Iterative non-linear least analysis for robust estimation of regional MBF with 82Rb PET

Objectives It is hard to estimate regional myocardial blood flow (MBF) with 82Rb. We developed the iterative non-linear least analysis which allowed robust quantification of regional MBF.

Methods Twenty volunteers underwent 15O-water PET to estimate reference MBF and 82Rb PET. The myocardial ROI curve, left ventricular ROI curve, the tissue fraction in the myocardial ROI and spillover into myocardium were determined as R(t), LV(t), A and VA, respectively. True curves of the blood Ca(t) and myocardium Ct(t) were estimated as follows; dCt(t)/dt=K1Ca(t)-k2Ct(t), R(t)=ACt(t)+VACa(t) and LV(t)=BCa(t)+(1-B)Ct(t). The value B denoted a recovery coefficient of the blood. These equations estimated A, VA, K1, k2 and Ct(t). Ct(t) was re-calculated (Ct(t)’) with these A, VA, K1, k2 estimates. Iterative non-linear least analysis was performed to minimize the error between Ct(t) and Ct(t)’ using different initial values. These estimates were compared with those from the conventional method. A simulation study was performed to compare the effects of errors in the tissue fraction on the K1 with the two methods.

Results Our method yielded more correlated regional K1 with MBF (r=0.89) than the original method (r=0.37) (p<0.001), and yielded less deviated regional K1 (CV=25%) than the original method (CV=43%) (p<0.001). The simulation study demonstrated that 10% underestimation in the perfusable tissue fraction yielded 24% and 34% overestimation on the K1 value with the new and original methods, respectively.

Conclusions We developed a new method to estimate regional K1 of 82Rb, which yielded well correlated K1 with 15O-water MBF.