Abstract
Purpose/Objective(s): In order to minimize the dose delivered to healthy tissue near amoving tumor during radiotherapy it is first necessary to accurately measure tumor position as a function of time. For example, a portal imager can be used to detect surrogate markers implanted around the tumor in order to track its motion with a moving collimator. Lung tumors can move at up to 30 mm/s, requiring a sampling rate of 30 frame/s to achieve mm accuracy. However the passive a-Si Flat Panel Imagers (FPIs) available with current linear accelerators operate at 2 - 10 frames/s, significantly slower than the required rate. Furthermore a-Si FPIs provide low image quality at their fastest frame rates and are susceptible to damage by the treatment beam, requiring replacement every 1 - 2 years. Emerging CMOS active pixel sensors use an addressable and partial read-out architecture to achieve significantly improved frame-rates relative to their passive counterparts. They are also capable of higher resolution, image quality and radiation-hardness. This study investigates the feasibility of using a CMOS APS to quickly and accurately track radio-opaque markers during radiotherapy. Materials/Methods: A custom CMOS imaging system was designed and constructed in collaboration with the MI3 consortium. The performance of this system was characterized and compared with an a-Si FPI. Four cylindrical gold markers of diameter 0.8 to 2 mm and length 8 mm were positioned on a motion-platform and moved according to the Lujan approximation to respiratory motion. Images were acquired using the megavoltage treatment beam at a range of frame and dose rates. The success rate of an automatic detection routine, absolute mean-error from the expected position and contrast-to-noise ratio of the marker images were then evaluated as a function of marker size, marker speed, frame rate and dose rate. Results: TheCMOSimager was found to offer improved resolution and signal-to-noise than the standard a-Si FPI at a comparable dose. The long integration time of the FPI resulted in marker images being too blurred to detect. The CMOS was able to detect the three largest markers 100% of the time and estimate their position to within 0.3 mm at 150 - 300 MU/min and 20 - 50 frame/s. However success rate declined with decreasing dose or frame rate. Conclusions: A CMOS megavoltage imaging system was found to offer superior signal-noise and resolution than the standard a-Si FPI. Furthermore the high speed of CMOS provided sub mm tracking of moving markers at a clinically acceptable dose rate and marker size.