Renyang Gu1, Michela Antonelli1, Pritesh Mehta 2, Ashik Amlani 3, Adrian Green3, Radhouene Neji 4, Sebastien Ourselin1, Isabel Dregely1, and Vicky Goh1
1School of Biomedical Engineering & Imaging Sciences, King's College London, London, United Kingdom, 2Biomedical Engineering and Medical Physics, University College London, London, United Kingdom, 3Radiology, Guy’s and St Thomas’ Hospitals, London, United Kingdom, 4Siemens Healthcare Limited, Frimley, United Kingdom
1School of Biomedical Engineering & Imaging Sciences, King's College London, London, United Kingdom, 2Biomedical Engineering and Medical Physics, University College London, London, United Kingdom, 3Radiology, Guy’s and St Thomas’ Hospitals, London, United Kingdom, 4Siemens Healthcare Limited, Frimley, United Kingdom
An uncertainty-aware 2D U-Net for skeletal marrow segmentation compensated for noisy ground-truth labels and improved network performance, particularly for vertebral and pelvic segmentation.
Figure 1. uU-Net architecture. Conv: convolution. IN: instance normalization. ReLU: rectified Linear Unit.
Figure 3. Representative pelvic MR images of three patients with generated overlays.
