A novel MLP-Mixer and channel attention based network for medical image segmentation

Accurate medical image segmentation is vital for clinical applications, aiding in precise diagnosis and treatment planning. While Multi-Layer Perceptrons (MLPs) were traditionally used for image classification, the emergence of the MLP-Mixer architecture has demonstrated their efficacy beyond classification tasks. In this paper, we introduce a novel network architecture that combines the strengths of MLP-Mixer and channel attention mechanisms, specifically tailored for medical image segmentation. Our proposed mixer-block, integrating the MLP-Mixer with a channel attention module, enhances the network’s capability to focus on relevant features and capture intricate details in medical images. By incorporating this mixer-block into the U-Net architecture, we achieve improved segmentation performance. The proposed architecture outperformed State-of-the-Art (SOTA) methods on BUSI dataset achieving a dice coefficient of 81.45%.