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

TL;DR

UNet++ introduces a nested, densely connected U-Net architecture with deep supervision that significantly improves medical image segmentation accuracy across various tasks.

Contribution

The paper proposes UNet++, a novel nested U-Net architecture with redesigned skip pathways for better semantic feature integration.

Findings

UNet++ outperforms U-Net and wide U-Net in segmentation accuracy.

Achieves an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net.

Demonstrates effectiveness across multiple medical imaging tasks.

Abstract

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.