Deep learning is revolutionizing the field of medical image segmentation. The U-shaped (Unet) model, with its encoder-decoder architecture and skip connections, has become the dominant architecture for this task. However, variations like Nested-Unet are necessary to address the challenge of adapting network size for datasets of varying scales. This study proposes a novel model based on a multilayer Unet architecture that leverages all output features within nodes of a Nested Unet network. We evaluate the proposed model on four key medical image datasets: electron microscopy (EM), nuclei, polyps (CVC-ClinicDB), and skin lesions. The model achieves impressive Dice scores: 94.08% for EM, 92.68% for nuclei, 87.20% for skin lesions, and 93.02% for the CVC-ClinicDB dataset. These outstanding results demonstrate the superior performance of our proposed model compared to recent advancements in the field. Remarkably, the model achieves this performance while maintaining a size that is half that of a traditional Unet model.