Toshiko Tanaka¹, Naohiro Okamoto², Ikuhiro Kida^1,2, and Masahiko Haruno^1,2
1National Institute of Information and Communications Technology, Suita Osaka, Japan, ²Osaka Universitiy, Suita Osaka, Japan

Previous studies suggested that the initial decrease in the BOLD signal reflects primary neuronal activity more than the later hemodynamic positive peak responses. We applied the hyper-alignment algorithm to 7T-BOLD timeseries during the facial expression discrimination task. and conducted the MVPA using the aligned data. We found decoding accuracies in the amygdala and superior temporal sulcus at 2 s after the face onset were significantly beyond baseline and the voxels contributing to the decoding accuracy displayed decreasing pattern in hemodynamics response, revealing that the initial decrease in 7T-BOLD signals contains finer information than thought previously.

Tsutomu Inaoka¹, Akihiko Wada², Rumiko Ishikawa¹, Tomoya Nakatsuka¹, Hisanori Tomobe¹, Masaru Sonoda³, Akinori Yamamoto¹, Ryousuke Sakai¹, and Hitoshi Terada^1
1Radiology, Toho University Sakura Medical Center, Sakura, Japan, ²Radiology, Juntendo University, Tokyo, Japan, ³Radiology, Seirei Sakura Citizen Hospital, Sakura, Japan

The DML model including fat-suppressed contrast generation, normal image restoration, and classification and determination models may make it possible to detect all abnormalities in knee joint MRI once.

Ting He¹, Yi Mao¹, Yao Wang¹, Lei Wang¹, Qinmei Kuang¹, Jie Xu¹, Yuqi Ji¹, Yujie He¹, Meimei Zhu¹, and Fuqing Zhou^1
1The First Affiliated Hospital, Nanchang University, Nanchang, China

Multiple sclerosis (MS) is the most common immune-mediated disease of the central nervous system. Early identification of MS lesions and its mimics is very important to help alleviate the tension between the benefits of early diagnosis of MS and inaccurate diagnosis that can have serious health and economic consequences.The radiomics feature model based on T2-weighted images (T2WIs) has obvious clinical value and high specificity in differentiating patients with multiple sclerosis and ischemic demyelination. The high specificity of radiomic model could improve the accuracy of the 2017 McDonald diagnostic criteria for MS, by differentiating it from its mimics-- ischemic demyelination.

Haoyuan Huang¹, Baoer Liu², Yikai Xu², and Wu Zhou^1
1School of Medical Information Engineering, Guangzhou University of Chinese Medicine, Guangzhou, China, ²Department of Medical Imaging Center, Nanfang Hospital, Southern Medical University, Guangzhou, China

The intravoxel incoherent motion (IVIM) model of DWI with IVIM parameters has been widely used in characterization. However, the optimal method to obtain the IVIM parameters is still being explored. In this work, we propose a synthetic-to-real domain adaptation method for fitting the IVIM parameters. Specifically, we use synthesized data to train the network to learn the accurate mapping of the b-value images to the parameter map, and design a discriminator to help the network gradually adapt the learned mapping to the real data. Experimental results demonstrate that the proposed method outperforms previously reported methods for fitting IVIM parameters.

Di Guo¹, Jiaying Zhan¹, Zhangren Tu², Yi Guo¹, Yirong Zhou², Jianfan Wu¹, Qing Hong³, Vladislav Orekhov⁴, and Xiaobo Qu^2
1School of Computer and Information Engineering, Xiamen University of Technology, Xiamen, China, ²Department of Electronic Science, Biomedical Intelligent Cloud R&D Center, Fujian Provincial Key Laboratory of Plasma and Magnetic Resonance, Xiamen University, Xiamen, China, ³China Mobile Group, Xiamen, China, ⁴Department of Chemistry and Molecular Biology, University of Gothenburg, Gothenburg, Sweden

Nuclear magnetic resonance (NMR) serves as an indispensable tool in revealing physical, chemical and structural information about molecules. We present a hypercomplex low rank approach to reconstruct hypercomplex NMR spectrum reconstruction. We first introduce an adjoint matrix operation to convert the hypercomplex signal into complex matrix and  then propose a low-rank model and algorithm to reconstruct hypercomplex signal. The experiment results demonstrate that the proposed method provides a fast and high-fidelity reconstruction of hypercomplex NMR data. Furthermore, we made the method available at an open-access and easy-to-use cloud computing platform.

