Ming Zhang¹, Jie Feng¹, and Hongjiang Wei^1
1School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China

Supervised deep learning methods for QSM reconstruction from tissue phase mainly rely on the ground truth susceptibility maps for training, which are not available for QSM. To address this issue, we propose a self-supervised deep learning method, TKD2TKD, for susceptibility reconstruction using artifact-contaminated TKD-TKD pairs acquired at different head orientations of one subject. The primary motivation of TKD2TKD is that the average of the network gradients by feeding TKD pairs will converge to the true gradient of that trained with high-quality averaged TKD images. The preliminary results suggested that TKD2TKD performed well in QSM reconstruction with improved susceptibility quantification accuracy.

Ruimin Feng¹, Steven Cao², Chunlei Liu², and Hongjiang Wei^1
1Biomedical Engineering, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China, ²Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, CA, United States

The existing reconstruction methods for susceptibility tensor imaging assume that frequency shifts are purely from magnetic susceptibility effects, which may result in noise and error propagation in the calculated susceptibility quantities. In this study, we proposed an extended asymmetric susceptibility tensor model that introduced the non-susceptibility-induced frequency shift, i.e. offset term. Simulation, ex vivo, and in vivo data were reconstructed using the proposed model and compared models. Experimental results show that the proposed method leads to more discernable fiber structures and more accurate fiber orientation estimation. This study inspires STI reconstruction from the perspective of better modeling frequency sources. 

Fan Yu¹, Qiuxuan Li¹, Cheng Zhao¹, Mo Zhang¹, Liangjie Lin², Jiazheng Wang², and Jie Lu^1
1Radiology and Nuclear medicine, Xuanwu Hospital, Capital Medical University, Beijing, China, ²Philips Healthcare, Beijing, China, Beijing, China

The microstructure and micro-perfusion changes in normal brain aging is barely understand. The objective of this study is to describe these microscopic changes in normal brain tissue by intravoxel incoherent motion diffusion-weighted MRI (IVIM-DWI). Preliminary results show decreased D, D* and increased f in elder subjects compare with the youngers.

Nan-Hao Chen¹, Li-Ping Chen², Chia-Wei Hsu², Chin-Hua Yang^1,2,3, and Hsu-Hsia Peng^1
1Department of Biomedical Engineering and Environmental Sciences, National Tsing Hua University, Hsinchu, Taiwan, ²National Taiwan University Hospital Hsinchu Branch, Hsinchu, Taiwan, ³Taoyuan General Hospital, Ministry of Health and Welfare, Taoyuan, Taiwan

The purpose of this study was to evaluate the differential associations among various myelin-related indices in white matter (WM) and grey matter (GM) to comprehend the alteration of microstructural integrity in brain tissue. We calculated mean, coefficient of variance (CV), and peak probability of ADC, apparent myelin water fraction (aMWF), and T1w/T2w ratio and found differential associations among myelin-related indices in GM and WM. In conclusion, measurements of ADC, aMWF, and T1w/T2w ratio as well as their correlations can help to understand characteristics of microstructural tissue integrity in GM and WM.

Tanxin Dong¹, Jie Song², Hong-Hsi Lee³, Qiyuan Tian^4,5, Susie Y. Huang^4,5,6, and Qiuyun Fan^4,5,7
1Department of Intelligent Medical Engineering, Medical College, Tianjin University, Tianjin, China, ²Department of Network and Information, Tianjin Huanhu Hospital, Tianjin, China, ³Massachusetts General Hospital, Charlestown, MA, United States, ⁴Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA, United States, ⁵Harvard Medical School, Boston, MA, United States, ⁶Harvard-MIT Division of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, ⁷Department of Biomedical Engineering, College of Precision Instruments and Optoelectronics Engineering, Tianjin University, Tianjin, China

High gradient diffusion MRI has recently been demonstrated to be useful in mapping gray matter microstructure, despite of a few caveats arising from the challenges in comprehensively modeling the complex tissue environments. One example was the assumption of impermeable membranes in the SANDI model and hence its limitation in the presence of water exchange depending on the diffusion time (Δ) used in the MR experiments. We found time-dependence in the signal decay in cortical regions for Δ=19ms~49ms, suggesting fast exchange effects, which might be the cause of the observed time-dependence of the fitting parameters using the standard SANDI model.

