-HT transporters in human brain. The original multilinear reference tissue model (MRTM0) was modified (MRTM) and used to estimate a clearance rate (k'2) from the cerebellum (reference). Then, the number of parameters was reduced from three (MRTM) to two (MRTM2) by fixing k2' . The resulting BP and R1 estimates were compared with the corresponding nonlinear reference tissue models, SRTM and SRTM2, and one-tissue kinetic analysis (1TKA), for simulated and actual [11C]DASB data. MRTM gave k'2 estimates with little bias (<1%) and small variability (<6%). MRTM2 was effectively identical to SRTM2 and 1TKA, reducing BP bias markedly over MRTMO from 12-70% to 1-4% at the expense of somewhat increased variability. MRTM2 substantially reduced BP variability by a factor of 2-3 over MRTM or SRTM. MRTM2, SRTM2 and 1TKA had R1 bias < 0.3% and variability at least a factor of 2 lower than MRTM or SRTM. MRTM2 allowed rapid generation of parametric images with the noise reductions consistent with the simulations. Rapid parametric imaging by MRTM2 should be a useful method for human [11C]DASB PET studies.