Robust Fusion of Probability Maps

The fusion of probability maps is required when trying to analyse a collection of image labels or probability maps produced by several segmentation algorithms or human raters. The challenge is to weight properly the combination of maps in order to reflect the agreement among raters, the presence of outliers and the spatial uncertainty in the consensus. In this paper, we address several shortcomings of prior work in continuous label fusion. We introduce a novel approach to jointly estimate a reliable consensus map and assess the production of outliers and the confidence in each rater. Our probabilistic model is based on Student's t-distributions allowing local estimates of raters' performances. The introduction of bias and spatial priors leads to proper rater bias estimates and a control over the smoothness of the consensus map. Image intensity information is incorporated by geodesic distance transform for binary masks. Finally, we propose an approach to cluster raters based on variational boosting thus producing possibly several alternative consensus maps. Our approach was successfully tested on the MICCAI 2016 MS lesions dataset, on MR prostate delineations and on deep learning based segmentation predictions of lung nodules from the LIDC dataset.