School of Safety Science and EngineeringAnhui University of Science and Technology Huainan 232000 China.
School of Computer Science and EngineeringAnhui University of Science and Technology Huainan 232000 China.
IEEE J Transl Eng Health Med. 2023 Jun 27;11:441-450. doi: 10.1109/JTEHM.2023.3289990. eCollection 2023.
In the past few years, U-Net based U-shaped architecture and skip-connections have made incredible progress in the field of medical image segmentation. U-Net achieves good performance in computer vision. However, in the medical image segmentation task, U-Net with over nesting is easy to overfit.
A 2D network structure TransU-Net combining transformer and a lighter weight U-Net is proposed for automatic segmentation of brain tumor magnetic resonance image (MRI).
The light-weight U-Net architecture not only obtains multi-scale information but also reduces redundant feature extraction. Meanwhile, the transformer block embedded in the stacked convolutional layer obtains more global information; the transformer with skip-connection enhances spatial domain information representation. A new multi-scale feature map fusion strategy as a postprocessing method was proposed for better fusing high and low-dimensional spatial information.
Our proposed model TransU-Net achieves better segmentation results, on the BraTS2021 dataset, our method achieves an average dice coefficient of 88.17%; Evaluation on the publicly available MSD dataset, we perform tumor evaluation, we achieve a dice coefficient of 74.69%; in addition to comparing the TransU-Net results are compared with previously proposed 2D segmentation methods.
We propose an automatic medical image segmentation method combining transformers and U-Net, which has good performance and is of clinical importance. The experimental results show that the proposed method outperforms other 2D medical image segmentation methods. : We use the BarTS2021 dataset and the MSD dataset which are publicly available databases. All experiments in this paper are in accordance with medical ethics.
在过去的几年中，基于 U-Net 的 U 形结构和跳跃连接在医学图像分割领域取得了令人瞩目的进展。U-Net 在计算机视觉中表现出色。然而，在医学图像分割任务中，嵌套过深的 U-Net 容易过拟合。
提出了一种结合 Transformer 和更轻量级的 U-Net 的 2D 网络结构 TransU-Net，用于自动分割脑肿瘤磁共振图像（MRI）。
轻量级 U-Net 架构不仅可以获取多尺度信息，还可以减少冗余的特征提取。同时，嵌入堆叠卷积层中的 Transformer 块可以获取更多的全局信息；具有跳跃连接的 Transformer 可以增强空间域信息表示。提出了一种新的多尺度特征图融合策略作为后处理方法，以更好地融合高低维空间信息。
我们提出的模型 TransU-Net 取得了更好的分割结果，在 BraTS2021 数据集上，我们的方法平均 Dice 系数为 88.17%；在公开的 MSD 数据集上进行肿瘤评估，我们的 Dice 系数为 74.69%；除了比较 TransU-Net 的结果外，还与之前提出的 2D 分割方法进行了比较。
我们提出了一种结合 Transformer 和 U-Net 的自动医学图像分割方法，具有良好的性能，具有临床意义。实验结果表明，所提出的方法优于其他 2D 医学图像分割方法。我们使用了公开的 BraTS2021 数据集和 MSD 数据集。本文中的所有实验都符合医学伦理。
