Revisiting Dual-Encoder Fusion and Optimal Transport Regularization in PET-CT Tumor Segmentation: A Matched-Control Study under Extended Training
AI Medicine · 30.10.2026 · 3 (2)
DOI: 10.53941/aim.2026.100012 ↗Kurz gefasst
Fachartikel in der Zeitschrift AI Medicine, online erschienen am 30.10.2026. Verfasst von Man Li und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Department of Science and Technology of Liaoning Province.
Abstract (Originalsprache)
Dual-encoder fusion and explicit cross-modal alignment are two common strategies for multimodal PET-CT segmentation, but their relative contributions remain unclear when training duration and model capacity are controlled. We therefore reassess dual-encoder fusion and Optimal Transport (OT) regularization in DpDNet using an extended 200-epoch protocol with convergence monitoring and five-fold cross-validation on 75 AutoPET-derived PET-CT cases. The dual encoder shows no significant Dice advantage over a parameter-matched, approximately compute-matched weight-shared two-stream encoder (\(\Delta\text{Dice} = 0.0076\), \(p = 0.61\)), but achieves higher Dice than a parameter-matched early-fusion encoder (\(\Delta\text{Dice} = 0.0465\), \(p = 0.016\)). Because the latter requires \(1.24\times\) the FLOPs, this difference cannot be attributed uniquely to fusion position. OT yields no significant late-training Dice improvement on either the dual encoder (\(\Delta\text{Dice} = 0.0091\), \(p = 0.58\)) or the weight-shared two-stream encoder (\(\Delta\text{Dice} = 0.0080\), \(p = 0.45\)), and a 20-epoch pilot advantage does not persist under the 200-epoch protocol. The previously observed pre-training–OT interaction is likewise not reproduced and remains exploratory. Voxel Dice does not distinguish the matched-capacity configurations, whereas lesion-level detection shows only nominal differences that do not survive multiple-comparison correction. These results indicate that conclusions about multimodal PET-CT segmentation depend strongly on control design, training duration, and evaluation metric.
Förderung
- National Natural Science Foundation of China
- Department of Science and Technology of Liaoning Province
Zitieren
APA
Li, M., Tang, Y., Zhao, X., Zhao, Y., & Wang, M. (2026). Revisiting Dual-Encoder Fusion and Optimal Transport Regularization in PET-CT Tumor Segmentation: A Matched-Control Study under Extended Training. AI Medicine, 3(2). https://doi.org/10.53941/aim.2026.100012
Vancouver
Li M, Tang Y, Zhao X, Zhao Y, Wang M. Revisiting Dual-Encoder Fusion and Optimal Transport Regularization in PET-CT Tumor Segmentation: A Matched-Control Study under Extended Training. AI Medicine. 2026;3(2). doi:10.53941/aim.2026.100012
BibTeX
@article{Li2026,
author = {Li, Man and Tang, Yongwei and Zhao, Xiaoqing and Zhao, Yangyang and Wang, Meng},
title = {Revisiting Dual-Encoder Fusion and Optimal Transport Regularization in PET-CT Tumor Segmentation: A Matched-Control Study under Extended Training},
journal = {AI Medicine},
year = {2026},
volume = {3},
number = {2},
doi = {10.53941/aim.2026.100012}
}
EUVIMED