Thyroid nodules are a common high-incidence disease of the thyroid. Therefore, the precise segmentation of thyroid nodules in ultrasound images is crucial for the clinical auxiliary diagnosis of thyroid diseases. However, the boundary between thyroid nodules and surrounding tissue is indistinct and exhibits high variability, which lead to unsatisfactory segmentation performance. To tackle this issue, we propose the U-shaped segmentation network (GWUNet) combining gated attention and improved wavelet transform. In this network, we introduce a gated attention mechanism module based on Swin-Transformer, which enables the network to fully learn the location information. Furthermore, we design a wavelet transform module based on a frequency domain selection mechanism. This module is capable of simultaneously learning features from both the frequency domain and the spatial domain, thereby enhancing the ability to extract texture features of thyroid nodules. The proposed network (GWUNet) is evaluated on two thyroid nodules datasets, including the hospital-collected dataset and the open dataset DDTI, and it is shown that it achieves better segmentation performance compared to other state-of-the-art networks. The code is available at https://github.com/Shuijing2018/GWUNet.