Overview
The CLAP 2027 Challenge consists of two consecutive tasks: Task 1: Annotation and Task 2: Prediction.
Task 1: Annotation
In the first task, participants are requested to manually annotate anatomical landmarks. Each participant will receive pelvic MRI datasets with bony segmentations from 24 subjects. An instruction manual describing the landmark annotation procedure will be provided as well. Participants are requested to manually annotate anatomical landmarks for the provided subjects. The manual annotation is expected to take around 15 minutes per subject (so approximately 6 hours in total). Both point-like landmarks and area-based attachments must be identified. The completed annotations must be submitted following the provided instructions.
Task 2: Prediction
In the second task, participants are requested to predict the same set of anatomical landmarks as in Task 1 on 12 previously unseen subjects. Manual intervention during landmark prediction is NOT permitted. A subset of the collected annotations from Task 1 will be provided as template or training data. Any automated landmark prediction method may be used (e.g. deep learning, atlas-based methods, or non-rigid registration) on any suitable data representation (e.g. images or surface meshes). The completed predictions must be submitted following the provided instructions.