Haribalan Kumar*^1,2,3, Robby Green*¹, Samantha Holdsworth^1,4, Daniel Cornfeld¹, Paul Condron¹, Eryn Kwon¹, Taylor Emsden¹, Julie Choisne³, Thor Besier³, Justin Fernandez³, Kat Gilbert³, Martyn Nash³, Soroush Safaei³, Gonzalo Maso Talou³, Davidson Taylor^1,5,6, Luke Rolfe^1,7, Lucy McHugh^1,8,9, Graham Hingangaroa Smith^{10,11,12,13,14,15}, Leigh Potter^11,16,17,18, Merryn Tawhai³, Kelly Burrowes³, Alys Clark³, Jiantao Shan³, Elias Soltani³, Erana Hogarth^1,17, Rinki Murphy⁴, Josh McGeown¹, Laura Carman³, Alan Wang^1,3, Maryam Tayebi^1,3, Ali Mirjalili⁴, Grace Cleland-Pottie^1,19,20, Mahyar Osanlouy³, Vickie Shim^1,3, Jerome Maller², Reweti Ropiha²¹, and Graham Wilson^1,8
1Mātai Medical Research Institute, Gisborne, New Zealand, ²General Electric Healthcare AUS/NZ, Melbourne, Australia, ³Auckland Bioengineering Institute, Auckland, New Zealand, ⁴Faculty of Medical and Health Sciences, University of Auckland, Auckland, New Zealand, ⁵Ngāi Tāmanuhiri, Tairāwhiti, New Zealand, ⁶Rongowhakaata, Tairāwhiti, New Zealand, ⁷Ngāti Pāhauwera, Hawke's Bay, New Zealand, ⁸University of Otago, Dunedin, New Zealand, ⁹Ngāi Tahu, Te Waipounamu, New Zealand, ¹⁰Kāti Māmoe, North Otago, New Zealand, ¹¹Mātai Ngā Māngai Māori, Gisborne, New Zealand, ¹²Te Aitanga a Hauiti, Tairāwhiti, New Zealand, ¹³Ngāti Apa, Rangitīkei, New Zealand, ¹⁴Massey University, Palmerston North, New Zealand, ¹⁵Ngāti Kahungunu, Wairarapa, New Zealand, ¹⁶Rongomaiwahine, Mahia, New Zealand, ¹⁷Ngāti Porou, Tairāwhiti, New Zealand, ¹⁸Ngāti Kahungunu, Wairoa, New Zealand, ¹⁹University of Cantebury, Christchurch, New Zealand, ²⁰Ngāpuhi, Te Tai Tokerau, New Zealand, ²¹Turanga Health, Tairāwhiti, New Zealand

Approaches to magnetic resonance (MR) image acquisition and computational modelling often exist in silo. The application of computational modelling to paediatric data is in its infancy. This collaborative study addresses this gap for multiple organ systems. State-of-the art MR imaging sequences were selected and refined to provide imaging data for the creation of subject-specific geometric models. The MR protocol and modelling outputs will underpin a longitudinal paediatric study set in Aotearoa New Zealand, generating normative data in healthy children, and identifying early biomarkers of pathological processes. This will enable the development of new personalised, predictive, and preventative models of care.

Shuai Shen^1,2, Wenjing Xu¹, Hongbing Ma³, Xiaoyi Lv², Guanxun Cheng⁴, Liwen Wan¹, Lei Zhang¹, Ye Li¹, Dong Liang¹, Xin Liu¹, Hairong Zheng¹, and Na Zhang^1
1Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, ²College of Software, xinjiang University, Xinjiang, China, ³Department of Electronic Engineering, Tsinghua University, Beijing, China, ⁴Department of Radiology, Peking university shenzhen hospital, Beijing, China

Atherosclerotic plaque is a major cause of ischemic stroke. Some arterial morphological features obtained from MR vessel wall images show great potential for identifying high-risk plaques. Deep learning has now been applied to the automatic segmentation of vessel walls to accurately and efficiently measure arterial morphological features. However, the accuracy of the existing segmentation methods is not yet high enough for clinical practical applications. This study proposed a new segmentation framework with custom convolutional trajectories for automatic segmentation of arterial vessel wall and the framework improved the accuracy of vessel wall segmentation.

