FADA: Fourier-Augmented Dual Attention Architecture for Enhanced Contextual Awareness in Medical Image Segmentation

Medical image segmentation plays a crucial role in ensuring accurate diagnosis and effective treatment strategies. This paper presents FADA, a Fourier-Augmented Dual Attention model, an advanced deep learning framework specifically designed to enhance segmentation performance in medical imaging. By integrating Fourier based transformations with dual attention mechanisms, FADA effectively captures both local and global contextual information, leading to more precise segmentation of anatomical structures. To assess its performance, FADA is compared against conventional segmentation models such as U-Net and Attention U-Net, using two widely recognized medical imaging datasets. U-Net serves as a foundational benchmark due to its established effectiveness while Attention U-Net enhances feature selection through attention mechanisms. Unlike these models, FADA leverages fourier transformations within its attention module, offering a more robust approach to capturing contextual dependencies. Experimental results demonstrate that FADA outperforms both U-Net and Attention U-Net in segmentation accuracy, while also exhibiting greater resilience to noise and improved computational efficiency. These findings highlight the potential of combining fourier analysis with dual attention to advance medical image segmentation, ultimately contributing to better clinical decision-making and treatment planning.