MDA-Net: Multi-Dimensional Attention-Based Neural Network for 3D Image Segmentation
This work addresses the problem of efficient 3D medical image segmentation for researchers and practitioners, representing an incremental improvement by combining attention mechanisms with existing U-Net architectures.
The paper tackles the challenge of balancing computational efficiency and accuracy in 3D image segmentation by proposing MDA-Net, a multi-dimensional attention network that integrates slice-wise, spatial, and channel-wise attention into a U-Net, resulting in high segmentation accuracy with low computational cost as demonstrated on MICCAI iSeg and IBSR datasets.
Segmenting an entire 3D image often has high computational complexity and requires large memory consumption; by contrast, performing volumetric segmentation in a slice-by-slice manner is efficient but does not fully leverage the 3D data. To address this challenge, we propose a multi-dimensional attention network (MDA-Net) to efficiently integrate slice-wise, spatial, and channel-wise attention into a U-Net based network, which results in high segmentation accuracy with a low computational cost. We evaluate our model on the MICCAI iSeg and IBSR datasets, and the experimental results demonstrate consistent improvements over existing methods.