Na Hu¹, Tianwei Zhang², Yifan Wu³, Biqiu Tang¹, Minlong Li¹, Qiyong Gong¹, Shi Gu², and Su Lui^1
1Huaxi MR Research Center (HMRRC), Department of Radiology, West China Hospital of Sichuan University, Chengdu, China, ²Department of Computer and Engineering, University of Electronic Science and Technology of China, Chengdu, China, ³Department of Bioengineering, University of Pennsylvania, Philadelphia, PA, United States

We aimed to develop a method of CT-to-MR image synthesis to assist in detecting hypoattenuating brain lesions in acute ischemic stroke. Emergency head CT images of 193 patients with suspected stroke and follow-up MR images were collected. A generative-adversarial-network model was developed for CT-to-MR image synthesis. With synthetic MRI compared to CT, sensitivity was improved by 116% in patient detection and 300% in lesion detection, and extra 75% of patients and 15% of lesions missed on CT were detected on synthetic MRI. Our method could be a rapid tool to improve readers’ detection of hypoattenuating lesions in AIS.

Silu Zhang¹, Zoltan Patay¹, Bogdan Mitrea², Angela Edwards¹, Lydia McColl Makepeace¹, and Matthew A. Scoggins^1
1Diagnostic Imaging, St. Jude Children's Research Hospital, Memphis, TN, United States, ²Activ Surgical, Boston, MA, United States

Diffuse intrinsic pontine glioma (DIPG) is a pediatric brain tumor with very poor prognosis. In this study, we identify two subtypes of DIPG based on radiomic features. The two subtypes show a significant difference in survival rates. Subtype 1 has a mean progression-free survival (PFS) and overall survival (OS) of 8.9 and 12.7 months, respectively. Subtype 2 has a mean PFS and OS of 5.7 and 9.1 months, respectively. Our results suggest that shape features and intensity features extracted from FLAIR and T1-post contrast predict the prognosis of DIPG.

Zihao Chen^1,2, Yuhua Chen^1,2, Ankur Saini³, William Devine³, Yibin Xie¹, Cecilia Lo³, Debiao Li^1,2, Yijen Wu³, and Anthony Christodoulou^1
1Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, United States, ²Department of Bioengineering, UCLA, Los Angeles, CA, United States, ³Department of Developmental Biology, University of Pittsburgh, Pittsburgh, PA, United States

Genetically engineered mouse models (GEMM) are indispensable in modeling human diseases. High resolution MRI with spatial resolution less than 100 μm has made incredible progress for phenotyping mouse embryos. However, it takes more than 10 hours' acquisition time to reach such high resolution, so reducing the scan time is of great need. Here we propose a deep learning based super-resolution approach for 3x3 super-resolution (SR) of mouse embryo images using raw k-space data. Our method can reduce the scan time by a factor of 9 while preserving the diagnostic details and shows better quantitative results than previous SR methods.

Moran Artzi^1,2,3, Erez Redmard³, Oron Tzemach³, Jonathan Zeltser³, Omri Gropper⁴, Jonathan Roth^2,5,6, Ben Shofty^2,5,7, Danil A. Kozyrev^5,7, Shlomi Constantini^2,5,7, and Liat Ben-Sira^2,8
1Sagol Brain Institute, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, ²Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel, ³Sagol School of Neuroscience, Tel Aviv University, Tel Aviv, Israel, ⁴The Iby and Aladar Fleischman Faculty of Engineering, Tel Aviv University, Tel Aviv, Israel, ⁵Department of Pediatric Neurosurgery, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, ⁶The Gilbert Israeli Neurofibromatosis Center, Tel Aviv University, Tel Aviv, Israel, ⁷The Gilbert Israeli Neurofibromatosis Center, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel, ⁸Division of Radiology, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel

A fused architecture contrived of 2 neural networks, pre-trained ResNet-50 CNN and tabular based network is proposed for the classification of Posterior fossa tumors (PFT) types. The study included data for 158 MRI of healthy controls and pediatric patients with PFT. The input data were T1WI+C, FLAIR and diffusion MRI, and tabular data (subject's age). The best classification results obtained by the fused CNN + tubular data architecture and based on diffusion images, achieved cross-validation accuracy of validation=0.88±0.04, test=0.87±0.02. Overall, the proposed architecture achieved a ~16% improvement in accuracy for the test data compared to CNN method for this dataset.

