Highlights from MICCAI 2021 with Kitware

Kitware is deeply engaged in the medical imaging research and clinical communities, as demonstrated by our many contributions to the 24th annual international conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 2021).

1. MONAI Bootcamp: Stephen Aylward, Ph.D., Senior Director of Strategic Initiatives at Kitware, gave the introductory talk “What is MONAI? Overview and Related Projects” to the 2nd annual MONAI Bootcamp. MONAI is an open source platform for medical image deep learning research and development. The 2021 MONAI Bootcamp hosted over 750 participants from 58 countries and over 530 organizations. The interactive tutorials presented during this 3-day MICCAI-endorsed event are now freely available.

2. New Project MONAI releases: MONAI v0.7, MONAILabel v0.2, MONAIDeploy v0.1, and MONAIStream were announced at MICCAI 2021. MONAI, the core toolkit of Project MONAI, has been updated to provide 10x improvement in training speed as well as a multitude of new network architectures, metrics, optimizers, and more. The new release of MONAILabel now provides AI-assisted annotation support for 3D Slicer and OHIF. The initial release of MONAIDeploy allows a trained MONAI model to be packaged as a re-deployable application in as little as 20 minutes.  MONAIStream is a new initiative for Project MONAI and is going to support real-time deep learning inference for image-guided interventions and real-time medical image analysis, e.g., for endoscopy and ultrasound applications. To provide your input on the design and capabilities of MONAIStream, please contribute to this survey: https://forms.office.com/r/MjBSmDikLK.

3. ITKPOCUS Release Presentation and Demonstration: Brad Moore, Ph.D., Senior Research and Development Engineer at Kitware, announced the initial release of ITKPOCUS and gave a live demonstration of it during the Advances in Simplifying Medical UltraSound workshop at MICCAI.  ITKPOCUS “does the dirty work” of streaming and standardizing point-of-care ultrasound images from Butterfly iQ, Sonivate, Clarius, Interson, and SonoQue ultrasound probes, so that related AI research projects can focus on algorithm innovation.

4. Advances in Simplifying Medical UltraSound (ASMUS) Workshop: Stephen also co-chaired this full-day workshop with Alison Noble, Ph.D. from Oxford. This workshop had over 100 participants and included 22 paper presentations, 2 keynote presentations, and 9 live demonstrations of cutting-edge ultrasound research applications. The workshop’s papers will be published as a book later this year.  

5. ITK v5.3rc1: Kitware is proud to announce the initial release candidate for ITK v5.3.  This ITK update features improved support for MONAI as well as a wide variety of performance improvements. For example, deformable image registration using b-spline sampling is improved by ~30%. Multi-threading parallelism in Python is improved by 5-10% by adding Threading Building Blocks to the cross-platform binaries. Additionally, compression time with zlib, used by common medical imaging file formats like NIFTI, NRRD, or MetaImage, is dramatically reduced through migration to zlib-ng.

6. MICCAI Special Interest Group on Medical Ultrasound (SIG-MUS): Stephen is a founding organizer and treasurer of this new MICCAI special interest group. This SIG provides an open forum for identifying, investigating, and solving key challenges and exploring future research directions for the next breakthroughs in medical ultrasound.

7. NIH Cancer Imaging Data Repositories for Biomedical Data Science Research Workshop: Stephen gave the invited talk, “MONAI: Open Science for Medical Imaging AI”, that explained (1) the motivation for and capabilities of MONAI; (2) how the wealth of clinical data on NCI’s The Cancer Imaging Archive (TCIA) and Imaging Data Commons (IDC) repositories could be easily integrated into MONAI training workflows; and (3) how itkWidgets can be used to interactively view the TCIA data in Jupyter notebooks to speed AI research projects:

TCIA: https://www.cancerimagingarchive.net/

IDC: https://datacommons.cancer.gov/repository/imaging-data-commons

itkWidgets: https://github.com/InsightSoftwareConsortium/itkwidgets

8. TOPGRAD Workshop: Stephen gave the invited talk, “How to win friends and influence people using Powerpoint” at the TOPGRAD workshop.  This workshop helps current and recently graduated graduate students accelerate their careers by giving them crash courses on writing proposals, giving presentations, working with mentors, and maintaining mental health.
9. MICCAI Young Scientist Publication Impact Award: Stephen served on this award committee with Marc Niethammer, Ph.D. (UNC) and William “Sandy” Wells, Ph.D. (Harvard) for the eighth year. This award is given to the paper that had a significant impact on the MICCAI community and that was presented by a young scientist at a MICCAI conference within the past five years. The winner of this award will be announced at the MICCAI awards ceremony on September 30, 2021.