Organ segmentation in CT volumes is an important pre-processing step in many computerassisted intervention and diagnosis methods. In recent years, convolutional neural networkshave dominated the state of the art in this task. However, since this problem presents achallenging environment due to high variability in the organ’s shape and similarity betweentissues, the generation of false negative and false positive regions in the output segmentationis a common issue. Recent works have shown that the uncertainty analysis of the modelcan provide us with useful information about potential errors in the segmentation. In thiscontext, we proposed a segmentation refinement method based on uncertainty analysisand graph convolutional networks. We employ the uncertainty levels of the convolutionalnetwork in a particular input volume to formulate a semi-supervised graph learning problemthat is solved by training a graph convolutional network. To test our method we refine theinitial output of a 2D U-Net. We validate our framework with the NIH pancreas datasetand the spleen dataset of the medical segmentation decathlon. We show that our methodoutperforms the state-of-the-art CRF refinement method by improving the dice score by1% for the pancreas and 2% for spleen, with respect to the original U-Net’s prediction.Finally, we perform a sensitivity analysis on the parameters of our proposal and discussthe applicability to other CNN architectures, the results, and current limitations of themodel for future work in this research direction. For reproducibility purposes, we make ourcode publicly available athttps://github.com/rodsom22/gcn_refinement.