MICCAI 2022 Daily – Tuesday

26 DAILY MICCAI Tuesday Center (DKFZ), Intel Labs, MLCommons, Sage Bionetworks, and Intel Health & Life Sciences that all shared the same vision and the same principles of transparency, reproducibility, fairness, and open- science. As such all the software infrastructure used and developed for the conduction of the challenge is open-source, and publicly available through their respective GitHub pages. The exact tools comprise the Cancer Imaging Phenomics Toolkit (CaPTk), the Federated Tumor Segmentation (FeTS) tool, the Open Federated Learning (OpenFL) library, MedPerf, Kaapana, and the Trustworthy Federated Data Analytics (TFDA) project. Our immediate next steps and goal are to make the FeTS challenge a continuous evaluation challenge through the SAGE Bionetworks’ Synapse platform, leveraging our open-source software infrastructure and offering annual snapshots of algorithmic performance at the MICCAI conference. Figure 1. The FeTS challenge problem definition. A-C denote the specific tumor regions considered for the performance evaluation of the participating algorithms and specifically highlight (from left to right): the enhancing tumor (ET - yellow) visible in a T1 post-contrast scan, surrounding the cystic/necrotic components of the core (panel A), the tumor core (TC – magenta) and the whole tumor (WT - cyan) visible in the corresponding T2 (panel B) and T2-FLAIR (panel C) scans, respectively. Panel D, depicts the combined segmentations generating the final tumor sub-region labels, as provided to the participants: enhancing core (yellow), necrotic/cystic core (red), and edema/invasion (green). Challenge - FeTS