Advanced Tumor Segmentation in PET/CT Imaging: A Training Strategy Study with nnU-Net for AutoPET III

2026-05-04

AI Summary

This study addresses the challenge of generalizing tumor segmentation in whole-body PET/CT imaging, where lesion variability in size, contrast, and anatomical distribution hinders robust performance. Building upon the nnU-Net framework, the authors integrate a ResNet encoder and systematically evaluate the synergistic effects of intensity normalization, batch-wise Dice loss optimization, and CraveMix data augmentation in the multi-center, multi-tracer setting of the AutoPET III challenge. This work presents the first comprehensive analysis of combined training strategies in this context, substantially reducing false positives and enhancing model robustness to lesion heterogeneity. The proposed method achieved a peak Dice score of 0.80 on the preliminary AutoPET III test set, securing third place in the final ranking.

Abstract

Tumor segmentation in whole-body PET/CT imaging is crucial for precise disease evaluation and treatment planning. However, it remains challenging due to variability in lesion size, contrast, and anatomical distribution. Relying on manual segmentation makes the process time-consuming and prone to intra- and inter-observer variability. This work presents a whole-body tumor segmentation method developed for the AutoPET III challenge, where the goal is to build models that generalize across tracers and multi-center data. We employ the nnU-Net framework with a ResNet-based encoder as our baseline and systematically investigate the impact of training strategies, including intensity normalization, batch dice optimization, and data augmentation using CraveMix. Our experiments show that these strategies significantly influence model performance, particularly in reducing false positives and improving robustness to lesion variability. The best-performing configuration achieves a Dice score of up to 0.80 on the preliminary test phase, and our method ranked third in the AutoPET III challenge. The code is publicly available here.

Problem

• Tumor Segmentation
• PET/CT Imaging
• Generalization
• Multi-center Data
• Lesion Variability

Innovation

• nnU-Net
• CraveMix
• intensity normalization
• batch Dice optimization
• whole-body tumor segmentation