Chaoxing Huang¹, Yurui Qian¹, Jian Hou¹, Baiyan Jiang^1,2, Queenie Chan³, Vincent Wong⁴, Winnie Chu¹, and Weitian Chen^1
1Department of Imaging and Interventional Radiology, The Chinese University of Hong Kong, Shatin, Hong Kong, ²Illuminatio Medical Technology Limited, Hong Kong, China, ³Philips Healthcare, Hong Kong, China, ⁴Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Shatin, Hong Kong

Quantification of liver  T1rho has gained interest in liver pathological study. Traditional fitting method requires acquisition of multiple  T1rho-weighted images and it can be affected by respiratory motion. We propose a physics-informed self-supervised mapping method by taking only one  T1rho-weighted image to do the mapping. Our preliminary experimental results show that our method has the potential to outperform the traditional multi-TSL acquisition method, particularly in the scenario of free-breathing MRI scan.

Dian Jiang^1,2, Zhanqi Hu³, Cailei Zhao⁴, Xia Zhao³, Jun Yang^1,2, Yanjie Zhu^2,5, Jianxiang Liao³, Dong Liang^1,2,5, and Haifeng Wang^2,5
1Research Centre for Medical AI, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, ²University of Chinese Academy of Sciences, Beijing, China, ³Department of Neurology, Shenzhen Children’s Hospital, Shenzhen, China, ⁴Department of Radiology, Shenzhen Children’s Hospital, Shenzhen, China, ⁵Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China

Distinguishing epilepsy drug treatment outcomes is crucial for treating children with tuberous sclerosis complex (TSC). Here, a deep-learning framework named AE-net was proposed to analyze epilepsy drug treatment outcomes using multi-contrast MRI data. Firstly, multi-contrast image-based models were respectively generated using the EfficientNet3D-B0 networks. Then, an averaging ensemble network was created as the final model. The proposed AE-net achieved the best AUC performance of 0.800 and sub-optimal AUC performance of 0.763 in the testing cohort, better than others. And the proposed method can predict epilepsy drug treatment outcomes to help clinical radiologists formulate more targeted treatments in the future.

Xinjun Han¹, Niange Yu², Qianjiang Xiao², Mingyang Gao², Dandan Zheng², and Zhenghan Yang^1
1Department of Radiology, Beijing Friendship Hospital, Capital Medical University, Beijing, China, ²Shukun (Beijing) Technology Co., Ltd, Beijing, China

Accurate segmentation of liver not only facilitates the subsequent quantitative assessment of the regions of interest but also benefits precise diagnosis, and surgical planning. These tasks are usually performed by radiologists via visual inspection and manual delineations, which are tedious, labor-intensive, time-consuming. Convolutional neural networks (CNNs) have shown promise for performing automated liver segmentation for CT examinations, but there is less research on MR images. In this study, we provide a 3D U-Net based model for robust whole-liver and Couinaud segment measurements to support the treatment decision-making process on MR images.

Siyao Du¹, Mengfan Wang¹, Shasha Liu¹, Xiaoqian Bian¹, Xinyue Chen¹, Liangcun Guo¹, Guoliang Huang¹, Ruimeng Zhao¹, Can Peng¹, Wenhong Jiang¹, Qinglei Shi², Xu Yan², Guang Yang³, and Lina Zhang^1
1Department of Radiology, The First Affiliated Hospital of China Medical University, Shenyang, China, ²MR Scientific Marketing, Siemens Healthineers Ltd., Beijing, China, ³Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai, China

 In this study, we built a support vector machine (SVM) model based on quantitative parameters of continuous-time random-walk (CTRW) diffusion model in predicting the human epidermal growth factor receptor-2 (HER-2) expression in breast invasive ductal carcinoma. An AUC of 0.753 was achieved, which may have a great potential in future clinical practice.

Cong Ning¹, Yue Ma¹, Di Yu¹, Dan Tong¹, Yi Zhu², and Ke Jiang^2
1Department of Radiology, The first hospital of jilin university, Changchun,Jilin, China, ²Philips Healthcare, Beijing, China

MRI has advantages in detecting acute ischemia and describing the core volume of the infarct without radiation, but the long scanning time affects clinical use. Compressed sense combined with artificial intelligence (CS-AI) technology accelerates the acquisition of imaging. This study aims to use the CS-AI technique to accelerate common sequences of Acute cerebral infarction and investigate the effect of different acceleration factors on image quality and diagnosis. The results show that CS-AI reconstruction scans reduce scanning time while maintaining image quality compared to conventional SENSE.