Tomoko Maekawa¹, Masaaki Hori^1,2, Katsutoshi Murata³, Thorsten Feiweier⁴, Kouhei Kamiya^1,2, Christina Andica¹, Akifumi Hagiwara¹, Shohei Fujita^1,5, Koji Kamagata¹, Akihiko Wada¹, and Shigeki Aoki^1
1Radiology, Juntendo University School of Medicine, Tokyo, Japan, ²Radiology, Toho University Omori Medical Center, Tokyo, Japan, ³Siemens Healthcare K.K., Tokyo, Japan, ⁴Siemens Healthcare GmbH, Erlangen, Germany, ⁵Radiology, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan

The purpose of our study was to investigate the utility of changes in diffusivity between short and long effective diffusion times (Δ_eff) in distinguishing extra-axial brain tumors. Patients with 12 meningiomas, 13 acoustic neuromas, and 11 pituitary adenomas were undergoing diffusion-weighted imaging with Δ_eff of 6.5 ms and 35.2 ms. The dependence of λ1, λ2, λ3, and MD values on Δ_eff was stronger in pituitary adenomas than in meningiomas and acoustic neuroma, suggesting differences in internal tissue structure. Our results showed that the use of DTI with shorter Δ_eff provides additional information about the microstructure of brain tumors.

Cece Dong¹, Qian Xu¹, Haiyun Wang¹, Gaofeng Shi¹, Qinglei Shi², Xu Yan², and Guang Yang^3
1Department of CTMRI, The Fourth Hospital of Hebei Medical University, Shijiazhuang, China, ²MR Scientific Marketing, Siemens Healthineers, Beijing, China, ³Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai, China

In this study, we evaluated and compared the values of quantitative parameters derived from the six DWI models in the diagnosis of HER-2 expression levels in patients with IDC of the breast, including conventional mono-exponential model, IVIM, DKI, SEM, FROC, and CTRW models. We found that the CTRW model showed higher sensitivity in the predicting HER-2 expression levels than the other models. Meanwhile, a logistic regression model was built by combining parameters of the advanced DWI models, which showed a high diagnostic performance and may potentially be used in clinical practice.

Miao Liu¹, Qi Wang¹, Gaofeng Shi¹, Li Yang¹, and Qinglei Shi^2
1The Fourth Hospital of Hebei Medical University, Shi Jiazhuang, China, ²MR Scientific Marketing , Siemens Healthineers Ltd., Beijing, China

In this study, through optimizing data enhancement, data normalization, dimension reduction and feature screening schemes, an auto encoder (AE) based on clinical and radiomics features of ADC map was established and demonstrated high value in predicting the prognosis of locally advanced cervical cancer LACC in concurrent chemo-radiotherapy. 

Yeefan Lee^1,2, Kuan-Hung Cho³, Chih-Hsing Tang², Chia-Wen Chiang³, Shih-Yen Lin⁴, Chen-Hsiang Kuan⁵, Chien-Yuan Lin⁶, Hsiao-Ling Lee⁶, and Li-Wei Kuo^3
1Department of Medical Imaging, National Taiwan University Hospital, Taipei, Taiwan, ²Department of Medical imaging, National Taiwan University Hospital Hsin-Chu Branch, Hsin-Chu, Taiwan, ³Institute of Biomedical Engineering and Nanomedicine, National Health Research Institutes, Miaoli County, Taiwan, ⁴Department of Computer Science, National Yang Ming Chiao Tung University, Hsinchu, Taiwan, ⁵Department of Surgery, National Taiwan University Hospital, Taipei, Taiwan, ⁶GE Healthcare, Taipei, Taiwan

Lymphedema is defined as accumulation of fluid and fibroadipose tissues due to disruption of lymphatic flow. Our results demonstrate a significant difference of diffusion measures between lymphedema and normal groups. DKI and DTI derivatives have similar tendency when comparing lymphedema patients with healthy subjects. The diffusion properties may aid the diagnosis of lymphedema and have the potential for evaluating the severity in postoperative follow-up.

Xin Tang¹ and Chenxi Hu^1
1The Institute of Medical Imaging Technology, School of Biomedical Engineering, Shanghai Jiao Tong University (SJTU), Shanghai, China

A multi-b-value DWI imaging method based on multi-shot EPIK with continuously changing b values and Locally Low-Rank (LLR) constraint is proposed. The multi-shot EPIK acquisition improves resolution of multi-b-value DWI, and acquisition of the central k-space in every b-value bypasses the need for careful initialization. The LLR constrained problem is solved by a newly developed ADMM algorithm with robust performance. Preliminary data in healthy subjects showed that the proposed method outperforms LLR-POCS and SPA-LLR for reconstruction at high undersampling rates. 

