RT Journal Article
SR Electronic
T1 Impact of non-rigid and local rigid image registration on PET/CT response assessments
JF Journal of Nuclear Medicine
JO J Nucl Med
FD Society of Nuclear Medicine
SP 2073
OP 2073
VO 52
IS supplement 1
A1 Floris van Velden
A1 Paul van Beers
A1 Linda Velasquez
A1 Wendy Hayes
A1 Adriaan Lammertsma
A1 Ronald Boellaard
YR 2011
UL http://jnm.snmjournals.org/content/52/supplement_1/2073.abstract
AB 2073 Objectives The aim of this study was to investigate the impact of various image registration strategies on FDG PET/CT response assessment in case volumes of interest (VOIs) are defined on the baseline scan only. Methods Double baseline (n=11) and both early and late response whole-body FDG PET/CT scans were acquired in subjects with advanced gastrointestinal malignancies. Test and retest studies were performed within 10 d. In total, more than 19 liver and lung lesions could be identified. Non-rigid and local rigid image registration techniques were applied to PET or CT. VOIs drawn on test scans were projected onto retest and response scans to derive: (1) Dice similarity coefficients (DSC), and (2) test-retest (TRT, %difference) and response based on SUVmax, SUVmean, volume and total lesion glycolysis (TLG). VOIs drawn independently on all scans were used to obtain reference TRT and response data. Results CT-based registration provided poor results and therefore only PET derived registration results are shown. SUVmax TRT and responses were similar for all registration methods. TRT of other parameters was better for non-rigid than for rigid image registration. Observed responses differed between the various image registration techniques. See the table for median values (%). Conclusions PET-based non-rigid image registration is best suited for image registration of test-retest studies. Assessment of response in SUV is less affected by registration strategy than that in TLG. Research Support CTMM AIRFORC