ESTRO 38 Abstract book

S577 ESTRO 38

recently demonstrated suitable for both Attenuation Correction in PET/MR and for pseudo CT conversion for head and neck applications [1]. To achieve an optimized MR-only workflow however, dedicated RT coils are needed to allow patient positioning in the MRI including RT fixation devices. To reach the desired coverage and image quality, current clinical practice often results in a bulky and uncomfortable set-up typically in form of a composition of different coils. Here we present a novel RT coil based on GE lightweight Air technology [2] and demonstrate its performance for standard head and neck MR imaging and ZTE based pseudo CT conversion. This highly flexible Air coil greatly simplifies the RT workflow and well adapts to varying patient position and size. Material and Methods A 3-tesla GE SIGNA MR scanner (GE Healthcare, Chicago, IL) and a prototype coil were used for standard MRI and ZTE data acquisition in phantoms and healthy volunteers. The coil consists of 22 channels of which 15 located in the face, allowing for 3 different coil modes: head only, head and neck, and chest only. ZTE data were processed with a Deep Learning (DL) method where a 3D convolutional neural network of the U-net architecture with 8-layers, Adam optimizer and RMSE cost function was adapted to perform pseudo CT computation via image regression. This method was trained on N=50 patients using matched pairs of ZTE and CT patient data sets using standard product surface coils [3]. Results Figure 1 shows the prototype coil and the coronal views of standard gradient echo (LAVA-Flex) and fast spin echo (FS) MR pulse sequences including Dixon type fat-water separation. The images clearly illustrate full head and neck coverage, including the shoulders, as required for RT planning.

Figure 2: top ZTE images with 1.5 isotropic resolution. Bottom: DL derived pseudo CTs Conclusion A silent, patient-friendly MR-only RT workflow compatible with all types of fixation devices has been enabled by a new lightweight and flexible Air coil prototype. These first results show that the coil has appropriate coverage and image quality to achieve pseudo CT conversion paving the way for an efficient and patient friendly MR-only RT workflow PO-1039 Spine SBRT with Halcyon™: Plan Quality, Delivery Accuracy, and Speed H. Petroccia 1 , I. Malajovich 1 , C. Wang 2 , K. Teo 1 , D. Lei 1 , T. Li 1 1 University of Pennsylvania, Radiation Oncology, Philadelphia, USA ; 2 Duke University, Radiation Oncology, Durham, USA Purpose or Objective A new dual-layer staggered 1cm MLC in Halcyon TM treatment platform (Varian Medical, Palo Alto, CA) has improved speed, leakage, and DLG compared to 120- Millennium (0.5cm) and High Definition (0.25cm) MLCs in the TrueBeam platform. Halcyon 2.0 with SX2 MLC has the ability to modulate both upper and lower MLC banks; while previously in version 1.0 the SX1 MLC uses the lower MLCs to modulate the fluence and the upper MLCs function as a backup jaw by moving to the most distally extended lower leaf pair. It is not clear how these two newly designed MLCs perform for spine SBRT cases and when compared to TrueBeam MLCs. In this study, we investigate the effects of 4 different MLC designs on spine SBRT treatments. Material and Methods 15 patients previously treated with spine SBRT were re- planned according to NRG-BR002 guidelines for all MLCs with a prescription of 3000cGy in 3 fractions, 6xFFF, 800 MU/min, and 3-arc VMAT technique. All plans were normalized to 90% of the target volume covered by prescription dose. Since the diameter of the spinal cord is comparable to the 1cm leaf width, this study evaluates the ability of the dual-layer staggered MLCs to modulate fluence using an 0.5cm effective leaf width by examining the percentage of shaping performed by the proximal MLC, the modulation complexity score (MCS), the total MU, and the gamma index passing rate based on delivery measurements using ArcCheck. Results Halcyon SX1 and SX2 plans are shown to have similar conformity and homogeneity index (SX2: CI 1.0±0.06, HI 0.27±0.05; SX1: CI 1.0±0.07, HI 0.29±0.05) as compared to the TrueBeam platform (HD: CI 0.95±0.03, HI 0.18±0.06; Millennium: CI 0.96±0.07, HI 0.21±0); however, the gradient measure as defined by Eclipse indicates the TrueBeam platform has steeper dose fall off than the Halcyon platform.

Figure 1 Top left: air coil prototype; Right: Water and Fat separation from LAVA-Flex for neck and shoulder coverage. Bottom: T2FS water; T2FS fat; T2cube acquisitions. Figure 2 shows ZTE images acquired with 1.5 mm isotropic resolution with corresponding, DL computed pseudo CT at the bottom. The skull and the spine are correctly captured, while a few false positive bone voxels are present in the neck and in the sinus region. A proper training on data acquired with the prototype coil with matched resolution is expected to produce superior results