Accuracy of image guidance for stereotactic liver radiotherapy using active breathing control and daily soft tissue imaging

Abstract

External skin marks or radiological bone structures have traditionally been used as positioning references in conventional radiotherapy (RT). However, these are poor surrogates for liver tumor position which is not fixed to the skeleton but is better correlated with diaphragm position. PTV margins must be large enough to account for uncertainty in tumour position using such localization methods. In addition, the liver moves due to breathing, requiring increased PTV margins and normal tissue irradiation. We have developed a hypofractionated 6 fraction stereotactic RT trial for liver cancer, using active breathing control (ABC) to immobilize the liver and on-line orthogonal imaging with repositioning prior to each radiation treatment in an effort to reduce geometric uncertainties and breathing motion. The purpose of this study is to report feasibility of this localization strategy for patients with liver cancer and to measure the accuracy of liver localization using ABC and on-line imaging compared to traditional setup based on skin marks. Eleven patients with liver cancer were treated with RT on study from August 2003 to February 2004. ABC was not used in three patients due to poor reproducibility with ABC (2) or patient intolerance (1). In these patients, the skeleton was used for localization with PTV margins accounting for breathing motion, leaving 8 patients treated with ABC and daily imaging. For the patients treated with ABC, following daily setup using skin marks, orthogonal megavoltage imaging was acquired at breath hold using ABC. A graticule provided a reference point at the beam central axis. The images were aligned to the planning CT digitally reconstructed radiographs using the diaphragm for caudal-cranial (CC) alignment and the skeleton for anterior-posterior (AP) and left-right (LR) alignment (see figure). Adjustments were made for positioning errors greater than a threshold of 3 mm and images were repeated. Off-line image alignments were repeated twice by 2 investigators to determine the final setup accuracy and the error in initial setup measurement prior to repositioning. Liver position during treatment was also measured using megavoltage movie loops acquired during delivery of each beam. Orthogonal imaging from 48 treatment fractions delivered using ABC in 8 patients were evaluated. Repositioning occurred in 45 of the 48 fractions (due primarily to offsets in the CC direction). Treatment was delivered using repeat breath holds from 10 to 25 seconds. The average treatment time was 25 minutes (range: 19 - 40 minutes). The use of on-line imaging and repositioning reduced random setup errors (σ) from 4.1 mm (CC), 2.8 mm (AP), and 2.5 mm (ML) to 2.6 mm (CC), 2.1 mm (AP) and 1.8 mm (ML). Systemic errors (μ) were reduced from -3.4 mm (CC), 2.8 mm (AP) and -1.3 mm (ML) to -0.3 mm (CC), 0.1 mm (AP) and 0.6 mm (ML). The average absolute offset in position was reduced from 5.1 mm (CC), 3.0 mm (AP) and 2.7 mm (ML) to 2.6 mm (CC), 2.0 mm (AP) and 1.9 mm (ML). Analysis of movie loops acquired during treatment is ongoing. Daily targeting of liver cancers using orthogonal megavoltage imaging to localize the liver for alignment is feasible and improves accuracy compared to conventional localization without daily imaging and repositioning. As the liver is a surrogate for actual tumor position, we are investigating kilovoltage cone beam CT for direct tumor targeting.