autoPET II - Automated Lesion Segmentation in PET/CT - Domain Generalization¶
📰 News¶
April 1st:
The new autoPET-III
challenge is live.
October 12th:
Dear autoPET enthusiasts,
we are pleased to announce that we are going to host an online challenge
session on the
9th Nov, 2 PM - 3:30 PM (UTC+0, Time conversion)
Please register for the session
here (to retrieve the Zoom link).
We will announce the winners of the challenge in this session and ask
the top teams to give a short presentation about their solution. We will
reach out to the top teams via separate mails in the next days.
September 19th:
Dear autoPET participants,
in order to fix any bugs that may appear in the final data, we strongly
recommend that you submit the first final container today. Some volumes
are small which might lead to problems when using skip-connections
without padding.
September 14th:
Dear autoPET participants,
the challenge deadline is approaching and we would like to encourage you
to submit your algorithm also to the final test set. Don‘t be
discouraged by the preliminary results as these are not
representative. For everyone who submits now, the chances are good as
not many teams have provided a final submission so far.
August 22nd:
The final test phase is now open for submissions. Please note
that in total only 2 submissions per team are allowed to the final
test phase. We encourage everyone to first test their Docker containers
via the preliminary test phase which will remain open till the end of
the challenge. The final leaderboard is private and the final ranking
will be announced at the MICCAI 2023 challenge session.
March 8th:
Website is online and challenge opens.
🎬 Introduction
Positron Emission Tomography / Computed Tomography (PET/CT) is an integral part of the diagnostic workup for various malignant solid tumor entities. Due to its wide applicability, Fluorodeoxyglucose (FDG) is the most widely used PET tracer in an oncological setting reflecting glucose consumption of tissues, e.g. typically increased glucose consumption of tumor lesions.
As part of the clinical routine analysis, PET/CT is mostly analyzed in a qualitative way by experienced medical imaging experts. Additional quantitative evaluation of PET information would potentially allow for more precise and individualized diagnostic decisions.A crucial initial processing step for quantitative PET/CT analysis is segmentation of tumor lesions enabling accurate feature extraction, tumor characterization, oncologic staging and image-based therapy response assessment. Manual lesion segmentation is however associated with enormous effort and cost and is thus infeasible in clinical routine. Automation of this task is thus necessary for widespread clinical implementation of comprehensive PET image analysis.Recent progress in automated PET/CT lesion segmentation using deep learning methods has demonstrated the principle feasibility of this task. However, despite these recent advances tumor lesion detection and segmentation in whole-body PET/CT is still a challenging task. The specific difficulty of lesion segmentation in FDG-PET lies in the fact that not only tumor lesions but also healthy organs (e.g. the brain) can have significant FDG uptake; avoiding false positive segmentations can thus be difficult. One bottleneck for progress in automated PET lesion segmentation is the limited availability of training data that would allow for algorithm development and optimization.
To promote research on machine learning-based automated tumor lesion segmentation on whole-body FDG-PET/CT data we host the autoPET-II challenge - as a successor of the autoPET challenge - and provide a large, publicly available training data set on TCIA:
AutoPET-II is hosted at the MICCAI 2023:
Challenge Aims
I.) Accurate detection and segmentation of FDG-avid tumor lesions in
whole body FDG-PET/CT. The specific challenge in automated segmentation
of FDG-avid lesions in PET is to avoid false-positive segmentation of
anatomical structures that have physiologically high FDG-uptake (e.g.
brain, kidney, heart, etc…) while capturing all tumor lesions.
II.) Robust behavior of the algorihtms in term of moderate changes in
acquisition protocol or acquisition site. This will be reflected by the
test data which will be drawn partly from the same distribution as the
training data and partly from a different hospital with a similar, but
slightly different acquisition setup.
Figure: Example case of fused FDG-PET/CT whole-body data. The right image shows the manually segmented malignant lesions.
📋 Task¶
Automatic tumor lesion segmentation in whole-body FDG-PET/CT on large-scale database of 1014 studies of 900 patients (training database) acquired on a single site for accurate and fast lesion segmentation and to avoid false positives (brain, bladder, etc.).
The challenge scope is:
- to focus on robustness across different environments and
- to relax the restriction regarding the use of external and additional data for algorithm development.
Testing will be performed on 200 studies (held-out test database). Test data will be drawn in part (1/4) from the same source and distribution as the training data. The majority of test data (3/4) however will consist of oncologic PET/CT examinations that were drawn from different sources reflecting different domains and clinical settings.
Prizes will be awarded in three categories:
- Award category 1: Highest overall score
- Award category 2: Highest robustness across domains
- Award category 3: Jury prizes for scientific contribution and creative engineering