Validation of an automatic contour propagation method for lung cancer 4D adaptive radiation therapy

The objective of this work is to validate a fast and reliable 4D-based adaptive treatment planning procedure for the Radiation Oncology clinic. A non-rigid B-Spline based registration is performed between CT scans at different respiratory phases. Radiotherapy contours drawn on a reference phase are than propagated to the target phase. We present results from two novel validation methods: a validation study using a 4D anthropomorphic computational phantom, and a patient data analysis comparing auto-rater versus inter-rater variability. For the phantom study, the overlap was greater than 90%, while the diameter variation was 5.1% at exhale and 3.4% at inhale. For the patient dataset, pairwise auto-rater Dice coefficient values exceeded inter-rater values for 9 of 15 structures. Though subject to inaccuracies caused by residual motion artifacts, and manual contouring variations, the reliability of the method suggests it could be soon introduced in the clinical practice.