37 posts found
5 Feb. 2024 by Khrystyna Faryna
Organizing a deep learning challenge in medical imaging is a commendable initiative, as it can help drive advancements in the field and bring together researchers and practitioners to tackle important healthcare problems. However, there are several key considerations you should keep in mind to ensure the success and integrity of the challenge.
29 Jan. 2024 by Miriam Groeneveld
The RSE team started the year with exciting updates! We've shifted algorithm job and challenge evaluation execution to Amazon SageMaker Training, removing one-hour time limits for algorithms. Editing and creating display sets is now easier, reducing reliance on our API. Cirrus now supports seamless display of videos, thumbnails, and text alongside medical images. We've improved 3D image display, eliminating the need to navigate after changing orientation. In educational reader studies, we introduced an "Instant Verification" option for faster participation. Users can now set default colors for reader study questions. Check out our blog for more details on these enhancements!
3 Jan. 2024 by Miriam Groeneveld
Our latest blog post brings thrilling enhancements to boost your experience! Challenge editors can open submission logs for direct participant access. The revamped pathology viewer now supports overlays with LUT support. Algorithm pages showcase Challenge performance insights, and job limits extend beyond one hour. To ensure responsible use, algorithm editors now face restrictions. Discover the efficient 3-point angle annotation for a more user-friendly approach. Dive into the blog for a closer look at these game-changing updates!
This cycle we continued our work on the client-side pathology viewer. We also implemented a first version of a Intensity-over-time chart feature. For challenge organizers, it is now possible to send messages directly to their participants! Check out the blog for these features and more!
In this cycle, we worked on improving the performance of our reader studies, enabled setting the viewing depth of annotations in 3D images, and continued our work on the client-side pathology viewer. One of our team members attended Miccai 2023 to meet with challenge organizers and receive feedback. Read the blog for more details!
This cycle, the RSE team dedicated significant efforts to support various MICCAI 2023 challenges. Additionally, we enhanced the Cirrus viewer's usability, enabling seamless switching between different hanging protocols within the viewer. Furthermore, we addressed a prior issue, reinstating the functionality to switch between the polygon and mask editor. In the context of mask creation, we introduced a new feature allowing you to conveniently hide the annotation you're editing for accuracy verification using a straightforward keybinding. Explore our latest blog post for more details!
This cycle the RSE team has started working on a new, experimental feature: Implementing a client-side viewer, for now specifically for pathology images. This should provide a faster, more responsive experience when viewing these images. Furthermore, the GitHub repository integration has been improved and bugs squashed. Challenge organizers can now combine results from different phases to create an overall leaderboard. Finally, we have added support for viewing overlapping semantic segmentations.
30 May 2023 by Miriam Groeneveld
This cycle, the RSE team has made improvements for pathology. It is now possible to upload DICOM-WSI files, they will be converted to tiff. Segmentations as tiff files are now supported and will be validated according to the segmentation interface. The annotation statistics plugin has had an overall make-over, with progress bar and caching improving general usability. Other features include a 3D brush for creating and editing masks, and improvements to the admin page for challenge organizers.
As the first rays of spring sunshine warm our faces and the flowers begin to bloom, we are excited to bring you the latest news on the grand challenge ahead.
Cycle report of the research-software engineers of March 2023, including: more statistics on Grand Challenge, updated algorithm-job permissions on Grand Challenge, upgrade the CIRRUS backbone MeVisLab version to v3.6, added a new reader-study workflow for accepting and rejecting findings in both CIRRUS and Grand Challenge.
15 Feb. 2023 by Miriam Groeneveld
This cycle work was done to make algorithm result viewing faster by reusing Cirrus sessions. We also made it possible to select existing images for trying out an Algorithm, creating a Display Set in Reader Studies and creating Archive Items.
In this year's first release, we added the option to reuse existing images to run an algorithm on or create an archive item or display set. Previously this was only possible with the python client. We extended our annotation types with the ellipse annotation, to be used in Reader studies or Algorithms. Under the hood, we are working on improving the viewing of pathology images, making them faster and more error-proof. To improve the reliability and ease of use of the GC-API client, we have added retries to certain requests in the client. This should reduce the number of errors that need to be handled by the user.
29 Nov. 2022 by Miriam Groeneveld
A new tool and some bug fixes for the viewer in this cycle, while most effort has gone to the development of Grand Challenge Connect, presented at RSNA! A lot of interesting and exciting developments, so make sure to check out the dedicated blogpost.
23 Nov. 2022 by James Meakin
Radboudumc will be at RSNA to present the latest developments of Grand Challenge, building on Amazon HealthLake Imaging. Visit us for a hands-on demo at the AWS Booth (6758) between 27-30 November 2022 at McCormick Place in Chicago.
Hi all! With this newsletter we want to update you on all progress made for Grand Challenge: upcoming Challenges, interesting new Algorithms and Blogposts, leaderboards of finished Challenges and new features of Grand Challenge.
The MICCAI Special Interest Group (SIG) for Challenges is hosting a webinar on How to run a challenge? on Oct 25th, 2022 at 1 pm GMT. Bram van Ginneken, Kiran Vaidhya Venkadesh, and Anindo Saha will present how to use Grand Challenge for organizing high-profile challenges. Jake Albrecht from Sage will present tips for challenge organizers on how to define a successful community challenge. Join us! Free registration is available at https://bit.ly/BIAS1025
Hi all! With this newsletter, we want to update you on all progress made for Grand Challenge: upcoming Challenges, interesting new Algorithms, Blogposts, leaderboards of finished Challenges, and new features of Grand Challenge.
AIROGS Challenge Report: AI models can be used for glaucoma screening, but do they know when they cannot?
The rationale behind the Artificial Intelligence for RObust Glaucoma Screening (AIROGS) Challenge, an overview of its results and our experiences with Grand Challenge's new Type 2 challenges.
Hi all! With this newsletter we want to update you on all progress made for Grand Challenge: new features, interesting new blogposts and algorithms, upcoming challenges and the leaderboards of finished challenges.
Can AI predict breast cancer recurrence via automated quantification of tumor-infiltrating lymphocytes?
TIGER is the first challenge on fully automated assessment of tumor-infiltrating lymphocytes (TILs) in hematoxylin and eosin (H&E) breast cancer histopathology slides. With TIGER, we released a training set of 390 whole-slide images and a total award of $13,000 in AWS Credits, which will be awarded to the winning teams. Together with its participants, we aim to find the best AI-based solutions for automating the assessment of the TILs and produce a “TILs score” that can predict the recurrence of breast cancer.
Last week, we opened STOIC2021: A COVID-19 AI challenge with 10,000 CT scans. Together with its participants, we aim to find the best solution for predicting who will develop severe COVID-19 from a chest CT scan. We will make the final solution easily accessible for everyone. In total, $20,000 in AWS Credits will be awarded to the winning teams.
How to build a global, scalable, low-latency, and secure machine learning medical imaging analysis platform on AWS
It is hard to imagine the future for medical imaging without machine learning (ML) as its central innovation engine. Countless researchers, developers, start-ups, and larger enterprises are engaged in building, training, and deploying machine learning solutions for medical imaging that are posed to transform today’s medical workflows and the future value of imaging in diagnosis and treatment.