Uncertainty quantification via localized gradients for deep learning-based medical image assessments.

Brayden SchottDmitry PinchukVictor Santoro-FernandesZan KlanecekLuciano RivettiAlison DeatschScott PerlmanYixuan LiRobert Jeraj

Published in: Physics in medicine and biology (2024)

In summary, this work introduces and validates a novel gradient-based UQ method for deep learning-based medical image assessments to enhance user trust when using deployed clinical models.

Keyphrases

deep learning
artificial intelligence
convolutional neural network
healthcare
machine learning
health information
social media