PapersBenchmarks
University College London NHS Trust
05 Feb 2024
computer-sciencecomputer-vision-and-pattern-recognitionmachine-learning

CLADE: Cycle Loss Augmented Degradation Enhancement for Unpaired Super-Resolution of Anisotropic Medical Images

UCLQueen Mary University of London logoQueen Mary University of LondonUniversity College London NHS TrustFrimley Health NHS Trust
Researchers at UCL developed CLADE, an unsupervised deep learning method that super-resolves anisotropic medical images by learning high-resolution features directly from the inherent anisotropy within the input volume, without needing paired training data. CLADE, which employs a modified CycleGAN with weight demodulation and a cycle-consistent gradient mapping loss, consistently outperformed state-of-the-art self-supervised methods in quantitative and qualitative image quality on abdominal MRI and CT.
There are no more papers matching your filters at the moment.