Encrypted login | home

Program Information

Assessing Segmentation Accuracy of Internal Volumes and Sub-Volumes in 4D PET/CT of Lung Tumors Using a Novel 3D Printed Phantom

D Soultan

D Soultan1*, J Murphy2 , C James3 , C Hoh4 , B Gill5 , V Moiseenko6 , L Cervino7 , (1) University of California-San Diego, San Diego State University, La Jolla, CA, (2) University of California,San Diego, La Jolla, California, (3) University of California,San Diego, La Jolla, California, (4) University of California,San Diego, La Jolla, California, (5) British Columbia Cancer Agency, Windsor, ON, (6) University of California, San Diego, La Jolla, CA, (7) University of California, San Diego, La Jolla, CA


SU-E-J-123 (Sunday, July 12, 2015) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: To assess the accuracy of internal target volume (ITV) segmentation of lung tumors for treatment planning of simultaneous integrated boost (SIB) radiotherapy as seen in 4D PET/CT images, using a novel 3D-printed phantom.

Methods: The insert mimics high PET tracer uptake in the core and 50% uptake in the periphery, by using a porous design at the periphery. A lung phantom with the insert was placed on a programmable moving platform. Seven breathing waveforms of ideal and patient-specific respiratory motion patterns were fed to the platform, and 4D PET/CT scans were acquired of each of them. CT images were binned into 10 phases, and PET images were binned into 5 phases following the clinical protocol. Two scenarios were investigated for segmentation: a gate 30-70 window, and no gating. The radiation oncologist contoured the outer ITV of the porous insert with on CT images, while the internal void volume with 100% uptake was contoured on PET images for being indistinguishable from the outer volume in CT images. Segmented ITVs were compared to the expected volumes based on known target size and motion.

Results: 3 ideal breathing patterns, 2 regular-breathing patient waveforms, and 2 irregular-breathing patient waveforms were used for this study. 18F-FDG was used as the PET tracer. The segmented ITVs from CT closely matched the expected motion for both no gating and gate 30-70 window, with disagreement of contoured ITV with respect to the expected volume not exceeding 13%. PET contours were seen to overestimate volumes in all the cases, up to more than 40%.

Conclusion: 4DPET images of a novel 3D printed phantom designed to mimic different uptake values were obtained. 4DPET contours overestimated ITV volumes in all cases, while 4DCT contours matched expected ITV volume values. Investigation of the cause and effects of the discrepancies is undergoing.

Contact Email: