The quality of such kilovoltage (kV) CBCT images, however, is degraded by megavoltage (MV) X-ray scatter from the treatment beam onto the imaging panel. This scatter introduces significant noise into the images, making them difficult to clinically interpret. To address this drawback, a UK research team has introduced a novel patient- and plan-specific MV scatter correction technique that should improve the quality of CBCTs recorded during treatment (Phys. Med. Biol. 57 3727).
"Our centre has been looking at the use of simultaneous CBCT during VMAT to reduce the fraction times, as potentially we only require one linac revolution to complete both treatment and imaging," explained Christopher Boylan, from the University of Manchester and the Christie NHS Foundation Trust. "We are also interested in seeing whether images taken during treatment, rather than before or after, offer a more accurate depiction of patient internal anatomy. In order to do this, we required a MV scatter correction that was effective, but simple to apply."
Comparing corrections
The correction method involves acquiring a set of reference CBCT images during VMAT delivery, taken without the kV imaging beam. The signal from these images arises solely from MV scatter onto the imaging panel. To perform the correction, the MV scatter image with the closest linac gantry angle is determined and subtracted from the corresponding kV/MV projection.
To test the technique's effectiveness, Boylan and colleagues examined a CATPhan 600 image-quality phantom, placed at the isocentre of an Elekta linac fitted with a Synergy CBCT system. They recorded reference images (without MV delivery) and then acquired CBCT images during delivery of three VMAT prostate plans.
For comparison, the researchers examined three other scatter correction methods, based on: a 1D uniform scatter map (the mean signal of each scatter image); a 2D scatter map acquired from a different phantom; and an analytical model derived from the treatment plan. The four correction methods were applied to the phantom images and signal-to-noise ratios (SNR) measured for standard CBCTs, uncorrected simultaneous CBCTs and the scatter-corrected CBCTs.
The effect of MV scatter reduced the low-contrast SNR from 3.2 for the standard CBCT to 2.0 in the simultaneous CBCT; and the high-contrast SNR from 11.3 to 3.0. The four scatter correction methods improved the SNRs by varying amounts, with the proposed full 2D scatter correction showing the largest SNR recovery.
Patient study
The researchers also applied the scatter correction to datasets of three VMAT prostate patients, who received simultaneous CBCTs as part of the team's ongoing SCART (simultaneous cone-beam during arc therapy) trial. The quality of the corrected images was assessed using a five-tiered scoring system, in which "1" represents a high-quality image with a clearly visible boundary between the prostate and surrounding soft tissue, and "5" is a clinically inadequate CBCT.
The image quality of the standard CBCT, the uncorrected simultaneous CBCT and the 2D scatter corrected image was assessed by four observers, and an average score calculated for each. CBCTs corrected using the 1D uniform scatter map and the analytical model were also assessed.
In all patients, the quality score for the uncorrected CBCTs was significantly worse than for the standard CBCTs (increased by an average of 1.4). Corrected images scored better than the uncorrected images, with the 2D scatter correction performing best and improving the score on average by 0.67. The 1D correction and analytical correction improved the average score by 0.54 and 0.5, respectively.
The researchers concluded that their simple MV scatter correction method can improve the quality of simultaneous CBCTs. Currently, any ensuing treatment corrections are made offline following delivery. However, the team is now investigating ways to perform the correction in real-time during treatment, which should be possible using a newer version of the commercial cone-beam software.
While the analytical model did not perform as well as methods based on scatter measurements, it did recover much of the image quality using only information from the plan. Such a scheme would be beneficial as it eliminates the need for measurements on the linac prior to treatment.
"We were encouraged by the results from the analytical model correction," said Boylan. "However, we need to do further work to determine whether patient characteristics such as size and shape can be included to better predict the MV scatter contribution. We hope to be able to do this once we have completed our SCART trial."
Boylan says that the trial should be finished by the end of this summer. "This will answer questions about the clinical relevance of images taken during treatment," he told medicalphysicsweb. "We also hope to use the data to investigate changes in patient anatomy during the treatment fraction."
• Related articles in PMB
A megavoltage scatter correction technique for cone-beam CT images acquired during VMAT delivery
C J Boylan et al Phys. Med. Biol. 57 3727
The use of a realistic VMAT delivery emulator to optimize dynamic machine parameters for improved treatment efficiency
C J Boylan et al Phys. Med. Biol. 56 4119
Intensity-modulated arc therapy: principles, technologies and clinical implementation
Cedric X Yu and Grace Tang Phys. Med. Biol. 56 R31
A generalized inverse planning tool for volumetric-modulated arc therapy
Daliang Cao et al Phys. Med. Biol. 54 6725