Dual structure-aware image filterings for semi-supervised medical image segmentation

Yuliang Gu; Zhichao Sun; Tian Chen; Xin Xiao; Yepeng Liu; Yongchao Xu; Laurent Najman

doi:10.1016/j.media.2024.103364

Article Dans Une Revue Medical Image Analysis Année : 2025

Dual structure-aware image filterings for semi-supervised medical image segmentation

(1, 2) , (1, 2) , (1, 2) , (1, 2) , (1, 2) , (1, 2) , (3)

1
2
3

Yuliang Gu

Fonction : Auteur

Artificial Intelligence Institute of Wuhan University [Wuhan City]

School of Computer Science

Zhichao Sun

Fonction : Auteur

Artificial Intelligence Institute of Wuhan University [Wuhan City]

School of Computer Science

Tian Chen

Fonction : Auteur

Artificial Intelligence Institute of Wuhan University [Wuhan City]

School of Computer Science

Xin Xiao

Fonction : Auteur

Artificial Intelligence Institute of Wuhan University [Wuhan City]

School of Computer Science

Yepeng Liu

Fonction : Auteur

Artificial Intelligence Institute of Wuhan University [Wuhan City]

School of Computer Science

Yongchao Xu

Fonction : Auteur
PersonId : 1428871
ORCID : 0000-0002-7253-3151

Artificial Intelligence Institute of Wuhan University [Wuhan City]

School of Computer Science

Laurent Najman

Fonction : Auteur
PersonId : 28
IdHAL : laurent-najman
ORCID : 0000-0002-6190-0235
IdRef : 087172712

Laboratoire d'Informatique Gaspard-Monge

Résumé

Semi-supervised image segmentation has attracted great attention recently. The key is how to leverage unlabeled images in the training process. Most methods maintain consistent predictions of the unlabeled images under variations (e.g., adding noise/perturbations, or creating alternative versions) in the image and/or model level. In most image-level variation, medical images often have prior structure information, which has not been well explored. In this paper, we propose novel dual structure-aware image filterings (DSAIF) as the image-level variations for semi-supervised medical image segmentation. Motivated by connected filtering that simplifies image via filtering in structure-aware tree-based image representation, we resort to the dual contrast invariant Max-tree and Min-tree representation. Specifically, we propose a novel connected filtering that removes topologically equivalent nodes (i.e., connected components) having no siblings in the Max/Min-tree. This results in two filtered images preserving topologically critical structure. Applying the proposed DSAIF to mutually supervised networks decreases the consensus of their erroneous predictions on unlabeled images. This helps to alleviate the confirmation bias issue of overfitting to noisy pseudo labels of unlabeled images, and thus effectively improves the segmentation performance. Extensive experimental results on three benchmark datasets demonstrate that the proposed method significantly/consistently outperforms some state-of-the-art methods. Source code is publicly available at https://github.com/GuGuLL123/DSAIF-SEMI.

Mots clés

Max-tree Min-tree Connected filtering Semi-supervised medical image segmentation

Domaines

Imagerie médicale Réseau de neurones [cs.NE]

Fichier principal

mia.pdf (2.86 Mo)

elsevier-logo.pdf (55.88 Ko)

fig_experiments_1_100_nature_la_dice_major_revision.pdf (15.02 Ko)

fig_experiments_1_100_nature_la_jaccard_major_revision.pdf (15.26 Ko)

fig_experiments_1_100_nature_pancreas_dice_major_revision.pdf (15.7 Ko)

fig_experiments_1_100_nature_pancreas_jaccard_major_revision.pdf (15.51 Ko)

fig_experiments_1_100_nature_prostate_dice_major_revision.pdf (15.12 Ko)

fig_experiments_1_100_nature_prostate_jaccard_major_revision.pdf (15.65 Ko)

fig_experiments_error_map.pdf (21.53 Ko)

fig_experiments_right_map_unlabeled_major_revision.pdf (20.84 Ko)

fig_experiments_unlabeled_dice1.pdf (21.62 Ko)

fig_experiments_unlabeled_dice2.pdf (21.63 Ko)

fig_experiments_visualiations_LA_10_1.pdf (58.02 Ko)

fig_experiments_visualiations_LA_10_2.pdf (57.48 Ko)

fig_experiments_visualiations_LA_10_3.pdf (57.34 Ko)

fig_experiments_visualiations_LA_10_4.pdf (57.28 Ko)

fig_experiments_visualiations_LA_10_5.pdf (55.76 Ko)

fig_experiments_visualiations_LA_10_6.pdf (54.71 Ko)

fig_experiments_visualiations_LA_5_1.pdf (57.26 Ko)

fig_experiments_visualiations_LA_5_2.pdf (55.55 Ko)

fig_experiments_visualiations_LA_5_3.pdf (54.49 Ko)

fig_experiments_visualiations_LA_5_4.pdf (55.71 Ko)

fig_experiments_visualiations_LA_5_5.pdf (55.24 Ko)

fig_experiments_visualiations_LA_5_6.pdf (54.88 Ko)

