Abstract:<section class="article-section__content"><h3 class="article-section__sub-title section1"> Purpose</h3><p>To evaluate two distortion correction techniques for diffusion‐weighted single‐shot echo‐planar imaging (DW‐ssEPI) on a 0.35T MR‐guided radiotherapy (MRgRT) system.</p></section><section class="article-section__content"><h3 class="article-section__sub-title section1"> Methods</h3><p>The effects of sequence optimization through enabling parallel imaging (PI) and selecting appropriate bandwidth on spatial distortion were first evaluated on the 0.35T MRgRT system using a spatial integrity phantom. Field map (FM) and reversed gradient (RG) corrections were then performed on the optimized protocol to further reduce distortion. An open‐source toolbox was used to quantify the spatial displacement before and after distortion correction. To evaluate ADC accuracy and repeatability of the optimized protocol, as well as impacts of distortion correction on ADC values, the optimized protocol was scanned twice on a diffusion phantom. The calculated ADC values were compared with reference ADCs using paired t‐test. Intraclass correlation coefficient (ICC) between the two repetitions, as well as between before and after FM/RG correction were calculated to evaluate ADC repeatability and effects of distortion correction. Six patients were recruited to assess the in‐vivo performance. The optimal distortion correction technique was identified by visual assessment. To quantify distortion reduction, tumor and critical structures were contoured on the turbo spin echo (TSE) image (reference image), the DW‐ssEPI image, and the distortion corrected images independently by two radiation oncologists. Mean distance to agreement (MDA) and DICE coefficient between contours on the reference images and the diffusion images were calculated. Tumor apparent diffusion coefficient (ADC) values from the original DW‐ssEPI images and the distortion corrected images were compared using Bland‐Altman analysis.</p></section><section class="article-section__content"><h3 class="article-section__sub-title section1"> Results</h3><p>Sequence optimization played a vital role in improving the spatial integrity, and spatial distortion was proportional to the total readout time. Before the correction, distortion of the optimized protocol (PI and high bandwidth) was 1.50±0.89 mm in a 100 mm radius and 2.21±1.39 mm in a 175 mm radius for the central plane. FM corrections reduced the distortions to 0.42±0.27 mm and 0.67±0.49 mm respectively, and RG reduced distortion to 0.40±0.22 mm and 0.64±0.47 mm respectively. The optimized protocol provided accurate and repeatable ADC quantification on the diffusion phantom. The calculated ADC values were consistent before and after FM/RG correction. For the patient study, the FM correction was unable to reduce chemical shift artifacts whereas the RG method successfully mitigated the chemical shift. MDA reduced from 2.52±1.29 mm to 1.11±0.72 mm after the RG correction. The DICE coefficient increased from 0.80±0.13 to 0.91±0.06. A Bland‐Altman plot showed that there was a good agreement between ADC measurements before and after application of the RG correction.</p></section><section class="article-section__content"><h3 class="article-section__sub-title section1"> Conclusion</h3><p>Two distortion correction techniques were evaluated on a commercial low‐field MRgRT system. Overall, the RG correction was able to drastically improve spatial distortion and preserve ADC accuracy.</p></section>

Fast geometric distortion correction using a deep neural network: Implementation for the 1 Tesla MRI‐Linac system

ReUINet: A fast GNL distortion correction approach on a 1.0 T MRI‐Linac scanner

Distortion-Corrected Image Reconstruction with Deep Learning on an MRI-Linac

A modular phantom and software to characterize 3D geometric distortion in MRI

AutoCorNN: An Unsupervised Physics-Aware Deep Learning Model for Geometric Distortion Correction of Brain MRI Images Towards MR-Only Stereotactic Radiosurgery

Comparison and evaluation of distortion correction techniques on an MR‐guided radiotherapy system

Intra-frame motion deterioration effects and deep-learning-based compensation in MR-guided radiotherapy

Characterization of real-time cine MR imaging distortion on 0.35 T MRgRT with concentric Cine imaging QA phantom

GDCNet: Calibrationless geometric distortion correction of echo planar imaging data using deep learning

Real‐time Radial Reconstruction with domain transform manifold learning for MRI‐guided radiotherapy

Clinical assessment of geometric distortion for a 0.35T MR-guided radiotherapy system

CT-guided PET Parametric Image Reconstruction Using Deep Neural Network Without Prior Training Data

PILOT: Physics-Informed Learned Optimized Trajectories for Accelerated MRI

NonCartesian MR image reconstruction with integrated gradient nonlinearity correction

A neural network approach for fast, automated quantification of DIR performance

Super-resolution neural networks improve the spatiotemporal resolution of adaptive MRI-guided radiation therapy

Automatic Tumor Segmentation With a Convolutional Neural Network in Multiparametric MRI: Influence of Distortion Correction

Open-source hardware and software for the measurement, characterization, reporting, and correction of geometric distortion in MRI

Data and Physics Driven Learning Models for Fast MRI -- Fundamentals and Methodologies from CNN, GAN to Attention and Transformers

A Deep Learning Approach for Automatic Segmentation during Daily MRI-Linac Radiotherapy of Glioblastoma

Improving accelerated 3D imaging in MRI-guided radiotherapy for prostate cancer using a deep learning method