An augmented correlation framework for the estimation of tumour translational and rotational motion during external beam radiotherapy treatments using intermittent monoscopic x-ray imaging and an external respiratory signal
- Publication Type:
- Journal Article
- Physics in Medicine and Biology, 2018, 63 (20)
- Issue Date:
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is currently unavailable due to the publisher's embargo.
The embargo period expires on 19 Oct 2019
© 2018 Institute of Physics and Engineering in Medicine. Increasing evidence shows that intrafraction tumour motion monitoring must include both six degrees of freedom (6DoF): 3D translations and 3D rotations. Existing real-time algorithms for 6DoF target motion estimation require continuous intrafraction fluoroscopic imaging at high frequency, thereby exposing patients to additional high imaging dose. This paper presents the first method capable of 6DoF motion monitoring using intermittent 2D kV imaging and a continuous external respiratory signal. Our approach is to optimise a state-augmented linear correlation model between an external signal and internal 6DoF motion. In standard treatments, the model can be built using information obtained during pre-treatment cone beam CT (CBCT). Real-time 6DoF tumor motion can then be estimated using just the external signal. Intermittent intrafraction kV images are used to update the model parameters, accounting for changes in correlation and baseline shifts. The method was evaluated in silico using data from 6 lung SABR patients, with the internal tumour motion recorded with electromagnetic beacons and the external signal from a bellows belt. Projection images from CBCT (10 Hz) and intermittent kV images were simulated by projecting the 3D Calypso beacon positions onto an imager. IMRT and VMAT treatments were simulated with increasing imaging update intervals: 0.1 s, 1 s, 3 s, 10 s and 30 s. For all the tested clinical scenarios, translational motion estimates with our method had sub-mm accuracy (mean) and precision (standard deviation) while rotational motion estimates were accurate to < and precise to . Motion estimation errors increased as the imaging update interval increased. With the largest imaging update interval (30 s), the errors were mm, mm and mm for translation in the left-right, superior-inferior and anterior-posterior directions, respectively, and , and for rotation around the aforementioned axes for both VMAT and IMRT treatments. In conclusion, we developed and evaluated a novel method for highly accurate real-time 6DoF motion monitoring on a standard linear accelerator without requiring continuous kV imaging. The proposed method achieved sub-mm and sub-degree accuracy on a lung cancer patient dataset.
Please use this identifier to cite or link to this item: