Assessment of lower extremities flow using dynamic Rb-82 PET: Acquisition protocols and quantification methods

53 Background: Quantitative assessment of lower extremity skeletal muscle flow is critical for managing patients with diabetes and peripheral arterial disease (PAD). However, reliable quantitative methods are not well established. In this study, we aim to investigate and optimize data acquisition protocols and quantitative data processing methods for dynamic Rb-82 PET imaging in an established porcine model of PAD through tracer kinetic modeling. Methods: Dynamic Rb-82 PET imaging was performed in five pigs following acute unilateral femoral artery occlusion using a 4-ring Siemens Biograph mCT scanner with continuous bed motion (CBM) and Jubilant Rb-82 generator, with additional pig and human studies ongoing. Rb-82 (518±37 MBq) was delivered using a constant activity delivery protocol over 45 seconds per injection. In each study, multiple sequential dynamic PET scans were acquired using several acquisition protocols that employed both a single bed position and/or CBM. With ongoing analysis for all protocols, we focus on reporting 3 protocols: 1) 7-min single bed position dynamic scan of the heart to derive input function from left ventricle (LV) blood pool (35 frames, 5s/frame for the first 90s, then 30s/frame); 2) 7-min single bed position dynamic scan of the legs (the same as above); and 3) 1.5-min single position scan (5s/frame × 24 frames) of the lower abdominal aorta (AA) followed by 5.5-min CBM scans (30s/frame ×11 frames) between AA and the legs, with input function derived from AA. Protocols 1 and 2 were performed under stable resting conditions, while acquisition protocol 3 was performed both at rest and during adenosine-induced vasodilation. Arterial blood activity was continuously sampled using an automated blood counter, and these data were used as the gold standard input function for each scan. A one-tissue compartmental model with blood volume term was used to quantify K1. Image derived arterial input functions from LV and AA were compared with those obtained through the continuous input function using IDL 8.0 and MatLAB 2020b. Results: High quality voxel-by-voxel parametric K1 images of the legs were generated. K1 values forskeletal muscle derived from Protocol 2 data using LV input function from Protocol 1 and 2 scans across the five pigs are 0.070±0.041 mL/min/cm3 and 0.030±0.012 mL/min/cm3 for the sample ROIs in the non-ischemic legs and ischemic legs, respectively (p<0.05). The image-derived input functions (IDIF) from LV are consistent with those of arterial blood samples. The peak input function derived from AA was consistently 60.0±0.3 % of the LV for all pigs, with lower terminal activity from AA, likely due to partial volume effect. The AA input functions from stress scans have similar peaks compared to those of the rest scans, but with slightly higher residual terminal activity. Further investigation is ongoing to quantify K1 based on data acquired on other acquisition protocols. Conclusions: It is feasible to quantify skeletal muscle blood flow in the lower extremities using dynamic Rb-82 PET. Optimal data acquisition protocols that take advantage of CBM, constant activity infusion, and an image derived input function, and tracer kinetic modeling methods need to be established to ensure accurate and reproducible quantification of lower extremities flows in setting of PAD.