NEWS!

  • 06.10.2023: The GRAIL 2023 virtual event has concluded! Thanks to all keynote speakers, to our presenting paper authors and to the attendees for making this an exciting and smooth event.
    We also announced the Best Paper Awardee, congratulations!
  • 11.09.2023: The registration form is ready. Please note that registration is mandatory to participate in the workshop.
  • 18.08.2023: The review period has been finalized! We accepted 9 papers to GRAIL, the program will be updated soon.
  • 20.07.2023: Paper acceptance is closed and assignment to reviewers has finished. The new review deadline is August 4th.
  • 04.07.2023: Date fixed: 4th Oct. 2023! Keynote speakers announced.
  • 20.05.2023: The GRAIL Website is up and running! Please note Important Dates.

SCOPE

GRAIL 2023 is the fifth international workshop on GRaphs in biomedicAl Image anaLysis, organised as a satellite event of MICCAI 2023 in Vancouver, Canada.

Graphs are powerful mathematical structures that provide a flexible and scalable framework to model unstructured information, objects and their interactions in a readily interpretable fashion. Graph theory provides a solid mathematical foundation for models and algorithms, such as spectral analysis, dimensionality reduction, and network analysis. Since 2017, geometric deep learning has married the field of graph signal processing with the flexibility and rapid advancements of deep neural architectures. Since then, applications of graph neural networks (GNNs) in medicine have been steadily increasing, ranging from medical imaging and shape understanding, brain connectomics, population models and patient multi-omics to discovery and design of novel drugs and therapeutics. In the leading computer vision and ML conferences (CVPR, ICLR, NeurIPS), GNNs have been among the hottest topics in 2022 and will likely increase in interest in 2023. With this workshop, we aim to provide a platform for understanding and application of graph-based models as versatile and powerful tools in biomedical image analysis and beyond. Our goal is to bring together scientists that develop graph-based models, and encourage their application to difficult clinical problems within a variety of biomedical data contexts. GRAIL 2022 featured keynote presentations from leading researchers in the graph community, and paper submissions on a wide range of topics, from brain connectomics and whole-slide image analytics, over semantic priors from biomedical knowledge graphs, explainable AI approaches for GNNs, to GNN-based genome alignment and their incorporation into multi-omics patient representations. For 2023, we expect a similarly exciting lineup of keynotes and paper submission. As an additional highlight, we will invite a tutorial-style keynote for hands-on demonstration of acceleration recipes in graph analytics and GNN frameworks like Deep Graph Library (DGL) or Pytorch-Geometric (PyG).

In more detail, the scope of methodology topics includes but is not limited to:

  • Deep/machine learning on graphs with regular and irregular structures
  • Probabilistic graphical models for biomedical data analysis
  • Signal processing on graphs for biomedical image analysis, including non-learning based approaches
  • Explainable AI (XAI) methods in geometric deep learning
  • Big data analysis with graphs
  • Graphs for small data sets
  • Semantic graph research in medicine: Scene graphs and knowledge graphs
  • Modeling and applications of graph symmetry and equivariance
  • Graph generative models
  • Combination of graphs with other SOTA domains (e.g. self-supervised learning, federated learning)

Applications covered include but are not limited to:

  • Image segmentation, registration, classification
  • Graph representations in pathology imaging and whole-slide image analysis
  • Graph-based approaches in intra-operative surgical support
  • Graph-based shape modeling and dimensionality reduction
  • Graphs for large scale patient population analyses
  • Combining multimodal/multi-omics data through graph structures
  • Graph analysis of brain networks and connectomics
  • REGISTRATION

    Please register via the form! Registration is mandatory to participate in the workshop. If you have any questions, please contact at grail.miccai@gmail.com.

    Preliminary Program (Timezone: Central European Time (CET))

