TY - GEN

T1 - NUDF

T2 - 19th IEEE International Symposium on Biomedical Imaging, ISBI 2022

AU - Sorensen, Kristine

AU - Camara, Oscar

AU - Backer, Ole De

AU - Kofoed, Klaus F.

AU - Paulsen, Rasmus R.

N1 - Funding Information: This work was supported by a PhD grant from the Technical University of Denmark - Department of Applied Mathematics and Computer Science (DTU Compute) and the Spanish Ministry of Science, Innovation and Universities under the Retos I+D Programme (RTI2018-101193-B-I00). Publisher Copyright: © 2022 IEEE.

PY - 2022

Y1 - 2022

N2 - Medical image segmentation is often considered as the task of labelling each pixel or voxel as being inside or outside a given anatomy. Processing the images at their original size and resolution often result in insuperable memory requirements, but downsampling the images leads to a loss of important details. Instead of aiming to represent a smooth and continuous surface in a binary voxel-grid, we propose to learn a Neural Unsigned Distance Field (NUDF) directly from the image. The small memory requirements of NUDF allow for high resolution processing, while the continuous nature of the distance field allows us to create high resolution 3D mesh models of shapes of any topology (i.e. open surfaces). We evaluate our method on the task of left atrial appendage (LAA) segmentation from Computed Tomography (CT) images. The LAA is a complex and highly variable shape, being thus difficult to represent with traditional segmentation methods using discrete labelmaps. With our proposed method, we are able to predict 3D mesh models that capture the details of the LAA and achieve accuracy in the order of the voxel spacing in the CT images.

AB - Medical image segmentation is often considered as the task of labelling each pixel or voxel as being inside or outside a given anatomy. Processing the images at their original size and resolution often result in insuperable memory requirements, but downsampling the images leads to a loss of important details. Instead of aiming to represent a smooth and continuous surface in a binary voxel-grid, we propose to learn a Neural Unsigned Distance Field (NUDF) directly from the image. The small memory requirements of NUDF allow for high resolution processing, while the continuous nature of the distance field allows us to create high resolution 3D mesh models of shapes of any topology (i.e. open surfaces). We evaluate our method on the task of left atrial appendage (LAA) segmentation from Computed Tomography (CT) images. The LAA is a complex and highly variable shape, being thus difficult to represent with traditional segmentation methods using discrete labelmaps. With our proposed method, we are able to predict 3D mesh models that capture the details of the LAA and achieve accuracy in the order of the voxel spacing in the CT images.

KW - computed tomography

KW - image segmentation

KW - left atrial appendage

KW - mesh modelling

KW - Unsigned distance fields

U2 - 10.1109/ISBI52829.2022.9761610

DO - 10.1109/ISBI52829.2022.9761610

M3 - Article in proceedings

AN - SCOPUS:85129621349

T3 - Proceedings - International Symposium on Biomedical Imaging

BT - ISBI 2022 - Proceedings

PB - IEEE Computer Society Press

Y2 - 28 March 2022 through 31 March 2022

ER -