Retrospective self-sorted 4D-MRI for the liver. Issue 3 (June 2018)

• Record Type: Journal Article
• Authors: van de Lindt, Tessa N.; Fast, Martin F.; van der Heide, Uulke A.; Sonke, Jan-Jakob
• Is Part Of: Radiotherapy and oncology. Volume 127:Issue 3(2018)
• Page Start: 474
• Page End: 480
• Publication Date: 2018-06
• Subjects: Respiratory motion -- 4D-MRI -- MRI-guidance -- Liver -- MR-Linac -- Self-gating; Oncology -- Periodicals; Radiotherapy -- Periodicals; Tumors -- Periodicals; Medical Oncology -- Periodicals; Neoplasms -- radiotherapy -- Periodicals; Radiotherapy -- Periodicals; Radiothérapie -- Périodiques; Cancérologie -- Périodiques; Tumeurs -- Périodiques; Electronic journals; 616.9940642
• DOI: 10.1016/j.radonc.2018.05.006
• Languages: English
• ISSNs: 0167-8140

Abstract

Abstract: Purpose: Daily MRI-guidance for liver radiotherapy is becoming possible on an MR-Linac. The purpose of this study was to develop a 4D-MRI strategy using an image-based respiratory signal with an acquisition-reconstruction time <5 min, providing T2-weighting for non-contrast enhanced tumor visibility. Materials and Methods: Images were acquired using an axial multi-slice 2D Turbo Spin Echo (TSE) sequence, repeated a variable number of times (dynamics). A self-sorting signal (SsS) was retrieved from the data by computing correlation coefficients between all acquired slices. Images were sorted into 10 phases and missing data were interpolated. The method was validated in a phantom and 10 healthy volunteers. The SsS, image quality (SSIM index: structural similarity index) and quantified liver motion were assessed as a function of the number of dynamics. Tumor visibility was demonstrated in two patients with liver metastasis on the Elekta Unity MR-Linac. Results: SsS was in good agreement with the reference navigator signal. Missing data increased from 0.4 ± 0.6% to 37.1 ± 6.6% for 60 to 10 dynamics. The SSIM index for the interpolated slices was ∼0.6. The RMSD of quantified motion was <1 mm in phantom experiments and in volunteers <1 mm for >10 dynamics. Conclusion: For 30 dynamics, acquisition-reconstruction time was <5 min and showed good performance in the validation experiments. The tumor was clearly visible in the patient images.