Hu Yin^1,2, Fangrong Zong^3,4, Xiaofeng Deng^1,2, and Jizong Zhao^1,2
1Department of Neurosurgery, Beijng Tiantan Hospital, Capital Medical University, Beijing, China, ²China National Clinical Research Center for Neurological Diseases, Beijing, China, Beijing, China, ³Institute of Biophysics, Chinese Academy of Sciences, Beijing, China, ⁴University of Chinese Academy of Sciences, Beijing, China, Bei Jing, China

Cerebellum has been proved to play an important role in no motor language function. The aims of this study including: building a new tractography atlas for cerebrum cerebellum pathway base on human connectome project (HCP) datasets which can be used in any study focused on these white matters. Some special tracking strategies were employed in the tracking process. The new tractography atlas was saved including a total of 11 tract templates and performed well in 30 healthy subjects. Both diffusion metrics and shape analysis metrics of these tracts were obtained.

Junjiao Hu¹, Huiting Zhang², Hu Guo³, Thomas Benkert⁴, Shan Jiang¹, Weijun Situ¹, and Jun Liu^1
1Department of Radiology,The Second Xiangya Hospital, Central South University, Changsha, China, ²MR Scientific Marketing, Siemens Healthineers, Wuhan, China, ³MR Application, Siemens Healthineers, Changsha, China, ⁴MR Application Predevelopment, Siemens Healthcare GmbH, Erlangen, Germany

Image quality and apparent diffusion coefficient (ADC) values of conventional single-shot spin-echo echo-planar imaging (SS-EPI) and prototype small field-of-view EPI (Zoomit_EPI) diffusion-weighted imaging in kidney tumor were compared. Results showed that Zoomit_EPI had better image quality, including fewer distortion artifacts and blurring (p < 0.01), and clearer edges of the lesions. ADC value of Zoomit_EPI had a good correlation with those of SS-EPI (R2 = 0.91), but Zoomit_EPI had significantly larger ADC than SS_EPI with 80.4*10-6 mm2/s.

Wei Mao¹, Xiaoqiang Ding², Caixia Fu³, Yuqin Ding¹, Thomas Benkert⁴, Robert Grimm⁴, Mengsu Zeng¹, and Jianjun Zhou^1
1Department of Radiology, Zhongshan Hospital, Fudan University, Shanghai, China, ²Department of Nephrology, Zhongshan Hospital, Fudan University, Shanghai, China, ³MR Application Development, Siemens Shenzhen Magnetic Resonance Ltd, Shenzhen, China, ⁴MR Application Predevelopment, Siemens Healthcare GmbH, Erlangen, Germany

The use of intravoxel incoherent motion (IVIM) and diffusion-weighted imaging (DWI) has increased for renal function and fibrosis assessments. However, debates have continued about which method is best. In this study, we used IVIM to assess the pathologic injury associated with IgA nephropathy and compared the results with the mono-exponential model of DWI. Both diffusion models were derived from advanced zoomed field of view (FOV) DWI acquisitions. IVIM had better diagnostic performance than the mono-exponential model in assessing the renal pathologic injury in patients with IgA nephropathy.

Zunquan Chen¹, Jiechao Wang¹, Zhigang Wu², Jianfeng Bao³, Jingliang Cheng³, Congbo Cai¹, and Shuhui Cai^1
1Department of Electronic Science, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China, ²MSC Clinical & Technical Solutions, Philips Healthcare, Shenzhen, China, ³Department of Magnetic Resonance Imaging, the First Affiliated Hospital of Zhengzhou University, Zhengzhou University, Zhenzhou, China

Diffusion tensor imaging (DTI) requires several diffusion-weighted image (DWI) acquisitions, leading to a long scan time. In this study, a deep-learning model based on multi-slice information sharing was implemented to obtain ultrafast DTI. This method uses the similarity of DWIs among neighboring slices to minimize the requirement of the number of diffusion-encoding directions. The results indicate that the proposed method can reconstruct high-quality diffusion and DTI-derived maps, exceptionally robust to noise in fractional anisotropy mapping even when only 3-direction DWIs.

Masami Yoneyama¹, Takashige Yoshida², Johannes M Peeters³, Jihun Kwon¹, Yasutomo Katsumata³, and Marc Van Cauteren^3
1Philips Japan, Tokyo, Japan, ²Tokyo Metropolitan Police Hospital, Tokyo, Japan, ³Philips Healthcare, Best, Netherlands

Deep learning constrained Compressed SENSE reconstruction (CS-AI) achieves significant improvement of image quality of whole-brain, ultra-thin-slice, high-resolution pseudo-3D diffusion imaging using single-shot EPI or single-shot TSE acquisition, compared with conventional SENSE and CS DWI.