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

We proposed a unified deep learning framework for predicting the motion deformation between different phases of 4D-MRI with simultaneous image quality enhancement. The network combines a coarse-to-fine unsupervised registration model to estimate the deformation vector fields (DVFs) in different image scales and a GAN-based enhancement network to restore anatomic features. Particularly, a prior knowledge of 4D-MRI is incorporated into the unified model, guiding an accurate DVF prediction and maintaining image topology. Both qualitative and quantitative results showed that the predicted DVFs and resultant 4D-MRI images achieved improved performance compared with the traditional method without modifications.

Yuya Saito¹, Koji Kamagata¹, Norihide Maikusa², Christina Andica¹, Wataru Uchida¹, Hayato Nozaki^1,3, Mana Owaki^1,3, Akifumi Hagiwara¹, Shohei Fujita^1,4, Toshiaki Akashi¹, Akihiko Wada¹, Shinsuke Koike², Masaaki Hori⁵, and Shigeki Aoki^1
1Department of Radiology, Graduate School of Medicine, Juntendo University, Tokyo, Japan, ²Center for Evolutionary Cognitive Sciences, Graduate School of Art and Sciences, The University of Tokyo, Tokyo, Japan, ³Department of Radiological Sciences, Graduate School of Human Health Sciences, Tokyo Metropolitan University, Tokyo, Japan, ⁴Department of Radiology, University of Tokyo, Tokyo, Japan, ⁵Department of Radiology, Toho University Omori Medical Center, Tokyo, Japan

When analyzing multi-site diffusion MRI (dMRI) data, metrics should be harmonized to remove the site effect. In this study, we applied the combined association test (ComBat), which uses regression of covariates with an empirical Bayes framework, for diffusion metrics based on diffusion tensor imaging (DTI) and neurite orientation dispersion and density imaging (NODDI). The results showed that ComBat can harmonize site-related effects in DTI and NODDI metrics based on multi-site dMRI while preserving subject biological information, such as sex differences and correlation with age. Thus, ComBat could be applied in large multi-site studies to identify subtle white matter changes.

Motohide Kawamura¹, Daiki Tamada¹, Kazuyuki Sato², Masahiro Hamasaki², Satoshi Funayama¹, Tetsuya Wakayama³, Utaroh Motosugi⁴, Hiroyuki Morisaka¹, and Hiroshi Onishi^1
1Department of Radiology, University of Yamanashi, Chuo, Japan, ²Division of Radiology, University of Yamanashi Hospital, Chuo, Japan, ³MR Collaboration and Development, GE Healthcare, Hino, Japan, ⁴Department of Radiology, Kofu-Kyoritsu Hospital, Kofu, Japan

PROPELLER sequence is useful because of its robustness to patient motion. Longer acquisition than FSE is a major drawback limiting its wider application in clinical practice. Here, we propose an accelerated T1 weighted PROPELLER of the brain using deep learning based parallel imaging (PI) reconstruction. Our method can unfold highly undersampled aliased images (PI factor = 7), enabling 2.3 times faster acquisition than full-sampling. A preliminary reader study with prospectively undersampled data showed that the proposed method significantly outperformed a conventional SENSE reconstruction in terms of streak artifact.

Jun Wang¹, Guangyao Liu¹, Pengfei Zhang¹, Kai Ai², Laiyang Ma¹, Wanjun Hu¹, and Jing Zhang^1
1Magnetic Resonance, Lanzhou University Second Hospital, Lanzhou, China, ²Philips Healthcare, Xi’an, China

Diagnosis and differential diagnosis of Inflammatory Bowel Disease (IBD) remains challenge since the particularity of small bowel. The objective of this study was to use the contrast enhanced MRI (CE-MRI) radiomics to distinguish subtypes of IBD patients. A total of 216 patients confirmed by pathology underwent small bowel CE-MRI. After the pipeline of radiomics, it was found that wavelet transformed texture features can effectively identify ulcerative colitis (UC) and Crohn’s disease (CD) with high performance. This work provides more detailed and microscopic details for the differential diagnosis of IBD subtypes based on the preliminary results.

