Probabilistic Atlases to Enforce Topological Constraints
This work addresses segmentation challenges in medical imaging for structures beyond standardized ones like the brain or heart, though it appears incremental by extending atlas-based methods to more complex cases.
The paper tackles the problem of segmenting complex anatomical structures with arbitrary shapes and poses by proposing an encoding-decoding CNN architecture that uses probabilistic atlases to enforce topological constraints, resulting in improved segmentation results.
Probabilistic atlases (PAs) have long been used in standard segmentation approaches and, more recently, in conjunction with Convolutional Neural Networks (CNNs). However, their use has been restricted to relatively standardized structures such as the brain or heart which have limited or predictable range of deformations. Here we propose an encoding-decoding CNN architecture that can exploit rough atlases that encode only the topology of the target structures that can appear in any pose and have arbitrarily complex shapes to improve the segmentation results. It relies on the output of the encoder to compute both the pose parameters used to deform the atlas and the segmentation mask itself, which makes it effective and end-to-end trainable.