Stochastic Segmentation Networks: Modelling Spatially Correlated Aleatoric Uncertainty
This addresses the need for reliable uncertainty estimation in medical imaging, where expert disagreements on boundaries are common, though it is an incremental improvement over existing probabilistic methods.
The paper tackled the problem of modeling aleatoric uncertainty in image segmentation, where multiple plausible solutions exist, by introducing stochastic segmentation networks (SSNs) that generate spatially coherent hypotheses; it demonstrated improved performance over state-of-the-art methods on medical imaging tasks like lung nodule and brain tumor segmentation.
In image segmentation, there is often more than one plausible solution for a given input. In medical imaging, for example, experts will often disagree about the exact location of object boundaries. Estimating this inherent uncertainty and predicting multiple plausible hypotheses is of great interest in many applications, yet this ability is lacking in most current deep learning methods. In this paper, we introduce stochastic segmentation networks (SSNs), an efficient probabilistic method for modelling aleatoric uncertainty with any image segmentation network architecture. In contrast to approaches that produce pixel-wise estimates, SSNs model joint distributions over entire label maps and thus can generate multiple spatially coherent hypotheses for a single image. By using a low-rank multivariate normal distribution over the logit space to model the probability of the label map given the image, we obtain a spatially consistent probability distribution that can be efficiently computed by a neural network without any changes to the underlying architecture. We tested our method on the segmentation of real-world medical data, including lung nodules in 2D CT and brain tumours in 3D multimodal MRI scans. SSNs outperform state-of-the-art for modelling correlated uncertainty in ambiguous images while being much simpler, more flexible, and more efficient.