1. Learning to Jointly Segment the Liver, Lesions and Vessels from Partially Annotated Datasets
- Author
-
Ali, Omar, Bone, Alexandre, Rohe, Marc-Michel, Vibert, Eric, Vignon-Clementel, Irene, Laboratoire Guerbet / Guerbet Research, Hôpital Paul Brousse, Université Paris-Saclay, SImulations en Médecine, BIOtechnologie et ToXicologie de systèmes multicellulaires (SIMBIOTX ), Inria Saclay - Ile de France, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Physiopathogénèse et Traitement des Maladies du Foie, Hôpital Paul Brousse-Université Paris-Saclay, and AP-HP. Université Paris Saclay
- Subjects
partially labeled data ,weighted loss function ,multi-task learning ,Semantic segmentation ,[SPI.SIGNAL]Engineering Sciences [physics]/Signal and Image processing - Abstract
International audience; The segmentation of the liver, lesions, and vessels from pre-opera-tive CT scans is of major importance in hepatic surgery planning. However, large databases with reference segmentations for these regions of interest remain unavailable, a challenge often encountered in medical image segmentation. In this work, we propose the FuSeloss, a novel loss function for multi-task learning on datasets with partial annotations. By employing the nnU-Net’s 3D self-configur-ing pipeline to calibrate and train a deep network for the joint segmentation of the liver, lesions, and vessels, we show how the FuSeloss allows to learn from the differently annotated IRCAD and LiTS datasets, improving the overall baseline segmentation performance. With the FuSe loss, the dice scores reached up to 95.9%, 70.6% and 60.0% for the liver, lesions, and vessels respectively.
- Published
- 2022
- Full Text
- View/download PDF