2.5D U-Net for Abdominal Multi-organ Segmentation

Accurate and efficient segmentation of multiple abdominal organs from medical images is crucial for clinical applications such as disease diagnosis and treatment planning. In this paper, we propose a novel approach for abdominal organ segmentation using the U-Net architecture. Our method addresses the challenges posed by anatomical variations and the proximity of organs in the abdominal region. To improve the segmentation accuracy, we introduce an attention mechanism into the U-Net architecture. This mechanism allows the network to focus on salient regions and suppress irrelevant background regions, enhancing the overall segmentation performance. Additionally, we incorporate 3D information by connecting three consecutive slices as 3-dimensional inputs. This enables us to exploit the spatial context across the slices while minimizing the increase in GPU memory usage. We evaluate our proposed method on the MICCAI FLARE 2023 validation dataset, the mean DSC is 0.3683 and the mean NSD is 0.3668.