fig_experiments_visualiations_Pancreas_10_1.pdf (56.9 Ko)

fig_experiments_visualiations_Pancreas_10_2.pdf (56.57 Ko)

fig_experiments_visualiations_Pancreas_10_3.pdf (56.71 Ko)

fig_experiments_visualiations_Pancreas_10_4.pdf (56.79 Ko)

fig_experiments_visualiations_Pancreas_10_5.pdf (57.08 Ko)

fig_experiments_visualiations_Pancreas_10_6.pdf (57.73 Ko)

fig_experiments_visualiations_Pancreas_5_1.pdf (56.96 Ko)

fig_experiments_visualiations_Pancreas_5_2.pdf (57.01 Ko)

fig_experiments_visualiations_Pancreas_5_3.pdf (56.83 Ko)

fig_experiments_visualiations_Pancreas_5_4.pdf (58.2 Ko)

fig_experiments_visualiations_Pancreas_5_5.pdf (57.92 Ko)

fig_experiments_visualiations_Pancreas_5_6.pdf (57.91 Ko)

fig_experiments_visualiations_Prostate_10_1.pdf (52.78 Ko)

fig_experiments_visualiations_Prostate_10_2.pdf (53.06 Ko)

fig_experiments_visualiations_Prostate_10_3.pdf (53.08 Ko)

fig_experiments_visualiations_Prostate_10_4.pdf (53.28 Ko)

fig_experiments_visualiations_Prostate_10_5.pdf (52.66 Ko)

fig_experiments_visualiations_Prostate_10_6.pdf (52.68 Ko)

fig_experiments_visualiations_Prostate_5_1.pdf (45.41 Ko)

fig_experiments_visualiations_Prostate_5_2.pdf (45.37 Ko)

fig_experiments_visualiations_Prostate_5_3.pdf (45.55 Ko)

fig_experiments_visualiations_Prostate_5_4.pdf (45.34 Ko)

fig_experiments_visualiations_Prostate_5_5.pdf (45.35 Ko)

fig_experiments_visualiations_Prostate_5_6.pdf (45.39 Ko)

fig_method_pipeline.pdf (155.93 Ko)

fig_method_synthetic_aug.pdf (145.97 Ko)

fig_method_synthetic_noaug1.pdf (253.49 Ko)

fig_method_synthetic_noaug1_major_revision.pdf (249.46 Ko)

fig_method_tree_transform_LA_1_1.pdf (18.24 Ko)

fig_method_tree_transform_LA_1_2.pdf (14 Ko)

fig_method_tree_transform_LA_1_3.pdf (10.08 Ko)

fig_method_tree_transform_LA_1_4.pdf (14.35 Ko)

fig_method_tree_transform_Pancreas_1_1.pdf (42.75 Ko)

fig_method_tree_transform_Pancreas_1_2.pdf (40.86 Ko)

fig_method_tree_transform_Pancreas_1_3.pdf (15.5 Ko)

fig_method_tree_transform_Pancreas_1_4.pdf (35.06 Ko)

fig_method_tree_transform_Prostate_1_1.pdf (15.49 Ko)

fig_method_tree_transform_Prostate_1_2.pdf (19.29 Ko)

fig_method_tree_transform_Prostate_1_3.pdf (17.77 Ko)

fig_method_tree_transform_Prostate_1_4.pdf (18.01 Ko)

fig_suppl_pipeline_conet_suppl3_major_revision.pdf (157.47 Ko)

fig_suppl_pipeline_mcnet_suppl1_major_revision.pdf (144.12 Ko)

fig_suppl_pipeline_mcnetp_suppl2_major_revision.pdf (151.17 Ko)

medima-logo.pdf (86.25 Ko)

medimafig01.pdf (19.91 Ko)

mia.spl (40 B)

Origine	Fichiers produits par l'(les) auteur(s)

Laurent Najman : Connectez-vous pour contacter le contributeur

https://hal.science/hal-04740759

Soumis le : vendredi 18 octobre 2024-10:40:30

Dernière modification le : jeudi 19 décembre 2024-16:50:04

Dates et versions

hal-04740759 , version 1 (18-10-2024)

Identifiants

HAL Id : hal-04740759 , version 1
DOI : 10.1016/j.media.2024.103364

Citer

Yuliang Gu, Zhichao Sun, Tian Chen, Xin Xiao, Yepeng Liu, et al.. Dual structure-aware image filterings for semi-supervised medical image segmentation. Medical Image Analysis, 2025, 99, pp.103364. ⟨10.1016/j.media.2024.103364⟩. ⟨hal-04740759⟩

Exporter

BibTeX XML-TEI Dublin Core DC Terms EndNote DataCite

Collections

ENPC CNRS UNIV-EIFFEL

23 Consultations

96 Téléchargements

Dual structure-aware image filterings for semi-supervised medical image segmentation

Résumé

Mots clés

Domaines

Dates et versions

Identifiants

Citer

Exporter

Collections

Altmetric

Partager