Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

Nested U-Net architecture

A nested U-Net architecture is a specialized convolutional neural network framework designed primarily for semantic and medical image segmentation tasks. Extending the standard encoder-decoder structure of the traditional U-Net, this design connects the encoder and decoder stages through a series of nested, densely connected intermediate convolution layers along redesigned skip pathways. These interconnected intermediate pathways bridge the semantic gap between the low-level spatial features from the encoder and the high-level contextual features from the decoder before feature fusion occurs. The architecture also frequently integrates deep supervision across intermediate decoder layers, enabling multi-scale feature aggregation, improved gradient flow during training, and model pruning to balance segmentation accuracy with computational efficiency during inference.

1 item

UNet++: A Nested U-Net Architecture for Medical Image Segmentation

UNet++: A Nested U-Net Architecture for Medical Image Segmentation

Zongwei Zhou, Md Mahfuzur Rahman Siddiquee, Nima Tajbakhsh, Jianming Liang

OrganizationsArizona State University

Why you should read this

Proposes UNet++, a deeply-supervised encoder-decoder architecture with nested, dense skip pathways that reduces the semantic gap between feature representations to achieve superior segmentation performance across diverse medical imaging modalities.

In this paper, we present UNet++, a new, more powerful architecture for medical image segmentation. Our architecture is essentially a deeply-supervised encoder-decoder network where the encoder and decoder sub-networks are connected through a series of nested, dense skip pathways. The re-designed skip pathways aim at reducing the semantic gap between the feature maps of the encoder and decoder sub-networks. We argue that the optimizer would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar. We have evaluated UNet++ in comparison with U-Net and wide U-Net architectures across multiple medical image segmentation tasks: nodule segmentation in the low-dose CT scans of chest, nuclei segmentation in the microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. Our experiments demonstrate that UNet++ with deep supervision achieves an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively.

Added

2026-09-07