Multi-subject atlas-based auto-segmentation reduces interobserver variation and improves dosimetric parameter consistency for organs at risk in nasopharyngeal carcinoma: A multi-institution clinical study.

To assess whether consensus guideline-based atlas-based auto-segmentation (ABAS) reduces interobserver variation and improves dosimetric parameter consistency for organs at risk (OARs) in nasopharyngeal carcinoma (NPC). Eight radiation oncologists from 8 institutes contoured 20 OARs on planning CT images of 16 patients via manual contouring and manually-edited ABAS contouring. Interobserver variation [volume coefficient of variation (CV), Dice similarity coefficient (DSC), three-dimensional isocenter difference (3D-ICD)] and dosimetric parameters were compared between the two methods of contouring for each OAR. Interobserver variation was significant for all OARs in manual contouring, resulting in significant dosimetric parameter variation (P<0.05). Edited ABAS significantly improved multiple metrics and reduced dosimetric parameter variation for most OARs; brainstem, spinal cord, cochleae, temporomandibular joint (TMJ), larynx and pharyngeal constrictor muscle (PCM) obtained most benefit (range of mean DSC, volume CV and main ICD values was 0.36-0.83, 12.1-84.3%, 2.2-5.0mm for manual contouring and 0.42-0.86, 7.2-70.6%, 1.2-3.5mm for edited ABAS contouring, respectively; range of dose CV reduction: 1.0-3.0%). Substantial objective interobserver differences occur during manual contouring, resulting in significant dosimetric parameter variation. Edited ABAS reduced interobserver variation and improved dosimetric parameter consistency, particularly for brainstem, spinal cord, cochleae, TMJ, larynx and PCM.