Xiaozhi Cao^1,2, Congyu Liao^1,2, Zheng Zhong¹, Erpeng Dai¹, Siddharth Srinivasan Iyer^1,3, Ariel J Hannum^1,4, Mahmut Yurt^1,2, Stefan Skare⁵, and Kawin Setsompop^1,2
1Department of Radiology, Stanford university, Stanford, CA, United States, ²Department of Electrical Engineering, Stanford university, Stanford, CA, United States, ³Department of Electrical Engineering and Computer Science, MIT, Cambridge, MA, United States, ⁴Department of Bioengineering, Stanford university, Stanford, CA, United States, ⁵Department of Clinical Neuroscience, Karolinska Institute, Stockholm, Sweden

In this work, a diffusion preparation was implemented in a 3D spiral-projection MRF framework to introduce additional diffusion weighting. Using MRF dictionary with diffusion terms, it enables whole-brain T₁, T₂, PD and additional diffusivity mapping with 1.25-mm isotropic resolution within 3min. To improve the image quality, cardiac gating and low-resolution navigator were also implemented to mitigate the signal variation caused by motion during diffusion encoding. Subspace reconstruction was used along with LLR regularization to improve the reconstruction conditioning as well as SNR.

Geonhui Son¹, Taejoon Eo¹, Yohan Jun¹, Hyungseob Shin¹, and Dosik Hwang^1
1Electrical and Electronic Engineering, Yonsei university, Seoul, Korea, Republic of

Training deep neural networks for medical imaging commonly requires large image datasets and paired label datasets. However, in the medical imaging research field, labeling costs are very expensive. To overcome this issue, we propose a data generation method which is StyleGAN2-based architecture that jointly generates multi-contrast magnetic resonance (MR) images and segmentation maps. The effectiveness of our generation model is validated in terms of segmentation performance for tumors. We demonstrate that the segmentation model only trained with the fake data generated from our method achieves comparable performance to that trained with real data.

Sebastian Mueller^1,2, Felix Glang¹, Kai Herz^1,2, Klaus Scheffler^1,2, and Moritz Zaiss^1,3
1High-field Magnetic Resonance Center, Max Planck Institute for Biological Cybernetics, Tuebingen, Germany, ²Department of Biomedical Magnetic Resonance, Eberhard Karls University Tübingen, Tuebingen, Germany, ³Department of Neuroradiology, University Hospital Erlangen, Erlangen, Germany

Discovery of MR contrast and/or conventional sequence parameter optimization usually requires a theoretical model to describe MR physics. Here we investigate if novel contrasts can be found by directly running numerical optimization on a real MRI scanner instead of a simulation. To this end, a derivative-free optimization algorithm is set up to repeatedly update and execute a parametrized sequence on the scanner and map the acquired signals to a given target contrast. As proof-of-principle, we show that this enables creatine concentration mapping by learning a CEST-prepared sequence, which is found solely based on known target concentrations in a phantom.

Felix Glang¹, Anton Nikulin^1,2, Jonas Bause¹, Rahel Heule^1,2, Theodor Steffen¹, Nikolai Avdievich¹, and Klaus Scheffler^1,2
1Magnetic Resonance Center, Max Planck Institute for Biological Cybernetics, Tübingen, Germany, ²Department of Biomedical Magnetic Resonance, Eberhard Karls University Tübingen, Tübingen, Germany

Previously, we have shown in simulations that electronically modulated time-varying receive sensitivities can improve parallel imaging reconstruction when fast modulations are applied during acquisition of k-space lines. Here, we demonstrate this concept experimentally with a prototype 8-channel reconfigurable receive coil, for which B₁^- modulation is achieved by fast switching PIN diodes in the receive loops. With this setup, MR measurements were performed in both phantom and human subject. Lower reconstruction errors and g-factors (~25% improvement for R=4) were observed for the case of rapidly switched sensitivities compared to conventional reconstruction with static sensitivities.

Daiki Tamada¹, Aaron S. Field^1,2, and Scott B. Reeder^1,2,3,4,5
1Radiology, University of Wisconsin-Madison, Madison, WI, United States, ²Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States, ³Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ⁴Emergency Medicine, University of Wisconsin-Madison, Madison, WI, United States, ⁵Medicine, University of Wisconsin-Madison, Madison, WI, United States

A novel method to obtain simultaneous 3D T1- and T2-weighted imaging with retrospective contrast adjustment was developed. Acquisitions are performed using a 3D RF phase-modulated GRE acquisition with a small RF phase increment. After phase correction, real and imaginary components are separated into T2- and T1-weighted images, respectively. Image contrast can also be modified by modulating additional phase into the signals retrospectively. Simulation and in vivo examples are shown. The results suggested the proposed method successfully provides arbitrary image contrasts with clinically acceptable quality and short acquisition time.

Artem Mikheev¹, Louisa Bokacheva¹, Jeff Lei Zhang², Hersh Chandarana¹, Sungheon Gene Kim³, and Henry Rusinek^1
1Department of Radiology, New York University School of Medicine, New York, NY, United States, ²School of Biomedical Engineering, ShanghaiTech University, Shanghai, China, ³Department of Radiology, Weill Cornell Medical College, New York, NY, United States

The arterial signal in DCE MRI often suffers from the inflow effect, which may cause large errors in the arterial concentration and compartmental model analysis. We implemented a constrained signal-to-concentration conversion using the subject’s cardiac output based on the Stewart-Hamilton principle, which limits the area under the arterial first pass peak. The constrained conversion significantly reduced the variation of the arterial concentration sampled at three levels along the abdominal aorta. The constrained arterial input function resulted in a significantly lower variability of the Tofts model K^trans in psoas muscle compared to the uncorrected input function. 

Ute Goerke¹, Eze Ahanonu², Mahesh Keerthivasan³, Ali Bilgin^2,4,5, Vibhas Deshpande⁶, and Maria Altbach^4,5
1Siemens Healthcare USA, Tucson, AZ, United States, ²Department of Electrical Engineering, University of Arizona, Tucson, AZ, United States, ³Siemens Healthineers USA, New York, NY, United States, ⁴Department of Biomedical Engineering, University of Arizona, Tucson, AZ, United States, ⁵Department of Medical Imaging, University of Arizona, Tucson, AZ, United States, ⁶Siemens Healthineers USA, Austin, TX, United States

In the inversion-recovery Look-Locker T1-mapping, the use of slice-selective inversion recovery pulse is explored to map the long T1-values of the spleen with as many slices as possible. The optimal slice gap in combination with interleaving the slices was determined. Artifact-free T1-maps of the spleen were obtained after parameter optimization with double the number of slices than using the standard approach in a single breath-hold. This new approach is an important step towards achieving complete coverage of abdominal organs within a single breath-hold.

Tzu-Cheng Chao¹, Dinghui Wang¹, and James G. Pipe^1
1Department of Radiology, Mayo Clinic, Rochester, MN, United States

Accurate fat fraction and T2* evaluation offers useful diagnostic information. A data driven B₀-update algorithm in conjunction with the Two-Point Dixon method is illustrated in this work, offering efficient computation to improve B₀ accuracy to separate fat and water in the multiple TE acquisition. The individual fat and water signals are then used for T2* mapping. The reconstructed parametric maps have comparable quality to those of the existing methods. The results also show that the proposed method requires shorter computation time and is more stable in the presence of an inaccurate reference field map.