Session S51.3

Quantitative Cardiac Dynamic Imaging of Small Animal PET Images Using Cluster Analysis

S Domenichelli, D D'Ambrosio, S Trespidi, C Nanni, V Ambrosini,
S Boschi, R Franchi, M Marengo, AE Spinelli*

Scuola di Specializzazione in Fisica Sanitaria
Bologna, Italy

Introduction: Positron emission tomography (PET) studies allow the visualization of radiotracer biodistribution in order to obtain quantitative in vivo measurements of physiological and biochemical processes. Quantitative PET imaging requires a dynamic scan in order to measure the arterial input function and the tissue time-activity (TAC) curves. By combining these two curves with adequate mathematical models it is possible to obtain useful physiological information such as the metabolic rate, perfusion, receptors density etc. Usually, manual delineation of Region of Interest (ROI) is used in order to obtain TAC extraction from a dynamic PET images. The objective of this work was to implement a method for automatic TAC delineation of small animal dynamic PET studies using cluster analysis (CA).
Materials and Methods: Cluster Analysis allows to group pixels having the same kinetic, in this work performances of three clustering algorithms were assessed. The user must supply a set of pixels that need to be grouped (images acquired at different time points) and the number of clusters. The choice of the correct number of clusters was performed by using a parsimony criteria. In order to test the proposed method a set of noisy rat cardiac images were simulated considering typical IF and myocardial FDG uptake, its performance was also evaluated with real dynamic microPET data. A figure of merit (FM) was calculated by taking the ratio of area of difference between the true and clustered curve to the area under true curve.
Results: The values of FM were calculated for several noise realizations and the mean values were respectively equal to: 8% and 3% for blood IF and myocardial uptake.
Conclusions: Results show that using CA it is possible to obtain accurate IF and TAC without the need of manual ROI delineation.

(Abstract Control Number: 264)