View Presentation Video

Keywords: Transferable skills: Reproducible research

The talk will present the importance of reproducibility in scientific research and how it has become a crucial part of scientific practices. The talk will also discuss the role of open data and source methods in promoting reproducibility and creating a more accessible and equitable scientific landscape. Finally, the talk will explain how implementing reproducible practices can improve the scientific output of labs in terms of both quality and efficiency.

View Presentation Video

View Presentation Video

Mathieu Boudreau^1,2, Agah Karakuzu¹, Julien Cohen-Adad^1,3, Madeline Carr^4,5, Mariya Doneva⁶, Seraina A. Dual^7,8, Daniel B. Ennis⁹, Alex Ensworth^10,11, Alexandru Foias¹, Véronique Fortier^10,12, Guillaume Gilbert¹³, Matthew Grech-Sollars^14,15, Lois Holloway^4,5, Siyuan Hu¹⁶, Oscar Jalnefjord^17,18, Peter Koken⁶, Anastasia Kolokotronis^10,19, Simran Kukran²⁰, Nam Lee²¹, Ives R. Levesque¹⁰, Dan Ma¹⁶, Burkhard Maedler²², Nyasha Maforo²³, Kévin Moulin^9,24, Jamie Near²⁵, Robba Rai^4,5, Ben Statton²⁶, Christian Stehning²², Chenyang Wang²⁷, Kilian Weiss²², Niloufar Zakariaei²⁸, Shuo Zhang²², and Nikola Stikov^1,2
1NeuroPoly, Polytechnique Montreal, Montreal, QC, Canada, ²Montreal Heart Institute, Montreal, QC, Canada, ³Unité de Neuroimagerie Fonctionnelle (UNF), Centre de recherche de l’Institut Universitaire de Gériatrie de Montréal (CRIUGM), Montreal, QC, Canada, ⁴Medical Physics, Ingham Institute for Applied Medical Research, Liverpool, Australia, ⁵Department of Medical Physics, Liverpool and Macarthur Cancer Therapy Centres, Liverpool, Australia, ⁶Philips Research, Hamburg, Germany, ⁷Stanford University, Stanford, CA, United States, ⁸Department of Biomedical Engineering and Health Systems, KTH Royal Institute of Technology, Stockholm, Sweden, ⁹Department of Radiology, Stanford University, Stanford, CA, United States, ¹⁰Medical Physics Unit, McGill University, Montreal, QC, Canada, ¹¹University of British Columbia, Vancouver, BC, Canada, ¹²Medical Imaging, McGill University Health Centre, Montreal, QC, Canada, ¹³MR Clinical Science, Philips Canada, Mississauga, ON, Canada, ¹⁴Department of Computer Science, University College London, London, United Kingdom, ¹⁵Lysholm Department of Neuroradiology, National Hospital for Neurology and Neurosurgery, University College London Hospitals NHS Foundation Trust, London, United Kingdom, ¹⁶Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, United States, ¹⁷Department of Medical Radiation Sciences, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden, ¹⁸Department of Medical Physics and Biomedical Engineering, Sahlgrenska University Hospital, Gothenburg, Sweden, ¹⁹Hôpital Maisonneuve-Rosemont, Montreal, QC, Canada, ²⁰Department of Surgery & Cancer, Imperial College London, London, United Kingdom, ²¹Department of Biomedical Engineering, Viterbi School of Engineering, University of Southern California, Los Angeles, CA, United States, ²²Philips GmbH Market DACH, Hamburg, Germany, ²³Philips Healthcare Germany, Hamburg, Germany, ²⁴CREATIS Laboratory, Univ. Lyon, UJM-Saint-Etienne, INSA, CNRS UMR 5520, INSERM, Lyon, France, ²⁵Centre d'Imagerie Cérébrale, Douglas Mental Health University Institute, Montreal, QC, Canada, ²⁶MRC, London Institute of Medical Sciences, Imperial College London., London, United Kingdom, ²⁷Department of Radiation Oncology - CNS Service, The University of Texas MD Anderson Cancer Center, Houston, TX, United States, ²⁸Department of Radiation Oncology, Henry Ford Cancer Institute, Detroit, MI, United States

