Name
Abstract | ||
|---|---|---|
We propose to adapt segmentation networks with a constrained formulation, which embeds domain-invariant prior knowledge about the segmentation regions. Such knowledge may take the form of simple anatomical information, e.g., structure size or shape, estimated from source samples or known a priori. Our method imposes domainin-variant inequality constraints on a network output of unlabeled target samples. It implicitly matches prediction statistics between target and source domains with permitted uncertainty of prior knowledge. We address our constrained problem with a differentiable penalty, fully suited for conventional gradient descent approaches, removing the need for computationally expensive Lagrangian optimization with dual projections. Unlike current two-step adversarial training, our formulation is based on a single loss in a single network, which simplifies adaptation by avoiding extra adversarial steps, while improving convergence and quality of training. The comparison of our approach with state-of-the-art adversarial methods reveals substantially better performance on the challenging task of adapting spine segmentation across different MRI modalities. Our results also show a robustness to imprecision of size priors, approaching the accuracy of a fully supervised model trained directly in a target domain. Our method can be readily used for various constraints and segmentation problems. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1007/978-3-030-32245-8_37 | Lecture Notes in Computer Science |
Keywords | DocType | Volume |
Image segmentation,Domain adaptation,Constrained CNNs | Conference | 11765 |
ISSN | Citations | PageRank |
0302-9743 | 0 | 0.34 |
References | Authors | |
0 | 5 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Mathilde Bateson | 1 | 0 | 0.68 |
| Hoel Kervadec | 2 | 25 | 2.94 |
| Dolz Jose | 3 | 91 | 16.76 |
| Hervé Lombaert | 4 | 0 | 0.68 |
| Ismail Ben Ayed | 5 | 678 | 52.28 |