3D ultrasound for prostate localization in radiation therapy: A comparison with implanted fiducial markers

Abstract

This study compares prostate localization using three-dimensional ultrasound (3D US) to a standard technique using implanted fiducial markers (FMs) for prostate image guided radiation therapy (IGRT). Two methods to determine prostate position on US were evaluated: Assisted segmentation (prospectively) and manual segmentation (retrospectively). Daily couch shifts to align the prostate into treatment position were measured using each technique. A total of 278 FM couch shifts and 255 and 218 corresponding assisted and manual segmentation US couch shifts were analyzed in each direction: Anterior-posterior, right-left, and superior-inferior. Ninety five percent "limits-of-agreement" (LOA) were used to analyze paired couch shifts and to determine if US can reliably replace FMs. We chose an error tolerance of +/- 3 mm for the LOA analysis. For FM vs assisted-segmentation US, 35.3%, 51.0%, and 48.2% of couch shifts (anterior-posterior, right-left, and superior-inferior, respectively) agreed within +/- 3 mm. Agreement improved using manual segmentation US (corresponding agreements were 45.3%, 64.1%, and 55.2%), however, results still lie markedly below the 95% we consider to indicate clinical equivalence. Based on these results, our experience indicates US cannot replace FMs for prostate IGRT, using either assisted or manual segmentation. US couch shifts showed considerably greater variability than FM measures and US image quality is shown to affect agreement. Planning target volume margins for use with the US system were found to be 15.8, 8.7, and 12.5 mm for assisted segmentation and 13.1, 7.6, and 9.8 mm for manual segmentation. Comparison of these margins to those reported in recent studies for use with FM IGRT indicate FMs offer greater sparing of the rectum and bladder than the US system.

Ultrasound

“…Lung weight (LW), volume (LV) and lung physical density (LD) were calculated from CT scans using a software developed by the authors and based on the ITK toolkit [5]. The software is based on region-growing algorithms providing fast semi-automatic segmentations that were refined manually when needed.…”

“…in the echo images and to determine what the ultrasound system has to provide as images. An approach of a demonstration using in vivo images instead of phantoms is given in [5].…”

Radiobiological Modelling in Radiation Oncology

“…Precise knowledge of the CTV location immediately prior to treatment is essential for safe and accurate margin reduction. Pinpointing the CTV at the time of treatment is the purpose of IGRT (Johnston et al 2008). …”

High-performance GPU-based rendering for real-time, rigid 2D/3D-image registration and motion prediction in radiation oncology

“…Methods for retrieving this additional image information range from ultrasound [6,7], magnetic resonance imaging [8] and, most important, kilovoltage and megavoltage x-ray imaging, either by means of planar radiography [9,10] or 3D modalities [11,12]. Non-rigid registration [13] techniques can be employed to simulate the deformation of the tumor [14,15] between treatment sessions.…”