Fully automated segmentation of [18F]FDG- and PSMA-PET/CT images via data-centric Deep-Learning.

Purpose To develop and validate LION (Lesion Identification in Oncological Nuclear imaging), an open-source PET-only tumor segmentation pipeline for [ ¹⁸ F]FDG and PSMA-targeted PET/CT, and to investigate how training data characteristics influence segmentation performance. Materials and Methods In this retrospective multicenter study, 5,209 [ ¹⁸ F]FDG PET/CT scans spanning 19 disease types and 2,046 PSMA-targeted PET/CT scans were used to train PET-only segmentation models. Tumor segmentation incorporated organs with physiological uptake as auxiliary classes to enable PET-only inference. Tumor Occurrence Maps (TOMs) quantified tumor spatial diversity across the training data. For [ ¹⁸ F]FDG, disease-specific and mixed-disease models trained on progressively larger subsets were compared to test whether increasing spatial diversity improves generalization. Scanner-related domain shift was analyzed using DINOv2 embeddings. Models were evaluated on multicenter holdout cohorts (616 [ ¹⁸ F]FDG across 4 diseases; 443 PSMA-targeted prostate cancer scans) and compared with three open-source tools. Results Organ context improved median Dice from 0.62 to 0.71 for [ ¹⁸ F]FDG and from 0.75 to 0.83 for PSMA. Spatial diversity measured by TOMs was strongly associated with Dice (Spearman ρ = 0.87, P < 0.001). A mixed-disease model trained on 500 patients matched the performance of a lymphoma specialist model trained on 3,031 cases. DINOv2 embeddings revealed scanner-induced domain shift, between same-disease cohorts. LION achieved median Dice scores of 0.71 ([ ¹⁸ F]FDG) and 0.85 (PSMA) and outperformed other open-source approaches on common holdout patients. Conclusion LION enables PET-only automated segmentation for [ ¹⁸ F]FDG and PSMA-targeted PET. Training data composition, particularly spatial diversity quantified by TOMs, was strongly associated with segmentation performance.