The evaluation of a deformable image registration segmentation technique for semi-automating internal target volume (ITV) production from 4DCT images of lung stereotactic body radiotherapy (SBRT) patients

Abstract

PurposeTo evaluate a deformable image registration (DIR) segmentation technique for semi-automating ITV production from 4DCT for lung patients, in terms of accuracy and efficiency. MethodsTwenty-five stereotactic body radiotherapy lung patients were selected in this retrospective study. ITVs were manually delineated by an oncologist and semi-automatically produced by propagating the GTV manually delineated on the mid-ventilation phase to all other phases using two different DIR algorithms, using commercial software. The two ITVs produced by DIR were compared to the manually delineated ITV using the dice similarity coefficient (DSC), mean distance between agreement and normalised DSC. DIR-produced ITVs were assessed for their clinical suitability and also the time savings were estimated. ResultsEighteen out of 25 ITVs had normalised DSC>1 indicating an agreement with the manually produced ITV within 1mm uncertainty. Four of the other seven ITVs were deemed clinically acceptable and three would require a small amount of editing. In general, ITVs produced by DIR were smoother than those produced by manual delineation. It was estimated that using this technique would save clinicians on average 28min/patient. ConclusionsABAS was found to be a useful tool in the production of ITVs for lung patients. The ITVs produced are either immediately clinically acceptable or require minimal editing. This approach represents a significant time saving for clinicians.