RT Journal Article
SR Electronic
T1 Determination of gross tumor volume using an adaptive region-growing approach
JF Journal of Nuclear Medicine
JO J Nucl Med
FD Society of Nuclear Medicine
SP 1440
OP 1440
VO 50
IS supplement 2
A1 Chang, Guoping
A1 Chang, Tingting
A1 Pan, Tinsu
A1 Clark, John
A1 Mawlawi, Osama
YR 2009
UL http://jnm.snmjournals.org/content/50/supplement_2/1440.abstract
AB 1440 Objectives It was previously shown that gross tumor volume (GTV) can be determined on PET images by using an adaptive region-growing algorithm followed by a dual-front active contour method (Med Phys 35(8):3711-3721). The aim of this abstract is to investigate the feasibility of determining GTV using the region-growing algorithm without the dual-front active contour step. Methods A NEMA/IEC phantom with 6 spheres (10-37mm) was imaged on a GE DRX scanner using different sphere-to-background ratio (SBR, 3:1, 5:1, 8:1 & 11:1) in both 2D&3D. All data were acquired for 3 min and images were reconstructed using OSEM (2iter, 21subs). For each sphere in each image, a rough region-of-interest (ROI) was manually drawn around the sphere, and the voxel with the maximum intensity was selected as the seed point for region growing. An algorithm similar to the one used in the original reference was then applied to this seed point, and a grown region was delineated for each threshold from 100% down to 1% in a step of 1%. This process generated a sharp demarcation on the resultant volume-threshold plot which was used to represent the sphere volume. The ratio between this measured volume and the actual volume was then calculated for each sphere. Results For all images, the measured volume was constantly bigger than the actual volume of each sphere, with an average measured-to-actual ratio of 2.06. When 2.06 was selected as a single correction factor for all the measured volumes, the average difference between the actual and the corrected measured volume was 10.2±4.5%. Conclusions PET GTV can be determined by only using the adaptive region-growing step while using a constant correction factor of 2.06.