MLMIA team wins MICCAI K2S MR reconstruction and segmentation challenge!

We are delighted to announce that MLMIA lab members Nikolas Morshuis and Paul Fischer won the MICCAI 2022 K2S Challenge in Singapore! The objective of this challenge was to segment knee structures directly from 8x undersampled 3D MR images. Our team found that optimising directly for segmentation quality not only produces the best segmentations, but also yields very intriguing intermediate MR reconstructions that deviate from the ground-truth contrast properties, but are very sharp and easy to segment.

More information can be found in this Twitter thread.

Machine Learning in Medical Image Analysis
Machine Learning in Medical Image Analysis
Bridging the gap between AI and clinical practice