02152nas a2200277 4500000000100000008004100001100001600042700001800058700001900076700002100095700001700116700001600133700001700149700001800166700001700184700001700201700002300218700002000241700001900261700001300280245004900293300001200342490000800354520149800362022001401860 2017 d1 aMcGale Paul1 aDuane Frances1 aAznar Marianne1 aBartlett Freddie1 aCutter David1 aDarby Sarah1 aJagsi Reshma1 aLorenzen Ebbe1 aMcArdle Orla1 aMyerson Saul1 aVivekanandan Sindu1 aWarren Samantha1 aTaylor Carolyn1 aRahimi K00aA cardiac contouring atlas for radiotherapy. a416-4220 v1223 a

BACKGROUND AND PURPOSE: The heart is a complex anatomical organ and contouring the cardiac substructures is challenging. This study presents a reproducible method for contouring left ventricular and coronary arterial segments on radiotherapy CT-planning scans.

MATERIAL AND METHODS: Segments were defined from cardiology models and agreed by two cardiologists. Reference atlas contours were delineated and written guidelines prepared. Six radiation oncologists tested the atlas. Spatial variation was assessed using the DICE similarity coefficient (DSC) and the directed Hausdorff average distance (d→H,avg). The effect of spatial variation on doses was assessed using six different breast cancer regimens.

RESULTS: The atlas enabled contouring of 15 cardiac segments. Inter-observer contour overlap (mean DSC) was 0.60-0.73 for five left ventricular segments and 0.10-0.53 for ten coronary arterial segments. Inter-observer contour separation (mean d→H,avg) was 1.5-2.2mm for left ventricular segments and 1.3-5.1mm for coronary artery segments. This spatial variation resulted in <1Gy dose variation for most regimens and segments, but 1.2-21.8Gy variation for segments close to a field edge.

CONCLUSIONS: This cardiac atlas enables reproducible contouring of segments of the left ventricle and main coronary arteries to facilitate future studies relating cardiac radiation doses to clinical outcomes.

 a1879-0887