Samantha J Ma^1,2, Mona Sharifi Sarabi², Kai Wang², Wenli Tan², Huiting Wu³, Lei Hao³, Yulan Dong³, Hong Zhou³, Lirong Yan², Yonggang Shi², and Danny JJ Wang^2
1Siemens Medical Solutions USA, Inc., Los Angeles, CA, United States, ²University of Southern California, Los Angeles, CA, United States, ³Department of Radiology, The First Affiliated Hospital of University of South China, Hunan, China

Given the inaccessibility of cerebral small vessels to existing clinical in vivo imaging technologies, early cerebral microvascular morphological changes in small vessel disease (SVD) can be difficult to evaluate. In this study, we trained a deep learning (DL)-based algorithm with 3T and 7T black-blood images on two vendor platforms to automatically segment lenticulostriate arteries (LSAs) in the brain. Our results show that black-blood imaging in conjunction with DL is a promising approach to enable quantitative morphometric analysis in patients with cerebral SVD.

Berardino Barile¹, Aldo Marzullo², Claudio Stamile³, Françoise Durand-Dubief⁴, and Dominique Sappey-Marinier^1,5
1CREATIS (UMR 5220 CNRS & U1206 INSERM), Université Claude Bernard Lyon 1, Villeurbanne, France, ²Department of Mathematics and Computer Science, University of Calabria, Rende, Italy, ³R&D Department, CGnal, Milan, Italy, ⁴Hôpital Neurologique, Hospices Civils de Lyon, Bron, France, ⁵MRI, CERMEP - Imagerie du Vivant, Bron, France

The Expanded Disability Status Scale (EDSS) monitors physical impairment in Multiple Sclerosis (MS). A Staking Ensemble model composed of 4 ML "boosting" models was used to predict EDSS using both white matter (WM) fiber-bundles and structural connectome data. This model provided excellent prediction results with an RMSE of 0.92 to 1.08. A counterfactual model was added to highlight the most important WM links and fiber-bundles in the prediction process. The accordance of the findings obtained with both data types confirmed the clinical interest of such methods for disability prediction using DTI data.

ChengLong Deng^1,2, BingChao Wu^1,2, QingJun Wang³, QingLei Shi⁴, Bei Guan^1,2, DaCheng Qu⁵, and YongJi Wang*^1,2,6
1Collaborative Innovation Center, Institute of Software, Chinese Academy of Sciences, Beijing, China, ²University of Chinese Academy of Sciences, Beijing, China, ³Department of Radiology, PLA 6th medical center, Beijing, China, ⁴MR Scientific Marketing, Siemens Healthcare, Beijing, China, ⁵School of Computer Science and Technology， Beijing Institute of Technology, Beijing, China, ⁶State Key Laboratory of Computer Science, Institute of Software, Chinese Academy of Sciences, Beijing, China

In this study, we optimized a Faster R-CNN algorithm through constraining anchor boxes generated by Region Proposal Network (RPN) based on prior knowledge, and evaluated the feasibility of the optimized model in detecting and differentiating Hashimoto's thyroiditis (HT) from papillary thyroid microcarcinomas (PTMC) based on high b-value (2000 sec/mm²) diffusion-weighted images that acquired with readout segmentation of long variable echo-trains (RESOLVE) sequence. The study indicated that our model based on high b-value (2000 sec/mm²) DWI images demonstrated great potential as a new inspection tool in the diagnosis of benign and malignant thyroid micronodules.

