Compartment model corrected blood input function estimation improves with iterative image reconstruction and cardiac gating

Objectives To evaluate that compartment model corrected blood input function (MCBIF) estimation (1), from small animal PET images in mice, in-vivo, improves with OSEM-MAP and cardiac gating (2) as compared to un-gated FBP reconstructed images.

Methods Imaging of mice (n=3) was initiated under isoflurane anesthesia with the microPET scanner right after FDG administration to capture the initial time points. All the images acquired for an hour were reconstructed using OSEM-MAP into 23 time bins and 3 cardiac bins and into 23 time bins with no gates using FBP reconstruction algorithm. Blood samples were also obtained for each of the mice at late time points where the spill-over (SP) of radioactivity from the myocardium to the blood pool is expected to be high. Time activity curves were generated by drawing regions of interest in the blood-pool (BP) and the myocardial wall. Optimization of MCBIF was performed in MATLAB by simultaneous estimation of the compartment model parameters and the blood input function with SP and partial volume (PV) corrections using extended least squares (ELS).

Results Figure 1A shows transverse PET images obtained from OSEM-MAP gated and FBP un-gated reconstructed images at the last time bin. Corresponding MCBIF estimates (figure 1B) are shown along with blood samples at 43 and 56 minutes post FDG administration. The resulting net myocardial FDG influx constant, Ki, was also computed.

Conclusions Lower SP of radioactivity from the myocardium to the BP and higher recovery of BP radioactivity (PV) from OSEM-MAP gated images (figure 1A) lead to improved MCBIF estimate (solid curve with triangles) as compared to that obtained from un-gated FBP images (solid curve with squares), especially at the late time point. Mean myocardial Ki (0.196±0.013) computed using the rate constants generated by optimization of blood input function from OSEM-MAP gated images agreed well with that obtained using arterial blood sampling (0.14±0.07)(1), as the difference did not reach statistical significance on a Student t-test (p=0.24)

Research Support NIH R21HL10262