Click here to

Session: MRI [Return to Session]

BEST IN PHYSICS (IMAGING): Parallel Imaging Stream for Multi-Purpose Real-Time Adaptive MRI-Guided Prostate Radiotherapy

T Bruijnen1,2*, P Borman1, T Schakel1, B Raaymakers1, (1) Department of Radiotherapy, University Medical Center Utrecht, Utrecht, UT, NL, (2) Computational Imaging Group for MRI diagnostics and therapy, Centre for Image Sciences, University Medical Center Utrecht, Utrecht, UT, NL


TH-C-TRACK 3-1 (Thursday, 7/29/2021) 1:00 PM - 2:00 PM [Eastern Time (GMT-4)]

Purpose: To develop a multiple-purpose 3D parallel imaging stream for real-time adaptive MRI-guided prostate radiotherapy. Here, we propose to modify the clinical 3D CINE to increase the temporal resolution, decrease the reconstruction latency and to include multiple image streams for improved motion detection. The CINE is modified in two ways: 1)the k-space data is sampled using a novel dual density Cartesian Acquisition with SPiral profile ordering (dd-CASPR), which enables viewsharing and multi-resolution image reconstruction; 2)the k-space data is rerouted and processed outside of the vendors software using an in-house built reconstruction framework (ReconSocket) to produce multiple image streams with low reconstruction latency.

Methods: The dd-CASPR CINE was implemented on a 1.5T MR-linac system (Elekta Unity, Elekta AB, Stockholm, Sweden) and connected to the ReconSocket to generate the three image streams:1)Real-time high frequency 1D navigators (projection) to detect extreme bulk motion at 3Hz.2)Real-time low-resolution(10mm³) imaging to detect large target motion (e.g. gas bubbles) at 0.5Hz.3)Real-time or retrospective high-resolution(2mm³) imaging for online replanning or dose accumulation at 0.15Hz.Images where reconstructed with parallel imaging using the BART toolbox on a GPU. The dd-CASPR CINE was evaluated using the Quasar motion phantom (Modus Medical Devices Inc., Canada) in which motion traces were compared. In addition, dd-CASPR was qualitatively evaluated in a volunteer scan, with deliberate onset of bulk motion.

Results: The k-space data was streamed in real-time using the ReconSocket(latency=17ms, maximum jitter<10ms). The phantom experiments showed a high correlation between the input trajectory and the three different image streams. The in-vivo scans showed high quality images with reconstruction times 25ms, 0.8s and 7s for image streams 1-3, respectively.

Conclusion: The proposed 3D dd-CASPR CINE produces multiple image streams with low reconstruction latency. Future work will focus on the integration of the multiple-image stream to facilitate start-and-stop real-time adaptive prostate radiotherapy.



    Reconstruction, MRI, Image Guidance


    IM- MRI : Motion management

    Contact Email