Evaluation of internal target volume of abdominal tumors using cine-MRI.

Introduction: The detailed anatomy visualization with magnetic resonance (MR)-guided radiotherapy is particularly attractive for abdominal treatments, but patient respiratory motion can compromise image quality. The "navigator technique" produces high-quality 3D images, triggered by diaphragm displacement, in exhale phase only. The gold standard for planning is 4D imaging, which visualizes the lesion for all breathing phases. When 4D imaging is not available, an alternative is using 3D imaging combined with motion information from cine-MR.

Methods: This work investigates two alternative internal target volume (ITV) generation methods and compares them with the original treatment 4DCT imaging ITV. Datasets were analyzed from 10 upper abdominal patients that originally had been treated with a 4DCT-based ITV. In addition to the 4DCT, these patients received an exhale MR and cine-MR scans prior to treatment. An MR-CT-compatible motion phantom was also used to compare the two alternative ITV methods with the clinical 4DCT method. The first ITV method uses "margins expansion" (ME method) asymmetrically. The second method duplicates the exhale gross tumor volume (GTV) and shifts it to the positions of the average inhale GTV and mid-position GTV. The ITV is the "Boolean combine" (BC method) of the three displaced GTVs. The ME and BC methods were compared with the clinical 4DCT method using the Dice similarity coefficient (DSC) to determine the impact of approximating the true GTV trajectory and neglecting deformation.

Results: The ITV DSC ranges were 73%-96% for the ME method and 76%96% for the BC method. The BC approach created smaller treatment volumes than the ME method and more closely resembled the 4DCT margin for cases with larger motion and a significant component in the anterior-posterior direction.

Conclusions: An exhale MR combined with cine-MR can be used to simply create an ITV for adaptive MR-guided radiotherapy. For small lesions with larger anterior motion, the Boolean Combine method is the more accurate method.