Towards bridging the distribution gap: Instance to Prototype Earth Mover's Distance for distribution alignment.

Zhou, Qin; Wang, Runze; Zeng, Guodong; Fan, Heng; Zheng, Guoyan (2022). Towards bridging the distribution gap: Instance to Prototype Earth Mover's Distance for distribution alignment. Medical image analysis, 82(102607), p. 102607. Elsevier 10.1016/j.media.2022.102607

[img] Text
1-s2.0-S1361841522002365-main.pdf - Published Version
Restricted to registered users only
Available under License Publisher holds Copyright.

Download (2MB)

Despite remarkable success of deep learning, distribution divergence remains a challenge that hinders the performance of many tasks in medical image analysis. Large distribution gap may deteriorate the knowledge transfer across different domains or feature subspaces. To achieve better distribution alignment, we propose a novel module named Instance to Prototype Earth Mover's Distance (I2PEMD), where shared class-specific prototypes are progressively learned to narrow the distribution gap across different domains or feature subspaces, and Earth Mover's Distance (EMD) is calculated to take into consideration the cross-class relationships during embedding alignment. We validate the effectiveness of the proposed I2PEMD on two different tasks: multi-modal medical image segmentation and semi-supervised classification. Specifically, in multi-modal medical image segmentation, I2PEMD is explicitly utilized as a distribution alignment regularization term to supervise the model training process, while in semi-supervised classification, I2PEMD works as an alignment measure to sort and cherry-pick the unlabeled data for more accurate and robust pseudo-labeling. Results from comprehensive experiments demonstrate the efficacy of the present method.

Item Type:

Journal Article (Original Article)

Division/Institute:

04 Faculty of Medicine > Faculty Institutions > sitem Center for Translational Medicine and Biomedical Entrepreneurship

UniBE Contributor:

Zeng, Guodong

Subjects:

600 Technology > 610 Medicine & health

ISSN:

1361-8415

Publisher:

Elsevier

Language:

English

Submitter:

Pubmed Import

Date Deposited:

13 Sep 2022 14:31

Last Modified:

05 Dec 2022 16:24

Publisher DOI:

10.1016/j.media.2022.102607

PubMed ID:

36075148

Uncontrolled Keywords:

Distribution alignment Earth Mover’s Distance Instance to prototype matching Semi-supervised classification Unpaired multi-modal segmentation

BORIS DOI:

10.48350/172770

URI:

https://boris.unibe.ch/id/eprint/172770

Actions (login required)

Edit item Edit item
Provide Feedback