Gating Window Optimization for Respiratory Phase-Gated Radiotherapy Using the Ultra-Fast Volumetric 4D-MRI Generated 3D Probability Distribution Map

Abstract

Respiratory phase-gated RT is clinically popular in reducing radiation dose to the surrounding normal tissues by managing the respiratory motion. Although the optimal setting is controversial, the gating windows are usually set at 30-50%, 30-60%, 40-60% or 30-70% (0% = inhale, 50% = exhale), and is usually based simply on the variation of the SI translational surrogate. In this study, we aim to optimize the gating window based on the volume of overlapping VOI with a probability of 90% (V90,i, i = respiratory phase) through an ultra-fast volumetric 4D-MRI technique. Free breathing volumetric abdominal MR images were acquired on 9 healthy subjects (34 ± 6 yrs) using a 1.5T MR-simulator (3D gradient echo sequence with 1.63 vol/sec, 56 axial slice/vol, voxel size = 2.7 x 2.7 x 4 mm3, 144 dynamics). Binary mask of each dynamic volumetric of the left kidney (LK) and right kidney (RK) was segmented. The images were binned into 10 respiratory phases based on liver position. For all phases, the binary mask of each volumetric time frame at phase i were summated prior to the normalization to the total number of dynamic volumes at phase i (resulting in a total of 10 probability distribution maps). For each phase i, V90, i and the SI translational range (RSI, i) of the liver position were evaluated. For each subject, the optimal gating window phase (Pw) and the gating window durations (Dw) was determined with two approaches based on (1) RSI (Pw = phases with minimum RSI, Dw = neighboring phases of Pw with RSI< 2mm); (2) dice similarity coefficient (DSC) (Pw = phase of maximum V90 (V90,max), DW = neighboring phases of Pw with DSC > 0.9). The mean RSI of the phase of V90,max for all VOIs, and the mean minimum RSI (RSI,min) were also reported. Comparison between RSI and DSC approaches was performed using signed rank test. Comparing between approaches, similar Pw at exhalation was obtained using RSI (50±11%) and DSC (LK:46±24%, p=1; RK:54±19%, p=0.69). The mean V90,max (LK=156±34mm3, RK=149±26mm3) was close to the mean organ volume (LK=160±34mm3; RK=154±27mm3). The mean RSI of the phase at V90,max (LK=2.48±1.92mm, p=0.38; RK=2.41±1.91mm, p=1) was similar to RSI,min (2.06±1.99mm). The choice of gating window using DSC was preferable because a highly repeatable volumetric coverage was obtained, while such volume information cannot be fully revealed by the RSI approach. A smaller Dw was obtained using RSI (14±5%) then DSC (LK:42±23%, p=0.016; RK:60±31%, p=0.016), though insignificant, might suggest a less effective gated-RT delivery based simply on SI displacement. Larger Dw was obtained based on DSC approach allows a more effective gated-RT delivery with the overlapping volume taken into consideration.