Faculty Bibliography

PURPOSE:Subject motion in MRI remains an unsolved problem; motion during image acquisition may cause blurring and artifacts that severely degrade image quality. In this work, we approach motion correction as an image-to-image translation problem, which refers to the approach of training a deep neural network to predict an image in 1 domain from an image in another domain. Specifically, the purpose of this work was to develop and train a conditional generative adversarial network to predict artifact-free brain images from motion-corrupted data. METHODS:*-weighted, FLASH magnitude, and phase brain images for 53 patients was used to generate complex image data for motion simulation. To simulate rigid motion, rotations and translations were applied to the image data based on randomly generated motion profiles. A conditional generative adversarial network, comprising a generator and discriminator networks, was trained using the motion-corrupted and corresponding ground truth (original) images as training pairs. RESULTS:The images predicted by the conditional generative adversarial network have improved image quality compared to the motion-corrupted images. The mean absolute error between the motion-corrupted and ground-truth images of the test set was 16.4% of the image mean value, whereas the mean absolute error between the conditional generative adversarial network-predicted and ground-truth images was 10.8% The network output also demonstrated improved peak SNR and structural similarity index for all test-set images. CONCLUSION:The images predicted by the conditional generative adversarial network have quantitatively and qualitatively improved image quality compared to the motion-corrupted images.

Accelerating Magnetic Resonance Imaging (MRI) by taking fewer measurements has the potential to reduce medical costs, minimize stress to patients and make MRI possible in applications where it is currently prohibitively slow or expensive. We introduce the fastMRI dataset, a large-scale collection of both raw MR measurements and clinical MR images, that can be used for training and evaluation of machine-learning approaches to MR image reconstruction. By introducing standardized evaluation criteria and a freely-accessible dataset, our goal is to help the community make rapid advances in the state of the art for MR image reconstruction. We also provide a self-contained introduction to MRI for machine learning researchers with no medical imaging background

PURPOSE:To develop and evaluate a rapid spherical navigator echo (SNAV) motion correction technique, then apply it for retrospective correction of brain images. METHODS:The pre-rotated, template matching SNAV method (preRot-SNAV) was developed in combination with a novel hybrid baseline strategy, which includes acquired and interpolated templates. Specifically, the SNAV templates are only rotated around X- and Y-axis; for each rotated SNAV, simulated baseline templates that mimic object rotation about the Z-axis were interpolated. The new method was first evaluated with phantom experiments. Then, a customized SNAV-interleaved gradient echo sequence was used to image three volunteers performing directed head motion. The SNAV motion measurements were used to retrospectively correct the brain images. Experiments were performed using a 3.0T whole-body MRI scanner and both single and 8-channel head coils. RESULTS:Phantom rotations and translations measured using the hybrid baselines agreed to within 0.9° and 1mm compared to those measured with the original preRot-SNAV method. Retrospective motion correction of in vivo images using the hybrid preRot-SNAV effectively corrected for head rotation up to 4° and 4mm. CONCLUSIONS:The presented hybrid approach enables the acquisition of pre-rotated baseline templates in as little as 2.5s, and results in accurate measurement of rotations and translations. Retrospective 3D motion correction successfully reduced motion artifacts in vivo.

PURPOSE/OBJECTIVE:To develop and evaluate a tool for accurate, reproducible, and programmable motion control of imaging phantoms for use in motion sensitive magnetic resonance imaging (MRI) appli cations. METHODS:In this paper, the authors introduce a compact linear motion stage that is made of nonmagnetic material and is actuated with an ultrasonic motor. The stage can be positioned at arbitrary positions and orientations inside the scanner bore to move, push, or pull arbitrary phantoms. Using optical trackers, measuring microscopes, and navigators, the accuracy of the stage in motion control was evaluated. Also, the effect of the stage on image signal-to-noise ratio (SNR), artifacts, and B0 field homogeneity was evaluated. RESULTS:The error of the stage in reaching fixed positions was 0.025 ± 0.021 mm. In execution of dynamic motion profiles, the worst-case normalized root mean squared error was below 7% (for frequencies below 0.33 Hz). Experiments demonstrated that the stage did not introduce artifacts nor did it degrade the image SNR. The effect of the stage on the B0 field was less than 2 ppm. CONCLUSIONS:The results of the experiments indicate that the proposed system is MRI-compatible and can create reliable and reproducible motion that may be used for validation and assessment of motion related MRI applications.