Po-Jui Lu^1,2,3, Muhamed Barakovic^1,2,3, Matthias Weigel^1,2,3,4, Reza Rahmanzadeh^1,2,3, Riccardo Galbusera^1,2,3, Simona Schiavi⁵, Alessandro Daducci⁵, Francesco La Rosa^6,7,8, Meritxell Bach Cuadra^6,7,8, Robin Sandkühler⁹, Jens Kuhle^2,3, Ludwig Kappos^2,3, Philippe Cattin⁹, and Cristina Granziera^1,2,3
1Translational Imaging in Neurology (ThINk) Basel, Department of Biomedical Engineering, University Hospital Basel and University of Basel, Basel, Switzerland, ²Neurology Clinic and Policlinic, Departments of Medicine, Clinical Research and Biomedical Engineering, University Hospital Basel and University of Basel, Basel, Switzerland, ³Research Center for Clinical Neuroimmunology and Neuroscience (RC2NB) Basel, University Hospital Basel and University of Basel, Basel, Switzerland, ⁴Division of Radiological Physics, Department of Radiology, University Hospital Basel, Basel, Switzerland, ⁵Department of Computer Science, University of Verona, Verona, Italy, ⁶Signal Processing Laboratory (LTS5), Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, ⁷Medical Image Analysis Laboratory, Center for Biomedical Imaging (CIBM), University of Lausanne, Lausanne, Switzerland, ⁸Department of Radiology, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, ⁹Center for medical Image Analysis & Navigation, Department of Biomedical Engineering, University of Basel, Allschwil, Switzerland

We applied an attention-based convolutional neural network to select discriminating diffusion measures derived from mathematical models of  multi-shell diffusion data in the classification of multiple sclerosis lesions. Further, we correlated the selected measures or their combinations with the Expanded Disability Status Scale (EDSS) and the serum level of neurofilament light chain (sNfL). Our results show that the combinations have stronger correlations with EDSS and sNfL than the individual measures. The proposed method might be useful for selecting the microstructural measures most discriminative of focal tissue damage and identifying the combination most related to clinical disability and neuroaxonal damage. 

Parna Eshraghi Boroojeni¹, Yasheng Chen², Paul K. Commean¹, Cihat Eldeniz¹, Udayabhanu Jammalamadaka¹, Gary B. Skolnick³, Kamlesh B. Patel³, and Hongyu An^1
1Mallinckrodt Institute of Radiology, Washington University in St. Louis, Saint louis, MO, United States, ²Department of Neurology, Washington University in St. Louis, Saint louis, MO, United States, ³Division of Plastic and Reconstructive Surgery, Washington University in St. Louis, Saint louis, MO, United States

Computed tomography (CT) scans are commonly used in pediatric patients with head trauma and craniosynostosis to identify skull fractures and sutures, respectively. However, the ionizing radiation associated with the CT scans increases the pediatric patients’ risk for cancer. We developed a deep learning-based method, which consists of two networks focusing on skull and head separately, to generate high-resolution pseudo-CT (pCT) from a radial MR scan. A Dice coefficient of 0.90 ± 0.02 was obtained in the bone.Moreover, a pCT mean absolute error (MAE) of 87.5 ± 4.4 HU was achieved.

Theodor Rumetshofer¹, Francesca Inglese², Jeroen de Bresser², Peter Mannfolk³, Olof Strandberg⁴, Markus Nilsson¹, Itamar Ronen², Andreas Jönsen⁵, Linda Knutsson^6,7, Tom Huizinga⁸, Gerda Steup-Beekman⁸, and Pia Sundgren^1,9,10
1Clinical Science Lund / Diagnostic Radiology, Lund University, Lund, Sweden, ²Department of Radiology, Leiden University Medical Center, Leiden, Netherlands, ³Department of Medical Imaging and Physiology, Skåne University Hospital, Lund, Sweden, ⁴Clinical Memory Research Unit, Department of Clinical Sciences, Malmö, Lund University, Lund, Sweden, ⁵Department of Rheumatology, Lund University, Skåne University Hospital, Lund, Sweden, ⁶Department of Medical Radiation Physics, Lund University, Lund, Sweden, ⁷Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ⁸Department of Rheumatology, Leiden University Medical Center, Leiden, Netherlands, ⁹Department of Clinical Sciences/Centre for Imaging and Function, Skåne University Hospital, Lund, Sweden, ¹⁰Lund University BioImaging Center, Lund University, Lund, Sweden

White Matter Hyperintensities (WMH) are common clinical neuroimaging brain markers. However, WMH in Systemic Lupus Erythematosus (SLE) are non-specific. For this purpose, we developed and unsupervised machine learning approach based on individual WMH distribution to unveil hidden MRI phenotypes. Cluster analysis was performed on a two-site SLE dataset with significant different WMH burden and MRI acquisition protocols. The resulting MRI phenotypes show a clear lesion pattern on distinct WM tracts. This approach reduces the influence of the total WMH burden and MRI acquisition parameters and improves WMH characterization in SLE.