Keywords: Quantitative Imaging, Relaxometry, Reproducibility, challenge

A collaborative reproducibility challenge was launched to explore if an imaging protocol independently-implemented at multiple centers can reliably measure T₁ using inversion recovery in a standardized quantitative MRI phantom (ISMRM/NIST). A total of 19 submissions were accepted, totalling 41 phantom T₁ mapping datasets. Errors relative to the temperature-corrected reference T₁ values were under 10% for the range of values expected in the human brain in vivo. All submitted phantom data, code, pipelines, and scripts were shared on open platforms. 

View Presentation Video

Nam G Lee¹, Grzegorz Bauman², Oliver Bieri², and Krishna S Nayak^3
1Biomedical Engineering, University of Southern California, Los Angeles, CA, United States, ²Department of Radiology, University of Basel Hospital, Basel, Switzerland, ³Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States

Keywords: Lung, Low-Field MRI

The reproducibility of scientific reports is crucial to advancing human knowledge. This abstract is a response to the 2023 ISMRM Challenge “Repeat it With Me: Reproducibility Team Challenge”. We reproduce the bSTAR sequence, a very short-TR, 3D half-radial dual-echo bSSFP sequence, providing banding-artifact free images within a large FOV. bSTAR imaging is attractive for various applications at low field and especially attractive for lung parenchyma imaging due to the prolonged T2’. We have successfully reproduced the bSTAR method, and figures with comparable image quality compared to published literature. We provide an open-source implementation using Pulseq and BART.

View Presentation Video

Shohei Fujita^1,2, Borjan Gagoski^3,4, Ken-Pin Hwang⁵, Marcel Warntjes^6,7, Kazumasa Yokoyama^8,9, Issei Fukunaga¹, Wataru Uchida¹, Yuya Saito¹, Rina Tachibana¹, Tomoya Muroi¹, Toshiya Akatsu¹, Akihiro Kasahara², Ryo Sato², Tsuyoshi Ueyama², Christina Andica¹, Akifumi Hagiwara¹, Koji Kamagata¹, Shiori Amemiya², Hidemasa Takao², Nobutaka Hattori⁹, Osamu Abe², and Shigeki Aoki^1
1Department of Radiology, Juntendo University, Tokyo, Japan, ²Department of Radiology, The University of Tokyo, Tokyo, Japan, ³Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children's Hospital, Boston, MA, United States, ⁴Department of Radiology, Harvard Medical School, Boston, MA, United States, ⁵Department of Imaging Physics, MD Anderson Cancer Center, Houston, TX, United States, ⁶SyntheticMR, Linköping, Sweden, ⁷Center for Medical Imaging Science and Visualization (CMIV), Linköping University, Linköping, Sweden, ⁸Tousei center for neurological diseases, Shizuoka, Japan, ⁹Department of Neurology, Juntendo University, Tokyo, Japan

Keywords: Quantitative Imaging, Precision & Accuracy, Cross-vendor

Multiparametric techniques compatible with multiple vendors to facilitate the pooling of data among different sites and vendors are desired. Here, we developed a vendor-standardized whole-brain multiparametric mapping scheme based on 3D-QALAS. Intra-scanner repeatability and inter-vendor reproducibility were evaluated on test-retest session data on five different 3T systems from four MRI vendors (GE, Philips, Siemens, and Canon). T1 and T2 relaxation times and proton density values derived from 3D-QALAS showed coefficient of variations of <4.0% across scanners from different vendors. Finally, we performed an inter-vendor validation on multiple sclerosis patients to assess the feasibility of the scheme in real-world clinical settings.

