Abstract: At present, there are common problems in the field of medical image segmentation, such as bloated network models and low automatic segmentation accuracy, which are difficult to apply in clinical diagnosis. To address the above problems, a lightweight pulmonary nodule segmentation model, Mobile-TransUnet is proposed. The Mobile-TransUnet model was mainly divided into three parts, which were composed of an encoder, a Transformer block and a decoder. In the encoder part, the linear bottleneck inverted residual algorithm and the attention mechanism were combined to replace the traditional convolutional blocks (Convolutional Neural Networks, CNNs). In the Transformer part, the extracted feature maps were encoded in position sequences. In the decoder part, bilinear interpolation was used. Experimental results show that Mobile-TransUnet has greatly improved segmentation accuracy and operating efficiency compared with TransUnet using standard convolution with the same number of layers, and is significantly better than most existing pulmonary nodule segmentation algorithms.