Attention U-Net: Learning Where to Look for the Pancreas
This addresses medical image segmentation for clinicians by providing a more efficient and accurate method, though it is incremental as it builds on existing U-Net architectures.
The authors tackled the problem of segmenting abdominal organs in CT scans by introducing attention gates that automatically focus on relevant structures, which improved U-Net's segmentation accuracy across datasets without adding significant computational cost.
We propose a novel attention gate (AG) model for medical imaging that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn to suppress irrelevant regions in an input image while highlighting salient features useful for a specific task. This enables us to eliminate the necessity of using explicit external tissue/organ localisation modules of cascaded convolutional neural networks (CNNs). AGs can be easily integrated into standard CNN architectures such as the U-Net model with minimal computational overhead while increasing the model sensitivity and prediction accuracy. The proposed Attention U-Net architecture is evaluated on two large CT abdominal datasets for multi-class image segmentation. Experimental results show that AGs consistently improve the prediction performance of U-Net across different datasets and training sizes while preserving computational efficiency. The code for the proposed architecture is publicly available.