Encrypted login | home

Program Information

Evaluation of a Knowledge-Based Atlas for Automated Cardiac Structures Quality Assurance for NCTN Clinical Trial NRG Oncology RTOG 0617 and 1308

M Huang

M Huang1*, D Guttmann2 , J Yu3 , C Cheng4 , H Geng5 , H Zhong6 , E Gore7 , J Bradley8 , Z Liao9 , K Forster10 , M Gillin11 , R Mohan12 , Y Xiao13 , (1) University of Pennsylvania, Philadelphia, PA, (2) University of Pennsylvania, Philadelphia, PA, (3) Georgetown University, Washington, DC, (4) University of Pennsylvania, Philadelphia, PA, (5) University of Pennsylvania, Bryn Mawr, Pennsylvania, (6) ,,,(7) Froedtert and the Medical College of Wisconsin, Milwaukee, WI, (8) Washington Univ. School of Medicine, Saint Louis, MO, (9) UT MD Anderson Cancer Center, Houston, Texas, (10) University of South Alabama, Dallas, TX, (11) MD Anderson Cancer Ctr., Houston, TX, (12) UT MD Anderson Cancer Center, Houston, TX, (13) University of Pennsylvania, Philadelphia, PA


WE-F-205-4 (Wednesday, August 2, 2017) 1:45 PM - 3:45 PM Room: 205

Purpose: To evaluate the feasibility of automated quality assurance of cardiac structures based on Atlas contours created from RTOG 0617, and corresponding dose metrics for the Pericardium and Heart for RTOG 1308 trial comparing photon and proton radiotherapy for non-small cell lung cancer.

Methods: Cardiac structures (pericardium, ventricles, and atria) were re-contoured by expert radiation oncologists with consistent contouring guidelines for over 500 cases from RTOG 0617. An Atlas-based auto-contouring database was created using 100 randomly selected cases. This atlas library was then used to generate cardiac contours for Pericardium structures for 25 cases from RTOG 1308 applying deformable registration. Also, all 25 cases were contoured manually. The geometrical discrepancies and dosimetric impact between these two sets of cardiac contours were evaluated. Geometrical difference indices included Dice coefficient (DIC), Jaccard similarity coefficient (JC), Mean Distance to Agreement (MDA), and Hausdorff distance (HD). The dosimetric indices included integral dose, V5Gy[%], V15Gy[%], V30Gy[%] , V45Gy[%], V60Gy[%], Mean, Max, Min Heart contours submitted from the institutions participating in the trial.

Results: Geometric differences between contours generated automatically and manually were 0.87±0.04 mm for DIC, and 0.78±0.07 mm for JC. MDA between the two volumes was 4.77±1.87 mm, and HD was 48.74±11.72 mm. The dosimetric differences to the pericardium volume and to heart volume in Mean, Max, Min, V5Gy[%], V15Gy[%], V30Gy[%], V45Gy[%], V60Gy[%] between pericardium and heart contours were 9.5±5.0Gy, 3.0±2.9Gy, 0±0.3Gy, 33.3±23.4%, 44.6±23.40%, 57.9±23.8%, 66.10±25.2%, and 73.3±26.3%, respectively. These dosimetric differences show that the pericardium has higher mean and percentage volume for given dose than heart structures.

Conclusion: It is feasible to use atlas library built from prior knowledge for automated cardiac contour QA for clinical trials. Consistency in contour definition guidelines for cardiac structures is essential, otherwise significant dosimetric discrepancies can arise.

Funding Support, Disclosures, and Conflict of Interest: This project was supported by grants U10CA180868 (NRG Oncology Operations), U10CA180822 (NRG Oncology SDMC), U24CA180803 (IROC), from the National Cancer Institute (NCI), Eli Lilly. This project is funded, in part, under a grant with the Pennsylvania Department of Health. The Department specifically disclaims responsibility for any analyses, interpretations or conclusions.

Contact Email: