Suixing Zhong¹, Ya Zhang¹, Yingying Ding¹, Xiaoyong Zhang², Jing Tan¹, Conghui Ai¹, Yan Jin¹, Hongbo Wang¹, Huimei Zhang¹, Miaomiao Li¹, Rong Zhu¹, and Shangwei Gu^1
1Department of Radiology, Yunnan Cancer Hospital, Third Affiliated Hospital of Kunming Medical University, Kunming, China, ²Clinical Science, Philips Healthcare, Chengdu, China

This study aimed to investigate the diagnostic value of intravoxel incoherent motion (IVIM) diffusion-weighted imaging and diffusion kurtosis imaging (DKI) in distinguishing metastatic from nonmetastatic lymph nodes (LNs) in cervical cancer patients. The clinical and imaging data of 102 patients with cervical cancer were collected prospectively, including 38 patients with LN metastasis and 64 patients without LN metastasis. The morphological parameters, diffusion parameters and tumor markers of primary tumors and LNs were measured and compared between the two groups. The results showed that the diagnostic efficiency of diffusion parameters were not as good as morphological parameters.

Na Liu¹, Wei qing Song¹, Wei yan Miao¹, Lian ai Liu¹, Jie Liang Lin², and Fang Xiao Xu^3
1the First Affiliated Hospital of Dalian Medical University, Da Lian, China, ²Philips Healthcare, Da Lian, China, ³Philips Healthcare, Bei Jing, China

IDEAL-IQ by GE and mDixon Quant by Philips are two MR fat quantification techniques that have been widely used for clinical disease evaluation and diagnosis in recent years. This study aims to compare performance of the two similar methods for fat quantification in the liver, pancreatic and lumbar vertebral. Results showed no significant difference between IDEAL-IQ and mDixon Quant on the liver, pancreas, and lumbar vertebral fat quantification in the same cohort of heathy volunteers

Na Liu¹, Wei Qing Song¹, Jie Liang Lin², Nan Wang¹, Lian ai Liu¹, and Wei Yan Miao^1
1the First Affiliated Hospital of Dalian Medical University, Da Lian, China, ²Philips Healthcare, BeiJing, China

Multivane (MV) imaging, based on the blade rotating acquisition in k space, is a novel sequence for motion reduction imaging. The MV acquisition percentage is critical parameter in MV imaging for balanced image quality and scan time. This study aims to assess the performance of compressed sensing for accelerated T2-weighted head MV imaging under altered MV acquisition percentages. Results indicated that scan time of head Multivane (MV) T2WI was significantly reduced by compressed sensing (CS) without reduced image quality.

Yinghui Wang¹, Shaohua Zhi¹, Haonan Xiao¹, Tian Li¹, and Jing Cai^1
1Department of Health Technology and Informatics, The Hong Kong Polytechnic University, Hong Kong, China

We developed and evaluated a deep learning technique for enhancing four-dimensional MRI (4D-MRI) image quality based on conditional adversarial networks. The quantitative and qualitative evaluative results demonstrated that the proposed model was able to reduce artifacts in low-quality 4D-MRI images and recover the details obtained from high-quality MR images, and performed better as compared with a state-of-the-art method. 

Junqi Xu¹, Qianfeng Wang¹, Xuchen Yu¹, Jiawei Han^1,2, Yimei Lu³, and He Wang^1,2
1Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University, ShangHai, China, ²Human Phenome Institute, Fudan University, ShangHai, China, ³Department of Radiology, Xinhua Hospital, Shanghai Jiao Tong University, ShangHai, China

This study proposed a new method to assess the stage of liver fibrosis by signal curves clustering and multiexponential decay T1ρ spectral analysis. Compared to single component in normal tissue, liver fibrosis tissue has two T1ρ components, in carbon tetrachloride (CCI­₄) rat model. Meanwhile, T1ρ tends to increase with fibrosis score. Due to the spatial distribution of clusters reveal the fibrosis region, this method might contribute to diagnosis and staging of liver fibrosis.

Abhilasha Indoria¹, Sachin Patalasingh¹, subhas Konar¹, and Jitender Saini^1
1NIMHANS, Bengaluru, India

This study assessed the impact of normalization scale and intensity discretization on first order radiomics features. Features were extracted from T2W images by varying the normalization parameter and bin width parameter of Pyradiomics library; Un-normalized (Without_normalization), normalized to scale1, normalized to scale 100 while keeping the bin width constant; and Bin width set to 20, bin width set to 40, bin width set to 60 while keeping the normalization scale set to 1. We found that the radiomics features were highly dependent on normalization scale and independent of bin width parameter.

hacer dasgin¹, basak soydas², Gul Yalcin Cakmakli³, Bilge Volkan Salanci², Eser Lay Ergun², Bulent Elibol³, and Kader K. Oguz^4
1Aysel Sabuncu Brain Research Center-UMRAM, Ankara, Turkey, ²Department of Nuclear Medicine, Hacettepe University, Ankara, Turkey, ³Department of Neurology, Hacettepe University, Ankara, Turkey, ⁴Department of Radiology, Hacettepe University, Ankara, Turkey

Genetic Parkinson’s Disease (gPD) displays distinct clinical, demographic and pathological features  from idiopathic Parkinson’s Disease (PD). The present study aims to search for alterations in the basal ganglia network (BGN) including substantia nigra (SN) and Subthalamic nucleus (STN) in a peculiar group of gPD using resting state (RS) fMRI. Genetic PD patients revealed less activation in the  basal ganglia, SN+STN pathway and reduced connectivity with ACC, MCC and insula, which may reflect frontal lobe and autonomic dysfunction in these patients.

Jian Dong¹, Yaping Wu¹, Yan Bai¹, and Meiyun Wang^1
1Department of Medical Imaging, Henan Provincial People's Hospital, Zhengzhou, China

In this study, we trained a fully automatic lesion segmentation model of venous malformations based on Unet++ structure and fat-saturated T2-weighted images. The results of automatic segmentation of lesions in different sequence directions and locations of the lesions are consistent with the results of manual segmentation by radiologists. The Dice coefficient on the test set reached 0.847. This fully automatic lesion segmentation model can provide support for subsequent automatic diagnosis studies related to venous malformations.

Yuting Li¹, Yuxuan Shang¹, Zhuhong Chen¹, Jingting Sun¹, and Xiaochen Wei^2
1Tangdu Hospital, Fourth Military Medical University, Xi'an, China, ²GE Healthcare, MR Research China, Beijing, Beijing, China

Featured alterations in dynamic brain functions, neurovascular coupling, or brain structure occurs at early-, intermediate- or late-stage of unilateral sudden sensorineural hearing loss (SSNHL) “when one door is half closed, another does not open yet”, respectively, and may serve as the neuroimaging markers for SSNHL.

Juan Yue^1,2,3, Na Zhao^1,2,3,4,5, Yang Qiao^1,2,3,5,6, Zi-Jian Feng^1,2,3, Yun-Song Hu^1,2,3, Yong Zhang ⁷, Yu-Feng Zang^1,2,3, and Jue Wang^8
1Center for Cognition and Brain Disorders, The Affiliated Hospital of Hangzhou Normal University, Hangzhou, China, ²Institute of Psychological Sciences, Hangzhou Normal University, Hangzhou, China, ³Zhejiang Key Laboratory for Research in Assessment of Cognitive Impairments, Hangzhou, China, ⁴Unit of Psychiatry, Department of Public Health and Medicinal Administration, & Institute of Translational Medicine, Faculty of Health Sciences, University of Macau, Macao SAR, China, ⁵Centre for Cognitive and Brain Sciences, University of Macau, Macao SAR, China, ⁶Faculty of Health Sciences, University of Macau, Macao SAR, China, ⁷MR Research, GE Healthcare, Shanghai, China, ⁸Institute of Sports Medicine and Health, Chengdu Sport University, Chengdu, China

Using multi-center resting-state fMRI databases under eyes closed (EC) and eyes open (EO), this study systematically investigated the intra- and inter-scanner reliability and validity of Wavelet-mALFF across multiple frequency bands. We found that Wavelet-mALFF outperformed FFT-mALFF in terms of intra- and inter-reliability and validity, particularly in the higher frequency band slow-2 (0.1992-0.25 Hz), where db2-mALFF is the top performer in Wavelet-mALFF. We propose that instead of FFT-ALFF, db2-ALFF can be employed to measure the spontaneous oscillations of local brain activity in future studies.

Wei Wang¹, Shuhui Cai¹, Lin Chen¹, Zhigang Wu², Congbo Cai¹, and Zhong Chen^1
1Xiamen University, Xiamen, China, ²MSC Clinical & Technical Solutions, Philips Healthcare, Xiamen, China

In multi-shot high-resolution MRI, the motion of patients often leads to serious degradation of imaging quality. In this study, a novel motion correction method based on deep learning and spatiotemporally encoded MRI was proposed to address this problem. The proposed method is robust to motion without utilizing extra scan or parallel reconstruction. The results of simulation and in vivo rat brain experiments demonstrate its efficacy in reducing image motion artifacts when subject movement exists.

Jinyang Wang¹, Changqing Wang¹, Scott B. Reeder^2,3,4,5,6, and Diego Hernando^2,3
1School of Biomedical Engineering, Anhui Medical University, Heifei, China, ²Radiology, University of Wisconsin-Madison, Madison, WI, United States, ³Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ⁴Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States, ⁵Medicine, University of Wisconsin-Madison, Madison, WI, United States, ⁶Emergency Medicine, University of Wisconsin-Madison, Madison, WI, United States

Monte Carlo modeling enables characterization of MR signals in various tissues, and has been applied to liver MR in the presence of fat. However, Monte Carlo modeling requires accurate information about the underlying tissue properties. In this work, we investigate the size and clustering of fat droplets in the liver using stereology and spatial statistics for three human liver biopsy samples with steatosis. Results show that the generalized gamma distribution function can accurately determine the size and location distributions of fat droplets. This may enable analysis of the underlying biophysical mechanisms between fat fraction and R2* from microscopic magnetic sensitivity.

Yao Zeng¹, Chenfei Ye², Junyan Sun³, Ting Ma¹, and Tao Wu^3
1Department of Electronic and Information Engineering, Harbin Institute of Technology at Shenzhen, Shenzhen, China, ²Peng Cheng Laboratory, Shenzhen, China, ³National Clinical Research Center for Geriatric Disorders, Xuanwu Hospital Capital Medical University, Beijing, China

  Since the heterogeneity of non-motor disorders in Parkinson's patients, subtyping research will facilitate the development of clinical diagnosis and treatment. We applied the sparse canonical correlation analysis method on clinical scales and resting-state brain fMRI data, to extract the typical variables for subtyping. In this data-driven study, we found that four PD subtypes with varying cognitive and psychological phenotypes demonstrate unique patterns of brain functional connectivity, which may underly the mechanism of PD heterogeneous non-motor symptoms.

Debdut Mandal¹, Lipeng Ning², and Yogesh Rathi^2
1Electronics and Electrical Communication Engineering, Indian Institute of Technology Kharagpur, Kharagpur, India, ²Harvard Medical School, Boston, MA, United States

Noninvasive measurement of axon radii has always been challenging using in vivo diffusion MRI (dMRI) data. Although dMRI allows estimation of the effective radius, a very high gradient strength scanner is required to acquire the data. In our proposed method, we overcome this limitation by using analytical methods to accurately predict the signal at high b-values from reasonably lower b-value data that can be obtained from a Prisma-like scanner. Our findings, both in synthetic data and in vivo dMRI data, show that estimating axon radii reliably is possible using dMRI data with a certain SNR level.

Qiyang Zhang^1,2, Zizheng Xiao³, Xu Zhang³, Yingying Hu³, Yumo Zhao³, Jingyi Wang⁴, Jiatai Feng⁴, Yun Zhou⁴, Yongfeng Yang¹, Xin Liu¹, Hairong Zheng¹, Wei Fan³, Dong Liang¹, and Zhanli Hu^1
1Lauterbur Research Center for Biomedical Imaging, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, ²National Innovation Center for High Performance Medical Devices, Shenzhen, China, ³Department of Nuclear Medicine, Sun Yat-sen University Cancer Center, Guangzhou, China, ⁴Central Research Institute, United Imaging Healthcare Group, Shanghai, China

Young children are more sensitive to radiation doses than adults, and their absorption of effective doses can be 4-5 times that of adults. When performing PET imaging, the use of low-dose imaging agents for high-quality imaging is of clinical importance. Here, we use artificial intelligence techniques combined with prior CT information to improve the quality of total-body PET/CT images in ultralow-dose pediatric FDG scans, and the results show that the equivalent quality of 600s acquisition data can be obtained using 15s acquisition.

Shenjun Zhong¹, Zhaolin Chen¹, and Gary Egan^1
1Monash Biomedical Imaging, Monash University, Australia, Melbourne, Australia

Robust latent representation of white matter streamlines are critical for parcellating streamlines. This work introduced a novel transformer-based siamese network with triplet margin loss, that learns to embed any lengths of streamlines into fixed-length latent representations. Results showed that a minimum of two layers of transformer encoders were sufficient to model streamlines with a very limited number of training data.

