Single-cell RNA-sequencing has been widely used to investigate the cell type composition of different tissues. Incorporating spatial information of individual cells can offer deeper insights into the mechanisms that govern cell differentiation and potential cell-cell interactions. The imaging-based spatial transcriptomics technology makes it possible to obtain the transcriptomics profiles and spatial information of single cells simultaneously. However, current imaging-based technologies are limited to profiling several hundred of genes, which makes it hard to perform analyses that are commonly done for single-cell RNA-sequencing data such as cell type annotation. Annotating individual cells is a key step to understand the cell type composition of a specific tissue of interest, and it is also important for downstream analyses of spatial data such as differential expression analysis and neighbourhood enrichment analysis. Here, we compared six different cell type annotation methods using publicly available 10x Xenium data of human HER2+ breast cancer samples. These six methods include one manual annotation method based on marker genes, four reference-based cell type prediction methods designed for single-cell RNA-sequencing data, and one reference-based method designed for sequencing-based spot-level spatial transcriptomics data. We compare and discuss the advantages and disadvantages of all six methods in terms of their computational performances.