Abstract
Even though many semantic segmentation methods exist that are able to perform well on many medical datasets, often, they are not designed for direct use in clinical practice. The two main concerns are generalization to unseen data with a different visual appearance, e.g., images acquired using a different scanner, and efficiency in terms of computation time and required Graphics Processing Unit (GPU) memory. In this work, we employ a multi-organ segmentation model based on the SpatialConfiguration-Net (SCN), which integrates prior knowledge of the spatial configuration among the labelled organs to resolve spurious responses in the network outputs. Furthermore, we modified the architecture of the segmentation model to reduce its memory footprint as much as possible without drastically impacting the quality of the predictions. Lastly, we implemented a minimal inference script for which we optimized both, execution time and required GPU memory
Original language | English |
---|---|
Number of pages | 8 |
Publication status | Published - 2021 |
Event | 24th International Conference on Medical Image Computing and Computer Assisted Intervention: MICCAI 2021 - Virtual, Strasbourg, France Duration: 27 Sept 2021 → 1 Oct 2021 |
Conference
Conference | 24th International Conference on Medical Image Computing and Computer Assisted Intervention |
---|---|
Abbreviated title | MICCAI 2021 |
Country/Territory | France |
City | Strasbourg |
Period | 27/09/21 → 1/10/21 |