Enhanced fundus image quality assessment using spatial deformable multiscale perception network

Fundus images play a crucial role in diagnosing and monitoring eye diseases. However, their quality is often compromised during acquisition, necessitating effective enhancement algorithms. Evaluating the quality of enhanced fundus images remains challenging. In this study, we propose a Spatial Deformable Multi-Scale Perception Network (SDMP-Net) specifically designed for Fundus Image Quality Assessment (FIQA). By integrating Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), SDMP-Net extracts both local and global features adaptively, capturing fine textures and structural details. Extensive experiments on enhanced fundus image dataset demonstrate that SDMP-Net achieves a Spearman rank-order correlation coefficient of 0.810 and a Pearson linear correlation coefficient of 0.803, outperforming state-of-the-art methods. Our work contributes to enhancing the reliability of fundus image enhancement and facilitating clinical diagnostics.