Bin Guo^1,2, Fugen Zhou¹, Muwei Li^2,3, Zhaohua Ding^2,4,5, and John C. Gore^2,3,5
1Image Processing Center, School of Astronautics, Beihang University, Beijing, China, ²Vanderbilt University Institute of Imaging Science, Nashville, TN, United States, ³Department of Radiology and Radiological Sciences, Vanderbilt University Medical Center, Nashville, TN, United States, ⁴Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, TN, United States, ⁵Department of Biomedical Engineering, Vanderbilt University, Nashville, TN, United States

Blood oxygenation-level dependent (BOLD) MRI signals have been reliably detected in white matter (WM) in both task and resting states in numerous studies. However, the relationship between WM BOLD signals and regional metabolism remains to be elucidated. In the present study, we investigated the relationship between resting state functional connectivity and glucose uptake in WM using simultaneous MRI and PET studies of human subjects. We find a significant correlation between these two measurements, suggesting that functional involvement of WM in neural activities was accompanied by an increase in glucose metabolism.

Yoichiro Ikushima^1,2, Shogo Tokurei¹, Hiroyuki Tarewaki³, Junji Morishita⁴, and Hidetake Yabuuchi^4
1Department of Radiological Science, Faculty of Health Sciences, Junshin Gakuen University, Fukuoka, Japan, ²Department of Health Sciences, Graduate School of Medical Sciences, Kyushu University, Fukuoka, Japan, ³Division of Radiology, Department of Medical Technology, Osaka University Hospital, Suita, Japan, ⁴Department of Health Sciences, Faculty of Medical Sciences, Kyushu University, Fukuoka, Japan

The QEMDIM, a no-reference image quality assessment (IQA) algorithm, calculates the quality scores based on the quality difference between an image and reference images. Therefore, scores can be affected by the quality of the reference images. We modified the QEMDIM by changing the scoring method to compensate for this drawback. Brain MR images of various qualities were scored by subjective IQA (our gold standard), QEMDIM, and the modified algorithm. Modified scores showed a higher correlation with subjective IQA scores than the QEMDIM scores, using reference images of varied qualities. Our modified algorithm would be more clinically advantageous than the QEMDIM.

Yaan Ge¹, Xiaolan Liu¹, Qingyu Dai¹, and Kun Wang^1
1GE Healthcare, Beijing, China

This study proposed an automatic no-reference image quality rating metrics (VSS score) based on SVR model, that not requiring clinical expert labeled data, simulating human visual sense and applicable to all anatomies and contrast. The feasibility of applying this rating metrics in DL-recon integrated rapid scan protocol automatic evaluation is demonstrated. The result shows VSS score is in good correlation with visual sense to image quality and outperformed BRISQUE and PIQUE rating algorithms.

Ghazaleh Jamshidi¹, Ali Abbasian Ardakani², Farshid Babapour Mofrad¹, Hamidreza Saligheh Rad^3,4, and Mahyar Ghafoori^5
11- Department of Medical Radiation Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran (Islamic Republic of), ²2- Department of Radiology Technology, School of Allied Medical Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran (Islamic Republic of), ³Tehran University of Medical Science, Tehran, Iran (Islamic Republic of), ⁴Quantitative MR Imaging and Spectroscopy Group, Research Center for Molecular and Cellular Imaging, Tehran University of Medical Sciences, Tehran, Iran (Islamic Republic of), ⁵3- Department of Radiology, Hazrat Rasoul Akram Hospital, School of Medicine, Iran University of Medical Sciences, Tehran, Iran (Islamic Republic of)

Multiparametric MRI (MP-MRI) has been widely used for detection of Prostate Cancer (PCa). In this study, we propose a new method using MP-MRI including T2-weighted (T2W) and dynamic contrast enhanced (DCE-) MRI for detection of PCa. 32 patients who had high prostate specific antigen (PSA) level recruited. We generated predictive models by extracting radiomics features and classifying benign and malignant lesions. The feature scores are evaluated with Relieff feature selection for each of the modalities. The fused classifier using decision template method showed the highest performance with accuracy, specificity, and sensitivity of 100.

