Reproducible quantification of cardiac sympathetic innervation using graphical modeling of...

Graphical methods of radiotracer kinetic modeling in PET are ideal for parametric imaging and data quality assurance but can suffer from noise bias. This study compared the Logan and Multilinear Analysis-1 (MA1) graphical models to the standard one-tissue-compartment (1TC) model, including correction for partial-volume effects, in dynamic PET-CT studies of myocardial sympathetic innervation in the left ventricle (LV) using [¹¹C]HED.

Test and retest [¹¹C]HED PET imaging (47 ± 22 days apart) was performed in 18 subjects with heart failure symptoms. Myocardial tissue volume of distribution (V_T) was estimated using Logan and MA1 graphical methods and compared to the 1TC standard model values using intraclass correlation (ICC) and Bland-Altman analysis of the non-parametric reproducibility coefficient (NPC).

A modeling start-time of t* = 5 min gave the best fit for both Logan and MA1 (R² = 0.95) methods. Logan slightly underestimated V_T relative to 1TC (p = 0.002), whereas MA1 did not (p = 0.96). Both the MA1 and Logan models exhibited good-to-excellent agreement with the 1TC (MA1-1TC ICC = 0.96; Logan-1TC ICC = 0.93) with no significant differences in NPC between the two comparisons (p = 0.92). All methods exhibited good-to-excellent test-retest repeatability with no significant differences in NPC (p = 0.57).

Logan and MA1 models exhibited similar agreement and variability compared to the 1TC for modeling of [¹¹C]HED kinetics. Using t* = 5 min and partial-volume correction produced accurate estimates of V_T as an index of myocardial sympathetic innervation.