68-6

27/104

the conventional static jaw technique, use of the JTT can reduce the dose to normal tissues3-5). To accurately deliver VMAT with JTT, jaws must move to the position defined in a treatment plan. If the jaw moves to an incorrect position, delivered dose distribution can differ from that calculated by the treatment planning system6). Matsubayashi et al. reported that a dose error to normal tissue can be 0.179% when the jaw position error is 1.0 mm, which can be minimized by routinely controlling the accuracy of jaw positions7). Faungrod et al. reported a method to detect jaw positions during VMAT with JTT using cine images acquired using an electronic portal imaging device (EPID) mounted on a linac8). Their method is useful for patient- specific quality assurance (QA).It is desirable to check jaw positions during JTT delivery as routine QA. In this work, we propose a new jaw position detection method for JTT by extending the conventional QA method for MLC and evaluate its accuracy for jaw position detection.Materials and MethodsOverviewOne of the most important aspects of quality Figure 1　Multileaf collimator (MLC) jaw pattern of the bidirectional picket fence (BPF) plan showing the MLC leaves and jaws in each segment. The plan consists of eight MLC segments with an MLC aperture of 2.0 mm. The MLC-aperture centers of the upper and lower leaves in each segment are 70 mm, 50 mm, 30 mm, 10 mm, −10 mm, −30 mm, −50 mm, and −70 mm for the upper leaves and −70 mm, −50 mm, −30 mm, −10 mm, 10 mm, 30 mm, 50 mm, and 70 mm for the lower leaves, respectively. The left-sided jaw (X1) and right-sided jaw (X2) are set to be 5.0 mm retracted from the nearest MLC tips. Arrows indicates the positions of the MLC-aperture centers.control for use of MLCs is the accuracy of posi-tioning. A common method to check MLC positions is the picket fence (PF) test9, 10). In the PF test, the several slits with the MLC aperture of 1-2 mm are delivered repeatedly with separations of a few centimeters. In the conventional PF test, the jaw edges in the MLC transmission region cannot be detected on the EPID image because the dose from the MLC aperture is high. To resolve this problem, a new MLC pattern was created, termed the bidi︲rectional picket fence (BPF) pattern as described in this study. Figure 1 shows the jaw and MLC posi-tions in the BPF pattern. In the figure, the upper, lower, left, and right directions correspond to the Y2, Y1, X1, and X2 directions in the Varian IEC scale. In the BPF pattern, the upper leaves of MLC move in the left-to-right direction, and lower leaves move in the opposite direction. The left-sided jaw (X1) moves from the left to right with the edge following the aperture of the upper MLC leaves in segments 1, 2, 3, and 4 and moves in the opposite direction with the edge preceding the aperture of the lower MLC leaves in segments 5, 6, 7, and 8. The right-sided jaw (X2) moves from the right to left with the edge following the aperture of 591