Extended Data Fig. 2: Example of a complete dataset.

Computed tomography (CT) volumes, CT-based organ segmentation masks and Fluorodeoxyglucose Positron Emission Tomography (FDG-PET/CT) volumes (or subsets of these volumes) were used as input data. Metabolically active tumor lesions were segmented manually (colored in blue) and automatically (colored in red). This particular dataset is drawn from the UKT test set and shows excellent alignment between manual lesion segmentation and automated lesion segmentation using an nnU-Net baseline with CT and FDG-PET as input.