68-6

28/104

Cine Image Acquisition with aS 1200 EPIDAll measurements were performed using the TrueBeam linac. Cine images were acquired with aS 1200 EPID. An aS 1200 EPID was an amorphous silicon detector array and had an active area of 43 cm × 43 cm with 1280 × 1280 pixels and a resolu-tion of 0.336 mm at a source to imager distance (SID) = 100 cm. All data were acquired in the cine imaging mode. The cine images were exported in the DICOM RT format from the Image Processing Service (IPS) in the TrueBeam service mode to verify the jaw positions.Jaw Detection with Image ProcessingJaw positions were verified by analyzing the acquired cine images in the following workflow for each segment. The analyses were performed in an in-house script in an image processing package (Fiji, National Institutes of Health, Bethesda, MD, USA)11). Figure 2 shows the workflow of the jaw detection method in this study. The signal of the jaw edge was much lower than the signal from the MLC-defined aperture because the jaw edge was always blocked by the MLC leaves in the BPF pattern (Figure 1). The cine images imported into Fiji were processed by the histogram equitation to enhance the image contrast (Figure 2b)12-14). Based on the enhanced histogram, Otsu thresholding15) was applied to binarize each image into a radia-tion-detected area and a radiation-blocked area (Figure 2b). After binarization, to remove small cavities and smooth any small and isolated areas, morphology processing (dilation and erosion)16) was performed (Figure 2d). The Canny edge detection method17) was used to extract the edges in the binarized image (Figure 2e). The binarized edge image was summed in the vertical direction to obtain a one-dimensional horizontal profile (Figure 2f). The profile had four peaks, with the inner two peaks corresponding to the jaw edges and the outer two corresponding to the edges of the MLC apertures. The peak positions of the inner two edges were recorded as the jaw positions. Several cine images were acquired for each segment. The image processing was applied to each acquired image and the mean and standard deviation (SD) of the jaw positions were calculated for each segment.the lower MLC leaves and moves in the opposite direction with the edge preceding the aperture of the upper MLC leaves in segments 5, 6, 7, and 8 (Figure 1). In the bidirectional pattern, the jaw edge in the MLC transmission region can be seen without the disturbance from the MLC slit because a sufficient gap is maintained between the edge of the right-sided (left-sided) jaw and the upper (lower) MLC slit in segments 1, 2, 3, and 4 (Figure 1). For segments 5, 6, 7, and 8, a sufficient gap is main-tained between the edge of the right-sided (left-sided) jaw and the lower (upper) MLC slit. We created the Digital Imaging and Communications in Medicine (DICOM) RT plan in which jaws were set 5.0 mm from the nearest MLC slit in the jaw- tracking mode. The BPF plan was imported into a linac (TrueBeam, Varian Medical Systems, Palo Alto, CA, USA) and cine images were acquired with aS 1200 EPID mounted on the linac. The cine images were analyzed to detect jaw positions as described in following text.This article does not contain any studies with human participants performed by any of the authors.Bidirectional Picket Fence PlanThe Millennium 120 MLC, which has 120 leaves, was mounted on the linac. A dynamic MLC plan in which the upper leaves (leaf numbers 31-60) moving from the left to right and lower leaves (leaf numbers 1-30) moving from the right to left was created using MLC Shaper (Varian Medical Systems). Both upper leaves and lower leaves were stopped every 20 mm with a 2-mm aperture. The total number of the stopping points, or segments, was 8 (Figure 1). The MLC file was imported into Eclipse Ver. 13.6 (Varian Medical Systems) and exported in the DICOM RT Plan format. The exported plan was converted into a jaw-tracking BPF plan by defining jaw positions on each MLC segments. The jaw positions were 5.0 mm retracted from the nearest leaf edges (Figure 1). The plans were created as the static gantry mode at angles of 0 degrees, 90 degrees, and 270 degrees, and the VMAT mode with 360-degree full gantry rotation. The energy of X-ray was 6 MV, the dose rate was 600 MU/min, and the delivered MU was 120 MU. The jaw-tracking BPF plan was irradiated in a dynamic mode.592