Research Repository

Disentangle, align and fuse for multimodal and semi-supervised image segmentation.

Chartsias, Agisilaos and Papanastasiou, Giorgos and Wang, Chengjia and Semple, Scott and Newby, David and Dharmakumar, Rohan and Tsaftaris, Sotirios (2021) 'Disentangle, align and fuse for multimodal and semi-supervised image segmentation.' IEEE Transactions on Medical Imaging, 40 (3). pp. 781-792. ISSN 0278-0062 (In Press)

[img]
Preview
Text
Chartsias, Papanstasiou, et al_IEEE TMI_acceptedversion_2020.pdf - Accepted Version

Download (16MB) | Preview

Abstract

Magnetic resonance (MR) protocols rely on several sequences to assess pathology and organ status properly. Despite advances in image analysis, we tend to treat each sequence, here termed modality, in isolation. Taking advantage of the common information shared between modalities (an organ's anatomy) is beneficial for multi-modality processing and learning. However, we must overcome inherent anatomical misregistrations and disparities in signal intensity across the modalities to obtain this benefit. We present a method that offers improved segmentation accuracy of the modality of interest (over a single input model), by learning to leverage information present in other modalities, even if few (semi-supervised) or no (unsupervised) annotations are available for this specific modality. Core to our method is learning a disentangled decomposition into anatomical and imaging factors. Shared anatomical factors from the different inputs are jointly processed and fused to extract more accurate segmentation masks. Image misregistrations are corrected with a Spatial Transformer Network, which non-linearly aligns the anatomical factors. The imaging factor captures signal intensity characteristics across different modality data and is used for image reconstruction, enabling semi-supervised learning. Temporal and slice pairing between inputs are learned dynamically. We demonstrate applications in Late Gadolinium Enhanced (LGE) and Blood Oxygenation Level Dependent (BOLD) cardiac segmentation, as well as in T2 abdominal segmentation.

Item Type: Article
Uncontrolled Keywords: Image segmentation; Biomedical imaging; Annotations; Training; Semantics; Decoding; Multimodal segmentation; disentanglement; magnetic resonance imaging
Divisions: Faculty of Science and Health
Faculty of Science and Health > Computer Science and Electronic Engineering, School of
SWORD Depositor: Elements
Depositing User: Elements
Date Deposited: 06 Nov 2020 11:30
Last Modified: 18 Aug 2022 12:25
URI: http://repository.essex.ac.uk/id/eprint/28978

Actions (login required)

View Item View Item