Disentangled representation learning in cardiac image analysis
- Submitting institution
-
The University of Essex
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 1476
- Type
- D - Journal article
- DOI
-
10.1016/j.media.2019.101535
- Title of journal
- Medical Image Analysis
- Article number
- 101535
- First page
- 101535
- Volume
- 58
- Issue
- -
- ISSN
- 1361-8415
- Open access status
- Deposit exception
- Month of publication
- July
- Year of publication
- 2019
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- Yes
- Number of additional authors
-
7
- Research group(s)
-
A - Artificial Intelligence (AI)
- Citation count
- 14
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- MIA is one of the highest-ranked journals in computer science and image analysis. We present a highly-cited, first deep learning model that disentangles medical images into spatial and non-spatial information. This model (SDNet) has unique abilities: maintains high segmentation accuracy and is able to quantify cardiac volumes, even when the amount of labelled images is ?6%. With unlabelled data being abundant worldwide, this work has a major influence on semi-supervised learning by other leading groups (Du, MIA,2020; Li, IEEE T Neur Net Lear,2020), and has been cited as a reference standard in high-impact surveys (Tajbakhsh, MIA,2020; Panayides, IEEE JBHI,2020).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -