Apr 16, 2012
Quantifying uncertainty in lung SBRT
Stereotactic body radiotherapy (SBRT) uses a small number of high-dose treatment fractions to deliver a prescribed radiation dose to a tumour. Such condensed treatment regimes demand accurate targeting; and targets that exhibit significant physiological motion, such as lung tumours, require particularly close attention.
Researchers at the Erasmus University Medical Centre in Rotterdam, the Netherlands have assessed the geometric uncertainties associated with SBRT treatment of T1 and T2 non-small cell lung cancer (NSCLC). Speaking at the 12th international conference on electronic patient imaging (EPI2k12), held last month in Sydney, Australia, physicist Mischa Hoogeman presented his group's findings.
Early-stage NSCLC patients at Erasmus MC's Daniel den Hoed Cancer Centre are treated with the robotic CyberKnife system, using the Synchrony respiratory tracking technology to correct for tumour motion. The centre uses a 5 mm gross tumour volume to planning target volume expansion.
Since introducing the technology to clinical use in 2005, the centre has treated around 700 patients, with results that have pleased clinicians. "Our clinical outcomes are excellent, with local control for T1 and T2 tumours of 96% at two years following treatment," said Hoogeman.
To detect tumour motion, the technology uses a combination of implanted fiducial markers, visualized using stereoscopic kilovoltage imaging, and LEDs placed on the patient's torso, tracked using a ceiling-mounted camera. A correlation model relates the breathing-induced LED motion to the motion of the fiducials. In this way, LED motion guides the robotic tracking of the tumour by the linear accelerator over the course of each breathing cycle. A second model predicts tumour motion based on the acquired data, sidestepping the latency of the robotic technology, which would otherwise result in a lag between tumour motion detection and the adaptation of the irradiation geometry.
The researchers investigated three main sources of uncertainty in their treatment approach, identified as the correlation model, the motion prediction model and the geometric stability of the fiducial markers. Based on the examination of data from 42 patients, they identified the markers as the biggest source of uncertainty.
The team examined CT scans taken eight, 11 and 13 days after the original treatment planning CT for displacements in the marker centre-of-mass. This parameter was selected as a relevant measure as the tracking technology uses the centre-of-mass to follow the tumour. Thirteen days following the planning CT, an average centre-of-mass migration of 1.7 ± 1.9 mm was observed. The group observed sub-millimetre errors due to both the correlation and prediction models. Based on these findings, the SBRT technique was confirmed as sufficiently accurate for clinical use.
About the author
Jude Dineley is a freelance science writer and former medical physicist based in Sydney, Australia.