There is a range of data types in biomedicines, including images, time series, various omics molecular data, etc. However, the data in this field typically experience problems with sample scarcity and heterogeneity. Addressing the heterogeneity of the data while making the optimal use of limited data is crucial for the biomedical field. The application of transfer learning in the biomedical is developing rapidly. This approach has the potential to address the dissimilarities between the source and target domains. It accomplishes this by acquiring the existing knowledge from the source domain and identifying the shared factors between the two domains. Depending on the source and target domains and their respective tasks, transfer learning can be used for three different application scenarios. In this review, we discuss the definition of transfer learning. Furthermore, we present several different application examples, which depict the insights of transfer learning applied in biomedicine related to varying scenarios.