Repository logo
 

Automatic contour propagation using deformable image registration to determine delivered dose to spinal cord in head-and-neck cancer radiotherapy.

Accepted version
Peer-reviewed

Type

Article

Change log

Authors

Yeap, PL 
Noble, DJ 
Harrison, K 
Bates, AM 
Burnet, NG 

Abstract

To determine delivered dose to the spinal cord, a technique has been developed to propagate manual contours from kilovoltage computed-tomography (kVCT) scans for treatment planning to megavoltage computed-tomography (MVCT) guidance scans. The technique uses the Elastix software to perform intensity-based deformable image registration of each kVCT scan to the associated MVCT scans. The registration transform is then applied to contours of the spinal cord drawn manually on the kVCT scan, to obtain contour positions on the MVCT scans. Different registration strategies have been investigated, with performance evaluated by comparing the resulting auto-contours with manual contours, drawn by oncologists. The comparison metrics include the conformity index (CI), and the distance between centres (DBC). With optimised registration, auto-contours generally agree well with manual contours. Considering all 30 MVCT scans for each of three patients, the median CI is [Formula: see text], and the median DBC is ([Formula: see text]) mm. An intra-observer comparison for the same scans gives a median CI of [Formula: see text] and a DBC of ([Formula: see text]) mm. Good levels of conformity are also obtained when auto-contours are compared with manual contours from one observer for a single MVCT scan for each of 30 patients, and when they are compared with manual contours from six observers for two MVCT scans for each of three patients. Using the auto-contours to estimate organ position at treatment time, a preliminary study of 33 patients who underwent radiotherapy for head-and-neck cancers indicates good agreement between planned and delivered dose to the spinal cord.

Description

Keywords

Automation, Head and Neck Neoplasms, Humans, Image Processing, Computer-Assisted, Observer Variation, Radiotherapy Planning, Computer-Assisted, Spinal Cord, Tomography, X-Ray Computed

Journal Title

Phys Med Biol

Conference Name

Journal ISSN

0031-9155
1361-6560

Volume Title

62

Publisher

IOP Publishing
Sponsorship
Cancer Research Uk (None)
Addenbrooke's Charitable Trust (ACT) (24/15 A/Noble)
Science and Technology Facilities Council (ST/I004297/1)
Science and Technology Facilities Council (ST/I005846/1)
Science and Technology Facilities Council (PP/E002730/1)
STFC (4158926)
Science and Technology Facilities Council (ST/G502304/1)
Science and Technology Facilities Council (ST/I000143/1)
Cancer Research Uk (None)
Science and Technology Facilities Council (PP/E000347/1)
Science and Technology Facilities Council (ST/M007227/1)
Cancer Research UK (C20/A20917)
CRUK, Singapore Government, Addenbrooke's Charitable Trust, Cambridge Biomedical Research Centre