Research on Focal Segmentation of Bone Scan Based on Swin Transformer

Bone scan is widely used in nuclear medicine to help doctors monitor the bone metastases of cancer. In view of the fact that it takes a long time for doctors to search for lesions based on their professional knowledge and experience. And it is easy to cause missed diagnosis and misdiagnosis due to subjective factors. A segmentation method based on deep learning was proposed for bone imaging lesions. Firstly, Swin Transformer is used as decoder to extract feature information in the image. Secondly, feature fusion is carried out step by step through the decoder to reduce the distraction of attention. Meanwhile, the local features of the lesions in the image are emphasized. Experimental results show that the proposed method achieves 97.65%, 77.81% and 35.59% of accuracy, dice coefficient and IOU in the data set constructed in this paper.