Yang Li¹, Shihan Ma¹, Qing Li², Bo Xiong³, Weiwei Wang⁴, Xinjun Sheng¹, and Xiangyang Zhu^1
1Shanghai Jiao Tong University, Shanghai, China, ²MR Collaborations, Siemens Healthineers Ltd., Shanghai, China, ³Siemens Digital Medical Technology (Shanghai) Co., Ltd., Shanghai, China, ⁴Huashan Hospital affiliated to Fudan University, Shanghai, China

To reconstruct the trajectories of forearm muscle fibers during dynamic contractions, we designed a customized MRI-compatible rig to help acquire MRI data of the forearm muscles under three wrist postures. A piecewise registration method was proposed to correct the misregistration between DTI and anatomical images. The reconstructed supinator was highly consistent with the cadaveric dissection reference. Diffusion and architectural parameters were calculated to characterize voluntary contractions. During the wrist flexion and extension, the fascicle length of the supinator was inversely related to the pennation angle. The diffusion parameters increased slightly from neutral position to the other two postures.

Deepa Anand¹, Dattesh Dayanand Shanbhag¹, and Rakesh Mullick^1
1Advanced Technology Group, GE Healthcare, Bangalore, India

Data diversity is a key ingredient for robust deep learning models, especially in the medical domain. We present a diversity visualization and quantification scheme which enables decisions on data selection different enough from already existing data. Out experiments amply validate the usefulness of the proposed diversity metric in terms of enhancement in accuracy of models resulting from using them in data selection decision process  with accuracy improvement from 3%->10% across different sites.

YAN TU HUANG¹, Peter Speier², Tom Hilbert^3,4,5, and Tobias Kober^3,4,5
1Magnetic Resonance, Siemens Shenzhen Magnetic Resonance Ltd., Shenzhen, China, ²Magnetic Resonance, Siemens Healthcare, Erlangen, Germany, ³Advanced Clinical Imaging Technology, Siemens Healthcare, Lausanne, Switzerland, ⁴Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, ⁵LTS5, École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland

Motion continues to be a major impediment in clinical MRI. Using Pilot Tone to correct for motion is highly attractive as it uses inexpensive hardware and is independent from sequence. Here, we propose a model-based approach to calculate motion parameters from Pilot Tone signal. To avoid non-convergence those models are prone to, we introduce predefined displacement fields with the Pilot Tone as auxiliary, resulting in a linear motion model with fewer unknowns. Validation is performed on numerically simulated data and in-vivo measurements, comparing different motion models. Image reconstructions with the Pilot Tone showed fewer residual motion artefacts and converged faster.

Rui Wu¹, Weiqiang Dou², and Aiyin Li^3
1Department of radiology, Shandong Provincial Qianfoshan Hospital, Cheeloo college of Medicine, Shandong University, Jinan, China, ²MR Research, GE Healthcare, Beijing, China, ³Department of radiology, The First Affiliated Hospital of Shandong First Medical University & Shandong Provincial Qianfoshan Hospital, Jinan, China

This study mainly aimed to build a multi-modality radiomics model for predicting lymph node metastasis (LNM) of rectal cancer (RC) preoperatively. 82 RC patients were enrolled in this study. Two, six, four and eleven optimal features were separately selected for venous phase CT, HR-T2WI, DWI and multi-modality by incorporating CT, HR-T2WI and DWI together. Together with MRI-reported LN status feature, clinical features, and the combination of multi-modality and clinical features, in total seven support vector machine (SVM) classification models were respectively built. Incorporated with clinical features, multi-modality radiomics model provided the most robust predictive performance in predicting LNM.

Yashi Nan¹, Jian Zhang¹, Jianming Ni², and Silun Wang^1
1Yiwei Medical Technology, Shenzhen, China, ²Department of Radiology, Wuxi No2. People’s Hospital, Affiliated Wuxi Clinical College of Nantong University, Wuxi, China

White matter hyperintensities (WMH) features may assist detecting diabetes-related MCI. We aimed to identify the correlations between WMH and MCI in T2D using brain magnetic resonance imaging. Fifty participants, matched for age, were included. Total and regional WMH volumes were calculated using automated segmentation approach. WMH patterns were compared between groups using a voxel-wise analysis. Results show that cognitive impairment was related to a higher prevalence of regional WMH. Subcortical WMH volumes in specific brain regions. The findings suggested that the presence of WMH in specific regions rather the WMH volumes appears to be correlated with the MCI in T2D.