Online Image-based Monitoring of Soft-tissue Displacements for Radiation Therapy of the Prostate

Purpose: Emerging prolonged, hypofractionated radiotherapy regimens rely on high-dose conformality to minimize toxicity and thus can benefit from image guidance systems that continuously monitor target position during beam delivery. To address this need we previously developed, as a potential add-on device for existing linear accelerators, a novel telerobotic ultrasound system capable of real-time, soft-tissue imaging. Expanding on this capability, the aim of this work was to develop and characterize an image-based technique for real-time detection of prostate displacements. Methods and Materials: Image processing techniques were implemented on spatially localized ultrasound images to generate two parameters representing prostate displacements in real time. In a phantom and five volunteers, soft-tissue targets were continuously imaged with a customized robotic manipulator while recording the two tissue displacement parameters (TDPs). Variations of the TDPs in the absence of tissue displacements were evaluated, as was the sensitivity of the TDPs to prostate translations and rotations. Robustness of the approach to probe force was also investigated. Results: With 95% confidence, the proposed method detected in vivo prostate displacements before they exceeded 2.3, 2.5, and 2.8 mm in anteroposterior, superoinferior, and mediolateral directions. Prostate pitch was detected before exceeding 4.7 degrees at 95% confidence. Total system time lag averaged 173 ms, mostly limited by ultrasound acquisition rate. False positives (FPs) (FP) in the absence of displacements did not exceed 1.5 FP events per 10 min of continuous in vivo imaging time. Conclusions: The feasibility of using telerobotic ultrasound for real-time, soft-tissue-based monitoring of target displacements was confirmed in vivo. Such monitoring has the potential to detect small clinically relevant intrafractional variations of the prostate position during beam delivery. (C) 2012 Elsevier Inc.