Zhaoyang Jin¹ and Qing-San Xiang^2
1Hangzhou Dianzi University, Hangzhou, China, ²University of British Columbia, Vancouver, BC, Canada

In this study, we exploited a sparsifying deep learning method and an inverse filtering reconstruction to obtain high quality complex MR images for under-sampled MRI data. This study allows much more flexible data representations for complex MRI data training, leading to significantly higher complex reconstruction quality for practical MRI applications.

Yue Ma¹, Linna Li², Yang Sun², Chang Zhai², Dan Tong², Yi Zhu³, and Ke Jiang^3
1Department of Radiology, The First Hospital of Jilin University, Changchun,Jilin, China, ²The First Hospital of Jilin University, Changchun,Jilin, China, ³Philips Healthcare,Beijing,China, Beijing, China

Compressed SENSE-AI reconstruction system was introduced with its sufficient noise removal to maintain high image quality while reducing scanning time. Intracranial and carotid vessel wall MR imaging (VW-MRI) are widely available in detecting and characterizing atherosclerosis occurring in the two vascular ranges, however, long scan time had so far limited high resolution VW-MR imaging. In this study, we investigated the impact of Compressed SENSE-AI acceleration factor on the diagnostic quality of magnetic resonance images and compared with standard vessel wall MR imaging protocol.

Fuliang Lin^1,2, Zhou Liu³, Qinglei Zhou², Pengyu Gao⁴, Jie Wen³, Meng Wang³, Ya Ren³, Dehong Luo³, Ye Li¹, Dong Liang¹, Xin Liu¹, Hairong Zheng¹, and Na Zhang^1
1Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, ²Zhengzhou University, Zhengzhou, China, ³Department of Radiology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital & Shenzhen Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Shenzhen, China, ⁴Henan University, Kaifeng, China

Breast cancer is the most common cancer in women with the highest incidence. Dynamic contrast enhanced MRI is one of the backbone sequences for breast cancer diagnosis. Accurate segmentation of breast lesions based on DCE-MRI images is helpful for clinically objective and quantitative evaluation of breast lesions. However, the commonly used manual segmentation method is subject to high inter-observer variability. In this study, a 3D automatic algorithm is proposed for segmentation of breast lesions in DCE-MRI. The results show that the proposed network can obtain accurate and automatic 3D segmentation of breast lesions and achieves better segmentation results than VNet.

Dicheng Chen¹, Wanqi Hu¹, Huiting Liu¹, Tianyu Qiu¹, Yihui huang¹, Liangjie Lin², Di Guo³, Jianzhong Lin⁴, and Xiaobo Qu^1
1Department of Electronic Science, National Institute for Data Science in Health and Medicine, Xiamen University, Xiamen, China, ²Healthcare, Philips, Beijing, China, ³School of Computer and Information Engineering, Xiamen University of Technology, Xiamen, China, ⁴Department of Radiology, The Zhongshan Hospital affiliated to Xiamen University, Xiamen, China

¹H Magnetic Resonance Spectroscopy (MRS) suffers low Signal-Noise Ratio (SNR) due to low concentrations of metabolites. To improve the SNR, the current mainstream is to do Signal Averaging with  repeated samplings but it is time-consuming. Therefore, we designed a novel denoising ReLSTM-Net to learn the mapping from the low SNR MRS  to the high SNR one  in the time-domain by a few in vivo measured data. Denoised spectra by the proposed method has higher accuracy and reliability in quantifying metabolites Glx, tCho and mI, compared with the state-of-art Low-Rank method.

Lifeng Mei^1,2, Sixing Liu^1,2, Guoxiong Deng^1,2, Shaojun Liu^1,2, Yali Zheng^1,2, and Mengye Lyu^1,2
1College of Health Science and Environmental Engineering, Shenzhen Technology University, Shenzhen, China, ²College of Applied Sciences, Shenzhen University, Shenzhen, China

Computer aided diagnosis (CAD) is widely considered an important application of deep learning in healthcare. However, brain data with lesion location labels are rare in MRI domain. Recently, Microsoft Research has released a dataset of clinical pathology annotations based on the raw images from fastMRI and named it fastMRI+. Here, fastMRI+ brain dataset was analyzed and used to train deep learning-based models for lesion detection. Some well-known object detection architectures such as YOLOv3, Faster-RCNN and YOLOX  were compared. Overall, this abstract established a baseline with improvement suggestions for future studies.