SU-E-J-42: Motion Adaptive Image Filter for Low Dose X-Ray Fluoroscopy in the Real-Time Tumor-Tracking Radiotherapy System

Purpose: In the real‐time tumor‐tracking radiotherapysystem, fiducial markers are detected by X‐ray fluoroscopy. The fluoroscopic parameters should be optimized as low as possible in order to reduce unnecessary imagingdose. However, the fiducial markers could not be recognized due to effect of statistical noise in low doseimaging.Image processing is envisioned to be a solution to improve image quality and to maintain tracking accuracy. In this study, a recursive image filter adapted to target motion is proposed. Methods: A fluoroscopysystem was used for the experiment. A spherical gold marker was used as a fiducial marker. About 450 fluoroscopic images of the marker were recorded. In order to mimic respiratory motion of the marker, the images were shifted sequentially. The tube voltage, current and exposure duration were fixed at 65 kV, 50 mA and 2.5 msec as low doseimaging condition, respectively. The tube current was 100 mA as high doseimaging. A pattern recognition score (PRS) ranging from 0 to 100 and image registration error were investigated by performing template pattern matching to each sequential image. The results with and without image processing were compared. Results: In low doseimaging, theimage registration error and the PRS without the image processing were 2.15±1.21 pixel and 46.67±6.40, respectively. Those with the image processing were 1.48±0.82 pixel and 67.80±4.51, respectively. There was nosignificant difference in the image registration error and the PRS between the results of low doseimaging with the image processing and that of high doseimaging without the image processing.Conclusions: The results showed that the recursive filter was effective in order to maintain marker tracking stability and accuracy in low dosefluoroscopy.