PURPOSE: To evaluate on-board digital tomosynthesis (DTS) for patient positioning vs. two-dimensional (2D) radiography and three-dimensional cone beam (CBCT). METHODS AND MATERIALS: A total of 92 image sessions from 9 prostate cancer patients were analyzed. An on-board image set was registered to a corresponding reference image set. Four pairs of image sets were used: digitally reconstructed radiographs vs. on-board orthogonal paired radiographs for the 2D method, coronal-reference DTS vs. on-board coronal DTS for the coronal-DTS method, sagittal-reference DTS vs. on-board sagittal DTS for the sagittal-DTS method, and planning CT vs. CBCT for the CBCT method. The registration results were compared. RESULTS: The systematic errors in all methods were <1 mm/1 degrees . When registering the bony anatomy, the mean vector difference was 0.21 +/- 0.11 cm between 2D and CBCT, 0.11 +/- 0.08 cm between CBCT and coronal DTS, and 0.14 +/- 0.07 cm between CBCT and sagittal DTS. The correlation between CBCT to DTS was stronger (coefficient = 0.92-0.95) than the correlation between 2D and CBCT or DTS (coefficient = 0.81-0.83). When registering the soft tissue, the mean vector difference was 0.18 +/- 0.11 cm between CBCT and coronal DTS and 0.29 +/- 0.17 cm between CBCT and sagittal DTS. The correlation coefficient of CBCT to sagittal DTS and to coronal DTS was 0.84 and 0.92, respectively. CONCLUSION: DTS could provide equivalent results to CBCT when the bony anatomy is used as landmarks for prostate image-guided radiotherapy. For soft tissue-based positioning verification, coronal DTS produced equivalent results to CBCT, but sagittal DTS alone was insufficient. DTS could allow for comparable soft tissue-based target localization with faster scanning time and a lower imaging dose compared with CBCT.