Attention-Guided U-Net for Cell Nucleus Segmentation in Microscopy Images

Cell nuclei segmentation is a fundamental step in computational pathology and biomedical image analysis, enabling downstream tasks such as disease diagnosis and drug discovery. However, existing DL based approaches often rely on heavy encoders or complex multi-branch designs, leading to large parameter counts and limited practicality in clinical settings. We propose a lightweight encoder–decoder architecture that achieves improved segmentation performance with significantly reduced model complexity. Our custom residual encoder leverages dilated convolutions and Squeeze-and-Excitation (SE) modules to capture rich contextual features, while a Spatial Pyramid Pooling bottleneck enhances multi-scale representation. In the decoder, enhanced attention gates selectively refine skip connections, resulting in sharper boundaries and more robust segmentation under challenging conditions, such as overlapping or irregular nuclei. Extensive experiments on three diverse benchmarks, Blood Cell Segmentation (BCS), Data Science Bowl (DSB), and Nuclei Instance Segmentation (NuInsSeg), demonstrate that the proposed model consistently outperforms or matches State-of-the-art (SOTA) models while using fewer parameters. Specifically, it achieves Dice scores of 0.9817 on BCS, 0.9262 on DSB, and 86.45 on NuInsSeg, surpassing SOTA models in most cases despite much smaller computational footprint. These results establish the proposed model as an effective and practical solution for real-world biomedical segmentation pipelines, offering a favorable balance between accuracy and efficiency.