Cross-Domain Semi-Supervised Organ Detection

Nian Li, MORTEZA GHAHREMANI, Bailiang Jian, Pascual Tejero Cervera, Benedikt Wiestler, Marcus Makowski, Christian Wachinger

Keywords: Cross-Domain;Semi-Supervised;Object Detection;Teacher-Student;Pseudo-label;Curriculum Learning

TL;DR: Cross-Domain Semi-Supervised Organ Detection

Abstract:

Domain adaptation for 3D organ detection in CT imaging is challenging due to variations in scanner types, imaging protocols, and overall acquisition conditions.
As supervised detection models require large, annotated datasets from diverse scanners and institutions, semi-supervised approaches have gained attention for their ability to leverage limited unlabeled target data. However, traditional semi-supervised methods typically fail to make effective use of the few labeled target samples and most often do not yield satisfactory results. To address this limitation, we introduce a novel cross-domain semi-supervised detection framework (CDSS-Det) built upon the Transformer-based Organ-DETR model. CDSS-Det is a cross-domain semi-supervised framework for 3D organ detection that addresses unreliable pseudo-labels and limited target supervision under domain shift. It introduces a curriculum-guided pseudo-labeling mechanism and domain-robust representation learning to enable effective knowledge transfer from a well-annotated source domain to a sparsely labeled target domain. Experiments on multi-domain CT datasets demonstrate that incorporating a small number of labeled target samples significantly boosts detection performance over conventional domain adaptation and semi-supervised methods. CDSS-Det consistently achieves higher mean Average Precision (mAP), with notable improvements in detecting small organs, and surpasses a fully supervised model trained solely on the labeled target domain by over 10\%. These results underscore the potential of CDSS-Det in efficiently leveraging both labeled and unlabeled target data in cross-domain organ detection, advancing annotation-efficient deep learning models in medical imaging.

Primary Subject Area: Transfer Learning and Domain Adaptation

Secondary Subject Area: Detection and Diagnosis

Registration Requirement: Yes

Reproducibility: https://github.com/ai-med/CDSS-Det

Visa & Travel: Yes

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Replace NNN with your OpenReview submission ID., Includes \documentclass{midl}, \jmlryear{2026}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package., Did not use the times package., Use the correct spelling and format, avoid Unicode characters, and use LaTeX equivalents instead., Any math in the title and abstract must be enclosed within $...$., Did not override the bibliography style defined in midl.cls and did not use \begin{thebibliography} directly to insert references., Avoid using \scalebox; use \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., No separate supplementary PDF uploads., Acknowledgements, references, and appendix must start after the main content.

Submission Number: 10