Three dimensional arrangement of proteins inside the axoneme is essential information to elucidate the flagellar/ciliary bending mechanism. Cryo-electron tomography provides structural information in situ at ~30 Å resolution and thus is a suitable method for flagella/cilia research. The biggest challenge in cryo-electron tomography is the low signal-to-noise ratio. Therefore, subtomogram averaging, an in silico processing step to detect identical structural units based on 3D image analysis, aligns them three dimensionally and average is required to improve the signal-to-noise ratio. When structural units are not exactly identical, they must be classified. In this chapter, we describe our strategy to extract, align, classify, and average molecular structures from cryo-electron tomograms of flagella/cilia, utilizing longitudinal periodicity and pseudo-ninefold symmetry in the axoneme.