End inspiration phase of 4D-CT, end-expiration phase of 4D-CT, and PET images of moving lung tumor. All images are coronal slices. Note the blurred appearance of the moving tumor in the PET image, which is acquired over several breathing cycles.
Experimental setup. (a) Motion platform and actuator were placed on a flat tabletop with controlling laptop. (b) Phantom is placed on the table flush against a piece of Lucite anchored to the table for reproducibility. Hatch marks were placed on pieces of tape on the base and moving parts of the platform to ensure motion is accurate from scan to scan.
Two spherical meshes are compared to the surface separation algorithm. From each sampling point on the “reference” mesh (in this case, the smaller sphere), the closest point on the other mesh surface is calculated.
Axial, sagittal, and coronal PET images of a moving sphere, inner , motion , . PET contours are volumes segmented at different percentages of maximum activity concentration. These contours are compared to the reference volume contoured on cine CT using the surface separation algorithm. The threshold volume which minimizes the sum of squared deviations is the optimal threshold for that sphere volume, motion extent, and source-to-background.
Optimal thresholds versus motion, source-to-background, and volume (denoted by different symbols). Threshold values are normalized to background. Error bars represent 1 standard deviation (3 measurements). Legend lists nominal sphere inner diameters.
Optimal threshold versus source-to-background for stationary spheres. Each line represents a different sphere volume as denoted in the legend. Note the linear nature of the relationship. Error bars are 1 standard deviation (three measurements).
Optimal threshold versus motion for source-to-. Each line represents a different sphere volume as denoted in the legend. Error bars are 1 standard deviation (three measurements).
Optimal threshold versus volume for source-to-. Each line represents a different motion extent (0–30 mm). Error bars are 1 standard deviation (three measurements).
Surfaces of regression function stated in Eq. (4). Each of the four surfaces displayed is calculated for (top) spheres of inner diameter 10, 17, 28, and 37 mm and (bottom) SBRs equal to 5:1, 20:1, 35:1, and 50:1.
PET/CT images of patient 1 with PET-segmented volume (thin contour) and reference cine CT volume (thick contour). (A) 4D-CT phase image at end-expiration. (B) 4D-CT phase image at end-inspiration. (C) Maximum intensity projection. (D) Coronal PET/CT image. (E) Transverse PET/CT image.
PET/CT images of patient 2 with PET-segmented volume (thin contour) and reference cine CT volume (thick contour). (A) 4D-CT phase image at end-expiration. (B) 4D-CT phase image at end-inspiration. (C) Maximum intensity projection. (D) Coronal PET/CT image. (E) Transverse PET/CT image.
Reference volume for stationary sphere (inner ) compared to 35% maximum activity concentration of same volume sphere at 30 mm motion extent and source-to-. Note that the 35% threshold underestimates the axial extent of the sphere and overestimates the sagittal extent of the sphere, consistent with the findings of Okubo et al. (Ref. 18). The 35% threshold, however, underestimates the full motion envelope of the tumor.
Nominal and actual volumes and SBRs.
Regression coefficients for regression model in Eq. (2).
Application of threshold function to three lung cancer patients. Threshold is normalized to background measurement. Mean deviation is determined with surface separation algorithm.
Article metrics loading...
Full text loading...