RESUMO
Any clinically-deployed image-processing pipeline must be robust to the full range of inputs it may be presented with. One popular approach to this challenge is to develop predictive models that can provide a measure of their uncertainty. Another approach is to use generative modelling to quantify the likelihood of inputs. Inputs with a low enough likelihood are deemed to be out-of-distribution and are not presented to the downstream predictive model. In this work, we evaluate several approaches to segmentation with uncertainty for the task of segmenting bleeds in 3D CT of the head. We show that these models can fail catastrophically when operating in the far out-of-distribution domain, often providing predictions that are both highly confident and wrong. We propose to instead perform out-of-distribution detection using the Latent Transformer Model: a VQ-GAN is used to provide a highly compressed latent representation of the input volume, and a transformer is then used to estimate the likelihood of this compressed representation of the input. We demonstrate this approach can identify images that are both far- and near- out-of-distribution, as well as provide spatial maps that highlight the regions considered to be out-of-distribution. Furthermore, we find a strong relationship between an image's likelihood and the quality of a model's segmentation on it, demonstrating that this approach is viable for filtering out unsuitable images.
Assuntos
Processamento de Imagem Assistida por Computador , Humanos , Probabilidade , IncertezaRESUMO
A fast, subcortical pathway to the amygdala is thought to have evolved to enable rapid detection of threat. This pathway's existence is fundamental for understanding nonconscious emotional responses, but has been challenged as a result of a lack of evidence for short-latency fear-related responses in primate amygdala, including humans. We recorded human intracranial electrophysiological data and found fast amygdala responses, beginning 74-ms post-stimulus onset, to fearful, but not neutral or happy, facial expressions. These responses had considerably shorter latency than fear responses that we observed in visual cortex. Notably, fast amygdala responses were limited to low spatial frequency components of fearful faces, as predicted by magnocellular inputs to amygdala. Furthermore, fast amygdala responses were not evoked by photographs of arousing scenes, which is indicative of selective early reactivity to socially relevant visual information conveyed by fearful faces. These data therefore support the existence of a phylogenetically old subcortical pathway providing fast, but coarse, threat-related signals to human amygdala.