Performance characteristics and quality assurance aspects of kilovoltage cone-beam CT on medical linear accelerator.

Research paper by Cheng B CB Saw, Yong Y Yang, Fang F Li, Ning J NJ Yue, ChuXiong C Ding, Krishna K Komanduri, Saiful S Huq, Dwight E DE Heron

Indexed on: 03 May '07Published on: 03 May '07Published in: Medical Dosimetry


A medical linear accelerator equipped with optical position tracking, ultrasound imaging, portal imaging, and radiographic imaging systems was installed at University of Pittsburgh Cancer Institute for the purpose of performing image-guided radiation therapy (IGRT) and image-guided radiosurgery (IGRS) in October 2005. We report the performance characteristics and quality assurance aspects of the kilovoltage cone-beam computed tomography (kV-CBCT) technique. This radiographic imaging system consists of a kilovoltage source and a large-area flat panel amorphous silicon detector mounted on the gantry of the medical linear accelerator via controlled arms. The performance characteristics and quality assurance aspects of this kV-CBCT technique involves alignment of the kilovoltage imaging system to the isocenter of the medical linear accelerator and assessment of (a) image contrast, (b) spatial accuracy of the images, (c) image uniformity, and (d) computed tomography (CT)-to-electron density conversion relationship were investigated. Using the image-guided tools, the alignment of the radiographic imaging system was assessed to be within a millimeter. The low-contrast resolution was found to be a 6-mm diameter hole at 1% contrast level and high-contrast resolution at 9 line pairs per centimeter. The spatial accuracy (50 mm +/- 1%), slice thickness (2.5 mm and 5.0 mm +/- 5%), and image uniformity (+/- 20 HU) were found to be within the manufacturer's specifications. The CT-to-electron density relationship was also determined. By using well-designed procedures and phantom, the number of parameter checks for quality assurance of the IGRT system can be carried out in a relatively short time.

More like this: