Gregor Köhler

TL;DR: MedNeXt as discussed by the authors is a Transformer-inspired large kernel segmentation network for medical image segmentation, which uses a 3D Encoder-Decoder network to preserve semantic richness across scales.

TL;DR: The Medical-Out-Of-Distribution-Analysis-Challenge (MOOD) is introduced as an open, fair, and unbiased benchmark for OoD methods in the medical imaging domain and shows that performance has a strong positive correlation with the perceived difficulty, and that all algorithms show a high variance for different anomalies, making it yet hard to recommend them for clinical practice.

TL;DR: In this article, a hierarchical multi-scale representation encoding including a spatio-temporal attention mechanism is proposed to learn a learned growth model that can be conditioned on an arbitrary number of observations, and produce a distribution of temporally consistent growth trajectories on a continuous time axis.