Population averaged image derived arterial input function in 15-O Water PET

Ziel/Aim:

Arterial input function (IF) determination is traditionally based on arterial blood sampling which is invasive and ill-suited for clinical routine. Alternatively, one can use a population averaged IF determined via blood samples (BSIF) in a separate study. We present a strategy for generation of a population averaged image derived IF (IDIF) in brain investigations.

Methodik/Methods:

An image subvolume enclosing the carotid arteries is manually defined. Further processing is fully automated: 1: peak phase detection of the IF, 2: threshold based, background aware delineation using peak phase optimized for small structure size (~ 5 mm), 3: time activity curve generation, 4: model free recovery coefficient (R) determination 5: recovery and spillover correction according to Cm = R x Ct + (1 - R) x B, Cm/Ct: measured/true signal/IF, B: local background. This method was applied to 48 15-O Water brain studies (acquired with an ECAT HR+ PET). The resulting IFs were averaged, leading to a population averaged image derived IF (IDIF). Gray matter perfusion was computed from the Kety-Schmidt model assuming a partition coefficient in gray matter of 0.9 (f1). Perfusion was also computed using a BSIF from a different institution (f2). f1 and f2 were compared.

Ergebnisse/Results:

f1 and f2 are strongly correlated (Pearson 0.999) and differences are small: (f2-f1)/f2 = (9.7 +/- 0.9)% (mean +/- sd), range 6.7% - 12.7%.

Schlussfolgerungen/Conclusions:

The presented method allows to generate an IDIF which reproduces the results achieved with BSIF with good accuracy. One significant advantage of the new approach is the ease of regeneration of the IDIF for each study group (e.g. after injection and acquisition protocol changes). Due to statistical fluctuations of vessel delineation individual IF determination is not yet possible with this method due to the modest reconstructed resolution of the HR+ data. This might become possible when using state-of-the-art scanners and reconstruction algorithms.