This repo is the official implementation of "A U-shaped Network based on Convolution Coupled Transformer for Segmentation of Peripheral and Transition Zones in Prostate MRI".
In this work, a U-shaped network based on the convolution coupled Transformer is proposed for segmentation of peripheral and transition zones in prostate MRI, named the convolution coupled Transformer U-Net (CCT-Unet). The convolutional embedding block is first designed for encoding high-resolution input to retain the edge detail of the image. Then the convolution coupled Transformer block is proposed to enhance the ability of local feature extraction and capture long-term correlation that encompass anatomical information. The feature conversion module is also proposed to alleviate the semantic gap in the process of jumping connection.
Model | Dataset | Resolution | #Params | Flops | Pretrain model |
---|---|---|---|---|---|
CCT-Unet | ProstateX | 224x224 | 7.023 G | 27.59 M | model |
einops==0.6.1
numpy==1.23.5
timm==0.6.13
torch==1.12.1
●A U-shaped Network based on Convolution Coupled Transformer for Segmentation of Peripheral and Transition Zones in Prostate MRI, Yifei Yan, Rongzong Liu, Haobo Chen, Limin Zhang, Qi Zhang
●G. Litjens, O. Debats, J. Barentsz, et al., “Computer-aided detection of prostate cancer in MRI,” IEEE Trans. Med. Imaging, vol. 33, no. 5, pp. 1083–1092, 2014.
●Y. Liu, K. Sung, G. Yang, et al., “Automatic Prostate Zonal Segmentation Using Fully Convolutional Network with Feature Pyramid Attention,” IEEE Access, vol. 7, pp. 163626–163632, 2019.