AutoPETIII: The Tracer Frontier. What Frontier?

2024-09-19

arXiv.org

AI Summary

To address poor generalizability in multi-tracer (FDG/PSMA) PET/CT lesion segmentation caused by unknown tracer types, this work proposes a tracer-agnostic, fully automatic segmentation method. The core innovation is a lightweight MIP-CNN dynamic routing network that adaptively identifies the tracer type based on maximum intensity projection (MIP) features and accordingly dispatches the corresponding nnUNetv2 six-fold ensemble model—without requiring manual tracer annotations. Evaluated on the AutoPET III Challenge, the method achieves state-of-the-art cross-tracer generalization, significantly outperforming competitors in Dice score. It represents the first approach to achieve high-accuracy, fully automatic lesion segmentation without prior tracer information, thereby eliminating dependence on tracer labels entirely.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-instance/Multi-view Learning
• Planning, Routing, and Scheduling: Activity and Plan Recognition

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Agentic search
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

For the last three years, the AutoPET competition gathered the medical imaging community around a hot topic: lesion segmentation on Positron Emitting Tomography (PET) scans. Each year a different aspect of the problem is presented; in 2024 the multiplicity of existing and used tracers was at the core of the challenge. Specifically, this year's edition aims to develop a fully automatic algorithm capable of performing lesion segmentation on a PET/CT scan, without knowing the tracer, which can either be a FDG or PSMA-based tracer. In this paper we describe how we used the nnUNetv2 framework to train two sets of 6 fold ensembles of models to perform fully automatic PET/CT lesion segmentation as well as a MIP-CNN to choose which set of models to use for segmentation.

Problem

• Automatic lesion segmentation on PET/CT scans
• Handling multiple tracer types without prior knowledge
• Developing ensemble models for robust segmentation performance

Innovation

• nnUNetv2 framework for PET/CT lesion segmentation
• Two sets of 6-fold ensemble models
• MIP-CNN for model selection based on tracer