Feasibility study for markerless tracking of lung tumors in stereotactic body radiotherapy.

PURPOSE To evaluate the feasibility and accuracy of a method for markerless tracking of lung tumors in electronic portal imaging device (EPID) movies and to analyze intra- and interfractional variations in tumor motion. METHODS AND MATERIALS EPID movies were acquired during stereotactic body radiotherapy (SBRT) given to 40 patients with 49 pulmonary targets and retrospectively analyzed. Tumor visibility and tracking accuracy were determined by three observers. Tumor motion of 30 targets was analyzed in detail via four-dimensional computed tomography (4DCT) and EPID in the superior-inferior direction for intra- and interfractional variations. RESULTS Tumor visibility was sufficient for markerless tracking in 47% of the EPID movies. Tumor size and visibility in the DRR were correlated with visibility in the EPID images. The difference between automatic and manual tracking was a maximum of 2 mm for 98.3% in the x direction and 89.4% in the y direction. Motion amplitudes in 4DCT images (range, 0.7-17.9 mm; median, 4.9 mm) were closely correlated with amplitudes in the EPID movies. Intrafractional and interfractional variability of tumor motion amplitude were of similar magnitude: 1 mm on average to a maximum of 4 mm. A change in moving average of more than ±1 mm, ±2 mm, and ±4 mm were observed in 47.1%, 17.1%, and 4.5% of treatment time for all trajectories, respectively. Mean tumor velocity was 3.4 mm/sec, to a maximum 61 mm/sec. CONCLUSIONS Tracking of pulmonary tumors in EPID images without implanted markers was feasible in 47% of all treatment beams. 4DCT is representative of the evaluation of mean breathing motion on average, but larger deviations occurred in target motion between treatment planning and delivery effort a monitoring during delivery.