View Presentation Video

Jessica E. Ringshaw^1,2,3, Layla E. Bradford^1,2, Marlie Miles^1,2, Carly Bennallick³, Niall J. Bourke³, Emil Ljungberg^3,4, Sean C.L. Deoni⁵, Steven C.R. Williams³, and Kirsten A. Donald^1,2
1Department of Paediatrics and Child Health, University of Cape Town, Cape Town, South Africa, ²Neuroscience Institute, University of Cape Town, Cape Town, South Africa, ³Department of Neuroimaging, King's College London, London, United Kingdom, ⁴Department of Medical Radiation Physics, Lund University, Lund, Sweden, ⁵Maternal, Newborn, and Child Health (MNCH) Discovery and Tools (D&T), Bill and Melinda Gates Foundation, Seattle, WA, United States

Keywords: Neuro, Low-Field MRI, Training, Capacity Development, Low- and Middle Income Countries

In a multi-site global collaboration (UNITY; Ultra-Low Field Neuroimaging In The Young) implementing the novel Hyperfine 64mT low-field MRI into research on relevant health priorities in low- and middle-income countries (LMICs), capacity development has been identified as a key objective. Following an initial training wave in foundational sites including South Africa, adapted workshops have been conducted in new consortium countries including Malawi, Ethiopia, and Ghana. This has informed the development of a strategic framework for context-specific training aimed at promoting the increased success of neuroimaging in research and clinical practice, and the sustainability of MRI technology in under-resourced settings.

View Presentation Video

Harsh Sinha¹ and Pradeep Reddy Raamana^2
1Intelligent Systems Program, University of Pittsburgh, Pittsburgh, PA, United States, ²Intelligent Systems Program, Department of Biomedical Informatics, Department of Radiology, University of Pittsburgh, Pittsburgh, PA, United States

Keywords: Data Acquisition, Data Processing, Protocol Compliance, Quality Control, Quality Assurance

Acquiring MR imaging data by multi-site consortia requires careful monitoring of MR physics protocols. Conventionally, protocol compliance has been an ad-hoc and manual process that is error-prone. Hence, it is often overlooked for the lack of realization that parameters are routinely improvised locally at different sites. Such variation and inconsistencies across acquisition protocols can reduce SNR, & statistical power and, in the worst case, may invalidate the results altogether. We present an open-source tool, called mrQA, to assess protocol compliance and demonstrate the lack of it by analyzing over 20 large open datasets. 

View Presentation Video

Thomas Kluge¹ and Christoph Forman^1
1Siemens Healthcare GmbH, Erlangen, Germany

Keywords: Data Acquisition, Pulse Sequence Design, Open Innovation; Reproducible Research

Open innovation and reproducible research have become increasingly important in recent years. Furthermore, software version dependent application development introduces significant efforts to developers and researchers in maintaining different versions of applications for a variety of scanner software releases. In this work, we propose a software-version and programming language independent interface for data acquisition on Siemens Healthcare MAGNETOM scanners. This is achieved by generic interfaces and streaming protocols. A wide range of development and deployment opportunities is enabled with a web-API. The use of the interface is exemplarily shown with a prototype FLASH sequence and demonstrated with a phantom scan.

View Presentation Video

Xiaoqing Wang^1,2, Berkin Bilgic^1,2, Daniel Gallichan³, Kwok-Shing Chan⁴, and José P. Marques^4
1Massachusetts General Hospital, Athinoula A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States, ²Department of Radiology, Harvard Medical School, Boston, MA, United States, ³Cardiff University Brain Research Imaging Centre, Cardiff, United Kingdom, ⁴Donders Institute for Brain, Cognition and Behavior, Radboud University, Nijmegen, Netherlands

Keywords: Pulse Sequence Design, Quantitative Susceptibility mapping, open-source, pulseq

We propose an open-source multi-echo GRE sequence and reconstruction algorithms for rapid acquisition and harmonization of QSM. This is implemented using Pulseq, where different sampling patterns are used across echoes to provide complementary information and facilitate the proposed MOdel BAsed (MOBA) reconstruction. MOBA estimates M0, R2* and frequency maps using nonlinear optimization and is implemented in BART to facilitate dissemination. B0 correction is performed using 1D navigators acquired during the crusher gradients in each repetition. Motion navigators inserted in the sequence permit robust estimation of motion parameters with time frames of ~13seconds, and will lend themselves to retrospective motion correction. 

