Unsupervised Adaptation from FDG to PSMA PET/CT for 3D Lesion Detection under Label Shift

2026-03-13

AI Summary

This work addresses the label shift problem—manifested as changes in lesion size distribution and count—arising from tracer differences in unsupervised domain adaptation from FDG PET/CT to PSMA PET/CT. To tackle this, we propose a novel self-training framework that explicitly models and compensates for label shift. The method alternates between optimizing the detector on the FDG source domain and retraining on pseudo-labels generated for the PSMA target domain, incorporating an exponential moving average mechanism to adaptively adjust anchor box shapes. Furthermore, a quota-based pseudo-label selection strategy, guided by lesion volume histogram binning, is introduced to enhance reliability. Evaluated on the AutoPET 2024 dataset, our approach significantly outperforms both the source-only baseline and conventional self-training methods, with improvements in AP and FROC metrics confirming the effectiveness of the proposed label shift modeling.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Learning to Search
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

In this work, we propose an unsupervised domain adaptation (UDA) framework for 3D volumetric lesion detection that adapts a detector trained on labeled FDG PET/CT to unlabeled PSMA PET/CT. Beyond covariate shift, cross tracer adaptation also exhibits label shift in both lesion size composition and the number of lesions per subject. We introduce self-training with two mechanisms that explicitly model and compensate for this label shift. First, we adaptively adjust the detection anchor shapes by re-estimating target domain box scales from selected pseudo labels and updating anchors with an exponential moving average. This increases positive anchor coverage for small PSMA lesions and stabilizes box regression. Second, instead of a fixed confidence threshold for pseudo-label selection, we allocate size bin-wise quotas according to the estimated target domain histogram over lesion volumes. The self-training alternates between supervised learning with prior-guided pseudo labeling on PSMA and supervised learning on labeled FDG. On AutoPET 2024, adapting from 501 labeled FDG studies to 369 $^{18}$F-PSMA studies, the proposed method improves both AP and FROC over the source-only baseline and conventional self-training without label-shift mitigation, indicating that modeling target lesion prevalence and size composition is an effective path to robust cross-tracer detection.

Problem

• unsupervised domain adaptation
• label shift
• 3D lesion detection
• cross-tracer adaptation
• PET/CT

Innovation

• unsupervised domain adaptation
• label shift
• 3D lesion detection
• pseudo-labeling
• cross-tracer PET/CT