Simplifying our understanding of SRTM

Objectives The simplified reference tissue model (SRTM) is a valuable tool for the quantitative estimation of binding potential (BP_ND) from dynamic PET. SRTM obviates the need for arterial blood data and maintains quantitative accuracy provided its model assumptions are met. SRTM assumes a single tissue compartment kinetically describes the reference and target tissues, the blood volume contribution is negligible, the non-displaceable distribution volume is homogeneous and the reference tissue is devoid of specific binding. We characterise the bias introduced when these assumptions are not met with a particular focus on invalid compartmental topologies.

Methods Computer simulations were used to simulate a wide range of 1- and 2-tissue compartmental (1TC & 2TC) behaviour in both the reference and target tissues by generating 25,000 paired realisations with different sets of rate constants (K₁,k₂,k₃,k₄) and a standard plasma input function. For each tissue the Model Order Metric (MOM) was calculated, to measure how closely the data conformed to a 1TC, as the integral of the absolute difference between the two exponential impulse response function of the 2TC model and the best one exponential approximation from the 1TC model. For each realisation pair, SRTM was applied to estimate BP_ND^Est and the true BP_ND was calculated via the ratio of the 2TC distribution volumes (BP_ND^True = (V_T^Tar - V_T^Ref)/ V_T^Ref) which then allowed for an estimate of the SRTM bias as BP_ND^Bias = (BP_ND^Est - BP_ND^True)/ BP_ND^True.

Results When MOM indicated 1TC for both tissues, consistent with SRTM assumptions, the BP_ND^Bias was negligible. For other configurations the average BP_ND^Bias exceeded 30% when two compartments were required for the reference tissue. The BP_ND^Bias was lower when 2 tissue compartments were required for the target. Specific binding in the reference region and blood contribution to the tissues both yield underestimates in BP_ND^Est.

Conclusions SRTM performs well when there is a single tissue compartment in the target region but is particularly prone to bias when 2 compartments are required in the reference.