Stabilizing 3D EPI time series by servo navigation and phase equalization exploiting repeated shots (PEERS)
Authors:
Malte Riedel,
Thomas Ulrich,
Samuel Bianchi,
Klaas P. Pruessmann
Abstract:
Purpose: To enable run-time head motion control and robust frequency corrections for 3D EPI fMRI. Methods: A short 3D orbital navigator (3 ms) is inserted into a 3D EPI sequence. A linear perturbation model is calibrated to estimate rigid motion and frequency parameters per shot. Rigid motion is corrected by scan geometry updates in run-time, while several techniques are investigated to stabilize…
▽ More
Purpose: To enable run-time head motion control and robust frequency corrections for 3D EPI fMRI. Methods: A short 3D orbital navigator (3 ms) is inserted into a 3D EPI sequence. A linear perturbation model is calibrated to estimate rigid motion and frequency parameters per shot. Rigid motion is corrected by scan geometry updates in run-time, while several techniques are investigated to stabilize navigator-based frequency corrections in the reconstruction. An additional method termed PEERS is proposed that exploits the repetitive structure of fMRI scans to fine-tune shot-wise phase and frequency estimates using the motion-corrected EPI data itself. Results: Servo navigation effectively reduces motion in the raw data of in-vivo fMRI scans in six subjects. PEERS provides high-precision frequency parameters for robust phase-corrected reconstructions in the phantom and in-vivo accounting for scanner drifts and slice encoding-related effects on EPI. In combination, servo navigation and PEERS achieve successful intra-volume corrections and consistent tSNR improvements of 8% on average throughout the brain. The two methods prove to be highly synergetic. Conclusion: Servo navigation achieves high-precision motion correction for 3D-EPI fMRI in run-time and, in synergy with PEERS, provides stable frequency corrections with short navigators even for long echo times. With its automatic self-calibration and no hardware requirements, servo navigation and PEERS enable effective plug-and-play motion correction for 3D fMRI.
△ Less
Submitted 6 May, 2025;
originally announced May 2025.
MR image reconstruction using deep density priors
Authors:
Kerem C. Tezcan,
Christian F. Baumgartner,
Roger Luechinger,
Klaas P. Pruessmann,
Ender Konukoglu
Abstract:
Algorithms for Magnetic Resonance (MR) image reconstruction from undersampled measurements exploit prior information to compensate for missing k-space data. Deep learning (DL) provides a powerful framework for extracting such information from existing image datasets, through learning, and then using it for reconstruction. Leveraging this, recent methods employed DL to learn mappings from undersamp…
▽ More
Algorithms for Magnetic Resonance (MR) image reconstruction from undersampled measurements exploit prior information to compensate for missing k-space data. Deep learning (DL) provides a powerful framework for extracting such information from existing image datasets, through learning, and then using it for reconstruction. Leveraging this, recent methods employed DL to learn mappings from undersampled to fully sampled images using paired datasets, including undersampled and corresponding fully sampled images, integrating prior knowledge implicitly. In this article, we propose an alternative approach that learns the probability distribution of fully sampled MR images using unsupervised DL, specifically Variational Autoencoders (VAE), and use this as an explicit prior term in reconstruction, completely decoupling the encoding operation from the prior. The resulting reconstruction algorithm enjoys a powerful image prior to compensate for missing k-space data without requiring paired datasets for training nor being prone to associated sensitivities, such as deviations in undersampling patterns used in training and test time or coil settings. We evaluated the proposed method with T1 weighted images from a publicly available dataset, multi-coil complex images acquired from healthy volunteers (N=8) and images with white matter lesions. The proposed algorithm, using the VAE prior, produced visually high quality reconstructions and achieved low RMSE values, outperforming most of the alternative methods on the same dataset. On multi-coil complex data, the algorithm yielded accurate magnitude and phase reconstruction results. In the experiments on images with white matter lesions, the method faithfully reconstructed the lesions.
Keywords: Reconstruction, MRI, prior probability, machine learning, deep learning, unsupervised learning, density estimation
△ Less
Submitted 19 December, 2018; v1 submitted 30 November, 2017;
originally announced November 2017.
