A Comparison of Methods for Functional Volume Delineation in PET: Simple Vs. Advanced Ones
J Wang1*, L Li2, W Lu3, S Tan4, (1) Huazhong University of Science and Technology, Wuhan, ,(2) Huazhong University of Science and Technology, Wuhan, ,(3) University of Maryland School of Medicine, Baltimore, MD, (4) Huazhong University of Science & Technology , Wuhan,SU-E-J-115 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
To examine the importance of taking into account partial volume effect (PVE) for functional volume delineation in PET.
Five acquisitions with different source-to-background ratio (SBR) from a PET phantom were used as test data in this study. The phantom contains six spheres of different diameters. A total of eight methods were used to delineate the spheres from each acquisition. Five among the eight are simple thresholding methods and the other three are advanced methods which were well-known and widely used in the image processing community. The five thresholding methods are the dynamic thresholding, iterative thresholding, thresholding with 42% and 50% of the maximum uptake and the Otsu method. The three advanced methods are active contours without edges (CV), geodesic active contours (GAC) and max-flow/min-cut graph cut. The first two thresholding methods take into account the partial volume effect for the delineation while the other six do not. Robustness of different methods with respect to SBR and object size was evaluated using dice similarity index (DSI), classification error (CE) and volume error (VE).
The first two thresholding methods outperformed all the other six methods at all SBR levels and for all object sizes according to DSI, CE and VE. The first two thresholding methods were also less sensitive to the change of object size.
It has been widely accepted that advanced methods should outperform simple thresholding ones for functional volume delineation in PET. The results indicated, however, that simple thresholding methods may not be inferior to advance methods once these thresholding methods take into account the PVE. It is thus important to consider the PVE for functional volume image delineation in PET.