Lu Wang¹, Xiaorui Xu¹, Liyuan Liang¹, and Hing-Chiu Chang^2
1Department of Diagnostic Radiology, University of Hong Kong, Hong Kong, China, ²Department of Biomedical Engineering, Chinese University of Hong Kong, Hong Kong, China

Deep learning solutions have been proposed to correct motion-induced artefact in free-breathing abdominal MRI that can overcome some challenges in conventional methods, such as requiring respiratory modeling, external motion monitoring devices, or extremely long computation time. However, ground truth data for training can only be acquired with breath-holding or using those conventional methods. In this study, we proposed to use FD-UNet for the motion correction of free-breathing abdominal MRI. The high-quality and artifact-free ground truth data were produced from repeated k-t-subsampling and greedy randomized adaptive search procedure (ReK-GRASP), without relying on respiratory modeling, external devices, or long computation time.

Jeryn Chang¹, Frederik Steyn^1,2,3,4, Shyuan Ngo^2,3,4,5, Robert Henderson^2,3,4, Christine Guo⁶, Steffen Bollmann^7,8, Jurgen Fripp⁹, Markus Barth^7,8, and Thomas B Shaw^2,7,8
1School of Biomedical Sciences, The University of Queensland, St Lucia, Australia, ²Department of Neurology, Royal Brisbane and Women's Hospital, Brisbane, Australia, ³Wesley Medical Research, The Wesley Hospital, Brisbane, Australia, ⁴Centre for Clinical Research, The University of Queensland, Brisbane, Australia, ⁵Australian Institute of Bioengineering and Nanotechnology, The University of Queensland, St Lucia, Australia, ⁶QIMR Berghofer Medical Research Institute, Brisbane, Australia, ⁷School of Information Technology and Electrical Engineering, The University of Queensland, St Lucia, Australia, ⁸Centre for Advanced Imaging, The University of Queensland, Brisbane, Australia, ⁹Health and Biosecurity, The Commonwealth Scientific and Industrial Research Organisation (CSIRO), Brisbane, Australia

Segmentation and volumetric analysis of the hypothalamus and fornix plays a critical role in improving the understanding of degenerative processes that might impact the function of these structures. We present Open-Source Hypothalamic-ForniX (OSHy-X) atlases and tool for multi-atlas fusion segmentation for 3T and 7T. The atlases are based on 20 manual segmentations, which we demonstrate have high interrater agreement. The versatility of the OSHy-X tool allows segmentation and volumetric analysis for different field strengths and contrasts. We also demonstrate that OSHy-X segmentation has higher Dice overlaps (3T and 7T inputs: p<0005, p<0.005) than a deep-learning segmentation method for the hypothalamus.

Yasuhiko Tachibana¹, Hiroki Tsuchiya¹, Riwa Kishimoto¹, Tokuhiko Omatsu¹, Shinichiro Mori¹, Takayuki Obata¹, and Tatsuya Higashi^1
1National institutes for Quantum science and Technology, Chiba, Japan

Deep-learning-based slice-by-slice noise reduction may not be suitable for low-SNR body DWI that contains insufficient information in the original single slice. Moreover, averaging multiple acquisitions after denoising to avoid this problem is insufficient because it causes blurring owing to a mismatch between acquisitions. Herein, we designed a neural network that utilises 4D convolution to incorporate adjacent slices and multiple acquisitions simultaneously for a slice to achieve adequate denoising. The results support the utility of the proposed method in comparison with the usual slice-by-slice method and averaging.

Shang Wan¹, Yi Wei¹, Hehan Tang¹, Lisha Nie², Xiaocheng Wei², and Bin Song^3
1Radiology, West China Hospital, Sichuan University, Cheng Du, China, ²GE Healthcare Beijing China, Beijing, China, ³West China Hospital, Sichuan University, Cheng Du, China

Early detection and diagnosis of hepatocellular carcinoma (HCC) is essential for patients’ prognosis, however, the imaging hallmarks for tumor detection and diagnosis has remained same for years despite the use of many new imaging methods. Thus, in this study, we aimed to prospectively evaluated the detection performance of various MR sequences and abbreviated MRI (aMRI) protocols in different clinical settings and further compared the different imaging criteria for the diagnosis of HCC using either extracellular contrast agent (ECA) or hepatobiliary specific contrast (HBSC) MR imaging, and the developed diagnostic criteria for detection and diagnosis of HCC may aid clinical diagnosis.