Graham Wright

Matthew Ng, Fumin Guo, Labonny Biswas et al.

Objective: Convolutional neural networks (CNNs) have demonstrated promise in automated cardiac magnetic resonance image segmentation. However, when using CNNs in a large real-world dataset, it is important to quantify segmentation uncertainty and identify segmentations which could be problematic. In this work, we performed a systematic study of Bayesian and non-Bayesian methods for estimating uncertainty in segmentation neural networks. Methods: We evaluated Bayes by Backprop, Monte Carlo Dropout, Deep Ensembles, and Stochastic Segmentation Networks in terms of segmentation accuracy, probability calibration, uncertainty on out-of-distribution images, and segmentation quality control. Results: We observed that Deep Ensembles outperformed the other methods except for images with heavy noise and blurring distortions. We showed that Bayes by Backprop is more robust to noise distortions while Stochastic Segmentation Networks are more resistant to blurring distortions. For segmentation quality control, we showed that segmentation uncertainty is correlated with segmentation accuracy for all the methods. With the incorporation of uncertainty estimates, we were able to reduce the percentage of poor segmentation to 5% by flagging 31--48% of the most uncertain segmentations for manual review, substantially lower than random review without using neural network uncertainty (reviewing 75--78% of all images). Conclusion: This work provides a comprehensive evaluation of uncertainty estimation methods and showed that Deep Ensembles outperformed other methods in most cases. Significance: Neural network uncertainty measures can help identify potentially inaccurate segmentations and alert users for manual review.

Tanushree Mitra, Graham Wright, Eric Gilbert

Today, social media provide the means by which billions of people experience news and events happening around the world. However, the absence of traditional journalistic gatekeeping allows information to flow unencumbered through these platforms, often raising questions of veracity and credibility of the reported information. Here we ask: How do the dynamics of collective attention directed toward an event reported on social media vary with its perceived credibility? By examining the first large-scale, systematically tracked credibility database of public Twitter messages (47M messages corresponding to 1,138 real-world events over a span of three months), we established a relationship between the temporal dynamics of events reported on social media and their associated level of credibility judgments. Representing collective attention by the aggregate temporal signatures of an event reportage, we found that the amount of continued attention focused on an event provides information about its associated levels of perceived credibility. Events exhibiting sustained, intermittent bursts of attention were found to be associated with lower levels of perceived credibility. In other words, as more people showed interest during moments of transient collective attention, the associated uncertainty surrounding these events also increased.