Optimizing Medical Image Segmentation with Advanced Decoder Design

TL;DR

This paper introduces Swin DER, an improved decoder design for medical image segmentation that enhances detail preservation and accuracy by optimizing upsampling, skip connections, and feature extraction modules, outperforming existing methods.

Contribution

It proposes a novel decoder architecture, Swin DER, with specific enhancements like learnable upsampling, attention-based skip connections, and deformable convolution, addressing the decoder's role in segmentation performance.

Findings

Swin DER outperforms state-of-the-art methods on Synapse and MSD brain tumor datasets.

The optimized decoder components improve segmentation detail and accuracy.

The proposed methods demonstrate significant performance gains in medical image segmentation.

Abstract

U-Net is widely used in medical image segmentation due to its simple and flexible architecture design. To address the challenges of scale and complexity in medical tasks, several variants of U-Net have been proposed. In particular, methods based on Vision Transformer (ViT), represented by Swin UNETR, have gained widespread attention in recent years. However, these improvements often focus on the encoder, overlooking the crucial role of the decoder in optimizing segmentation details. This design imbalance limits the potential for further enhancing segmentation performance. To address this issue, we analyze the roles of various decoder components, including upsampling method, skip connection, and feature extraction module, as well as the shortcomings of existing methods. Consequently, we propose Swin DER (i.e., Swin UNETR Decoder Enhanced and Refined) by specifically optimizing the design of these three components. Swin DER performs upsampling using learnable interpolation algorithm called offset coordinate neighborhood weighted up sampling (Onsampling) and replaces traditional skip connection with spatial-channel parallel attention gate (SCP AG). Additionally, Swin DER introduces deformable convolution along with attention mechanism in the feature extraction module of the decoder. Our model design achieves excellent results, surpassing other state-of-the-art methods on both the Synapse and the MSD brain tumor segmentation task. Code is available at: https://github.com/WillBeanYang/Swin-DER