G Angelis¹*, B Zwan^1,2, A Briggs¹, D Nguyen³, P Keall³, J Booth^1,4, (1) Northern Sydney Cancer Centre, Royal North Shore Hospital, Sydney, Australia, (2) School of Mathematical and Physical Sciences, University of Newcastle, Newcastle, Australia, (3) ACRF Image X Institute, The University of Sydney, Sydney, NSW, Australia, (4) School of Physics, University of Sydney, Sydney, Australia

• 
  G Angelis

(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: Accurate tracking of implanted fiducial markers can provide a wealth of information regarding tumor location during treatment. However, identification of these markers on kilovoltage images can be challenging, due to image quality, marker location/size or obstructions from other radio-opaque objects in the field of view. This works presents an automated methodology to identify these fiducial markers from a sequence of kilovoltage images and extract 6 degrees of freedom motion regarding the tumor’s motion during treatment.

Methods: The proposed methodology utilizes CBCT projection images to extract the templates of fiducial constellation at regular projection intervals around the patient. Since the constellation is subject to minor deformations due to internal motion, the extracted constellation is used as a first order approximation to identify the centroid position of the constellation on the kilovoltage images using a template matching approach. Using the template and a probability density map about the most likely position of each fiducial relative to the constellation, the centroid position of each individual fiducial is extracted. Finally, the 6-dof motion is acquired using an inter-dimensional correlation framework.

Results: Preliminary results on two pancreatic cancer patients (2 fractions/patient) showed that the proposed method was able to continuously track all visible markers in 100% of kilovoltage projection images (~400 projections in total, with maximum motion range ~8mm and occasional obstructions of stents). Fiducial identification was not affected by apparent constellation deformations between CBCT and kilovoltage images.

Conclusions: An automated methodology to provide real-time target tracking during radiotherapy using kilovoltage images has been developed that is robust to intrafraction deformation. Further assessment of the proposed methodology using more clinical cases is underway, as well its extension to track internal anatomical landmarks instead of implanted markers.

Keywords

Image Guidance, Image Analysis

Taxonomy

IM/TH- Image Analysis (Single Modality or Multi-Modality): Image processing

Contact Email