    3 - 3:40 p.m. Keynote 1:
    Dr. Matthias Fey
    PyTorch Geometric - Practical Realization of Graph Neural Networks.
    3:40 -3:45 p.m. Break
    3:45 – 4:30 p.m. Oral session 1:
    • 3:45 - 4 p.m. :
      SCOPE: Structural Continuity Preservation for Retinal Vessel Segmentation.
      Azade Farshad
    • 4 - 4:15 p.m. :
      Multi-level Graph Representations of Melanoma Whole Slide Images for Identifying Immune Subgroups.
      Lucy Godson
    • 4:15 - 4:30 p.m. :
      Extended Graph Assessment Metrics for Regression and Weighted Graphs.
      Tamara T. Mueller
    4:30 -4:40 p.m. Break
    4:40 - 5:20 p.m. Keynote 2:
    Dr. Dorina Thanou
    Informed machine learning for biology and medicine: A graph representation perspective.
    5:20 - 5:25 p.m. Break
    5:25 – 6:10 p.m. Oral session 2:
    • 5:25 - 5:40 p.m. :
      Multi-Head Graph Convolutional Network for Structural Connectome Classification.
      Anees Kazi
    • 5:40 - 5:55 p.m. :
      Self Supervised Multi-View Graph Representation Learning in Digital Pathology.
      Vishwesh Ramanathan
    • 5:55 - 6:10 p.m. :
      Prior-RadGraphFormer: A Prior-Knowledge-Enhanced Transformer for Generating Radiology Graphs from X-Rays.
      Yiheng Xiong
    6:10 - 6:15 p.m. Break
    6:15 - 6:55 p.m. Keynote 3:
    Prof. Dr. Ranjie Liao
    Graph Neural Networks Meet Transformers: Some Insights in Representation Learning and Generative Modeling.
    6:55 - 7 p.m. Break
    7 – 7:45 p.m. Oral session 3:
    • 7 - 7:15 p.m. :
      Tertiary Lymphoid Structures Generation through Graph-based Diffusion.
      Manuel Madeira
    • 7:15 - 7:30 p.m.:
      A Comparative Study of Population-Graph Construction Methods and Graph Neural Networks for Brain Age Regression.
      Kyriaki-Margarita Bintsi
    • 7:30 - 7:45 p.m.:
      Heterogeneous Graphs Model Spatial Relationship between Biological Entities for Breast Cancer Diagnosis.
      Akhila Krishna K
    7:45 - 8:00 p.m. Closing event
    Dr. rer. nat. Seyed-Ahmad Ahmadi

    Keynote Speakers

    Dr. Dorina Thanou
    Ecole Polytechnique Federal Lausanne (EPFL)

    Senior researcher and lecturer at EPFL, leading the Intelligent Systems for Medicine and Health research pillar, under the Center for Intelligent Systems. Her expertise includes spectral graph theory, graph-based signal processing and geometric deep learning for data representation and analysis, with a particular focus on the design of interpretable and robust models for healthcare.

    Dr. Matthias Fey
    Kumo.AI

    Matthias Fey is a founding engineer at Kumo.ai where we works on making state-of-the-art Graph Neural Network solutions readily available to large-scale data warehouses. Previously, he obtained his PhD at the TU Dortmund University, Germany. His main area of research lies in the development of new deep learning methods that can be directly applied to unstructured data such as graphs, point clouds and manifolds. Furthermore, he is the creator of the PyG (PyTorch Geometric) library, which aims to bundle many of the proposed methods in this area to make research more accessible, omparable and reproducible, and is a core member on the Open Graph Benchmark (OGB) team.

    Prof. Dr. Renjie Liao
    University of British Columbia (UBC)

    Assistant Professor at ECE, UBC, Vancouver. His research focuses on probabilistic and geometric deep learning and their intersection with vision, self-driving and healthcare. He has worked as a Visiting Faculty Researcher at Google Brain and as a Senior Research Scientist at Uber ATG.

    Submit a paper

    Submissions should be done through the GRAIL CMT portal

    This year we will accept papers in the following format:

    • Complete papers: Papers describing original research with MICCAI length (max 8 pages text + max 2 pages references, supplemental material allowed). Papers are eligible for the "Best Paper Award" and will be published in the GRAIL Open Proceedings and Lecture Notes in Computer Sciences (LNCS) series. Submissions should be anonymous, and formatted following the LNCS Style. Style. All complete papers will be peer-reviewed by 3 members of the program committee, with a double-blinded review process. The selection of the papers will be based on the significance of results, technical merit, relevance and clarity of presentation.

    GRAIL Best Paper Award

    The GRAIL 2023 Best Paper Award is awarded to:
    Vishwesh Ramanathan and Anne Martel
    for their paper:
    Self Supervised Multi-View Graph Representation Learning in Digital Pathology.
    On behalf of the organizing committee, we congratulate the authors for their excellent paper contribution to GRAIL 2023.

    Important dates

    We have registered with the CMT system and will publish the conference results as Springer Proceedings. Please use the LNCS templates. We accept full papers (max 8 pages text + max 2 pages references). Submit your paper here.
    07 July 2023 Abstract submission deadline (Intention to submit), 23:59 pm, PST
    14 July 2023 Submission deadline, 23:59 pm, PST
    04 August 2023 Reviews due, 23:59 pm, PST
    11 August 2023 (preliminary) Notification of acceptance, 23:59 pm, PST
    18 August 2023 (preliminary) Camera ready paper versions due, 23:59 pm, PST

    Organising Committee


    Past Events


    Sponsor

    t.b.a.