Published Paper Details:

Deep learning architectures have transformed biomedical image segmentation, enhancing accuracy and efficiency in medical diagnostics. This research work presents RES-UNET, integrating ResNet50v2 and UNet for intricate pattern recognition in medical imaging. ResNet50v2 as the encoder captures hierarchical features, while UNet's decoder reconstructs high-resolution segmentation masks with preserved spatial details. Feature Pyramid Networks (FPN) enrich multi-scale feature fusion, enhancing segmentation accuracy. A hybrid loss function combining counter-aware, focal, and generalized dice losses optimizes model robustness. Experimental results demonstrate RES-UNET achieves 92% Dice Similarity Coefficient, 86% Jaccard Index, 94% sensitivity, and 93% specificity, surpassing traditional methods. RES-UNET shows promise for precise biomedical image analysis, offering significant advancements in clinical diagnostics and treatment planning.