View Presentation Video

Divya Ramakrishnan¹, Leon Jekel², Matthew Sala¹, Manpreet Kaur³, Anastasia Janas¹, Gabriel Cassinelli Petersen⁴, Khaled Bousabarah⁵, MingDe Lin⁶, Sara Merkaj⁷, Marc von Reppert⁸, and Mariam Aboian^1
1Department of Radiology, Yale School of Medicine, New Haven, CT, United States, ²University of Essen, Essen, Germany, ³Ludwig Maximilians Universität (LMU), Munich, Germany, ⁴University of Goettingen, Goettingen, Germany, ⁵Visage Imaging, Dusseldorf, Germany, ⁶Department of Radiology and Biomedical Imaging, Yale University, New Haven, CT, United States, ⁷University of Ulm, Ulm, Germany, ⁸University of Leipzig, Leipzig, Germany

Keywords: Tumors, Machine Learning/Artificial Intelligence, Brain Metastasis

While there are many machine learning (ML) algorithms for brain metastasis (BM) detection and segmentation, very few have been validated on external datasets. There is a critical need for open access BM datasets for development and validation of more robust algorithms. Here, we present the Yale Brain Metastasis database of 290 patients with annotated segmentations of BM on T1 post-gadolinium and associated survival information. A subset of 228 patients have FLAIR segmentations, clinical features, and qualitative imaging features. Open access of this database will greatly aid in the development and validation of new AI algorithms for BM detection and segmentation.

View Presentation Video

Mengye Lyu¹, Lifeng Mei¹, Sixing Liu¹, Shoujin Huang¹, Yi Li¹, Kexin Yang¹, Yilong Liu², and Ed X. Wu^3,4
1College of Health Science and Environmental Engineering, Shenzhen Technology University, Shenzhen, China, ²Guangdong-Hongkong-Macau Institute of CNS Regeneration, Key Laboratory of CNS Regeneration (Ministry of Education), Jinan University, Guangzhou, China, ³Laboratory of Biomedical Imaging and Signal Processing, the University of Hong Kong, Hong Kong, China, ⁴Department of Electrical and Electronic Engineering, the University of Hong Kong, Hong Kong, China

Keywords: Machine Learning/Artificial Intelligence, Low-Field MRI, Open dataset

We release a new raw k-space dataset M4Raw acquired by the low-field MRI. Currently, it contains multi-channel brain data of 180 subjects each with 18 slices x 3 contrasts (T1w, T2w, and FLAIR). Moreover, each contrast consists of two or three repetitions (a.k.a. NEXs), leading to more than 25k trainable slices in total, which can be used in various ways by the low-field MRI community. It can be accessed via http://github.com/mylyu/M4Raw.

View Presentation Video

Naoto Fujita¹, Suguru Yokosawa², Toru Shirai², and Yasuhiko Terada^1
1Graduate School of Science and Technology, University of Tsukuba, Tsukuba, Japan, ²FUJIFILM Healthcare Corporation, Tokyo, Japan

Keywords: Image Reconstruction, Machine Learning/Artificial Intelligence, Deep Learning Reconstrcution

Deep neural networks (DNNs) for MRI reconstruction often require large datasets for training, but in clinical settings, the domains of datasets are diverse, and the degree to which deep neural networks are the robustness of DNNs to domain differences between training and testing datasets has been an open question. Here, we evaluated the robustness of four open-source multicoil networks to differences in the domain. We found that model-based networks exhibit higher robustness than data-driven networks and that robustness varies across network architectures, even within model-based networks. Our results provide insight into what network architectures are effective for generalization performance.