68-6

29/104

Picket Fence TestThe images obtained in all segments were accu-mulated to create a PF-like image to verify MLC leaf positions. Figure 3 shows the cine image in each segment (left) and the accumulated PF-like image (right). The PF-like images were analyzed with the picket fence module of Pylinac18, 19). We evaluated the passing rate, the maximum error, and the average error of the position of MLC leaves in the PF-like image with 0.5-mm leaf tolerance value.Validation Jaw Position Accuracy with Static- Jaw and Error-Added BPF PlanTo verify the validity of our method of jaw detec-Figure 2　Workflow to detect jaw positions from cine images for segment 8. The one-dimensional horizontal profile (f) was obtained by summing up the binarized edge image (e) in the vertical direction. The inner two peaks correspond to the jaw edges and the outer two peaks correspond to the edges of the multileaf collimator (MLC) apertures.Figure 3　Creating a multileaf collimator (MLC) picket-fence-like image from all segment frames of bidirectional picket fence (BPF) plan.tion, BPF plans with static jaw positions were created. The X1 and X2 jaw positions in this plan were −90.0 mm and +90.00 mm, respectively. Cine images were acquired with the static-jaw plans and analyzed using the same method as the jaw- tracking mode to detect jaw positions. Plans that had artificial errors on the jaw and leaf positions were also created to evaluate the accuracy of jaw position detection and sensitivity against posi-tioning error of jaws and leaves. The artificial errors for the jaw and leaf position were 0.5 mm or 1.0 mm. The obtained cine images with the error-added plans were also analyzed in the same way as described above to detect the jaw and leaf positions.593