Abstract
In brain structural segmentation, multi-atlas strategies are increasingly being used over single-atlas strategies because of their ability to fit a wider anatomical variability. Patch-based label fusion (PBLF) is a type of such multi-atlas approaches that labels each target point as a weighted combination of neighboring atlas labels, where atlas points with higher local similarity to the target contribute more strongly to label fusion. PBLF can be potentially improved by increasing the discriminative capabilities of the local image similarity measurements. We propose a framework to compute patch embeddings using neural networks so as to increase discriminative abilities of similarity-based weighted voting in PBLF. As particular cases, our framework includes embeddings with different complexities, namely, a simple scaling, an affine transformation, and non-linear transformations. We compare our method with state-of-the-art alternatives in whole hippocampus and hippocampal subfields segmentation experiments using publicly available datasets. Results show that even the simplest versions of our method outperform standard PBLF, thus evidencing the benefits of discriminative learning. More complex transformation models tended to achieve better results than simpler ones, obtaining a considerable increase in average Dice score compared to standard PBLF.
| Original language | English |
|---|---|
| Pages (from-to) | 143-155 |
| Number of pages | 13 |
| Journal | Medical Image Analysis |
| Volume | 44 |
| DOIs | |
| Publication status | Published - 2018 Feb 1 |
Fingerprint
Keywords
- Brain MRI
- Embedding
- Hippocampus
- Multi-atlas segmentation
- Neural networks
- Patch-based label fusion
ASJC Scopus subject areas
- Radiological and Ultrasound Technology
- Radiology Nuclear Medicine and imaging
- Computer Vision and Pattern Recognition
- Health Informatics
- Computer Graphics and Computer-Aided Design
Cite this
Learning non-linear patch embeddings with neural networks for label fusion. / for the Alzheimer's Disease Neuroimaging Initiative.
In: Medical Image Analysis, Vol. 44, 01.02.2018, p. 143-155.Research output: Contribution to journal › Article
}
TY - JOUR
T1 - Learning non-linear patch embeddings with neural networks for label fusion
AU - for the Alzheimer's Disease Neuroimaging Initiative
AU - Sanroma, Gerard
AU - Benkarim, Oualid M.
AU - Piella, Gemma
AU - Camara, Oscar
AU - Wu, Guorong
AU - Shen, Dinggang
AU - Gispert, Juan D.
AU - Molinuevo, José Luis
AU - González Ballester, Miguel A.
PY - 2018/2/1
Y1 - 2018/2/1
N2 - In brain structural segmentation, multi-atlas strategies are increasingly being used over single-atlas strategies because of their ability to fit a wider anatomical variability. Patch-based label fusion (PBLF) is a type of such multi-atlas approaches that labels each target point as a weighted combination of neighboring atlas labels, where atlas points with higher local similarity to the target contribute more strongly to label fusion. PBLF can be potentially improved by increasing the discriminative capabilities of the local image similarity measurements. We propose a framework to compute patch embeddings using neural networks so as to increase discriminative abilities of similarity-based weighted voting in PBLF. As particular cases, our framework includes embeddings with different complexities, namely, a simple scaling, an affine transformation, and non-linear transformations. We compare our method with state-of-the-art alternatives in whole hippocampus and hippocampal subfields segmentation experiments using publicly available datasets. Results show that even the simplest versions of our method outperform standard PBLF, thus evidencing the benefits of discriminative learning. More complex transformation models tended to achieve better results than simpler ones, obtaining a considerable increase in average Dice score compared to standard PBLF.
AB - In brain structural segmentation, multi-atlas strategies are increasingly being used over single-atlas strategies because of their ability to fit a wider anatomical variability. Patch-based label fusion (PBLF) is a type of such multi-atlas approaches that labels each target point as a weighted combination of neighboring atlas labels, where atlas points with higher local similarity to the target contribute more strongly to label fusion. PBLF can be potentially improved by increasing the discriminative capabilities of the local image similarity measurements. We propose a framework to compute patch embeddings using neural networks so as to increase discriminative abilities of similarity-based weighted voting in PBLF. As particular cases, our framework includes embeddings with different complexities, namely, a simple scaling, an affine transformation, and non-linear transformations. We compare our method with state-of-the-art alternatives in whole hippocampus and hippocampal subfields segmentation experiments using publicly available datasets. Results show that even the simplest versions of our method outperform standard PBLF, thus evidencing the benefits of discriminative learning. More complex transformation models tended to achieve better results than simpler ones, obtaining a considerable increase in average Dice score compared to standard PBLF.
KW - Brain MRI
KW - Embedding
KW - Hippocampus
KW - Multi-atlas segmentation
KW - Neural networks
KW - Patch-based label fusion
UR - http://www.scopus.com/inward/record.url?scp=85037974131&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85037974131&partnerID=8YFLogxK
U2 - 10.1016/j.media.2017.11.013
DO - 10.1016/j.media.2017.11.013
M3 - Article
C2 - 29247877
AN - SCOPUS:85037974131
VL - 44
SP - 143
EP - 155
JO - Medical Image Analysis
JF - Medical Image Analysis
SN - 1361-8415
ER -