Target definition of moving lung tumors in positron emission tomography: correlation of optimal activity concentration thresholds with object size, motion extent, and source-to-background ratio.
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
PURPOSE: Hardware integration of fluorodeoxyglucose positron emission tomography (PET) with computed tomography (CT) in combined PET/CT scanners has provided radiation oncologists and physicists with new possibilities for 3-D treatment simulation. The use of PET/CT simulation for target delineation of lung cancer is becoming popular and many studies concerning automatic segmentation of PET images have been performed. Several of these studies consider size and source-to-background (SBR) in their segmentation methods but neglect respiratory motion. The purpose of the current study was to develop a functional relationship between optimal activity concentration threshold, tumor volume, motion extent, and SBR using multiple regression techniques by performing an extensive series of phantom scans simulating tumors of varying sizes, SBR, and motion amplitudes. Segmented volumes on PET were compared with the "motion envelope" of the moving sphere defined on cine CT. METHODS: A NEMA IEC thorax phantom containing six spheres (inner diameters ranging from 10 to 37 mm) was placed on a motion platform and moved sinusoidally at 0-30 mm (at 5 mm intervals) and six different SBRs (ranging from 5:1 to 50:1), producing 252 combinations of experimental parameters. PET images were acquired for 18 min and split into three 6 min acquisitions for reproducibility. The spheres (blurred on PET images due to motion) were segmented at 1% of maximum activity concentration intervals. The optimal threshold was determined by comparing deviations between the threshold volume surfaces with a reference volume surface defined on cine CT. Optimal activity concentration thresholds were normalized to background and multiple regression was used to determine the relationship between optimal threshold, volume, motion, and SBR. Standardized regression coefficients were used to assess the relative influence of each variable. The segmentation model was applied to three lung cancer patients and segmented regions of interest were compared with those segmented on cine CT. RESULTS: The resulting model and coefficients provided a functional form that fit the phantom data with an adjusted R2 = 0.96. The most significant contributor to threshold level was SBR. Surfaces of PET-segmented volumes of three lung cancer patients were within 2 mm of the reference CT volumes on average. CONCLUSIONS: The authors successfully developed an expression for optimal activity concentration threshold as a function of object volume, motion, and SBR.
