DLUNet: Semi-supervised Learning based Dual-Light UNet for Multi-organ Segmentation

TL;DR

DLUNet introduces a semi-supervised dual-light UNet architecture that leverages both labeled and unlabeled data for efficient multi-organ segmentation in CT scans, reducing computational costs and improving accuracy.

Contribution

It proposes a novel semi-supervised dual-light UNet framework with consistent learning and efficient design for multi-organ segmentation.

Findings

Achieved an average DSC of 0.8718 on validation data.

Reduced computational cost through separable convolution and residual concatenation.

Effective use of unlabeled data improves segmentation performance.

Abstract

The manual ground truth of abdominal multi-organ is labor-intensive. In order to make full use of CT data, we developed a semi-supervised learning based dual-light UNet. In the training phase, it consists of two light UNets, which make full use of label and unlabeled data simultaneously by using consistent-based learning. Moreover, separable convolution and residual concatenation was introduced light UNet to reduce the computational cost. Further, a robust segmentation loss was applied to improve the performance. In the inference phase, only a light UNet is used, which required low time cost and less GPU memory utilization. The average DSC of this method in the validation set is 0.8718. The code is available in https://github.com/laihaoran/Semi-SupervisednnUNet.