Communication Dans Un Congrès Année : 2025

SCONet: Convolutional Occupancy Networks for Multi-Organ Segmentation

Résumé

Convolutional neural networks are the de facto standard for 3D multi-organ segmentation but still exhibit significant limitations, especially regarding their computational cost, with high running time and often prohibitive memory footprint for large 3D volumes. To overcome these limitations, we propose to replace the image voxel grid with a more compact point cloud representation. Recently, in the field of 3D object reconstruction, networks learning implicit functions from an input point cloud, such as Convolutional Occupancy Networks (ConvONet), have proven their good surface representation capabilities. We therefore propose SCONet (Segmentation Convolutional Occupancy Network), a lightweight ConvONet-based network adapted to the specific task of multi-organ segmentation. SCONet takes as input a point cloud extracted from the original volume with a standard contour detection algorithm, and enriches it with geometric and photometric features. Thanks to its ability to query per organ occupancy probabilities for any point in space, SCONet can be used to predict a multi-organ segmentation map at arbitrary resolution. We evaluate our method on an abdominal CT image dataset and compare its performances with those of discrete and implicit baselines. Our implementation is available at https://github.com/maylis-j/SCONet.

Fichier principal
Vignette du fichier
m91361-jouvencel final.pdf (5.49 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05066849 , version 1 (14-05-2025)

Licence

Identifiants

Citer

Maylis Jouvencel, Razmig Kéchichian, Julie Digne, Sébastien Valette. SCONet: Convolutional Occupancy Networks for Multi-Organ Segmentation. 2025 IEEE 22nd International Symposium on Biomedical Imaging (ISBI), Apr 2025, Houston, United States. pp.1-5, ⟨10.1109/ISBI60581.2025.10980745⟩. ⟨hal-05066849⟩
481 Consultations
484 Téléchargements

Altmetric

Partager

  • More